PATIENT SAFETY IN THE ACUTE HEALTHCARE CHAIN

PATIENT SAFETY IN THE ACUTE HEALTHCARE CHAIN

I s i t s a f e r @ h o m e ?

Is it safer@hom

e?

L o u i s e S . v a n G a l e n

Louise S. van Galen

PATIENT SAFETY IN THE ACUTE HEALTHCARE CHAIN

PATIENT SAFETY IN THE ACUTE HEALTHCARE CHAIN

I s i t s a f e r @ h o m e ?

Is it safer@hom

e?

L o u i s e S . v a n G a l e n

Louise S. van Galen

Cover/layout: Off Page / P.P.P. Bos

Printed by: Off Page, Amsterdam

ISBN: 978-94-6182-799-9

Copyright © L.S. van Galen

VRIJE UNIVERSITEIT

Patient Safety in the Acute Healthcare Chain: Is it Safer@home?

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad Doctor aan

de Vrije Universiteit Amsterdam,

op gezag van de rector magnificus

prof.dr. V. Subramaniam,

in het openbaar te verdedigen

ten overstaan van de promotiecommissie

van de Faculteit der Geneeskunde

op dinsdag 27 juni 2017 om 15.45 uur

in de aula van de universiteit,

De Boelelaan 1105

door

Louise Sandra van Galen

geboren te Hengelo

promotor: prof.dr. M.H.H. Kramer

copromotor: dr. P.W.B. Nanayakkara

Leescommissie:

dr. J. Car

prof.dr. M.E. Holland

prof.dr. H.A.H. Kaasjager

prof.dr. J. Klein

prof.dr. D.A. Legemate

prof.dr. C. Wagner

TABLE OF CONTENTS

Chapter 1 Introduction and general outline of the thesis 9

The starting point of the hospital care chain: novel ways of organizing care

Chapter 2 Acute medical units: The way to go? A literature review 27Eur J Intern Med 2017;39:24-31

Chapter 3 Measurement of generic patient reported outcome 47 measures (PROMs) in an acute admission unit: A feasibility study

Acute Med 2016;15(1):13-9

Prevention of serious adverse events on the clinical wards

Chapter 4 Delayed recognition of deterioration of patients in general wards is 61 mostly caused by human related monitoring failures: A root cause analysis of unplanned ICU admissions

PLoS One 2016;11(8):e0161393

Chapter 5 A protocolised once a day modified early warning score (MEWS) 79 measurement is an appropriate screening tool for major adverse events in a general hospital population

PLoS One 2016;11(8):e0160811

The use of quality indicators to assess patient safety

Chapter 6 Exploring the preventable causes of unplanned readmissions using 93 root cause analysis: Coordination of care is the weakest link

Eur J Intern Med 2016;30:18-24

Chapter 7 Hospital readmissions: A reliable quality indicator? 111Ned Tijdschr Geneeskd 2016;160:A9885

Chapter 8 Physician consensus on preventability and predictability of 117 readmissions based on standard case scenarios

Neth J Med 2016;74(10):434-42

Chapter 9 Patients’ and providers’ perceptions of the preventability of 131 hospital readmission: A prospective, observational study in 4 European countries

BMJ Qual Saf. 2017; In press

Chapter 10 Hospital Standardized Mortality Ratio: 153 A reliable indicator of quality of care?

Submitted

Chapter 11 Summary of main results, general discussion and future perspectives 165

Chapter 12 Dutch summary 181

Appendix List of publications 193List of scientific presentations 194Author affiliations 195Word of thanks 198Curriculum vitae / biography 200

CHAPTER 1

INTRODUCTION AND GENERAL OUTLINE OF THE THESIS

‘Primus non nocere (in the first place, do not harm)’ Hippocrates

11

INTRODUCTIONThe (acute) healthcare chains in Europe are increasingly under pressure due to the rise in demand

for care caused by the change in demographics (aging populations) and policy changes in

the healthcare system. This growing demand for healthcare is also reflected by the expenditure

in healthcare in the Netherlands. Currently, 90 billion euros are spend yearly on healthcare,

this is 40% more than in 2005.1 While emergency presentations have decreased, emergency

department (ED) admissions have risen greatly in the last decades, and the adjusted mean

length of hospital stay in the Netherlands has also increased by 8.5%.2,3 The ageing patient

often has more comorbidities and uses more medicine when compared to a younger patient.

In addition, due to overloaded general practitioners, budget cuts for nursing homes and

the introduction of obligatory deductible excess these patients tend to stay at home longer

without receiving the care they need. Because of this, they eventually present themselves

to the hospital with more complicated problems which are often not straightforward and

therefore require more diagnostics and therapy.4,5 This leads to congestion in the whole care

chain with resultant negative implications for the patients. As a result, the 1,1 million people

employed in the healthcare field in the Netherlands face a daily burden trying to make patient

flow as effective as possible.6 Healthcare workers in hospitals are constantly struggling to

find the balance between increasing number of ED admissions and a relative shortage of

beds. Furthermore, general practitioners (GPs) and primary care facilities such as home nurses

are also overloaded trying to organize appropriate care at home using the limited resources

available. This accumulation in medical teams’ workload in and outside the hospital results in

problems in communication and coordination within the acute healthcare chain. Inadequate

care coordination and poor teamwork take place both across sites of care as between providers7

and can result in poor handovers and inadequate teamwork.8,9 These issues are relevant to

focus on since they may lead to (serious) adverse events and complications, endangering

patient safety in hospitals.8,10 It is known, that in every ten patients, one is still harmed while

receiving hospital care.11

THE HEALTHCARE SYSTEM IN THE NETHERLANDS The Dutch healthcare system often receives positive appraisal. Its overall ranking on healthcare

quality, access and efficiency compared to other countries is high.12,13 The philosophy of

the system is based on some universal principles: ‘access to care for all, solidarity through

medical insurance (which is compulsory and available to all) and high-quality health services.’6

It is divided into three compartments: long-term care for chronic conditions, supplementary

care (i.e. dental care and physiotherapy) and basic and essential medical care (general

practitioner (GP), visits to hospital admissions and specialist appointments or procedures). This

last compartment consists of ‘planned’ care and ‘unplanned acute’ care.

Managed competition in healthcare market

The system has undergone some changes in the last decades. Since 2000, the government has

attempted to promote efficiency through ‘managed competition’, by introducing healthcare

insurers that consumers can choose from. These insurance companies are monitored by

12

the government but operate independently. The Netherlands has 9 insurance companies, of

which the four largest share 90% of the total.14 In the last decade, the increased amount for

obligatory deductible excess has led to a restriction to care.14,15 This ‘own risk’ is the amount

paid for covered health services before the insurance plan starts to pay. As a result, this

increase has unleashed a cascade effect on the insured, since the amount of supplementary

insurance has steadily decreased in the last 10 years.14 This financial burden for patients results

in less visits to a physician when needed, as well as not undergoing recommended diagnostics

or treatment. This causes a patient population that is more severely ill than needed if these

patients had timely consulted a doctor, since reasonably ‘simple’ problems could potentially

evolve in complicated syndromes.16

Patient flow through the healthcare chain (see infographic inside front cover)

Most of the times the patient flows start with the GP, the GP is seen as the gate-keeper of

the Dutch healthcare system. The Dutch GP-system consists of about 5000 GP offices and is

originally a well-designed system. Patients requiring hospital care mostly need a GP-referral

before they are seen by specialists. An exception is the patients who live in nursing homes

and are cared for by a nursing home physician or geriatric specialist on location. On average,

a patient contacts his GP 4.4 times a year, children from 5-17 years visit their GP 2.3 times,

patients above 85 visit 13.2 times.17 When a patient falls ill he usually visits the GP first, if

the GP assesses the patient as needing acute care he will refer the patient to the ED, mostly

after contacting emergency/acute physicians in the hospital for referral. If the problem is not

acute, the patient might be send home with a therapy initiated by the GP, or when the patient

needs non-acute specialized care he is referred to a specialist outpatient clinic.

If a patient becomes acutely ill without having time or the physical capability to visit the GP,

an ambulance may be called upon. This could for example be the case in an out-of-hospital

cardiac arrest on the streets or a severely septic patient who is in immediate need of life support

at home. Patients in the ED are mostly seen by emergency doctors or residents of the specialty

they were referred for. From the ED, patients can be sent home for problems which do not

require admission. Sometimes it is judged that better care can be provided in another hospital,

or in case of a social problem transfer to a nursing home might be most suitable. If a patient

does need in-hospital diagnostics and/or treatment he/she will be admitted. This can be either

to an acute medical unit (AMU) or directly on to a (specialist) clinical ward. On these wards

the patient is cared for by medical specialists, nurses, and other supportive healthcare providers

such as occupational therapists and geriatric nurses. If patients are severely ill, they might

also directly be admitted to an intensive care unit (ICU) or a unit likewise. During admission,

patients may revive rapidly and can be discharged soon, either from the AMU or from another

clinical ward. However, if patients deteriorate during their admission on a regular ward, an

unplanned ICU admission or even death might be inevitable.

At discharge from hospital patients are often given a follow-up appointment and their GP has

to receive a discharge letter from a hospital physician in order to be updated on the situation

13

of his patient and the course of events during hospital admission. Not all patients can be

discharged to their pre-existent (home) situation. Sometimes a transfer to another hospital,

a nursing home or a rehabilitation centre might be required temporarily, or in some cases

transfer to a nursing home permanently. Also, extra care such as professional ADL-support

might sometimes be needed when sending the patient back to their previous home situation.

Unfortunately, some discharged patients are readmitted unexpectedly to hospital, with

readmission rates varying from 10-30% internationally.18 Reasons for these readmissions can

for example be recurrence of or unresolved earlier complaints, or lack of social support in

the (new) home situation.

The text below describes an example of a patient moving through hospital and the difficulties

faced within the chain, which was used to illustrate the problem in a recent Dutch article.19

An 80-year old woman fell at home during the evening and was not able to get up by herself. The next morning a homecare nurse found her on the floor. In the past years, this patient hardly visited her general practitioner. She had suffered from some cognitive deterioration, but with the support of her family (two sons), current home situation was still acceptable. At the time of her fall, her sons were on vacation.

The home-care nurse called the GP immediately. He visited the patient and after seeing her he wanted to refer her to the hospital for further diagnostics and treatment. The ambulance services were busily engaged and after waiting for hours an ambulance picked her up at 5.15 pm. The GP had already left then and could not provide a face-to-face handover. After patient arrived at the overcrowded ED she was seen firstly by a medical student (an intern), followed by an emergency physician, and lastly by an internal medicine resident. The patient was diagnosed with delirium and had a low sodium, probably caused by a so called ‘tea and toast’-diet. Patient was unable to recall the exact medication she currently used. Initially, there were no hospital beds available to admit patient to hospital. Finally, at 11.30 pm a bed became available and patient was admitted. After a few days in hospital, patient recovered quickly and wished to go back home. She reported to the ward nurses that things at home were going well before the admission, and she did not need any additional help. This was a relief for the (transfer) nurses, since the originally intended nursing home did not have a place at that moment for her anyway. She was discharged to her initial home situation. It was however agreed that extra professional home care would be needed and provided twice a week. In addition, the GP’s practice nurse was advised to visit her occasionally.

A few days later, patient was readmitted to hospital, again in a confused state with low sodium. This time it was decided, in agreement with the GP among others, to move patient to a nursing home. Patients was discharged and she was advised not to continue with her diuretics because of the recurring hyponatremia. A week after discharge patient presented once again to the ED. Patient was found in bed, unresponsive. The letter from the ward doctor with advice on fluid restriction and discontinuation of diuretics was sent to the GP in time, but had not reached the nursing home doctor.

CHANGING DEMOGRAPHY The proportion of patients of 65 year and older is rising, in 2017 18% of the Dutch population is

over 65 years. In 2050, it is estimated this percentage will rise to 26%.8,20 The healthcare system

is forced to adapt to this development since preventable adverse events occur significantly

more often in older patients.21 In addition, older patients tend to have multiple comorbidities

and polypharmacy and often have atypical clinical presentations leading to more diagnostics

and as a result 75% of these patients spend too long in the emergency department (ED).5,22

14

The high workload caused by this increased patient flow and a decrease of human resources

due to budget cuts leads to the expense of time for patients. This is important especially in

this older population where communication plays an important role in the treatment. Patient’s

expectations and wishes should be spoken of thoroughly in order to prevent overtreatment

and unnecessary diagnostics.16 Lack of proper communication can have an effect on patient

safety sometimes even resulting in an unwanted ICU-admission or CPR in the elderly.23,24

Another problem faced by the Dutch healthcare system in the care for elderly is the recent

governmental budget cuts on chronic care. In 2012 the Netherlands spend three times more

on chronic care than Germany, and two times more than France, and the peak of the aging

population was yet to be reached.25 These cuts resulted in a decrease in numbers of nursing

homes available. To date, this gap in needed care has not been adequately taken over by

homecare nurses since the organization of home care by the local authorities to compensate

has fallen behind. Because of this, professional help is getting more scarce and the home-

caregivers are put under pressure to provide the necessary care and are forced to take

responsibility for the care of the patient. Ideally, general practitioners should provide care

for these patients. However, it is often difficult for a GP, who takes care of 2000 patient on

average, to manage this complex care at home for probably more than one patient. Since there

is no professional support for these patients, many patients remain at home until the situation

is unsustainable and are ultimately admitted to the ED, often quite severely ill. Older patients

are admitted more often than their younger equals and when admitted they often stay longer

in hospital than required. This results in stagnation of patient flow within the chain.19

PATIENT SAFETY As previously stated, the increasing burden on the acute healthcare chain can potentially lead to

serious adverse events within the chain.26 Iatrogenic harm was already described by Hippocrates

2400 years ago (‘iatros’=doctor, ‘genesis’=origin), but in the last decades the importance of

patient safety has been increasingly emphasized.8 Adverse events bring unintended, potentially

preventable care-related harm, and can compromise patient safety. Patient safety is defined as:

‘preventing errors and adverse effects to patients associated with health care’.11 These serious

advents such as in-hospital mortality, unplanned readmissions, and ICU admissions are also

often the result of a lack in adequate communication and coordination between professionals,

poor handover and inefficient teamwork.8,9,27,28 It is reported that patients can be handed

over 15 times in a 5-day hospitalization, and individual doctors participate in more than 3000

handovers per month.29 In addition, medical teams that are overloaded tend to perform poor

on patient safety outcomes.30

Since the first report ‘To err is human’ of the Institute of Medicine in 1999 which revealed

that 7-13% of all health-related harm contributes to the death of a patient, many countries

have started similar investigations.8,31 In 2004 the WHO started the Patient Safety Program,

and today 140 countries have worked to address challenges of unsafe care. In developed

countries it is estimated that 1 in ten patients (8-12%) is harmed while receiving hospital

15

care.11 This rate has stagnated and has not changed for more than 50 years.32 Statistics have

exposed that solely reducing the rate of adverse events in the European Union would prevent

more than 750.000 harm-inflicting errors per year leading to over 3.2 million fewer days of

hospitalization, 260.000 fewer incidents of permanent disability and 95000 fewer deaths per

year, and thus providing a relief of the current pressure on the healthcare system.11 It must

be remarked that most failures do not arise from lack of knowledge or inadequate skills of

the physicians, but rather from problems in the systems of care in place at time.33

Patient safety in the Netherlands

In the Netherlands, patient safety has also been an important subject of attention. EMGO-NIVEL

has investigated triggers for care-related harm since 2004. Due to multiple interventions and

improvement strategies such as implementing quality indicators34 (unplanned readmissions,

unintended long length of stay, HSMR), checklists35 (SBAR), and new monitoring instruments36

(early warning scores) the incidence of potentially preventable death was reduced from 5.5%

in 2008 to 2.6% in 2012.37

However, not all these new implementations are found beneficial within the healthcare

chain. An unintended result of these implementations is the resistance it causes among

healthcare workers on the frontline, who find most measures time-consuming, exhaustive, too

administrative, and unconnected to their actual work on the floor.32 They claim the government

is battling the complexity of hospital organizations with even more complex quality and safety

tools. As a result attention has shifted from care for the patient to care for the system.38

The Dutch Hospital Association has revealed that the number of health indicators is currently

up to more than 3400.8,38 In other European countries some of these indicators, such as

unplanned 30-day readmissions, are already used as a quality tool to rate and reimburse

hospitals, and financial penalties may also be attached.18,39 The intensive bureaucracy also has

resulted in nurses spending 60-70% of their time performing administrative actions, of which

they consider more than half as useless.38,40

Current situation in patient safety

In order to avoid harm to patients current caveats in improving patient safety must

be addressed:

→ Specialization and sub specialization within hospitals result in multiple

healthcare providers taking care of an individual patient, who only feel responsible

for certain aspects of the patient, rather than the patient as a whole. Due to this,

patients sometimes see from 50-200 healthcare workers during their treatment.8

Since these groups of professionals are more divided than ever, but still deal with

the same patient, more handovers are needed to get all required information to

other healthcare workers. More handovers weakens appropriate communication

and coordination.9 Handovers between shifts and disciplines should therefore be

a key element for appropriate improvement.28

16

→ Fragmentation in healthcare processes (the GP’s office, the ED, the AMU,

the ward, the ICU, transfer locations, nursing home etc.) results in inefficient and

discontinuity in patient flow since the transfer between these fragments is not

appropriately synchronized. Interdisciplinary interventions to streamline the patient

flow between all these partners, may reduce the incidence of adverse events.28,41

→ Hospitals are often trying to assess patient safety issues independently, and are

learning too little from each other’s work. Hospital partnerships could for example

play a critical role for technical-knowledge exchange between healthcare workers.

Also, co-development of solutions might result in rapidly evolving efficient global

health systems.42

→ The rise of electronic help devices and E-health to perform more care outside

the hospitals warrants adequate and smooth implementation, without impairing

patient safety.

→ Since most impulses improving patient safety have been external impulses, initiated

by supra-professional and regulatory bodies for example by the government or

the inspection of healthcare, they have caused an insufficient bureaucratic system,

which removes/takes away the attention from the real problems in the primary

process on the workfloor.8,28 These politicians, regulators, policy makers among

others, are too far from the clinical frontline, basing their efforts and solutions on

what they imagine everyday clinical work to be. However, these insights are all

based on second or third-hand accounts of how work on the frontline is actually

done, relying on data often analysed with delays.32 Efficient patient flow providing

a safer environment should be investigated from a practical and healthcare worker

point of view in order to find connection with the safety issues which are faced in

the field on a daily basis. Research on healthcare systems should focus on systems

‘that are real rather than ideal’.8,32

→ A culture change providing more patient participation, attitude changes in

professionals and more value-based leadership is needed to shift the attention

on patient safety from extrinsic to intrinsic motivation. Intrinsic motivation entails

activities undertaken because of internal motivation and provide immediate effect

after the action itself whereas extrinsic motivation makes people undertake actions

to get a reward or to dodge a penalty from an external party, without actually

being interested its immediate potential effect. High levels of ward’s shared values,

beliefs and behaviours plus an individual’s perception of culture are associated with

organization-wide reductions in adverse events.43

→ The focus in patient safety is mainly on the things that go wrong, instead of

the things that do go right. This fixation on errors is reactive and encourages a ‘find

and fix’ approach, without changing the mindset and culture on the workfloor.32

It is essential to learn from the far more frequent cases where things go right

and develop ways to support, augment and encourage these. We need to shift

our perspectives: from ‘as few things as possible go wrong’ to ‘as many things as

possible go right’.26

17

→ The current quality indicators used often do not measure real ‘quality’ or ‘safety’.

Of the 3400 quality indicators in the Netherlands, hardly any are developed

that actually measure outcomes of quality of care. The added value of all these

benchmarks and registration which should be performed to report them does not

measure up to the burden they cause in daily clinical life.38,44

→ Most importantly, even though we are aiming to improve patient care, we often

forget to include them in this development.

Patient involvement

In an age where patients are increasingly active as partners and demanders of care and in

which patient feedback is a compulsory part it seems surprising that patients’ (and their

carers’) views are not more routinely requested. In order to retrieve insights in how to do

this properly we must frequently converse with patients, especially those who have suffered

through unintended harm.8 The World Health Organization has also emphasized this by stating

that patient and community engagement and empowerment are the key. Their experiences

and perspectives are valuable resources for identifying needs, measuring progress and

evaluating outcomes.11

GENERAL OUTLINE OF THE THESISThe aim of this thesis is to provide insight and potential improvement strategies in safety issues

within the chain using novel ways to investigate these matters.

The fragments of the chain focused on in this thesis:

The starting point of the hospital care chain: novel ways of organizing care

Originally, patients enter the hospital through the ED, and if needed are admitted to the regular

clinical wards in the hospital. Over the past few decades shifts in this model have occurred.

Acute medical units (AMUs) have been introduced as a novel way to potentially improve

the efficiency of patient flow within the healthcare chain. AMUs are defined as hospital wards

specifically staffed and equipped to receive medical inpatients presenting with acute medical

illness from the ED, outpatient clinics and/or the community for expedited multidisciplinary

and medical specialist assessment, care and treatment for up to a designated period (typically

between 24 and 72 h) prior to discharge or transfer to other wards.45 In the Netherlands,

AMUs do not limit their case-mix to medical patients only, other specialties such as surgery and

neurology also admit patients to these units. AMUs form a link between the ED and the regular

wards in the hospital. Implementation of an AMU could lead to better and a more efficient

use of available beds. In addition, an AMU potentially reduces complications and unintended

errors by stimulating multidisciplinary teamwork, simplifying logistics and acceleration of

diagnostics and therapy.46-48

18

Over the past years their implementation has rapidly evolved, especially in the UK and Australia.

Since 2000, AMUs (or similar units) have also started up in the Netherlands. However, since

no national Dutch guidelines or implementation standards have been created, it is still hard to

conclude whether patient safety benefits from this (re)organization. In addition, no national

registration of centres with such a unit are easily available, making it hard for hospitals to find

and learn from each other and exchange experiences and pitfalls faced during implementation.

In chapter 2 of this thesis a systematic literature overview is provided describing the effects

of an acute medical unit on patient outcomes. In addition, it gives an overview of the current

situation in the Netherlands and makes recommendations that could be used to formulate

national AMU guidelines.

Even though research on the effectiveness of these units has been performed extensively,

the perspectives of patients on acute medical units have been underreported. The wards are

often busy and operate 24 hours a day, generating a constant buzz and a potential turmoil

for patients. It might be worthwhile talking to the patients in order to gather information

about their experienced quality of care and the feeling of safety in such a ward. Patients are

predominantly the initiator of their own healthcare pathway and their motives and experiences

should therefore be investigated in order to construct more streamlined care. In addition, since

the AMU caters to a heterogeneous patient population with a variety of pathology, it might

be interesting to test whether the planned patient-experience orientated research is feasible

in such a diverse group.

In chapter 3, we present the results of a feasibility study using Patient Reported Outcome

Measures (PROMs) to measure the perceived quality of care and feeling of safety in an AMU.

Prevention of serious adverse events on the clinical wards

Serious adverse events and adverse outcomes on clinical wards such as cardiac arrest, death

and unplanned ICU admissions endanger patient safety if they are the result of healthcare

processes. They are often preceded by signs of clinical deterioration, such as changes in vital

parameters as pulse rate, respiratory rate and level of consciousness.49,50

Unplanned ICU admissions from the ward

An unplanned ICU admission of an inpatient to the Intensive Care Unit from a general ward

is considered as a SAE, and is associated with poor long term survival, especially in older

patients.51 Therefore, it is important to investigate the root causes of these admissions, in order

to improve detection of these deteriorating patients and to potentially avoid unplanned ICU

admissions to enhance patient safety. A useful tool to analyse root causes of adverse events is

the PRISMA-tool (Prevention and Recovery Information System for Monitoring and Analysis).52

The main goal of this method is to build a database of incidents and process deviations by

creating causal trees, from which conclusions can be drawn to suggest countermeasures. It is

19

accepted by the World Alliance of Safety of the World Health Organization and has shown to

provide effective starting point for improvement in quality and safety of care.52-54

In chapter 4 we used a record review study to analyse the healthcare worker-, organizational-,

technical-, disease- and patient-related causes that contribute to late detection and treatment

of deteriorating patients in the general words using PRISMA-analysis.

Timely recognition of deterioration

To increase patient safety on nursing wards by timely detecting and treating the deteriorating

patients to prevent adverse events, rapid response systems (RRs) have been introduced. These

RRSs consist of an afferent limb residing a track and trigger system (TTS) such as the Modified

Early Warning Score (MEWS), and an efferent limb residing a rapid intervention team (RIT)

consisting of trained ICU personnel who will deliver treatment at the bedside when required.55

The effectiveness of these systems depends on timely recognition of these patients, and

requires proper implementation of both limbs. The TTS such as the MEWS, mostly used by

nurses to systematically monitor the patients in the wards, should be adequately implemented

in order to prevent potential adverse events.

In addition to the PRISMA-analysis, chapter 4 also assessed the adherence and effectiveness

of an already implemented track and trigger system (TTs) to investigate if patient deteriorations

were identified timely.

In chapter 5 we determined protocol adherence of a recently reimplemented rapid response

system in a prospective real-life setting, with a main focus on the effectuation of the afferent

limb, assessed by the use of MEWS on the wards. We also investigated the ability of this once

a day protocolised MEWS measurement to predict patient outcomes associated with poor

patient safety: in-hospital mortality, hospital length of stay, cardiac arrests, ICU-admissions and

unplanned 30-day readmissions.

The use of quality indicators to assess patient safety

Readmissions

Since 2015 the percentage of unplanned readmissions within 30 days after discharge of

a clinical admission in the same hospital was added as an official quality indicator to already

existing indicators, unintended long length of stay and in-hospital mortality.34

Currently, these indicators are regarded as a ‘major adverse events’ and are supposed to

reflect the quality of healthcare in Dutch hospitals. Policy makers in the USA and the UK are

already using the percentage of readmissions as a marker of quality and hospital safety; in

these countries a ‘high’ readmission rate can result in financial penalties.18,39 Empirical Dutch

research into readmissions is lacking, and policy makers are still struggling with the realization

of a reliable indicator – to date it does not distinguish between planned and unplanned

20

readmissions, and does not integrate ‘preventability’. The problem is that preventability is not

easily incorporated in this quality indicator, but should be the focus of improvement.

Readmissions are not easy to investigate, since they are often multi-causal, and their causes

do not solely lie in inadequate hospital care. It is however important to address this issue since

the increasing burden on healthcare might pressure doctors to discharge patients too soon,

potentially causing more unplanned readmissions since the actual problems may not have

been resolved during initial admission.30 In addition, it is known that in an older population,

which will make up an increasing amount of our patient-mix in the next years, mortality

increases if these patients are readmitted frequently.56 When studying a readmission the whole

process and decision-making in the chain should be looked upon, instead of concentrating

solely on the patient and his comorbidities.

Chapter 6 aimed to identify organizational-, technical-, healthcare worker- and patient-

related causes that contributed to readmission using previously described PRISMA-analysis. In

addition, we assessed how many of the investigated readmissions were deemed preventable.

The commentary presented in chapter 7 argues why readmissions cannot be regarded as

a reliable way of assessing quality of healthcare within a hospital in its current form.

Uniform factors related to preventable readmissions have not yet been identified.57,58 In

addition, consensus definition of the preventability of a readmission has not been established.

The current gold standard predominantly used is the opinions of physicians determining if

a readmission is preventable. Subsequently, factors that could predict preventable readmissions

are extracted from these findings.58-60

In chapter 8 we performed an international study to assess if there is any consensus between

physicians regarding the predictability and preventability of medical readmissions

As previously stated, in comparison to extensive US data, no large-scale European studies

have been performed about the predictability and preventability of readmissions. Furthermore,

patient and healthcare worker opinions, the stakeholders that are really confronted with this

problem on the frontline, are not frequently addressed to elucidate their vision.

In chapter 9, we describe the first prospective observational study of 1398 unplanned medical

readmissions to 15 centres in 4 European countries aiming to investigate the opinions of

readmitted patients, their carers, nurses and physicians on the predictability and preventability

of the readmissions. We also investigated the contributing factors that could potentially

predict (preventable) readmissions using their opinions.

21

Hospital Standardized Mortality Ratio (HSMR)

In order to be transparent about the quality of care, Dutch hospitals are compulsory to

publish their HSMR since 2014.61 It is one of the numerous healthcare indicators used in

the Netherlands, but together with unplanned readmissions and unexpected long length

of stay, these three elementary quality indicators ought to rate and represent the quality

of care within Dutch hospitals. The HSMR represents the ratio between the observed and

expected deaths. The expected deaths are calculated with the use of a statistical model

that corrects for certain factors such as age, socio-economic status and comorbidity.62 It is

questionable if the indicator in its current form really portrays the quality of a hospital, since

it is hard to extract a complete dataset which also incorporates the clinical illness in these

deceased patients.

Therefore, in a cohort of deceased patients with the diagnosis pneumonia, in chapter 10,

we examined whether the HSMR model under- or overestimates clinical disease severity. In

addition, we rated the completeness of the data collected by administrators to calculate HSMR.

Chapter 11 of this thesis provides a summary of the main results, a general discussion and

elucidates on future perspectives. A Dutch summary of this thesis can be found in chapter 12.

22

REFERENCES1. Centraal bureau voor de statistiek. Statline zorguitgaven; kerncijfers. http://statline.cbs.nl/Statweb.

Published December 2016. Accessed January 2, 2017.

2. Gaakeer MI, van den Brand CL, Gips E, et al. [National developments in Emergency Departments in the Netherlands: numbers and origins of patients in the period from 2012 to 2015]. Ned Tijdschr Geneeskd. 2016;160(0):D970.

3. Consultancy C. Ligduurmonitor Nederlandse ziekenhuizen 2013 & 2014. Trendbreuk in de ligduur. http://www.coppa.nl/wp-content/uploads/2016/01/Ligduurmonitor-2013-2014.pdf. Published December 2015. Accessed August, 2016.

4. Schrijver EJ, Toppinga Q, de Vries OJ, Kramer MH, Nanayakkara PW. An observational cohort study on geriatric patient profile in an emergency department in the Netherlands. Neth J Med. 2013;71(6):324-330.

5. Samaras N, Chevalley T, Samaras D, Gold G. Older patients in the emergency department: a review. Ann Emerg Med. 2010;56(3):261-269.

6. Ministry of Health Welfare and Sport. Healthcare in the Netherlands. http://investinholland.com/nfia_media/2015/05/healthcare-in-the-netherlands.pdf. Published January 2015. Accessed January 2, 2017.

7. Commonwelth Fund Commission. Why not the best? Results from the National Scorecard on U.S. Health System Performance. http://www.commonwealthfund.org/publications/fund-reports/2011/oct/why-not-the-best-2011. Published 2011. Accessed January 2, 2017.

8. Dillman R, Wagner C, Schellekens W, Klein J, Jager M, van der Grinten T. Patientveiligheid. Utrecht: de Tijdstroom uitgeverij. 2016.

9. Greenstein EA, Arora VM, Staisiunas PG, Banerjee SS, Farnan JM. Characterising physician listening behaviour during hospitalist handoffs using the HEAR checklist. BMJ Qual Saf. 2013;22(3):203-209.

10. Ong M, Bostrom A, Vidyarthi A, McCulloch C, Auerbach A. House staff team workload and organization effects on patient outcomes in an academic general internal medicine inpatient service. Arch Intern Med. 2007;167(1):47-52.

11. World Health Organization. Patient Safety. http://www.euro.who.int/en/health-topics/Health-systems/patient-safety/patient-safety. Accessed January 3, 2017.

12. David K, Stremikis K, Squires D, Schoen C. Mirror, Mirror on the Wall. How the Performance of the U.S. Health Care System Compares Internationally. http://www.commonwealthfund.org/~/media/files/publications/fund-report/2014/jun/1755_davis_mirror_mirror_2014.pdf. Published 2014. Accessed January 2, 2017.

13. Bjornberg A. Euro Health Consumer Index. www.healthpowerhouse.com/files/EHCI_2015_report.pdf Published January 2016. Accessed January 24, 2017.

14. Romp M, Merkx P. Zorgthermometer Vektis. Verzekerden in beeld 2015. http://www.vektis.nl/downloads/Publicaties/2015/Zorgthermometer%20nr14/#1/z. Published April 2015. Accessed January 3, 2017.

15. Heijne S. Waarom de zorgpremie in 2017 met gemiddeld 9% stijgt. NtvG. 01-2017.

16. van den Berg M, de Boer D, Gijsen R, et al. RIVM Zorgbalans 2014. De prestaties van de Nederlandse gezondheidszorg. http://www.gezondheidszorgbalans.nl/ dsresource? type=pdf&disposition= inline&objectid= rivmp:259835&versionid= &subobjectname=. Published September 2014. Accessed January 2, 2017.

17. de Hoon S, Verberne L, Davids R, et al. Zorg door de huisarts. NIVEL Jaarcijfers en trendcijfers 2011-2015. http://www.nivel.nl/sites/default/files/jaarrapport_huisarts_2015.pdf. Published October 2016. Accessed January 2, 2017.

18. Greysen SR, Stijacic Cenzer I, Auerbach AD, Covinsky KE. Functional impairment and hospital readmission in Medicare seniors. JAMA Intern Med. 2015;175(4):559-565.

19. van der Horst H. Medisch contact. http://www.medischcontact.nl/archief-6/Tijdschriftartikel/153800/Spoedhulp-voorouderen-in-het-gedrang.html. Published May 2016. Accessed January 2, 2017.

23

20. VUmc Netwerk Acute Zorg. Ontwikkelingen aanbod acute patiënten SEH’s RsHsQQ, 2014 en 2015 ROAZ-regio VUmc. https://www.vumc.nl/afdelingen-themas/719134/8513847/8514115/20150929_RapportAcuteZorgQ11.pdf. Published September 2015. Accessed August, 2016.

21. Merten H, Zegers M, de Bruijne MC, Wagner C. Scale, nature, preventability and causes of adverse events in hospitalised older patients. Age Ageing. 2013;42(1):87-93.

22. Vegting IL, Alam N, Ghanes K, et al. What are we waiting for? Factors influencing completion times in an academic and peripheral emergency department. Neth J Med. 2015;73(7):331-340.

23. Singal RK, Sibbald R, Morgan B, et al. A prospective determination of the incidence of perceived inappropriate care in critically ill patients. Can Respir J. 2014;21(3):165-170.

24. Cardona-Morrell M, Kim J, Turner RM, Anstey M, Mitchell IA, Hillman K. Non-beneficial treatments in hospital at the end of life: a systematic review on extent of the problem. Int J Qual Health Care. 2016;28(4):456-469.

25. Stellinga M, Weeda F. Interview E. Schippers in NRC Handelsblad 08-09-2012.

26. Hollnagel E, Wears RL, Braithwaite J. From Safety-I to Safety-II: A White Paper. https://www.england.nhs.uk/signuptosafety/wp-content/uploads/sites/16/2015/10/safety-1-safety-2-whte-papr.pdf. Published 2015, Accessed January 10, 2017.

27. Arora VM, Farnan JM. Inpatient Service Change: Safety or Selection? JAMA. 2016;316(21):2193-2194.

28. Pannick S, Beveridge I, Wachter RM, Sevdalis N. Improving the quality and safety of care on the medical ward: A review and synthesis of the evidence base. Eur J Intern Med. 2014;25(10):874-887.

29. Vidyarthi AR, Arora V, Schnipper JL, Wall SD, Wachter RM. Managing discontinuity in academic medical centers: strategies for a safe and effective resident sign-out. J Hosp Med. 2006;1(4):257-266.

30. Averbukh Y, Southern W. The impact of the number of admissions to the inpatient medical teaching team on patient safety outcomes. J Grad Med Educ. 2012;4(3):307-311.

31. Institute of Medicine. To err is human: building a safer health system. http://www.nationalacademies.org/hmd/~/media/Files/Report%20Files/1999/To-Err-is-Human/To%20Err%20is%20Human%201999%20%20report%20brief.pdf. Published November 1999. Accessed December 28, 2016.

32. Braithwaite J, Wears RL, Hollnagel E. Resilient health care: turning patient safety on its head. Int J Qual Health Care. 2015;27(5):418-420.

33. Francis R. Report of the Mid Staffordshire NHS Foundation Trust Public Inquiry. London. Stationery Office. 2013.

34. Basisset Kwaliteitsindicatoren 2014. Utrecht: Inspectie voor de gezondheidszorg, 2015. Accessed January, 2016.

35. Veiligheidsmanagementsysteem. Communicatie tussen hulpverleners volgens het SBAR-proces. VMS Veiligheidsprogramma. 02-2009.

36. Subbe CP, Gao H, Harrison DA. Reproducibility of physiological track-and-trigger warning systems for identifying at-risk patients on the ward. Intensive care medicine. 2007;33(4):619-624.

37. Langelaan M, de Bruijne M, Baines R, et al. NIVEL Monitor Zorggerelateerde Schade 2011/2012. Dossieronderzoek in Nederlandse ziekenhuizen. http://www.nivel.nl/sites/default/files/bestanden/monitor_zorggerelateerde_schade_2011_2012.pdf. Published 2013, Accessed December 28, 2016.

38. Poortmans J. Onderzoek - Zorgkwaliteit in ziekenhuizen. Driemaaldaags een pijnschaal. De Groene Amsterdammer. 2016(49).

39. Balla U, Malnick S, Schattner A. Early readmissions to the department of medicine as a screening tool for monitoring quality of care problems. Medicine (Baltimore). 2008;87(5):294-300.

40. Actal Eindrapportage. Onderzoen naar de ervaren regeldruk onder verpleegkundigen in het ziekenhuis. 114-bijlage-Eindrapportage-EY-ervaren-regeldruk-verpleegkundigen.pdf. Published September 2012. Accessed January 3, 2017.

24

41. O’Leary KJ, Buck R, Fligiel HM, et al. Structured interdisciplinary rounds in a medical teaching unit: improving patient safety. Arch Intern Med. 2011;171(7):678-684.

42. Giesen P, Smits M, Huibers L, Grol R, Wensing M. Quality of after-hours primary care in the Netherlands: a narrative review. Ann Intern Med. 2011;155(2):108-113.

43. Mardon RE, Khanna K, Sorra J, Dyer N, Famolaro T. Exploring relationships between hospital patient safety culture and adverse events. J Patient Saf. 2010;6(4):226-232.

44. Beaver K, Tysver-Robinson D, Campbell M, et al. Comparing hospital and telephone follow-up after treatment for breast cancer: randomised equivalence trial. BMJ. 2009;338:a3147.

45. Scott I, Vaughan L, Bell D. Effectiveness of acute medical units in hospitals: a systematic review. Int J Qual Health Care. 2009;21(6):397-407.

46. Cooke MW, Higgins J, Kidd P. Use of emergency observation and assessment wards: a systematic literature review. Emerg Med J. 2003;20(2):138-142.

47. Group V. De Acute Opname Afdeling ‘Het wat, waarom en hoe’. 2011.

48. Reid LE, Dinesen LC, Jones MC, Morrison ZJ, Weir CJ, Lone NI. The effectiveness and variation of acute medical units: a systematic review. Int J Qual Health Care. 2016.

49. Ludikhuize J, Smorenburg SM, de Rooij SE, de Jonge E. Identification of deteriorating patients on general wards; measurement of vital parameters and potential effectiveness of the Modified Early Warning Score. J Crit Care. 2012;27(4):424 e427-413.

50. Hillman K, Parr M, Flabouris A, Bishop G, Stewart A. Redefining in-hospital resuscitation: the concept of the medical emergency team. Resuscitation. 2001;48(2):105-110.

51. Ridley S, Jackson R, Findlay J, Wallace P. Long term survival after intensive care. BMJ. 1990;301(6761):1127-1130.

52. van der Schaaf TW HM. PRISM-Medical. A Brief Description. Eindhoven University of Technology, Faculty of Technology Management, Patient Safety Systems: Eindhoven. 2005.

53. van Vuuren W SC, van der Schaaf TW. The Development of an Incident Analysis Tool For the Medical Field. Eindhoven University of Technology: Eindhoven. 1997.

54. van Wagtendonk I, Smits M, Merten H, Heetveld MJ, Wagner C. Nature, causes and consequences of unintended events in surgical units. Br J Surg. 2010;97(11):1730-1740.

55. Alam N, Hobbelink EL, van Tienhoven AJ, van de Ven PM, Jansma EP, Nanayakkara PW. The impact of the use of the Early Warning Score (EWS) on patient outcomes: a systematic review. Resuscitation. 2014;85(5):587-594.

56. Zanocchi M, Maero B, Martinelli E, et al. Early re-hospitalization of elderly people discharged from a geriatric ward. Aging clinical and experimental research. 2006;18(1):63-69.

57. Jackson AH, Fireman E, Feigenbaum P, Neuwirth E, Kipnis P, Bellows J. Manual and automated methods for identifying potentially preventable readmissions: a comparison in a large healthcare system. BMC medical informatics and decision making. 2014;14:28.

58. van Walraven C, Jennings A, Forster AJ. A meta-analysis of hospital 30-day avoidable readmission rates. Journal of evaluation in clinical practice. 2012;18(6):1211-1218.

59. Cakir B, Gammon G. Evaluating readmission rates: how can we improve? Southern medical journal. 2010;103(11):1079-1083.

60. Meisenberg BR, Hahn E, Binner M, et al. ReCAP: Insights Into the Potential Preventability of Oncology Readmissions. J Oncol Pract. 2016;12(2):153-154.

61. Keesman A. Publicaties HSMR en SMR’s verplicht. https://www.hsmr.nl/publicatie-hsmr-en-smrs-verplicht/ Published 2013. Accessed January 5, 2017.

62. van der Laan. J, de Bruin. A, van den Akker-Ploemacher J, Penning C, Pijpers F. HSMR 2014 methodological report. http://www.hsmr.nl/wp-content/uploads/2016/01/2015hsmrmethodologicalreport2014.pdf. Published 2015. Accessed March, 2016.

THE STARTING POINT OF THE HOSPITAL CARE CHAIN: NOVEL WAYS OF ORGANIZING CARE

CHAPTER 2

ACUTE MEDICAL UNITS: THE WAY TO GO? A LITERATURE REVIEW

Louise S. van Galen & Eline M. Lammers | Linda J. Schoonmade | Nadia Alam | Mark H. Kramer | Prabath W. Nanayakkara

Eur J Intern Med 2017;39:24-31

‘Focus on systems that are real rather than ideal’ Jeffrey Braithwaite

28

ABSTRACT Background

The acute healthcare chains in the Netherlands are increasingly under pressure because of

rising emergency department (ED) admissions, relative bed shortages and government policy

changes. In order to improve acute patient flow and quality of care through hospitals, an

acute medical unit (AMU) might be a solution, as demonstrated in the UK. However, limited

information is available concerning AMUs in the Netherlands.

Therefore the aims of this study were to:

1. Systematically provide an overview of current international literature regarding

the effectiveness of AMUs

2. Give an overview of the current situation in the Netherlands

3. Make recommendations that could be used for future Dutch AMU guidelines

Methods

A systematic literature search was performed searching 3 electronic databases: PubMed,

Cochrane and EMBASE. All 106 hospitals in the Netherlands were contacted, inquiring about

the status of an ED, the AMU or future plans to start one.

Results

The literature search resulted in 31 studies that met inclusion criteria. In general, these studies

reported significant benefits on number of admissions, hospital length of stay (LOS), mortality,

other wards and readmissions. Among the Dutch hospitals with an Ed, 33 out of 93 Dutch

hospitals implemented an AMU or similar ward, these are however organized heterogeneously.

Following current trends, more AMUs are expected to be realized in the future.

Conclusion

In order to improve the current strain on the Dutch acute healthcare system, an AMU could

potentially provide benefits. However, a uniform guideline is warranted to optimize and

compare quality of care throughout the Netherlands.

Keywords

Acute medical unit, Healthcare quality, Implementation guidelines

Highlights

1. An Acute Medical Unit (AMU) could improve patient flow and hospital quality

of care.

2. Current literature shows an overall beneficial effect after implementation.

3. In the Netherlands one-third of the hospitals have implemented an AMU.

4. Existing AMUs are designed heterogeneously, this calls for a uniform guideline.

29

INTRODUCTIONAcute healthcare chains in the Netherlands have been increasingly under pressure in the last

decade. Due to the aging population and the policy changes in the Dutch healthcare system

the hospitals have seen a steady rise in emergency admissions. In addition, these patients

present themselves with a higher severity of illness because this population is increasingly

older and often have multiple comorbidities and polypharmacy.1 This strain on the emergency

departments (ED) increases ED length of stay (LOS) which in turn is related to a higher

morbidity and mortality.2,3 Older patients tend to have atypical clinical presentations leading

to more diagnostics and 75% spend too long in the ED.4 A recent report has stated that,

for the first time in years, the adjusted mean hospital length of stay in the Netherlands has

increased by 8,5 %.5 In addition to increasing admissions, a relative shortage of hospital beds

in combination with a high percentage of beds not being used efficiently have also contributed

to this increase.5 Therefore, the introduction of novel ways to improve the efficiency of

the whole acute care chain is needed.

Acute medical units (AMU), which have widely been implemented in the United Kingdom

(UK) and Australia, might offer a solution to the current burden on acute healthcare chains

in the Netherlands. AMUs or units with similar names [i.e. medical assessment units (MAU),

emergency assessment units (EAU)] are defined as hospital wards specifically staffed and

equipped to receive medical inpatients presenting with acute medical illness from the ED,

outpatient clinics and/or the community for expedited multidisciplinary and medical specialist

assessment, care and treatment for up to a designated period (typically between 24 and 72

hours) prior to discharge or transfer to other wards.6 Frequently admitted medical conditions

are for example: sepsis, acute coronary syndrome, pneumonia, and COPD.7 An implementation

of an AMU by reorganizing already available hospital beds could potentially lead to a better

and a more efficient use of available beds. Other possible advantages of implementing an

AMU is to improve the quality of care by stimulating multidisciplinary (team)work, simplifying

logistics, and accelerating diagnostics and therapy. In addition, an AMU potentially reduces

complications and unintended errors.8

Since 2000, an increasing number of hospitals have started implementing AMUs in

the Netherlands. However, there are no national guidelines setting uniform standards for

creating such a ward and therefore these AMUs have been established heterogeneously.

Also, no value-based indicators have been developed to evaluate the quality of healthcare in

these AMUs.

Previous work has already shown the effectiveness of an AMU on patient outcomes and

certain quality indicators.6,9,10 Since, in the coming years more AMUs are expected to start up

in the Netherlands in order to cope with the increasing acute patient flow, a clear guideline

could provide a gold standard from which the quality of these units could be standardized

and measured.

30

Therefore, our primary aim is to [1] Provide a systematic overview of current knowledge in

international literature about the effectiveness of AMUs, [2] give an overview of the current

situation in the Netherlands, and [3] make recommendations that could be used to formulate

national AMU guidelines.

METHODSSearch strategy

In order to find the correct key terms used to identify units similar to ‘acute medical units’, we

scanned previously written literature overviews and documentations on the matter. To identify

all relevant publications search terms included controlled terms from MeSH in PubMed, EMtree

in EMBASE.com plus free text terms for The Cochrane Library. Terms expressing ‘acute medical

units’ and their synonyms identified were used. The full search strategies and synonyms

used for all databases can be found in the Supplementary Information (Appendix A: search

strategy). Afterwards, a systematic search was performed in the bibliographic databases

PubMed, EMBASE.com and The Cochrane Library (via Wiley) from inception to 08-08-2016 in

collaboration with a medical librarian (LS). The search was performed without date, language

or publication status restriction. The references of the identified articles were searched for

relevant publications.

Selection process

A total of 1909 records were excluded from the analysis primarily, because these articles

did not address the effect of implementation of an AMU as their main subject (for example,

articles about HIV testing or the use of antibiotics in the AMU obtained during the search were

excluded). Both reviewers (LG, EL) independently screened all remaining articles identified

with the search on title and abstract. If considered necessary, the whole article was checked

for eligibility. After individual screening, included articles were compared and differences in

opinions between the reviewers were resolved through consensus. Inclusion criteria were (i)

the study should report the outcome measures on an AMU (ii) the AMUs or similar wards

described should have a minimum LOS of 24 h and should conform to the definition given

in the introduction [designated hospital wards specifically staffed and equipped to receive

medical inpatient presenting with acute medical illness from the ED, outpatient clinics and/

or the community for expedited multidisciplinary and medical specialist assessment, care and

treatment for up to a designated period (typically between 24-72 h) prior to discharge or

transfer to other wards] (iii) the AMU must admit medical patients because the AMUs were

originally designed for medical patients, and most available literature was on medical patients

(iv) the study should include patients from 14 years and above (v) the unit in the study must

be functional 7 days a week, 24 hours a day. Exclusion criteria were (i) no full text available

(in English) (ii) certain publication types: editorials, letters, legal cases, annual hospital reports,

etc. (iii) AMUs which solely admitted surgical patients, (iv) psychiatric AMUs.

Literature assessment

After identifying suitable studies both reviewers independently assessed the papers for their

quality using the NICE guidelines for qualitative studies and cohort studies.11 No controlled

31

trials were found. An item ‘clinical relevance’ was added to individually assess the relevance of

the paper in answering the primary objective (varying in four categories from ‘poor’ to ‘highly’

relevant). A standardized data worksheet was used for data extraction with the following

variables: study design, ward description, study population, time period, aim, methods,

primary endpoint, secondary endpoint, clinical relevance and overall study quality (NICE). This

worksheet was filled out after reaching consensus by both reviewers during their meetings.

Search results

In total the literature search generated 3386 references, excluding duplicates. Flowchart 1

(Fig. 1) illustrates our search and selection process. After selection 100 articles remained and

2 extra articles were added after reference checks. After assessment by both reviewers, a total

of 31 studies were suitable for inclusion. Appendix B shows an overview of all included articles

in a table with the variables selected for data extraction.

Situation in the Netherlands

To evaluate the current situation in the Netherlands, all 106 Dutch hospital locations were

telephoned and inquired 1) whether they had an ED 2) if yes, whether they had an AMU 3)

if yes, since when and 4) how their AMU was organized. If hospitals did not have an AMU,

they were asked if they had a similar construction such as a short stay unit, an observation

unit or an acute ambulatory unit, all being located on or near the ED. Subsequently, they were

asked if they had plans to implement an AMU in the future. The staff members involved with

the management were inquired about the organization of the AMU if available. The results

were processed anonymously. To assure no hospitals were missed a registered list of all

hospitals was used. For this analysis we only included hospitals that delivered (unplanned)

acute care, and thus had an ED. Fused hospitals were counted and assessed separately if they

both had a separate ED.

RESULTSStudy design

The studies included in this review were published in different countries; twelve originate from

the UK7,12-22, seven from Australia23-29, seven from Ireland30-36, one from Australia and New-

Zealand37, one from Canada38, one from Denmark39, one from Hong Kong40 and one from

the Netherlands41. The data of all studies were collected between 1993 and August 2016. 27

of the 30 studies had a quantitative design, three had a qualitative design.16,18,26

The most common types of quantitative study designs were comparisons of the situation

before and after establishment of the AMU (thirteen in total) and comparison of the AMU

to other medical wards/clinical teaching units (four in total). Furthermore, three studies used

questionnaires to assess the outcomes in a cross-sectional design.19,21,35 Brand et al. (2010)

and Suthers et al. (2012) made two comparisons: pre-AMU vs. AMU and AMU vs. ward.23,27

Elder et al. (2015) compared standard care to physician at triage and the combination of

physician at triage with the medical assessment unit.24 Relihan et al. (2009) compared

32

Figure 1 | Flowchart search and selection procedure of studies

the acute medical assessment unit to international benchmarking data.35 Sullivan et al. (2013)

compared different AMUs to emergency admissions in other specialties and to scheduled

admissions.21 The qualitative studies all used either focus groups, semi-structured interviews

or questionnaires. Two had a cross-sectional design and one compared the situation before

and after implementation of the AMU.18

All three articles published by Boyle et al. described the situation before and after reorganization

of the ED and the AMU into one emergency assessment unit.12-14. Seven articles described

the same 59-bedded acute medical assessment unit in an Irish hospital.30-36 Pradhan et al.

(2015) and McNeill et al. (2011) were multicentre analyses that includes different AMUs,

55 in the UK and 32 in Australia and New-Zealand.19,37 Sullivan et al. (2013) performed

a national study of different AMUs within the UK (the exact amount of AMUs is unknown).21

In order to determine factors that could potentially affect outcome measures, three studies

used a logistic regression model.30,34,36 In one study univariate and loglinear analyses were

33

performed.32 Multivariate logistic regression and multiple linear regression were used in two

different studies.21,27 One longitudinal cohort study performed by Van Galen et al. (2016) used

a Bonferroni correction to compare between different time periods.41

Study population

All AMUs included in this review organized care for ‘medical’ patients. Medical patients are

patients admitted for the following specialties: cardiology, gastroenterology, internal medicine,

nephrology, neurology, oncology, pulmonary medicine, geriatrics, and rheumatology. In

addition, the emergency assessment unit described in the three articles by Boyle et al. also

consist of a surgical and a paediatric unit.12-14 Furthermore, Hadden et al. (1996) described

an AMU dedicated to both medical and surgical patients.17 For these four articles we only

report on medical patients (Appendix 2). Van Galen et al. described an acute admission unit

with medical and surgical patients, but no results for the separate group of medical patients

were reported and therefore could not be displayed separately.41 Figure 2 shows the Dutch

AMU model, compared to the traditional model of emergency admissions and the patient flow

model mostly used on the AMU in the United Kingdom.

Becket et al. and Vork et al. described patients in the AMU who were referred from their

general physician (GP) or the ED.7,39 The acute admission unit described by van Galen et

al. (2016) received patients mainly from the ED but also from the outpatients clinic.41 In

all other included studies, patients were referred from the ED only or the process was not

described clearly.

  Figure 2 | Models of acute admissions (Traditional model, AMU in the Netherlands, AMU in the UK)

34

Ward description

The AMUs included in this paper have heterogeneous designs. The units differ in name, location

(next to ED or not), number of beds, number of nurses and physicians and the maximum LOS on

the unit before discharge home or to another medical ward. When reported, the number and

type of physicians on rotation differed significantly. In addition, the frequency of supervision

and number of ward rounds varied.

Assessed NICE quality and clinical relevance

Cohort studies were assessed for selection, attrition, performance and detection bias

according to the NICE-criteria.11 Boyle et al. (2012), Coary et al. (2014), McNeill et al. (2011),

Moloney et al. (2005), Slatyer et al. (2008), Suthers et al. (2012) and Rooney et al. (2008)

were rated the highest quality cohort studies included in this review.13,26,30,34,36,37 Hadden et

al. (1996) and Yong et al. (2011) were rated the lowest quality cohort studies included.17,29

Qualitative studies were scored for the quality of their study methodology, data collection and

analysis. The qualitative study with the best rating was Hanlon et al. (1997) and the study

with the lowest rating was McErlain Burns et al. (1997).16,18 Basey et al. (2016), Hadden et al.

(1997) and Yong et al. (2011) were assessed as poorly relevant in contribution to answering

the primary objective of this paper,17,22,29 while Coary et al. (2014), Conway et al. (2014),

McNeill et al. (2011), Moloney et al. (2005) , Rooney et al. (2008) and Suthers et al. (2012)

were assessed as highly relevant.27,30,31,34,36,37

Patient outcomes*(only significant outcomes are reported, other outcomes can be found in table 2/appendix B)

Number of emergency admissions

Two articles from the Irish acute medical assessment unit describe an increase of 9,7% (5038

vs. 5528 admissions) and 4.4%(5476 vs. 5715 admissions) in the number of acute medical

admissions after establishment of an AMU.32,34 Man Lo et al. (2014) in Hong Kong found

a 14.4% decrease in the average emergency admissions rate after reorganization into an

emergency medicine ward (absolute figures unknown).40

In the UK, Boyle et al. (2008) found a 16.3% decrease in the number of acute medical admissions

after reorganization of a separate AMU and ED into one emergency assessment unit, from

3667 admissions during a 3-month period before reorganization to 3068 afterwards.14 No

other reorganization structure as such was described elsewhere.

Length of stay (LOS)

St. Noble t al. (2008), Downing et al. (2008) and Williams et al. (2000) found a decrease

in mean hospital LOS after reorganization into an AMU from 9.3 to 7.8 days, from 5.5 to

4.6 days and from 3.8 to 2.6 days respectively.15,20,28 Moloney et al. (2005, 2006 and 2007),

Conway et al. (2014), Vork et al. (2011), Boyle et al. (2008) and Suthers et al. (2012) reported

a decrease in median hospital LOS after establishment of an AMU: from 6 days to 5 days, from

7.1 to 6.9 days, from 4.1 to 3.8 days, from 7 to 5 days and from 6.8 days to 5.2 days.14,27,31-34,39

35

Hanlon et al. (1997) reported no significant difference in median hospital LOS.18 Man Lo et

al. (2014) reported a significant decrease in hospital LOS (excluding the time spent in the ED)

after establishment of an AMU: from 5.1 days to 4.1 days.40 In addition, a study reporting on

a medical assessment and coordination Unit in Australia showed that patients that solely stay

in this unit have the shortest LOS (1.9 days), followed by patients that only stay in the medical

ward (4.9 days). Patients that stay in the Medical Assessment and Coordination Unit first and in

the medical ward afterwards have the longest LOS (9.2 days). St. Noble et al. (2008) reported

an increased trend of patients discharged in less than 24 h and 48 h after implementation

of an AMU.20 In contrast, Yong et al. (2011) showed a decrease in the number of patients

discharged within 72 h.29

Mortality

The acute medical assessment unit used in the analysis from Ireland showed a significant

decrease in 30-day mortality and annual mortality after establishment this unit from 8.8% to

5.6% and from 12.6% to 7%, respectively.36 Coary et al. (2014) also calculated the multivariate

estimate odds ratio (OR) of in-hospital 30-day mortality on this Irish acute medical assessment

unit and found an OR of 0.67 (95% CI; 0.71-0.82).30 In addition, Conway et al. (2014)

calculated the relative risk ratio (RRR) of mortality by episode and by patient in this cohort

and saw a decrease of both.31 An Australian study performed by Brand et al. (2010) obtained

a significantly lower in-hospital mortality rate in their medical assessment and planning Unit

(3.2%) compared to the rate in the medical ward (7.6%), however there was no significant

difference between overall in-hospital mortality before and after implementation.23 Boyle et

al. (2011) saw a lower actual all-cause mortality rate than the expected all-cause mortality,

whereas Yong et al. (2011) measured no significant difference, the all-cause mortality persisted

at 2%.12,29 Boyle et al. (2012) also calculated the hospital standardized mortality ratio (HSMR)

and had the lowest ratio of the included UK hospitals in 3 out of 4 years.13 Suthers et al. (2012)

reported no significant change in mortality either (measurement of mortality undefined).27

Effect on other wards

Beneficial effects on other wards have been described by Downing et al. (2008) who reported

a significant decrease in the number of medical patients in non-medical beds after reorganization

(38 vs. 11 patients).15 Man Lo et al. in Hong Kong found an average monthly decrease of 187

patients (15%) admitted to other medical wards than the AMU.40 Hadden et al. (1997) showed

a lower percentage of patients that needed transfer to an outpatient clinic from the short stay

ward in comparison to the medical ward.18 Another beneficial effect, solely described by Yong

et al. (2011), is the decrease of patients that needed transfer to the intensive care unit (ICU)

(3% before vs. 2% after reorganization).29. Besides positive effects, Abenhaim et al. (2000)

reported a percentage of 20% of patients who required transfer from the medical short stay

unit to ‘ordinary wards’ because of the severity of their disease.38 Also, in an Australian study

comparing 32 AMUs, McNeill et al. (2011) stated that 10%-19% of potential AMU patients

required transfer to another medical ward because of limited capacity.37

36

Emergency Department LOS / outpatient waiting times

Two studies reported a significant decrease in median ED LOS after establishment of an AMU,

3.1 h to 2.9 h and 8.7 h to 8.0 h respectively.24,27 Ong et al. (2012) found a significantly shorter

ED LOS for medical assessment unit patients than for medical ward patients (4.9 h vs. 6.5

h).25 Hanlon et al. (1997) saw no significant change in outpatient waiting times.18 Elder et al.

(2014) showed a significantly increased compliance to the 4-h target in the ED from 53.3%

before the medical assessment unit to 73% after implementation.24 However, McNeill et al.

(2011) reported that in more than 50% of the included units patients had to wait more than

4 hours on a daily basis because of limited AMU capacity.37 Basey et al. (2016) argued that

68% of patients were treated within the 4-h target.22 In addition, Moloney et. al (2005 and

2006) showed a decrease in the number of patients waiting in the ED after establishment of

the acute medical assessment unit, 9 patients before implementation, vs. 8 afterwards.32,34

Readmission rate

Five studies describe the effect of an AMU on the readmission rate. Two studies recorded no

change in 28-day readmission rate after reorganization into an AMU.15,33 Brand et al. (2010)

reported a decrease in the number of readmissions within 28 days in the medical hospital

population after establishment of the medical assessment and planning unit (MAPU) (6 %

readmitted vs. 5.4% readmitted), yet there was no significant difference between the MAPU

and the non-MAPU group.23Abenhaim et al. (2000) obtained a lower 30-day readmission rate

in the AMU than in the ward (9.6% vs. 13.9%), however this was not statistically significant.38

The only study performed in Denmark included in this literature review, states a significant

decrease in 30-day readmission rate from 19.8% before implementation of the Medical

Admission Unit to 14.6% afterwards.39

Staff and patient perspectives

The qualitative studies in this review mostly reported on the staff and patient experiences.

One study from the UK presented positive reactions to the AMU from both GPs and patients.

Furthermore, they also reported 64% of patients who were in the hospital before and

after the establishment of the AMU indicated improvement.16 Another study from the UK,

performed by Hanlon et al. (1997) also presented positive feedback from patients and staff.

However, medical staff were significantly more concerned about blocked beds in other clinical

wards than before implementation and nurses expressed that they felt significantly more

stressed in the AMU.18 A cross-sectional study performed in Ireland showed that the staff rated

their working environment in the AMU higher than the international benchmarking data.35

Van Galen et al. (2016) on a Dutch AMU found that patients felt safe in the AMU.41 Slatyer

et al. (2013) yet reported that patients and family members often leave the hospital with

limited understanding of their health problems.26 Sullivan et al. (2013) showed that the AMU

scored significantly lower in patient reviews than emergency admissions in other specialties or

scheduled admissions.21

37

Situation in the Netherlands

Of the 106 locations of hospitals in the Netherlands, 93 had an emergency department. Out

of these, more than one third had an acute medical unit or a comparable unit such as an

acute admission unit (AAU) or acute admission and diagnostic unit (AADU). Characteristics

of the Dutch hospitals with and without an AMU are described in Table 1. Eleven hospitals

reported to have another type of ward which was comparable to an AAU but did not meet

the definition of an AMU as previously mentioned. These were often wards functioning as

short stay units (<24 hours) which were located on or very near the ED entitled ‘short-stay unit’

or ‘observatorium’. Emergency physicians were mostly responsible for these wards. However,

the concept of the AMUs are fundamentally different from a short-stay unit. The short stay

units do not change existing system of patient flow, and can therefore be considered as

‘temporary normal care beds’ since they are simply a place where patients admitted through

the ED are waiting for a clinical bed and are temporarily placed for no longer than 12 to 24 h.

Therefore, these units do not share the essential principles of an AMU which are intensive and

active supervision, multi-disciplinary teamwork, and rapid diagnostics and therapy.

The first AMU in the Netherlands was implemented in 2000. Fig. 3 shows the trend in AMUs

implementation since 2000. These AMUs had no uniform organization. Firstly, the location

and number of beds differed. Most AMUs were placed on a separate ward near the ED and

the diagnostic facilities. The number of beds differed from 4 to 54 beds. Secondly, the AMUs

reported different specialties which were able to admit to the ward, diverging from all

medical and surgical specialties to only internal medicine patients. Most AMUs do not admit

cardiology patients, pregnant patients and children. Thirdly, different daily schedules were

described, varying from one doctor being available only in the morning, to intensive consultant

supervision during the day. Two other hospitals reported the start of an AMU in the next year,

five are currently talking about implementation but do not have any concrete plans. Three

reported to have already investigated this reorganization but decided not to go ahead, because

they expected that AMU would not be an effective addition to the current organization of

their hospitals.

Table 1 | Characteristics of Dutch hospitals with and without AMU

Hospitals with an ED 93

Hospitals with an AMU 33

Hospitals without an AMU but with SSU* 11

Beds AMU 887

Beds SSU 101

Mean beds AMU (SD) 28 (13)

Mean beds SSU (SD) 11 (6)

Mean Maximum Expected LOS (in hours) goal AMU (SD) 50,9 (10,4)

Mean Maximum Expected LOS (in hours) goal SSU (SD) 26,2 (11,8)

Mean clinical beds hospital (SD) 424 (255)

*SSU: Short Stay Unit or similar wards with a LOS <24 hours, usually on the ED

38

  33 

Table 1 | Characteristics of Dutch hospitals with and without AMU 

Hospitals with an ED  93 

Hospitals with an AMU  33 

Hospitals without an AMU but with SSU*  11 

Beds AMU  887 

Beds SSU  101 

Mean beds AMU (SD)   28 (13) 

Mean beds SSU (SD)   11 (6) 

Mean Maximum Expected LOS (in hours) goal AMU (SD)  50,9 (10,4) 

Mean Maximum Expected LOS (in hours) goal SSU (SD)  26,2 (11,8) 

Mean clinical beds hospital (SD)  424 (255) 

*SSU: Short Stay Unit or similar wards with a LOS <24 hours, usually on the ED  

 

Figure 3 | Trend AMUs Netherlands 2000‐2016 

 

 

DISCUSSION   The  primary  aim  of  this  review was  to  systematically  provide  an  overview  of  current  literature  about  the effectiveness of AMUs in terms of different patient outcomes. Most literature included in this review originated from three countries (the UK, Ireland and Australia), in which AMUs have already been implemented on a large scale. Overall, beneficial effects of implementing an AMU were reported. According to the current international literature, implementation of an AMU with reorganization of the existing beds mostly results in higher numbers of  emergency medical  admissions.32,34  In  contrast,  one  study  by Man  Lo  et  al.  describe  a  decrease  in  the amount of acute medical admissions. One of the aims of the reorganization to an emergency medicine ward described  in  this  article  was  to  reduce  the  number  the  of  unnecessary  admissions  which  could  be  an explanation for this decrease.40 Another important finding emerging from our study was that not only hospital LOS  was  significantly  reduced  when  introducing  the  AMU,  but  also  in‐hospital  mortality  and  28‐day readmission rates decreased.  In most studies, ED LOS also decreased significantly after  implementation of an AMU. In addition, several positive effects of the implementation of an AMU on other wards were reported: a 

0

5

10

15

20

25

30

35

2000 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Num

ber o

f AMUs

Figure 3 | Trend AMUs Netherlands 2000-2016

DISCUSSION The primary aim of this review was to systematically provide an overview of current literature

about the effectiveness of AMUs in terms of different patient outcomes. Most literature

included in this review originated from three countries (the UK, Ireland and Australia), in

which AMUs have already been implemented on a large scale. Overall, beneficial effects

of implementing an AMU were reported. According to the current international literature,

implementation of an AMU with reorganization of the existing beds mostly results in higher

numbers of emergency medical admissions.32,34 In contrast, one study by Man Lo et al. describe

a decrease in the amount of acute medical admissions. One of the aims of the reorganization

to an emergency medicine ward described in this article was to reduce the number the of

unnecessary admissions which could be an explanation for this decrease.40 Another important

finding emerging from our study was that not only hospital LOS was significantly reduced

when introducing the AMU, but also in-hospital mortality and 28-day readmission rates

decreased. In most studies, ED LOS also decreased significantly after implementation of an

AMU. In addition, several positive effects of the implementation of an AMU on other wards

were reported: a decrease of medical patients who were placed on non-medical beds, and

a decrease of transfers to other medical wards, outpatient clinics and the ICU.15,17,40 Patient

and staff’s perspectives were positive in most cases, for example improvements in admission

procedures resulting in more time for personal care and better cooperation between and

within the wards were mentioned.16,18 However, some studies reported coordination failures

such as insufficient communication to patients regarding their underlying health problems

at discharge from the AMU.26,37 In general, conclusions derived from studies performed in

the UK and Ireland were uniformly positive, but results from Australia were more mixed.

There is limited evidence available from the Netherlands. Although data from three university

hospitals in the Netherlands have demonstrated up to 14% (from 7000 to 8000 admissions)

increase in the number of emergency admissions after implementation of an AMU (personal

communication), these data have not been published yet.

39

The secondary aim of this study was to provide an overview of the current situation concerning

AMUs in the Netherlands. Through phone interviews we found that in one-third of the Dutch

hospitals where acute care is provided an AMU was present. We expect that if this trend

persists, almost half of the Dutch hospitals will have an AMU in approximately 10 years.

The AMUs in the above mentioned hospitals did have very different organizations, varying in

number of beds and organization of the ward. They had a goal of expected mean length of

stay of 50.9 h on average (SD 10,4 h) with a maximum length of stay between 48 and 72 h.

Our investigation has also shown that probably due to changing patient flow in acute care

chains (more elderly and sicker patients), many hospitals are working towards implementation

of an AMU, with the goal of improving patient flow.

Our study demonstrates that both internationally and in the Netherlands the AMUs are

very heterogeneously organized. Dutch hospitals started their AMUs independently without

proper (national) guidelines or support from national institutions, resulting in dissimilar

ward structures. In order to compare the efficiency and clinical outcomes among the AMUs,

a uniform organization is warranted. An AMU must be designated with the aim of improving

patient flow and efficiency, stimulate multidisciplinary healthcare, increase supervision and

take away borders between different specialties in the hospital.

In the Netherlands, however, we have limited data to conclude whether the currently

implemented AMUs are optimally organized. If we are aiming to tackle the current rise in

emergency admissions, and make patient flow more efficient through introduction of AMUs,

introducing uniform national guidelines may help hospitals to introduce the AMUs in a uniform

and efficient way. An important issue already stated by the Vreeland Group, a commercial

group involved in the implementation of AMUs in some hospitals, in their acute admission unit

booklet, is that when implementing an AMU, one must not forget to restructure the wards

around it.8 Other wards will be spared of acutely sick patients being admitted constantly, but

adequate communication between the AMU and these wards and enough beds in these wards

is required in order to guarantee efficient patient flow from the AMU. When this flow is not

adequate, patients may need to wait longer in the AMU and may even lead to increased LOS in

the ED, as already described on a MAU in Australia: half of the 32 interviewed AMUs patients

had to wait more than 4 hours in the ED because of limited MAU capacity.37 Also, a longer

LOS could potentially harm patient safety, since these units are not designed to treat/care for/

cater to patients longer than 48-72 h. The establishment of an AMU is an integral process

within the hospital. The reorganization required to create an AMU leads to a redistribution of

patient groups, and when this process is not performed with absolute care, this may negatively

affect certain patient groups. This is also demonstrated in the article from Denmark, where

cardiology patients were negatively affected by reorganization, with an increased mortality

rate as a result.39

40

Recommendations

On the basis of these findings we will attempt to make a few recommendations to improve

the implementation of an AMU in the Netherlands.

Before realizing an AMU one should have a clear vision about the purpose of an AMU. An AMU

should provide timely high quality of care for the acutely ill patient, in a suitable environment

to achieve this. This requires several pre-implementation decisions: the location of the AMU,

the number of beds, the facilities it should have, the amount and type of staffing, a plan for

proper training of the staff, the type of protocols required, the specialties that can admit to

the AMU and details about the medical and logistical responsibilities. These matters ought to

be profoundly thought of and talked about with every specialty and other partners involved

before reorganization. In addition, the hospital management must strongly support the idea.

The location of an AMU should be near the ED, with easy access to diagnostic facilities such

as radiology. Bell et al. (2008) already reported that in some hospitals in the UK, because

of organizational issues, the AMU was not close to ‘the front door’ of the hospital, which

resulted in less productive use of the unit.42

In order to arrange the ward as effectively as possible, an adequate number of beds is

needed. To decide how many beds should be in the ward, one must have an idea of the acute

healthcare flow through the hospital. Mostly, acute health-care is more predictable than

regular (chronic) hospital care, and the number of acute admissions is fairly constant per

hospital throughout the year. Also, approximately 50% of the patients admitted to the AMU

go home after two days of admission. In addition to introducing an AMU with new care

concepts, the planned admission schemes (for example for chemotherapy) can be adapted in

such a way that the congestion of beds in the regular wards can be reduced or prevented so

that the flow of patients needing transfers can be achieved without delay. This withholds that

these regular wards should at least adapt some of the concepts such as intensifying supervision

and multidisciplinary care so that flow in these wards can also be improved. Congestion in

the AMU leading to longer LOS is undesirable and can affect patient safety because the AMUs

are not designed or equipped to provide chronic care to patients. Therefore, Scott et al. (2011)

proposed rigorous business rules around patient entry clearly defining processes of admission

and discharge of eligible patients, as success factors that could influence AMU operations – not

specifically describing in which the effect of these factors could be measured in.6 In order to

promote patient flow and increase bed occupancy, post-discharge services such as professional

homecare or ambulatory treatment should be easily accessible.

The teams staffing the AMU should consist of dedicated professionals who are able to work

in multidisciplinary teams and deal with a high turnover of patients. The staffing should be

continuous seven days a week with a clear rotation schedule with ward rounds twice daily. It

is highly recommended for the supervisors to work closely with their juniors in order to speed

up the decision making process and thereby patient flow. Enough staff is needed to ensure

41

an adequate patient; nurse ratio, which could prevent delay of diagnostics and inadequate

monitoring. Allied support teams such as physiotherapists, geriatric nurses, should be easily

accessible and see the patient within 24 hours when necessary to prevent unnecessary delay

in patients’ recuperation. The acute medical unit is very suitable to consult other specialties

easily, because patients from multiple specialties are admitted to these units and therefore

the physicians from these specialties are often present during ward rounds. New concepts such

as a short (maximum 15 minutes) multidisciplinary meeting every morning with al involved

specialties (all supervisors, their juniors, nurses and allied health support teams) to shortly

discuss the medical and logistical bottlenecks of all the patients present at the AMU, will be

of help not only to enhance teamwork but also to improve the flow of patients. In the VU

University Medical Centre (VUmc) this is called the ‘standing session’ in which every morning

the diagnosis, therapeutic plan and predicted LOS for each patient on the AMU are briefly

presented. In this way every specialty knows about the patients they might be consulted

for, and the doctors and nurses are informed about the availability of beds on the wards.

Also, the team can easily identify with whom they will/can work together to take care of

their patient which improves communication and coordination around the patient. Because

of the high turnover, handovers should take place systematically at every change of shift from

team to the next team, in order to provide continuous care. Handover systems such as SBAR

(situation, background, assessment, recommendation) should be introduced and propagated

vigorously to make sure that the handovers are performed uniformly.43

The original design of the AMU was used for medical patients only, however to improve

the cooperation of specialty colleagues in early review and maximum use of available beds one

can debate whether surgical specialties should also admit patients to the AMU. Many hospitals

in the Netherlands have already chosen this construction. This construction may be beneficial,

because when all medical beds are occupied, one can use a surgical bed for a medical patient

and the other way around. One must however ensure that nurses are adequately trained in

caring for both surgical and medical patients.

Before implementation adequate training for the staff on the ward is required. Previous

literature has reported that AMUs in the past have found difficulties recruiting nurses and

allied health staff with appropriate levels of acute assessment skills.6 Nurses are not specially

trained for working on an AMU and the resources and time to train them should be allocated.

A recent study performed in Ireland has done this by training their nurses with a program

called ‘Nursing the Acute Medical Patient’ and it has made significant contributions to both

staff development and the delivery of evidence-based care in the AMU.44 Also, since the AMU

is a clinical setting in which many diagnostics are performed on divergent types of patients,

it is a useful place to train young doctors and medical trainees. A location for them such as

a ‘skills lab’ or a library on or near the AMU should be thought of beforehand.

As previously mentioned, in order to improve quality and compare different AMUs throughout

the Netherlands a uniform guideline for an AMU is needed. In the UK, the Society of Acute

42

Medicine has guidelines to help hospitals starting up and instruct them about minimal

requirements.45 These have been proven very effective, and at present hospitals in the UK

without an AMU can hardly be found. Also, once implemented, standardized evidence-

based protocols have been described as improving patient flow and quality of care.6,42

These guidelines also offer an adequate way to assess quality of care and advise on

improvements constructively.

Limitations

The limitations of our literature review must be acknowledged. Firstly, one-third of the articles

included, extracted their data from one hospital in Ireland and another hospital in the UK.

These results therefore may not be extrapolated to the all AMUs in these countries. Secondly,

the designs of the AMUs were rather heterogeneous, which makes it hard to deduce which

elements of the AMU correlate strongly to better patient outcomes. Thirdly, little is known

about the precise effect of an AMU on bed occupancy on other medical wards , this information

is essential as sparing these wards from a continuous stream of acutely ill patients and thereby

improving patient care in these units is also one of the underlying motives to reorganize into

an AMU.

CONCLUSIONThis study has shown that current literature proves the AMU to be an effective model to

provide acute care and improve efficiency in the care chain thereby improving the patient

flow in acute care chain. However, the current situation in the Netherlands shows that

AMUs are individually implemented and are therefore heterogeneously organized. To

optimize the effectiveness and compare the quality of care in our AMUs, a clear national

guideline is needed to provide a gold standard, especially since the trend shows that

more and more AMUs will be developed in the coming years. Very few studies have been

performed on the effectivity of AMUs in the Netherlands and more research is needed

to investigate the efficacy of AMUs in Dutch setting so that national guidelines can

be formulated.

CONFLICT OF INTEREST STATEMENTThe authors have nothing to disclose.

FUNDINGThis research did not receive any specific grant from funding agencies in the public, commercial,

or not-for-profit sectors.

SUPPLEMENTARY INFORMATION Appendix A. ‘Search strategy’

Appendix B. ‘Systemic overview included studies’

43

REFERENCES1. VUmc Netwerk Acute Zorg. Ontwikkelingen aanbod acute patiënten SEH’s RsHsQQ, 2014 en 2015

ROAZ-regio VUmc. https://www.vumc.nl/afdelingen-themas/719134/8513847/8514115/20150929_RapportAcuteZorgQ11.pdf. Published September 2015. Accessed August, 2016.

2. Plunkett PK, Byrne DG, Breslin T, Bennett K, Silke B. Increasing wait times predict increasing mortality for emergency medical admissions. Eur J Emerg Med. 2011;18(4):192-196.

3. Schrijver EJ, Toppinga Q, de Vries OJ, Kramer MH, Nanayakkara PW. An observational cohort study on geriatric patient profile in an emergency department in the Netherlands. Neth J Med. 2013;71(6):324-330.

4. Samaras N, Chevalley T, Samaras D, Gold G. Older patients in the emergency department: a review. Ann Emerg Med. 2010;56(3):261-269.

5. Consultancy C. Ligduurmonitor Nederlandse ziekenhuizen 2013 & 2014 ‘Trendbreuk in de ligduur’. December 2015.

6. Scott I, Vaughan L, Bell D. Effectiveness of acute medical units in hospitals: a systematic review. Int J Qual Health Care. 2009;21(6):397-407.

7. Beckett DJ, Raby E, Pal S, Jamdar R, Selby C. Improvement in time to treatment following establishment of a dedicated medical admissions unit. Emerg Med J. 2009;26(12):878-880.

8. Vreeland Group. De Acute Opname Afdeling ‘Het wat, waarom en hoe’. Published 2011. Accessed August 2016.

9. Cooke MW, Higgins J, Kidd P. Use of emergency observation and assessment wards: a systematic literature review. Emerg Med J. 2003;20(2):138-142.

10. Reid LE, Dinesen LC, Jones MC, Morrison ZJ, Weir CJ, Lone NI. The effectiveness and variation of acute medical units: a systematic review. Int J Qual Health Care. 2016.

11. National Institute for Health and Care Excellence Guidelines. www.nice.org.uk Accessed August, 2016.

12. Boyle A, Fuld J, Ahmed V, Bennett T, Robinson S. Does integrated emergency care reduce mortality and non-elective admissions? A retrospective analysis. Emerg Med J. 2012;29(3):208-212.

13. Boyle AA, Ahmed V, Palmer CR, Bennett TJ, Robinson SM. Reductions in hospital admissions and mortality rates observed after integrating emergency care: a natural experiment. BMJ Open. 2012;2(4).

14. Boyle AA, Robinson SM, Whitwell D, et al. Integrated hospital emergency care improves efficiency. Emerg Med J. 2008;25(2):78-82.

15. Downing H, Scott C, Kelly C. Evaluation of a dedicated short-stay unit for acute medical admissions. Clin Med (Lond). 2008;8(1):18-20.

16. McErlain-Burns T, Byrne C, Smith C. Admissions wards. Admitting a need for change. Health Serv J. 1997;107(5542):33.

17. Hadden DS, Dearden CH, Rocke LG. Short stay observation patients: general wards are inappropriate. J Accid Emerg Med. 1996;13(3):163-165.

18. Hanlon P, Beck S, Robertson G, et al. Coping with the inexorable rise in medical admissions: evaluating a radical reorganisation of acute medical care in a Scottish district general hospital. Health Bull (Edinb). 1997;55(3):176-184.

19. Pradhan S, Ratnasingham D, Subbe CP, Stevenson S, Ward D, Cooksley T. Society for Acute Medicine Benchmarking Audit 2015: The Patient Perspective. Acute Med. 2015;14(4):147-150.

20. St Noble VJ, Davies G, Bell D. Improving continuity of care in an acute medical unit: initial outcomes. QJM. 2008;101(7):529-533.

21. Sullivan P, Harris ML, Bell D. The quality of patient experience of short-stay acute medical admissions: findings of the Adult Inpatient Survey in England. Clin Med (Lond). 2013;13(6):553-556.

22. Basey AJ, Kennedy TD, Mackridge AJ, Krska J. Delays and interruptions in the acute medical unit clerking process: an observational study. JRSM Open. 2016;7(2):2054270415619323.

44

23. Brand CA, Kennedy MP, King-Kallimanis BL, Williams G, Bain CA, Russell DM. Evaluation of the impact of implementation of a Medical Assessment and Planning Unit on length of stay. Aust Health Rev. 2010;34(3):334-339.

24. Elder E, Johnston AN, Crilly J. Improving emergency department throughput: An outcomes evaluation of two additional models of care. Int Emerg Nurs. 2016;25:19-26.

25. Ong BS, Van Nguyen H, Ilyas M, Boyatzis I, Ngian VJ. Medical Assessment Units and the older patient: a retrospective case-control study. Aust Health Rev. 2012;36(3):331-335.

26. Slatyer S, Toye C, Popescu A, et al. Early re-presentation to hospital after discharge from an acute medical unit: perspectives of older patients, their family caregivers and health professionals. J Clin Nurs. 2013;22(3-4):445-455.

27. Suthers B, Pickles R, Boyle M, Nair K, Cook J, Attia J. The effect of context on performance of an acute medical unit: experience from an Australian tertiary hospital. Aust Health Rev. 2012;36(3):320-324.

28. Williams AG, Jelinek GA, Rogers IR, Wenban JA, Jacobs IG. The effect on hospital admission profiles of establishing an emergency department observation ward. Med J Aust. 2000;173(8):411-414.

29. Yong TY, Li JY, Roberts S, Hakendorf P, Ben-Tovim DI, Thompson CH. The selection of acute medical admissions for a short-stay unit. Intern Emerg Med. 2011;6(4):321-327.

30. Coary R, Byrne D, O’Riordan D, Conway R, Cournane S, Silke B. Does admission via an acute medical unit influence hospital mortality? 12 years’ experience in a large Dublin hospital. Acute Med. 2014;13(4):152-158.

31. Conway R, O’Riordan D, Silke B. Long-term outcome of an AMAU--a decade’s experience. QJM. 2014;107(1):43-49.

32. Moloney ED, Bennett K, O’Riordan D, Silke B. Emergency department census of patients awaiting admission following reorganisation of an admissions process. Emerg Med J. 2006;23(5):363-367.

33. Moloney ED, Bennett K, Silke B. Effect of an acute medical admission unit on key quality indicators assessed by funnel plots. Postgrad Med J. 2007;83(984):659-663.

34. Moloney ED, Smith D, Bennett K, O’Riordan D, Silke B. Impact of an acute medical admission unit on length of hospital stay, and emergency department ‘wait times’. QJM. 2005;98(4):283-289.

35. Relihan E, Glynn S, Daly D, Silke B, Ryder S. Measuring and benchmarking safety culture: application of the safety attitudes questionnaire to an acute medical admissions unit. Ir J Med Sci. 2009;178(4):433-439.

36. Rooney T, Moloney ED, Bennett K, O’Riordan D, Silke B. Impact of an acute medical admission unit on hospital mortality: a 5-year prospective study. QJM. 2008;101(6):457-465.

37. McNeill GB, Brand C, Clark K, et al. Optimizing care for acute medical patients: the Australasian Medical Assessment Unit Survey. Intern Med J. 2011;41(1a):19-26.

38. Abenhaim HA, Kahn SR, Raffoul J, Becker MR. Program description: a hospitalist-run, medical short-stay unit in a teaching hospital. CMAJ. 2000;163(11):1477-1480.

39. Vork JC, Brabrand M, Folkestad L, Thomsen KK, Knudsen T, Christiansen C. A medical admission unit reduces duration of hospital stay and number of readmissions. Dan Med Bull. 2011;58(8):A4298.

40. Lo SM, Choi KT, Wong EM, et al. Effectiveness of Emergency Medicine Wards in reducing length of stay and overcrowding in emergency departments. Int Emerg Nurs. 2014;22(2):116-120.

41. Galen LS, der Schors W, Damen NL, Kramer M, Wagner C, Nanayakkara P. Measurement of generic patient reported outcome measures (PROMs) in an acute admission unit: A feasibility study. Acute Med. 2016;15(1):13-19.

42. Bell D, Skene H, Jones M, Vaughan L. A guide to the acute medical unit. Br J Hosp Med (Lond). 2008;69(7):M107-109.

43. Veiligheidsmanagementsysteem. Communicatie tussen hulpverleners volgens het SBAR-proces. VMS Veiligheidsprogramma. 02-2009.

45

44. Mitchell L, Bury E, Leonard O. Meeting the learning needs of AMU nurses through collaborative working. Br J Nurs. 2016;25(6):314-318.

45. The Society of Acute Medicine. Quality Standarts and Toolkits. http://www.acutemedicine.org.uk/resources/quality-standards/ Accessed August, 2016.

CHAPTER 3

MEASUREMENT OF GENERIC PATIENT REPORTED OUTCOME MEASURES (PROMS) IN AN ACUTE

ADMISSION UNIT: A FEASIBILITY STUDY

Louise S. van Galen | Wouter van der Schors | Nikkie L. Damen | Mark H. Kramer | C. Wagner | P.W.B. Nanayakkara

Acute Med 2016;15(1):13-9

‘There is far too little thought about what it means to be a patient’ Edith Schippers

48

ABSTRACTObjective

Measuring patient-reported outcome measures (PROMs) is a challenge in Acute Admission

Units (AAUs), where patients present with a variety of pathologies. Generic PROMs may be

used to measure the quality of care in this population. The main objective of this study was to

assess the feasibility of measuring generic PROMs in a Dutch AAU.

Design

Longitudinal cohort study

Setting

An AAU of a tertiary hospital in Amsterdam, the Netherlands

Participants

123 patients admitted to the AAU during 5 weeks in May and June 2015

Methods

Patients admitted to the AAU were asked to fill out a questionnaire relating to three time points:

7 days before, during, and within 2 weeks after admission. Additionally, patients were asked

to report on their experienced level of safety on the AAU and the contribution of the AAU to

their recovery.

Results

There were significant trends in generic PROMs for all three domains. Physical functioning

decreased during hospital admission and almost fully returned to the previous level after

discharge. Satisfaction with social role and anxiety significantly decreased over time.

Conclusions

Measuring generic PROMs in the AAU is feasible. The analysis of the PROMs took little effort

and results could be reported back to the healthcare workers on the AAU quickly. Patients

appreciated being asked about their own perceived health and the quality of care. Given that

this is the first study focusing on PROMs in AAU patients in the Netherlands, future studies with

larger sample sizes, and from other nations are needed to further investigate PROMs in this

patient group to establish International reference values.

Keywords

PROMs, Acute Admission Unit, patient experience, measurement of quality, patient safety

Keypoints

1. Measuring and analysis of generic PROMs in a Dutch AAU is feasible.

2. Results could be reported back to the healthcare workers on the AAU quickly.

3. Patients appreciated being asked about their own perceived health and the quality

of care.

4. There were significant trends in generic PROMs for all domains.

49

INTRODUCTIONAlthough widely recognised in the UK, Acute Admission Units (AAU) are a relatively new

concept in the Netherlands. The AAU in VUMC is a designated ward where a multidisciplinary

medical and surgical team manages patients for up to 72 hours, following emergency

admission via the Emergency Department (ED) or outpatient clinic. The AAU is staffed and

equipped to receive medical and surgical inpatients needing treatment for acute illnesses,

along similar lines to models described in the UK.1 In the Netherlands, where about two

million patients rely on acute care at the emergency department every year, a large number

of (academic) hospitals have implemented an AAU.2 In contrast to other countries in Europe

such as the United Kingdom, where typically only medical patients are treated, the AAU in

the Netherlands also houses beds for a broader set of specialities, including surgery, trauma,

orthopaedics, and urology.

Because of this heterogeneity it can be a challenge to measure patient experiences and quality

of care in an AAU universally. In an age where growing attention is paid to the patients’

perspective on quality of care and treatment effectiveness, evaluating ‘patient-related

outcome measures’ (PROMs) has become increasingly important. To assess PROMs, standard

measurement tools have been developed to capture patients’ perceptions of their general health

and/or specific diseases and conditions.3,4 PROMs can provide information on how a patient

appreciates his or her own health status, as well as the experienced impact of a (surgical)

procedure or a prescribed treatment.5 PROMs are increasingly utilised alongside clinical quality

indicators to enhance shared decision-making between patients and clinicians.2,6,7

The National Institution of Health (NIH) in the United States has developed the Patient-

Reported Outcomes Measurement Information System. (PROMIS®)8-10 to enable validated and

uniform implementation PROMs in clinical practice. Disease-specific PROMs 11 are designed

to evaluate the impact of a specific condition on patients’ functioning.8,11-16 Generic PROMs,

enable various aspects of functioning, such as self-care, physical functioning, and anxiety to

be measured.3,10

Until now, the use of PROMs in the Netherlands has been largely restricted to orthopaedic

patients, and implementation in other departments is still in early stages.17,18 Several

International studies have been performed using and validating PROMs in a more heterogeneous

population.5,19-21 In the Netherlands, however, no study has investigated the feasibility of

the generic PROMs in a general Dutch hospital population (regardless of specialty or diagnosis

at admission).3

It seems logical that, given the heterogeneity of AAU patients, generic PROMs are used to

assess quality of care based on patients’ experiences.

50

The main objectives of this longitudinal cohort study were:

1. To assess the feasibility of measuring generic PROMs in a Dutch Acute

Admission Unit;

2. To investigate whether generic PROMs could be used to measure physical functioning,

anxiety and satisfaction with social role in a general Dutch AAU population.

Secondary aims were to evaluate aspects of other key elements of quality: the feeling of safety

and overall experiences in an AAU through the patient’s perspective.

METHODS Study design

Between May and June 2015, in a period of 5 weeks (day shifts Monday to Friday), patients

admitted to the AAU were invited to participate in this study by the researcher. 123 patients

aged 18 or over were recruited as soon as possible upon their admission and included in

the study. Following the morning handover one of the two professionally trained researchers

(WS or LG) recruited the patients. Patients were not asked to participate if they were unable

to give informed consent as judged by researcher, for example because of severity of illness,

impaired cognition, language barriers, or strict isolation measures. After obtaining informed

consent, participants were requested to complete the questionnaire. First, they were asked

to judge in retrospect (T0) the situation 7 days prior to receiving care on AAU and second,

the situation at the moment of completing the questionnaire (T1) (Table I). Depending on

the capability and preference of the patient, the coordinating researchers either asked

questions himself or let the patient fill out the questionnaire on his or her own. During and

after completing the questionnaire, the researcher documented the time taken to finish

the questionnaire and monitored any extra comments or observations made by the patient.

In this way, practical feasibility was observed. Additionally, participants were asked if they had

any remarks on the questionnaire or advice for improvement.

Ten days after the discharge from AAU the patient was called (T2) to obtain a phone interview

provided the patient granted permission to contact him through the phone during the index

interview. Patients were excluded from telephone follow-up if they were still in hospital at T2,

having been transferred to another clinical area on discharge from AAU. During this telephone

call, participants were once again asked to assess their health at that moment and the level of

satisfaction regarding the care received during their admission (T2). The local Ethics Committee

approved the study protocol.

Questionnaire development

The questionnaire comprised questions from three domains of the Dutch version of

the PROMIS Profile 29.0 (approved and translated by the Dutch-Flemish PROMs group in 2014),

supplemented by a few short questions on the evaluation of patient safety, general functioning

and overall experience on the ward (see table I). PROMIS 29.0 items are rated on a 5-point

Likert scale.22 The following domains of the PROMIS 29.0 were selected: anxiety, physical

51

functioning, and satisfaction with social role. These domains were selected on the basis of

their appropriateness and suitability within the patient population, after consulting researchers

specialised in working with and translating PROMs, epidemiologists, and supervising clinicians.8

Completion of the questionnaire took around 5-10 minutes per patient.

As is shown in figure 2, during the telephone follow-up 10 days after discharge from the AAU

(T2), the same 12 items from the PROMIS 29.0 were asked. In addition, 4 other questions

about the perception of safety, contribution of AAU towards recovery, a question about

general functioning and an item from the Net Promotor Score (NPS) were asked.23 The NPS is

based on the question ‘Would you recommend the hospital to friends and family?’ The score

is calculated by subtracting the percent of patients rating the hospital 0-6 from the percent

recommending the hospital with a score of 9 or 10. In total, the questionnaire at T2 contained

16 questions. Appendix I provides the complete questionnaire.

Data analyses

Categorical outcome measures were presented as frequencies and percentages, whereas

continuous variables were presented by mean and standard deviation and/or median and

interquartile range. Since the PROMs variables were not normally distributed, we chose a non-

parametric test over the T-test for paired samples. We used Wilcoxon signed-rank test to

compare the PROMs scores on T0, T1 and T2. To control for the family wise error rate of

the three comparisons (T1 vs T0, T2 vs T1 and T2 vs T1) separately within each domain we

used a Bonferroni correction and set the two-sided significance level at p=0,05/3=0,017.

Satisfaction with social role was compared between T0 and T2 using a Wilcoxon rank-sum test

using a two-sided significance level of 0,05 since there was no measurement of satisfaction

with social role on T1. All statistical analyses were performed with Statistical Package for Social

Sciences (SPSS) version 20.

Figure 1 | Patient characteristics

52

Figure 2 | Patient inclusion

53

RESULTSPatient characteristics

The total study population consisted of 79 men (61.7%) and 49 women (38.3%) with a mean

age of 58.8 years (SD 18.7, range 46-72). The mean length of stay in the AAU was 45 hours

and 48 minutes (SD 32 hours). A total of 99 patients entered the AAU through the ED (80.5%).

Other patient flow characteristics are shown in Figure I. Admitting speciality varied widely, thus

creating a cross-section of an overall hospital patient population. Internal medicine (n=19,

15.5%), gastroenterology (n=17, 13.8%), and traumatology (n=15, 12.2%) were the largest

specialty groups, followed by pulmonary medicine (n=13, 10.6%), orthopaedics (n=11, 8.9%),

nephrology (11, 8.9%), urology (n=8, 6.5%), general surgery (n=7, 5.7%), haematology (n=7,

5.7%), neurology (n=6, 4.9%), and vascular surgery (n=2, 1.6%). A total of 33 patients had

been admitted into a hospital within the 30 days prior to the AAU admission, indicating

a readmission rate of 26.8% in this population. For 6 patients this AAU admission was their

first time in hospital (4.9%), whereas for 87 patients this admission was the first time on an

AAU (72.5%).

Feasibility

Figure II shows patient inclusion. A total of 123 patients completed the questionnaire at T1.

During the study period, a total of 195 patients were admitted to the AAU. Seventy-two

patients were not approached for inclusion on T1 since they were not eligible. The main

reasons for not approaching for inclusion were severity of illness (n=29, 41.4%), logistic

Table 1 | Content questionnaire at T0, T1, T2

54

11,0 (5,7)

5,4 (2,6)

12,4 (6,4)

6,7 (3,3)

8,5 (4,8)

11,8 (6,1)

7,1 (3,6)

12,9 (6,1)*

0 2 4 6 8 10 12 14

Satisfaction with Social Role

Anxiety

Physical Function

T0 (7 days prior to admission) T1 (on AMU) T2 (after discharge)

reasons (being away for surgery etc.) (n=23, 32.9%) and language barriers (n=9, 12.9%). At

T1, 125 patients were approached for inclusion by coordinating researchers. The vast majority

of patients (n=123, 98.4%) participated, only 2 (1.6%) refused to participate. More than half

of the included patients (N=65, 52.8%) also completed the telephone questionnaire within 10

days after discharge from the Acute Admission Unit.

A few patients (5.7%, n=7) did not complete the total questionnaire at T1. Finishing

the questionnaire took 6 (SD 2.9) minutes on average. However, patients who took

the opportunity to share their worries and feelings with the coordinating researcher needed

longer time to finish their questionnaire which explains the wide range in time. A group of

105 patients understood all the questions asked (85.4%), whereas 14 participants (11.4%)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

To what extent has the AMU contributed to your recovery?(T2, N=64)

Did you feel safe during your stay on the AMU? (T2, N=65)

Do you feel safe on the AMU? (T1, N=123)

Not at all A little bit Somewhat Quite a bit Very much

Figure 3 | Physical functioning, anxiety and satisfaction with social role*Mean score (standard deviation)

Figure 4 | Feelings of safety and contribution to recovery on the AAU

55

needed additional explanation. For various reasons, the majority of patients (n=106, 86.2%)

chose to obtain help in completing the questionnaire, resulting in the researcher reading

out the questions. However, the researcher judged that about half of them were physically

able to fill out the questionnaire themselves (n=51, 48.1%). During telephone conversations

patients firstly expressed their appreciation for someone calling them after discharge and

they valued the fact that their unanswered questions after hospital admission were answered.

In general, all patients understood the follow-up questions appropriately and were able to

answer the questions.

Eighteen patients (14.6%) refused telephone follow-up. Thirty-three (31.4%) other patients

did give permission for telephone follow-up but were not included at T2. This is because at T2

a third of the patients included at T1 were not immediately discharged home (n=33, 31.4%),

because they either received additional care on other wards or health institutions, and therefore

not eligible to be included (figure II). At T2, three patients could not be included because of

they were not reachable at given numbers (7.5%). In addition, 4 patients in this group died

during hospital admission (10%). None of the participants had remarks or suggestions for

improvement of the questionnaire at either T1 or T2.

Patient reported outcome measures

At three time points (T0, T1, and T2) patients were asked to report on the PROMs domains

physical functioning, anxiety and satisfaction with social role. The mean scores and standard

deviations on these scales are given in Figure III. There was a significant decrease in physical

functioning between T0 and T1 (p<0.001), followed by a significant increase between T1 and

T2 (p<0,001). Patients almost fully recovered to their pre-admission level, however physical

functioning was still significantly lower at T2 when compared to T0 (p = 0,008). Satisfaction

with social role also significantly decreased between T0 and T2 (p=0,01). Experienced anxiety

also declined between T1 and T2 (p=0.007) and between T0 and T2, (p=0.000). 

At T2, 96.9% (n=63) of the patients felt they were in a safe environment on the AAU (‘quite

a bit’ to very much’) – figure III. In addition 60.0% (n=39) of the patients indicated that their

stay at the ward contributed to their recovery (‘quite a bit’ to ‘ very much’). The average Net

Promotor Score at T2 (a 0-10 score) was 39,7%. This score implies that around 40.0% of

patients at T2 would recommend the hospital with a 9 or 10. This is in line with the overall

hospital Net Promotor Score of 40.0 in this hospital.

DISCUSSIONSummary and interpretations

The results of this study suggest that measuring generic PROMs on the AAU is feasible. Overall,

patients were positive about the questionnaire, resulting in a high response rate. The questions

asked were easily interpretable and applicable to the diversity of patients on the AAU. Because

the questionnaire could be filled out on paper or with assistance of coordinating researcher

56

it is applicable to the diverse population on the AAU. Since filling out the questionnaire was

not time consuming (approximately 6 minutes), this also can be seen as an indication of its

feasibility. For telephone follow-up (approximately 5 minutes) most patients also showed their

readiness to be approached again. Patients were positive about the questionnaire since it gave

them the feeling the hospital was really trying to take their perspectives on perceived care into

account. Also it gave them the opportunity to share their experiences and were told where

to look for answers to certain questions. The fact that PROMs took little effort and results

could be reported back to the healthcare workers on the AAU quickly was also an important

finding. This study also shows that generic PROMs were suitable for all patients on the AAU;

no adaptations were necessary for specific patient groups. Results indicated significant trends

in PROMs for all three domains. Physical functioning decreased during stay and almost fully

returned to old level after discharge, and anxiety decreased significantly over time. These

findings suggest that generic PROMs are able to illustrate changes in patient perspectives on

own perceived wellbeing before, during and after hospital admission.

This study also aimed to evaluate other aspects of key elements of quality: feelings of safety

and overall patient experiences on the AAU through patients’ perspective. In general, most

patients showed a positive response regarding their feeling of safety during their stay on

the AAU and regarding the contribution of the AAU to their recovery.

Relation to other evidence

To our knowledge, this was the first study performed in the Netherlands to investigate generic

PROMs in an Acute Admission Unit. Black (2013) noted that emergency admissions present

a challenge in assessing patients’ experiences of care since PROMs are often only available

after the acute event.3 The current study addresses this issue, by assessing PROMs on an

AAU. In comparison to most PROMs research, where focus is mostly on effectiveness of

care using PROMIS®, this is the first study that adds safety and experience. Grosse (2012)

reported that poor experience of safety has an adverse impact on the effectiveness of care

and therefore these key dimensions should not be ignored.24 Previous work has demonstrated

that generic PROMs can highlight physical or psychological problems which might otherwise

be overlooked.5 Generic PROMs could therefore function as a starting point to discuss more

appropriate care for the individual patient, thus enhancing shared-decision making.

A problem arising whilst assessing the PROMS in the Netherlands is that, in contrast to other

countries, no Dutch reference values are available to enable comparison.19 International data,

such as the Net Promotor Score could be obtained25, although comparisons across different

Nations can be problematic.26 The UK ‘Friends and Family test’, for example, which assesses

the likelihood of the patient recommending an NHS hospital, uses a different scoring system.23,25

Limitations

As already stated by McKenna (2011) in a well-developed measure, patients will only be asked

questions that are relevant. In a heterogeneous group of patients, addressing these relevant

57

areas can be a challenge because these issues may minimize respondent acceptability and

increase missing data. In this study we decided to take three specific domains since it was

difficult to capture other areas of concern applicable to all patient populations. This may have

led to our missing data on areas that could be of concern to a specific patient on the AAU.27

Another limitation of this study is the direct approach of the coordinating researcher who in

many cases assisted with the completion of the questionnaire. In practice there would not

be a researcher available on every AAU. Since listening to the questionnaire being read out

might lead to different scores than personal viewing of the questionnaires (one is probably

more inclined to give extreme answers once one hears 5 answers rather than seeing them on

paper) this might lead to less nuance/other results. In addition, we excluded the patients who

were unable to give informed consent, for example because of severity of illness. This may

have influenced the case-mix of the group. Finally, a general caveat of this study is the loss

in follow-up at T2. Since a substantial group of the patients were still in hospital after their

initial admission on the AAU, it can be challenging to create an appropriate time frame to

ask questions for T2. Future studies should investigate measuring T2 at a later time point to

achieve more complete follow-up information.

CONCLUSION AND IMPLICATION FOR PRACTICE The current study is the first study performed in the Netherlands to assess PROMs on the AAU.

The most important finding emerging from this study is that the measurement of generic

PROMS in the Acute Admission Unit is feasible. This study has shown that, in a heterogeneous

patient population PROMs may be used to measure the wellbeing and feeling of safety through

the patient’s perspective. Patients appreciated the fact that their view on these matters was

taken into account contributing to the high percentage of overall patient participation.

The analyses of the PROMs took little effort and results could be reported back to the healthcare

workers on the AAU quickly.

Future studies with larger sample sizes are warranted to establish reference values and ideal

follow-up time to compare PROMs throughout the Netherlands. 

58

REFERENCES1. Scott I, Vaughan L, Bell D. Effectiveness of acute medical units in hospitals: a systematic review. Int J

Qual Health Care. 2009;21(6):397-407.

2. Dutch national Institute for Public Health and Environment. Acute Care. Accessed August 2015.

3. Black N. Patient reported outcome measures could help transform healthcare. BMJ. 2013;346:f167.

4. Marshall S, Haywood K, Fitzpatrick R. Impact of patient-reported outcome measures on routine practice: a structured review. J Eval Clin Pract. 2006;12(5):559-568.

5. Valderas JM, Kotzeva A, Espallargues M, et al. The impact of measuring patient-reported outcomes in clinical practice: a systematic review of the literature. Qual Life Res. 2008;17(2):179-193.

6. Stiggelbout AM, Van der Weijden T, De Wit MP, et al. Shared decision making: really putting patients at the centre of healthcare. BMJ. 2012;344:e256.

7. Lamberts MP, Drenth JP, van Laarhoven CJ, Westert GP. [Outcome of treatment reported by patients: instrument to reduce variations in clinical practice]. Ned Tijdschr Geneeskd. 2013;157(7):A5369.

8. Terwee CB, Roorda LD, de Vet HC, et al. Dutch-Flemish translation of 17 item banks from the patient-reported outcomes measurement information system (PROMIS). Qual Life Res. 2014;23(6):1733-1741.

9. Cella D, Yount S, Rothrock N, et al. The Patient-Reported Outcomes Measurement Information System (PROMIS): progress of an NIH Roadmap cooperative group during its first two years. Med Care. 2007;45(5 Suppl 1):S3-S11.

10. PROMIS ADULT PROFILE INSTRUMENTS, A brief guide to the PROMIS Profile instruments for adult respondents 2014: http://www.assessmentcenter.net/documents/PROMIS%20Profile%20Scoring%20Manual.pdf Accessed August 2015.

11. Spiegel BM, Hays RD, Bolus R, et al. Development of the NIH Patient-Reported Outcomes Measurement Information System (PROMIS) gastrointestinal symptom scales. Am J Gastroenterol. 2014;109(11):1804-1814.

12. Zorg Instituut Nederland https://www.zorginstituutnederland.nl/kwaliteit/toetsingskader+en+register/proms Accessed July, 2015.

13. Orbai AM, Bingham CO, 3rd. Patient reported outcomes in rheumatoid arthritis clinical trials. Curr Rheumatol Rep. 2015;17(4):28.

14. Rolfson O, Karrholm J, Dahlberg LE, Garellick G. Patient-reported outcomes in the Swedish Hip Arthroplasty Register: results of a nationwide prospective observational study. J Bone Joint Surg Br. 2011;93(7):867-875.

15. Flynn KE, Dew MA, Lin L, et al. Reliability and construct validity of PROMIS measures for patients with heart failure who undergo heart transplant. Qual Life Res. 2015.

16. Irwin DE, Gross HE, Stucky BD, et al. Development of six PROMIS pediatrics proxy-report item banks. Health Qual Life Outcomes. 2012;10:22.

17. Working on quality with PROMS: http://www.netqhealthcare.nl/patient-reported-outcome-measures/ Accessed August, 2015.

18. Zwaap J, Derksen J, Enzing J. Analysis for consultation care in knee and hip arthritis. Dutch Care Institute ( In Dutch). 2014.

19. Jensen RE, Potosky AL, Reeve BB, et al. Validation of the PROMIS physical function measures in a diverse US population-based cohort of cancer patients. Qual Life Res. 2015.

20. Sullivan P, Harris ML, Bell D. The quality of patient experience of short-stay acute medical admissions: findings of the Adult Inpatient Survey in England. Clin Med. 2013;13(6):553-556.

21. Cleary PD, Edgman-Levitan S, Roberts M, et al. Patients evaluate their hospital care: a national survey. Health Aff (Millwood). 1991;10(4):254-267.

22. Elaine IE, Seaman CA. Likert scales and data analyses. Quality Progress 40.7 2007: 64-65.

59

23. The Net Promoter Score and System: http://www.netpromoter.com/why-net-promoter/know. Accessed July, 2015.

24. Grosse Frie K, van der Meulen J, Black N. Relationship between patients’ reports of complications and symptoms, disability and quality of life after surgery. Br J Surg. 2012;99(8):1156-1163.

25. Friends and Family Test data http://www.england.nhs.uk/ourwork/pe/fft/friends-and-family-test-data/ [Consulted 20-10-2015].

26. Hamilton DF, Lane JV, Gaston P, et al. Assessing treatment outcomes using a single question: the net promoter score. Bone Joint J. 2014;96-B(5):622-628.

27. McKenna SP. Measuring patient-reported outcomes: moving beyond misplaced common sense to hard science. BMC Med. 2011;9:86.

PREVENTION OF SERIOUS ADVERSE EVENTS ON THE CLINICAL WARDS

CHAPTER 4

DELAYED RECOGNITION OF DETERIORATION OF PATIENTS IN GENERAL WARDS IS MOSTLY CAUSED BY

HUMAN RELATED MONITORING FAILURES: A ROOT CAUSE ANALYSIS OF UNPLANNED ICU ADMISSIONS

“UNPLANNED ICU ADMISSIONS: BEWARE OF THE MONITORING FAILURES”

Louise S. van Galen & Patricia W. Struik | Babiche E. Driesen | Hanneke Merten | Jeroen Ludikhuize | Johannes I. van der Spoel | Mark H. Kramer | Prabath W. Nanayakkara

PLoS One 2016;11(8):e0161393

‘Never let a good crisis go to waste’  Winston Churchill

62

ABSTRACTBackground

An unplanned ICU admission of an inpatient is a serious adverse event (SAE). So far, no in

depth-study has been performed to systematically analyse the root causes of unplanned

ICU-admissions. The primary aim of this study was to identify the healthcare worker-,

organisational-, technical,- disease- and patient- related causes that contribute to acute

unplanned ICU admissions from general wards using a Root-Cause Analysis Tool called

PRISMA-medical. Although a Track and Trigger System (MEWS) was introduced in our hospital

a few years ago, it was implemented without a clear protocol. Therefore, the secondary aim

was to assess the adherence to a Track and Trigger system to identify deterioration on general

hospital wards in patients eventually transferred to the ICU.

Methods

Retrospective observational study in 49 consecutive adult patients acutely admitted to

the Intensive Care Unit from a general nursing ward. 1. PRISMA-analysis on root causes

of unplanned ICU admissions 2. Assessment of protocol adherence to the early warning

score system.

Results

Out of 49 cases, 156 root causes were identified. The most frequent root causes were

healthcare worker related (46%), which were mainly failures in monitoring the patient. They

were followed by disease-related (45%), patient-related causes (5%), and organisational root

causes (3%). In only 40% of the patients vital parameters were monitored as was instructed

by the doctor. 477 vital parameter sets were found in the 48 hours before ICU admission, in

only 1% a correct MEWS was explicitly documented in the record.

Conclusions

This in-depth analysis demonstrates that almost half of the unplanned ICU admissions from

the general ward had healthcare worker related root causes, mostly due to monitoring failures

in clinically deteriorating patients. In order to reduce unplanned ICU admissions, improving

the monitoring of patients is therefore warranted.

Keywords

Root cause analysis, Unplanned ICU admission, Early warning score, Patient safety,

Healthcare quality

63

INTRODUCTIONDespite steady improvement in hospital care and programs to reduce harm in hospitalised

patients, serious adverse events (SAEs) are still common.1-4 In one out of ten patients,

the occurrence of an SAE during their admission contributes to permanent disability or to

death.5 Previous studies have documented that some adverse events are probably preventable

and that they are often related to errors in management 4. An unplanned admission to

the Intensive Care Unit (ICU) of an inpatient from a general ward is considered as a SAE.

According to the NICE (National Intensive Care Evaluation) criteria, an unplanned ICU admission

is defined as an admission that could not have been deferred without risk for at least 12 hours.6

It is known that unexpected ICU admission leads to a poor long term survival, especially in

older patients.7

Hence, it is important to detect deteriorating patients timely and treat them early in order to

prevent an eventual ICU admission or to transfer them to the ICU on time to improve clinical

outcomes. One of the ways to achieve this is by implementing a Rapid Response System

(RRS).8,9 These systems, composed of an afferent and efferent limb, are specifically designed

to enable early recognition and management of deteriorating patients on general wards.

The efferent limb consists of trained ICU personnel forming a Rapid Intervention Team (RIT)

who deliver immediate treatment to deteriorating patients at the bedside after being called in

by ward clinical staff. The clinical staff can detect patients deterioration early by using a Track

and Trigger Systems (TTS) such as Modified Early Warning Score (MEWS).10,11 This afferent limb

plays a key-role in limiting unplanned ICU admission, since early detection of deteriorating

patients could potentially prevent this.

However, the effectiveness of these TTSs in preventing unplanned ICU admissions remains

unclear.8,12 In addition, (root) causes leading to delayed detection of these deteriorating

patients in the wards are mostly unknown. The circumstances leading to delayed detection of

deteriorating patients in the wards are probably multi-causal. Current literature has identified

certain patient characteristics associated with unplanned admissions into the ICU, such as

age, and having a surgical procedure.13 Also, certain iatrogenic events such as disease induced

by a drug prescribed or environmental events (falls) have shown to cause ICU admissions.14,15

These studies, however, tell comparatively little about the healthcare worker related and

organisational/system related root causes, which, if known, can be used to improve early

detection of deteriorating patients and potentially avoid unplanned ICU-admissions.

A useful tool analysing these types of root causes is the PRISMA-tool (Prevention and Recovery

Information System for Monitoring and Analysis). The main goal of the PRISMA method is

to build a database of incidents and process deviations by creating causal trees, from which

conclusions may be drawn to suggest optimal countermeasures. This in-depth analysis method

has been accepted by the World Alliance for Patient Safety of the World Health Organisation

and has shown to provide effective starting points for improvement in quality of care.16-19

64

No study has yet been performed to systematically analyse and identify the root causes

leading to late detection and treatment of deteriorating patients in wards. To formulate

possible improvement and prevention strategies with the aim of reducing the number of ICU

admissions insight into these causes is essential. Also, an optimally implemented RRS could

contribute to increased patient safety by recognising those in need of extra care without any

unnecessary delay.

Therefore, the main aims of this retrospective record review study were to (1.) analyse

the healthcare worker-, organisational-, technical-, disease- and patient- related causes that

contribute to late detection and treatment of deteriorating patients in the general wards

using PRISMA-medical analysis, and (2.) assess the adherence to and effectiveness of an

already implemented TTS on general wards in the early recognition of deteriorating patients

transferred to the ICU.

MATERIAL AND METHODSStudy design

This retrospective, record review study included unplanned ICU admissions from general

wards in the VU University Medical Centre in Amsterdam (VUmc), The Netherlands. This is

an academic 733-bedded medical centre with approximately 50,000 admissions per year.

The Adult Intensive Care Unit consists of 22 beds.

We decided to include the first 50 consecutive patient records of 2015 meeting the inclusion

criteria to explore the causes of unplanned ICU-admissions, since previous studies have shown

that around 50 PRISMA-analyses are credible and provide a well-founded causal-profile.16,19

The following criteria were used for inclusion: all patients on general wards aged 18 years and

older who were admitted to the ICU unplanned according to the NICE criteria (“an admission

that could not have been deferred without risk for at least 12 hours”).6 Only the first ICU

admission of the patient was included.

Excluded from the study sample were: patients admitted on the ICU immediately from

the emergency department, the operation room, medium care, high care, or coronary care;

patients transferred from other hospitals; patients with a planned ICU admission (i.e. after

surgery). The Ethics committee of VU University Medical Centre, Amsterdam, approved

the study and necessity for informed consent was waived. Information from the clinical records

was anonymized and de-identified prior to analysis.

The RRS used in the VUmc is the MEWS, validated by Subbe et al. in 2007, which was

implemented in the VUmc in 2014.11 The MEWS used at the VUmc consists of an easy-to-use

algorithm of seven parameters: respiratory rate, saturation rate, heart frequency, systolic blood

pressure, temperature, consciousness, urine production (Fig. 1). Also, 1 point can be added if

a nurse is worried about the patient. If a vital parameter was not documented in the system,

65

this parameter was considered to be normal, therefore 0 points were given. When a patient

on a general ward has a score of 3 or higher, the nurse is expected to call the ward doctor

on duty or can immediately call the RIT team (Rapid Intervention Team). However, the nurses

were not required to perform MEWS on daily basis on set time points and the MEWS was

mostly performed on indication which probably led to late detection of the deteriorating

patient. It was decided to reintroduce the MEWS protocol in 2015, this study was performed

before reimplementation.

Assessment

For analysis doctor’s charts, nurse’s charts and electronic patient files including all test results

were available. Using a standardised abstraction form for each individual ICU-admission,

patient characteristics such as the age, mortality, APACHE scores and circumstances under

which ICU-admission took place (such as length of stay in the wards and admission speciality)

were collected (S1 File).20-22 Vital parameters measured and the use of the MEWS-protocol in

the 48 hours before the acutely unplanned ICU admission were also systematically registered

into this form.

Two medically and PRISMA-trained investigators (BD, PS) reviewed each case extensively

and filled out these chart abstraction forms. This analysis was a labour intensive process

(approximately 90 minutes average per investigator per chart). Subsequently they composed

individual causal trees; thereafter consensus was reached in a structured meeting with a third

independent experienced medical and PRISMA reviewer (LG), from which a composite root

causal tree was constructed. Finally, all cases and their causal trees were discussed with two

senior physicians (PN, JL) and a psychologist with a special interest in PRISMA-analysis (HM),

which resulted in the final root causal trees.

PRISMA-analysis

Fig. 2 shows three examples of causal trees from this study. In general the PRISMA-method

can be used to examine latent factors (technical and organisational), active failures (human)

and other factors (patient-related and other). A PRISMA-analysis starts with an incident, in this

study the unplanned ICU-admission was seen as the incident. This incident was then placed

at the top of the causal tree and was the starting point for the analysis. The next step was to

identify the direct causes underlying the unplanned ICU-admission. Direct causes are revealed

by asking ‘why’ this incident has occurred. Information from the records was used to identify

these direct causes. Below every direct cause the indirect causes were inserted. By constantly

asking ‘why’ an event had taken place, relevant indirect causes were systematically exposed.

If no more objective information was acknowledged as a cause, the last noted indirect cause

was labelled as a root cause and was placed at the bottom of the causal tree. The analysis

also stopped when the underlying cause lied outside of the hospital. The root causes were

then classified using the Eindhoven Classification Model (ECM), which is the corresponding

taxonomy of PRISMA-medical to classify root causes. The ECM is based on the skill-rules-

knowledge-based behavioural model of Rasmussen and on the system approach to human

66

error of Reason.23-25 The ECM was also used as a foundational component in the conceptual

framework for the International Classification for Patient Safety.26,27 Table 1 shows

the classification of the root causes according to the Eindhoven Classification Model.28 Since

previous studies have shown that progression of disease is a valuable addition to this model

applied in medical settings, this was added to the root causes for the current study.16

Statistical analysis

Using IBM SPSS Statistics, Chicago, USA, Version 22.0 descriptive characteristics and

frequencies were calculated. Categorical outcome measures are presented as frequencies and

percentages. For continuous variables we chose to use median and interquartile ranges since

none of them were normally distributed.

RESULTSPatient characteristics

PRISMA-analysis was performed on 49 available cases, since one chart was unavailable for

review during the study period. Table 2 displays baseline patient characteristics. The median

age was 69 years (range 34-90), and both sexes were represented almost equally (47% vs

53%). Nineteen patients died during their hospital admission, resulting in 39% in-hospital

mortality. The three scores for ICU-mortality (SAPS II, APACHE II, APACHE IV) had a median of

51, 24, and 95 resulting in a predicted mortality of 50%, 55%, 36% respectively. The median

time between hospital admission and ICU admission was 88 hours and 34 minutes (range

1h38 min – 733h). Most patients were transferred to the ICU between midnight and six

o’clock in the morning. Sixty-nine percent of the patients (n=34) had a ‘Do not resuscitate’-

policy (DNR), for three patients the DNR-policy was not clear (6%).

Figure 1 | MEWS protocol in VUmc

67

Figure 2 | Three examples of root causal trees**HKK: Human-related knowledge behaviour, HRI: Human-related intervention, PRF: Patient-related factor, HRM: Human-related monitoring, HRV: Human-related verification, DRF: Disease-related factor, HSS: Human-related skills-based.

68

Tab

le 1

| D

escr

iptio

n of

cat

egor

ies

of t

he E

indh

oven

Cla

ssifi

catio

n m

odel

: PRI

SMA

- m

edic

al V

ersi

on 17

,18

Mai

n c

ateg

ory

Sub

cate

go

ryC

od

eD

escr

ipti

on

Exam

ple

s (i

f av

aila

ble

) *

Tech

nica

lEx

tern

alT-

exTe

chni

cal f

ailu

res

beyo

nd t

he c

ontr

ol o

f th

e or

gani

zati

on.

Not

ava

ilabl

e

Des

ign

TDFa

ilure

s to

poo

r de

sign

of

equi

pmen

t et

c.N

ot a

vaila

ble

Con

stru

ctio

nTC

Cor

rect

des

ign

inap

prop

riat

ely

cons

truc

ted

or p

lace

d.N

ot a

vaila

ble

Mat

eria

lsTM

Mat

eria

l def

ects

not

cla

ssifi

ed u

nder

TD

or

TC.

Not

ava

ilabl

e

Org

aniz

atio

nal

Exte

rnal

O-e

xFa

ilure

s at

an

orga

niza

tion

al le

vel b

eyon

d th

e co

ntro

l and

re

spon

sibi

lity

of t

he in

vest

igat

ing

team

. N

ot a

vaila

ble

Tran

sfer

of

know

ledg

eO

KFa

ilure

res

ulti

ng f

rom

inad

equa

te m

easu

res

to t

rain

or

supe

rvis

e ne

w o

r in

expe

rien

ced

staf

f.

Not

ava

ilabl

e

Prot

ocol

sO

PFa

ilure

s re

lati

ng t

o th

e qu

alit

y or

ava

ilabi

lity

of a

ppro

pria

te

prot

ocol

s.N

ot f

ollo

win

g pa

in t

reat

men

t pr

otoc

ol a

fter

sur

gery

Man

agem

ent

prio

riti

esO

MIn

tern

al m

anag

emen

t de

cisi

ons

whi

ch r

educ

e fo

cus

on

pati

ent

safe

ty w

hen

face

d w

ith

confl

icti

ng p

rior

itie

s.N

o be

ds a

vaila

ble

at IC

U

Cul

ture

OC

Failu

re d

ue t

o at

titu

de a

nd a

ppro

ach

of t

he t

reat

ing

orga

niza

tion

.W

ard

whe

re v

ital

par

amet

ers

are

not

freq

uent

ly

take

n si

nce

‘no

one

does

it’

Hum

anEx

tern

alH

-ex

Hum

an f

ailu

res

beyo

nd t

he c

ontr

ol o

f th

e or

gani

zati

on/

depa

rtm

ent

Into

xica

tion

of

too

high

dos

age

med

icat

ion

pres

crib

ed o

utsi

de h

ospi

tal c

are

(by

GP)

Kno

wle

dge-

base

d be

havi

our

HK

KFa

ilure

of

an in

divi

dual

to

appl

y th

eir

know

ledg

e to

a n

ew

clin

ical

sit

uati

onN

o ad

equa

te d

iagn

osti

cs

No

phys

ical

exa

min

atio

n do

ne

Qua

lifica

tion

sH

RQA

n in

appr

opri

atel

y tr

aine

d in

divi

dual

per

form

ing

the

clin

ical

tas

kN

ot a

vaila

ble

Co-

ordi

nati

onH

RCA

lack

of

task

co-

ordi

nati

on w

ithi

n th

e he

alth

care

tea

mN

o co

ordi

nati

on o

f hy

pert

ensi

on t

reat

men

t

Veri

ficat

ion

HRV

Failu

re t

o co

rrec

tly

chec

k an

d as

sess

the

sit

uati

on b

efor

e pe

rfor

min

g in

terv

enti

ons

DN

R po

licy

not

adeq

uate

ly d

iscu

ssed

Inte

rven

tion

HRI

Failu

re r

esul

ting

fro

m f

ault

y ta

sk p

lann

ing

or p

erfo

rman

ceN

o di

agno

stic

s an

d ad

equa

te t

reat

men

t de

liriu

m

Mon

itor

ing

HRM

Failu

re t

o m

onit

or t

he p

atie

nt’s

pro

gres

s or

con

diti

onN

o ev

alua

tion

of

vita

ls a

fter

cha

ngin

g tr

eatm

ent

Vit

als

not

mon

itor

ed a

nd a

ctio

n un

dert

aken

aft

er

repo

rted

det

erio

rati

on

69

Tab

le 1

| (c

ontin

ued)

Mai

n c

ateg

ory

Sub

cate

go

ryC

od

eD

escr

ipti

on

Exam

ple

s (i

f av

aila

ble

) *

Skill

s-ba

sed

HSS

Failu

re in

per

form

ance

of

high

ly d

evel

oped

ski

llsO

bstr

ucti

ve le

sion

tra

chea

not

rec

ogni

zed/

mis

sed

by

radi

olog

ist

on C

T

Pati

ent

Pati

ent-

rela

ted

PRF

Failu

res

rela

ted

to p

atie

nt c

hara

cter

isti

cs o

r co

ndit

ions

, w

hich

are

bey

ond

the

cont

rol o

f st

aff

and

influ

ence

clin

ical

pr

ogre

ss

Mon

itor

ing

not

adeq

uate

bec

ause

pat

ient

ref

used

C

AD

Dis

ease

-rel

ated

DRF

Failu

res

rela

ted

to t

he n

atur

al p

rogr

ess

of d

isea

se w

hich

are

be

yond

con

trol

of

pati

ent,

its

care

rs a

nd s

taff

Tum

our

prog

ress

ion

in v

ena

cava

infe

rior

Bilia

ry p

ancr

eati

tis

XU

ncla

ssifi

able

XM

edic

atio

n w

as s

till

bein

g do

sed

prop

erly

Toxi

c re

acti

on c

hem

othe

rapy

*A t

able

wit

h ov

ervi

ew o

f al

l roo

t ca

uses

is p

rovi

ded

in S

1 Ta

ble.

70

Root causes

In total, from 49 unplanned ICU causal trees, 155 root causes were identified after PRISMA-

analysis. See Fig. 2 for three of the composed causal trees. Twelve unplanned ICU admissions

had one root cause (22%), 7 (14%) had two root causes, 11 (22%) three root causes, 12 four

root causes (25%) and 2 cases had 5 root causes (4%). Twelve percent of the cases had six or

more root causes (n=6). Almost half of the root causes identified were disease-related (n=70,

45%). Forty-four cases had at least one disease-related root cause. These disease-related root

causes comprise failures which are related to the natural progression of a disease which are

beyond control of patient, its carers and staff, for example neutropenic sepsis in a patient

with chemotherapy for an hematologic disease. Table 1. illustrates the subcategories used

in PRISMA-analysis with examples from our study. The distribution of all the root causes is

illustrated in Fig. 3a.

Figure 3 | Distribution root causes3a Main categories root causes. 3b Healthcare worker (HCW) root causes.

71

Almost half of the root causes classified were human-related (n=71, 46%), followed by

patient-related causes (n=7, 5%), and organisational root causes (n=5, 3%). Human related

causes are causes which are healthcare worker related. Most of the healthcare worker related

causes (34%) were monitoring failures (HRM), these are. failures in monitoring the patient’s

progress or condition, e.g. not monitoring vital parameters in a patient who is clinically

deteriorating. Fig. 3b shows the subcategorization of healthcare worker root causes. Human

intervention causes were accountable for almost one-third (27%) of the causes. They represent

failures resulting from faulty task planning or performance, for example the case where no

intervention was started after the nurse repeatedly mentioned patient’s vital signs worsening.

The remaining human-related root causes were verification (HRV: 11,3%), coordination (HRC:

7%), and skills-based (HSS: 4,2%) failures. Finally, the organisation root causes were those

caused by management issues, such as lack of beds on the ICU, or not handling certain

treatment according to the institutional protocol. The two unclassifiable root causes are shown

in table 1. No technical related root causes (T) were identified in this PRISMA-analysis.

Use of Track and Trigger System/Early Warning Score

Table 3 shows that in 86% (n=42 patients) of the cases orders were given by the doctors for

vital parameter monitoring. This was done as agreed in only half of the cases (n=20, 49%).

In the 48 hours before ICU admission, a total of 477 vital parameter sets were measured in

the 49 patients (i.e. 1 vital set could include: a blood pressure, pulse, breathing frequency

and temperature), with a median of 6 sets measured per patient. Of these 477 sets, an

explicit MEWS was calculated and documented correctly, according to protocol, in only 6

of these measurements in 4 patients (1%). After recalculation by the researcher based upon

the available vital parameters, 207 (43%) of the vital sets gave a critical MEWS score of 3 or

higher. 46 patients had at least one critical MEWS of 3 or higher during the 48 hours before

ICU admission. According to MEWS protocol, this implied that a doctor had to be called. In

125 of the measured sets, in 42 patients, this was done (although the MEWS score was often

not calculated and documented), and after this the doctor started an action upon this in 117

of these 125 phone calls. In total, 67 of all measurements, in 92% of the patients (n=45), led

to a call to the RIT at least once.

DISCUSSIONThe current study is the first to systematically investigate the root causes of unplanned ICU

admissions. The most important finding from this study was that almost half of the root causes

contributing to unplanned ICU admission were human (healthcare worker) related. These

causes predominantly included human monitoring and intervention failures, indicating flaws

in monitoring the patients progress or condition and faulty task planning or performance. This

illustrates a potential for improvement. The other half of the root causes were disease-related,

comprising the root causes related to the natural progression of the disease, which was to be

expected in this overall severely ill patient population, as reflected in their high mortality rates.

The secondary aim was to further investigate monitoring by analysing the use of a TTS on

a general ward in the early recognition of deteriorating patients. Results showed that in the 48

72

Table 2 | Patient characteristics

N (%) 49 (100%)

Age - median(range) 69 (34-90)

Male 23 (47%)

Deceased during admission 19 (39%)

Admission specialty :

Cardiology

Gastroenterology

Haematology

Internal medicine

ENT-diseases

Pulmonary medicine

Neurology

Nephrology

Oncology

Traumatology

Vascular surgery

1 (2%)

5 (10%)

7 (14%)

12 (24%)

1 (2%)

9 (18%)

3 (6%)

3 (6%)

3 (6%)

4 (8%)

1 (2%)

Polypharmacy* 36 (73)

Length of stay before unplanned transfer ICU in hours – median (range) 88 h 34 m (1h38m-733h)

Time unplanned ICU admission

24.00 – 06.00

06.00 – 12.00

12.00 – 18.00

18.00 – 24.00

17 (35%)

12 (24%)

11 (22%)

9 (18%)

DNR-policy before ICU admission

No restriction

Do not resuscitate, do ventilate

Do not resuscitate, do not ventilate

No ICU admission

Not clear

34 (69%)

9 (18%)

3 (6%)

0 (0%)

3 (6%)

SAPS II – median(range)21 51 (18-110)

APACHE II – median (range)20 24 (6-45)

APACHE IV – median (range)22 95 (36-186)

hours before deterioration only in 1% of the measured vital sets an explicit correct ‘MEWS’

was reported, although in 43% of the measurements patients had a critical score. This is an

important clinical finding, since it seems essential that recognition of critical scores by hospital

staff is present in order to improve monitoring.29

Monitoring failures emerging from the PRISMA-analysis were diverse. One example is a nurse

who documented in the chart that a patient repeatedly complains about shortness of breath,

did not undertake an action to measure vital parameters or request a physician for assessment.

73

Another example was unclear handovers resulted in confusion about how intensively the patient

needed monitoring, resulting in inadequate and insufficient monitoring measurements.

Interventional problems are exemplified by not performing adequate diagnostics and treatment

in a severely immunosuppressed patient with a suspected pneumonia. In our study, an unclear

DNR-policy was also found to be responsible for an unnecessary unplanned ICU admission

in a few cases, since these patients did not want to be admitted to the ICU but this was not

known or discussed on the ward. These are human-related verification failures (HRV).

The above results seem to be consistent with earlier research indicating that unplanned ICU

admissions are not only caused by the underlying disease, but are also potentially caused by

suboptimal care on the ward, and by inadequate assessment and monitoring of a patients

status.30,31 Other studies have shown that certain patient characteristics such as older age, being

male and having a higher co-morbidity increase the chance of unplanned ICU admissions.32 In

addition to above mentioned non-modifiable factors, our study results provides implications

for practice to potentially reduce unplanned ICU admission by improving patient monitoring.

One method for improving the recognition of these patients, is the implementation of TTSs.

Although the conclusions on effectiveness of TTSs in reducing clinical endpoints are still not

uniform, when properly followed they are effective in identifying deteriorating patients.33,34

The effectiveness depends on appropriate implementation, compliance and an effective clinical

Table 3 | Use of Track and Trigger System

Vital parameters documentation Frequency (%), N=49 (100%)

Orders were given for vital monitoring* 42 (86%)

Vital monitoring performed as agreed 20 (41%)

Registration of ICU admission in nurse’ charts 32 (65%)

Registration of ICU admission in doctors’ charts 38 (78%)

MEWS DocumentationTotal vital set measurements done in 48 hours before ICU admission in 49 patients

N = 477 (100%)

Number of vital set measurement done per patient in 48 hours before ICU admission – median(range)

6 (1-22)

Doctor called after vital parameters measured 174 (36%) 42 patients

Doctor started an action after being called 164 (34%) 42 patients

Evaluation after 60 minutes of started action 96 (20%) 41 patients

RIT-call 69 (14%) 45 patients

MEWS calculated and documented correctly in charts 6 (1%) 4 patients

Critical MEWS (after recalculation by researcher using vital set measurements ≥3)

207 (43%) 46 patients

Doctor called at a critical-MEWS according to the recalculated score by the researcher

125 (26%) 42 patients

74

response.35 The current study has shown that the MEWS score in the studied population was

registered according to protocol in very few patients. For the other patients, a score was

calculated by the researchers based on the available vital parameters. After recalculation 43%

(207 out of 477) of the measurements were critical, and in these cases in 60% (125 out of

207 measurements) a doctor was called in although an explicit MEWS was not documented

in the charts. If the doctor was notified, in most cases they started an action, which shows

the relevance of these critical measurements. These results show that the implementation

of MEWS at the time of the study was insufficient. When we implemented the protocol

a few years back the nurses and doctors were requested to perform MEWS when they

found patients to be sick and in need of critical care. The protocol did not require the MEWS

to be taken daily on set times to identify deterioration early. The added value of the daily

measurements has been demonstrated.35 We therefore implemented a new protocol in which

it was compulsory to measure the MEWS in all patients at least once a day in the morning.

If the MEWS is >3 the doctor has to be called within 30 minutes. The clinical staff was (re)

trained aiming to chance their mind-set about the importance of MEWS. A few weeks after

this implementation the protocol adherence improved to 89%, underlining the importance of

robust implementation.36

This in-depth study has provided analysis of the root causes of unplanned ICU admissions to

the full extent. To our knowledge, this is the first study performed that explores the underlying

causes of unplanned ICU admissions in such a way that it provides clear insight in the areas for

potential improvement in the healthcare system. Hence, it equips us with valuable information

about areas in need of attention. From this, tools for implementing process improvement in

hospitals can be initialised. A limitation of the study was its retrospective nature, which means

that researchers had to rely on written information in the patient records, instead of actively

interviewing the persons involved. Therefore valuable information might have been missed

and this could potentially have led to an underestimation of the factors leading to delayed

recognition. For future studies it is therefore recommended to combine this research method

with other methods, such as collecting qualitative data using interviews. Also, by assessing

these cases only on the basis of information available on paper, analysis might not elucidate all

the organisational factors which are not typically written down and are latent.

CONCLUSIONThis study is the first one that has systematically provided insight in the root causes of unplanned

ICU admissions. Half of the root causes identified were healthcare worker related failures,

mainly resulting from monitoring and interventional failures. The monitoring of patients might

be improved by a properly functioning Track and Trigger System. Results in this study however,

have shown that this warrants improvement.

75

CONFLICTS OF INTERESTS None

SUPPORTING INFORMATION CAPTIONSS1 File. Data collection sheet unplanned ICU admissions.S1 Table. Overview root causes.S1 Dataset. ICU admissions.

76

REFERENCES1. de Vries EN, Ramrattan MA, Smorenburg SM, Gouma DJ, Boermeester MA. The incidence and

nature of in-hospital adverse events: a systematic review. Qual Saf Health Care. 2008;17(3):216-223.

2. Wilson RM, Runciman WB, Gibberd RW, Harrison BT, Newby L, Hamilton JD. The Quality in Australian Health Care Study. Med J Aust. 1995;163(9):458-471.

3. Vlayen A, Verelst S, Bekkering GE, Schrooten W, Hellings J, Claes N. Incidence and preventability of adverse events requiring intensive care admission: a systematic review. J Eval Clin Pract. 2012;18(2):485-497.

4. Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med. 1991;324(6):377-384.

5. Zegers M, de Bruijne MC, Wagner C, et al. Adverse events and potentially preventable deaths in Dutch hospitals: results of a retrospective patient record review study. Qual Saf Health Care. 2009;18(4):297-302.

6. Arts D, de Keizer N, Scheffer GJ, de Jonge E. Quality of data collected for severity of illness scores in the Dutch National Intensive Care Evaluation (NICE) registry. Intensive Care Med. 2002;28(5):656-659.

7. Ridley S, Jackson R, Findlay J, Wallace P. Long term survival after intensive care. BMJ. 1990;301(6761):1127-1130.

8. Alam N, Vegting IL, Houben E, et al. Exploring the performance of the National Early Warning Score (NEWS) in a European emergency department. Resuscitation. 2015;90:111-115.

9. Hillman K, Parr M, Flabouris A, Bishop G, Stewart A. Redefining in-hospital resuscitation: the concept of the medical emergency team. Resuscitation. 2001;48(2):105-110.

10. Moon A, Cosgrove JF, Lea D, Fairs A, Cressey DM. An eight year audit before and after the introduction of modified early warning score (MEWS) charts, of patients admitted to a tertiary referral intensive care unit after CPR. Resuscitation. 2011;82(2):150-154.

11. Subbe CP, Gao H, Harrison DA. Reproducibility of physiological track-and-trigger warning systems for identifying at-risk patients on the ward. Intensive Care Med. 2007;33(4):619-624.

12. McGaughey J, Alderdice F, Fowler R, Kapila A, Mayhew A, Moutray M. Outreach and Early Warning Systems (EWS) for the prevention of intensive care admission and death of critically ill adult patients on general hospital wards. Cochrane Database Syst Rev. 2007(3):CD005529.

13. Quinn TD, Gabriel RA, Dutton RP, Urman RD. Analysis of Unplanned Postoperative Admissions to the Intensive Care Unit. J Intensive Care Med. 2015.

14. Mercier E, Giraudeau B, Ginies G, Perrotin D, Dequin PF. Iatrogenic events contributing to ICU admission: a prospective study. Intensive Care Med. 2010;36(6):1033-1037.

15. Darchy B, Le Miere E, Figueredo B, Bavoux E, Domart Y. Iatrogenic diseases as a reason for admission to the intensive care unit: incidence, causes, and consequences. Arch Intern Med. 1999;159(1):71-78.

16. Fluitman KS, van Galen LS, Merten H, et al. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Intern Med. 2016.

17. van der Schaaf TW HM. PRISM-Medical. A Brief Description. Eindhoven University of Technology, Faculty of Technology Management, Patient Safety Systems: Eindhoven. 2005.

18. van Vuuren W SC, van der Schaaf TW. The Development of an Incident Analysis Tool For the Medical Field. Eindhoven University of Technology: Eindhoven. 1997.

19. van Wagtendonk I, Smits M, Merten H, Heetveld MJ, Wagner C. Nature, causes and consequences of unintended events in surgical units. Br J Surg. 2010;97(11):1730-1740.

20. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13(10):818-829.

77

21. Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study. JAMA. 1993;270(24):2957-2963.

22. Zimmerman JE, Kramer AA, McNair DS, Malila FM. Acute Physiology and Chronic Health Evaluation (APACHE) IV: hospital mortality assessment for today’s critically ill patients. Crit Care Med. 2006;34(5):1297-1310.

23. Rasmussen J. Skills, rules and knowledge: signals, signs and symbols and other distinctions in human performance models. IEEE Trans Systems Man Cybernetics. 1983;13:257-266.

24. Reason JT. Human Error. Cambridge University Press. 1990.

25. Reason JT. Managing the Risk of Organisational Accidents. Asgate Aldershot. 1997.

26. Runciman W, Hibbert P, Thomson R, Van Der Schaaf T, Sherman H, Lewalle P. Towards an International Classification for Patient Safety: key concepts and terms. Int J Qual Health Care. 2009;21(1):18-26.

27. Sherman H, Castro G, Fletcher M, et al. Towards an International Classification for Patient Safety: the conceptual framework. Int J Qual Health Care. 2009;21(1):2-8.

28. Runciman WB, Williamson JA, Deakin A, Benveniste KA, Bannon K, Hibbert PD. An integrated framework for safety, quality and risk management: an information and incident management system based on a universal patient safety classification. Qual Saf Health Care. 2006;15 Suppl 1:i82-90.

29. Storm-Versloot MN, Verweij L, Lucas C, et al. Clinical relevance of routinely measured vital signs in hospitalized patients: a systematic review. J Nurs Scholarsh. 2014;46(1):39-49.

30. Bapoje SR, Gaudiani JL, Narayanan V, Albert RK. Unplanned transfers to a medical intensive care unit: causes and relationship to preventable errors in care. J Hosp Med. 2011;6(2):68-72.

31. McGloin H, Adam SK, Singer M. Unexpected deaths and referrals to intensive care of patients on general wards. Are some cases potentially avoidable? J R Coll Physicians Lond. 1999;33(3):255-259.

32. Frost SA, Alexandrou E, Bogdanovski T, Salamonson Y, Parr MJ, Hillman KM. Unplanned admission to intensive care after emergency hospitalisation: risk factors and development of a nomogram for individualising risk. Resuscitation. 2009;80(2):224-230.

33. Ranji SR, Auerbach AD, Hurd CJ, O’Rourke K, Shojania KG. Effects of rapid response systems on clinical outcomes: systematic review and meta-analysis. J Hosp Med. 2007;2(6):422-432.

34. Chan PS, Khalid A, Longmore LS, Berg RA, Kosiborod M, Spertus JA. Hospital-wide code rates and mortality before and after implementation of a rapid response team. JAMA. 2008;300(21):2506-2513.

35. Ludikhuize J, Borgert M, Binnekade J, Subbe C, Dongelmans D, Goossens A. Standardized measurement of the Modified Early Warning Score results in enhanced implementation of a Rapid Response System: a quasi-experimental study. Resuscitation. 2014;85(5):676-682.

36. van Galen LS, Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. PLoS One. 2016;11(8):e0160811.

CHAPTER 5

A PROTOCOLISED ONCE A DAY MODIFIED EARLY WARNING SCORE (MEWS) MEASUREMENT IS AN

APPROPRIATE SCREENING TOOL FOR MAJOR ADVERSE EVENTS IN A GENERAL HOSPITAL POPULATION

“A MEWS IN THE MORNING, A VERY GOOD WARNING’’

Louise S. van Galen & Casper C. Dijkstra | Jeroen Ludikhuize | Mark H.H. Kramer | Prabath W.B. Nanayakkara

PLoS One 2016;11(8):e0160811

‘Sometimes the protocol is a too tight harness which traps medical specialists’ Dana Ploeger

80

ABSTRACTBackground

The Modified Early Warning Score (MEWS) was developed to timely recognise clinically

deteriorating hospitalised patients. However, the ability of the MEWS in predicting serious

adverse events (SAEs) in a general hospital population has not been examined prospectively.

The aims were to (1) analyse protocol adherence to a MEWS protocol in a real-life setting

and (2) to determine the predictive value of protocolised daily MEWS measurement on SAEs:

death, cardiac arrests, ICU-admissions and readmissions.

Methods

All adult patients admitted to 6 hospital wards in October and November 2015 were included.

MEWS were checked each morning by the research team. For each critical score (MEWS ≥ 3),

the clinical staff was inquired about the actions performed. 30-day follow-up for SAEs was

performed to compare between patients with and without a critical score.

Results

1053 patients with 3673 vital parameter measurements were included, 200 (19.0%) had

a critical score. The protocol adherence was 89.0%. 18.2% of MEWS were calculated wrongly.

Patients with critical scores had significant higher rates of unplanned ICU admissions [7.0%

vs 1.3%, p < 0.001], in-hospital mortality [6.0% vs 0.8%, p < 0.001], 30-day readmission

rates [18.6% vs 10.8%, p < 0.05], and a longer length of stay [15.65 (SD: 15.7 days) vs 6.09

(SD: 6.9), p < 0.001]. Specificity of MEWS related to composite adverse events was 83% with

a negative predicting value of 98.1%.

Conclusions

Protocol adherence was high, even though one-third of the critical scores were calculated

wrongly. Patients with a MEWS ≥ 3 experienced significantly more adverse events. The negative

predictive value of early morning MEWS < 3 was 98.1%, indicating the reliability of this score

as a screening tool.

Keywords

Early warning score, Protocol adherence, Patient safety, Serious adverse events, Rapid

intervention teams, Track and trigger systems

81

INTRODUCTIONSerious adverse events (SAEs) in hospitalised patients are preceded by signs of clinical

deterioration in up to 80% of the patients.1 Therefore, changes in vital parameters such as

pulse rate, respiratory rate, and level of consciousness are often considered as predictors of

SAEs such as cardiac arrest, death and unplanned intensive care unit (ICU) admissions.1,2 To

improve timely detection and treatment of deteriorating patients on nursing wards, rapid

response systems (RRSs) have been introduced.3-5 RRSs consist of two different components:

an afferent limb consisting of track and trigger systems (TTS) such as Modified Early Warning

Score (MEWS) and an efferent limb, a rapid intervention team (RIT) consisting of trained ICU

personnel who will deliver immediate treatment to deteriorating patient at the bedside.

Some studies have demonstrated positive effects of implementing TTSs such as MEWS on

patient outcomes.6 On the basis of these results TTSs have been introduced in many hospitals

to increase patient safety.7,8 Firstly introduced in 1997 by Morgan et al. the TTS functions as

the afferent limb and is designed to detect deterioration early.9 Since this first introduction

multiple early warning bedside monitoring tools have been developed and implemented

internationally.10,11 These TTSs are used to detect deterioration and call upon a team to monitor

and treat patients to prevent further deterioration.12 In the VU university medical centre

(VUmc), RRS with an afferent limb consisting of a TTS (MEWS) and an efferent limb consisting

of a rapid intervention team (RIT) was introduced a few years ago. Because the afferent

limb of the system (RIT) did not function optimally, it was decided to reintroduce the MEWS

protocol in 2015 and (re)train the clinical staff aiming to change their mindset and improve

protocol adherence.

The effectiveness of a RRS is not only decided by the quality of the RIT but also by an

appropriate implementation and use of the TTS such as the MEWS.8,13 Unfortunately, very

few prospective studies have yet been performed investigating the compliance to any TTS

protocol in a real-life setting. In addition, although Smith et al. (2008) demonstrated MEWS

as a predictor for clinical outcomes retrospectively, prospective studies investigating the ability

of the MEWS to predict relevant clinical outcomes in a general in-hospital population are

lacking.14 In addition, no previous studies have investigated the association between MEWS

and the chance of 30-day readmissions. Positive association of MEWS with these endpoints

can be used to convince doctors and nurses about the value of MEWS as a prediction tool and

thereby increase their protocol adherence.

Therefore, the main aim of this study was to determine the protocol adherence mainly to

the afferent limb but also to the efferent limb in a real-life setting. The secondary aims were

to investigate the ability of once a day MEWS measurement to predict patient outcomes:

in-hospital mortality, hospital length of stay, cardiac arrests, ICU-admissions and 30-day

readmissions. Ultimate goal was to provide the hospital staff more insights into the value

of the MEWS in predicting outcomes in their own patient population and thereby increase

the awareness and protocol adherence.

82

MATERIALS AND METHODSThis prospective study was conducted in a large urban university medical centre (VUmc) with

approximately 50,000 admissions per annum in the Netherlands.

Patient selection

In the 7-week inclusion period from the 8th October until the 30th of November, all adult

patients who were in hospital at 08.00 at the date of inclusion on five wards (acute admission

unit, general surgery, internal medicine, trauma surgery, vascular surgery/urology/nephrology

ward) were included. Due to logistical reasons patients from the pulmonary ward were

included from the 1st of November. Patients 18 years and older with at least one overnight stay

were included. The Ethics committee of VU University Medical Centre Amsterdam, approved

the study and necessity for informed consent was waived.

MEWS protocol in our institution

In our hospital all vital parameter measurements are stored in an automatic electronic system.

According to the hospital wide protocol, every morning at the end of the nightshift or at

the beginning of the dayshift, nurses were requested to determine the MEWS using vital

parameter measurements recorded in this electronic system. Although MEWS measurements

could be repeated any time during the day on indication by the nurses and doctors, only

these early morning scores were used for analysis. The MEWS consists of an easy-to-use

algorithm of seven parameters (Fig. 1).15 The range for the MEWS is between 0 and 19.

During the implementation of the protocol staff was trained extensively and the protocol card

containing the protocol was distributed. MEWS was calculated by hand and electronically

documented in patients’ charts. A total score of 3 or higher was considered as a critical score.

Once a patient reaches a critical MEWS (≥ 3) nurses were requested to contact the doctor in

charge immediately. The doctor must then assess the patient within 30 minutes and draft

a plan for treatment, evaluate this after 60 minutes or call a RIT team. The RIT may also directly

be called by the nurses or the doctor at the outset.

Data collection

Charts of all included patients were checked by the coordinating investigator (CD) to obtain

the patients’ MEWS and to determine whether scores were documented and calculated

correctly. The MEWS were perceived as documented if MEWS was explicitly reported in

nurses charts’. If a vital parameter was not documented in the system, this parameter was

considered to be normal. Scores were recalculated by CD using available data in the charts.

If a patient had a critical score, charts were examined to find out what actions has been

taken, subsequently the nurses and doctors were asked about their actions. If no action

was undertaken the investigator inquired the staff about the reasons. If during recalculation

a patient had a MEWS of ≥ 3 and this was not explicitly documented by the nurse, nurses were

still asked about their actions. However, if a patient had a MEWS of ≥ 3 during recalculation

by the CD and this was wrongly calculated and documented by the nurse as a MEWS < 3,

no questions were asked. At the end of the inclusion period all answers were categorised.

83

Figure 1 | Mews and protocol in VUmc

If more than one action was taken, the most serious action was used in the categorisation.

For patients who were in hospital for multiple days the highest reached MEWS, labelled as

‘MaxScore’, was taken for predictive analysis.

Follow-up

All patients admitted during the study period were followed up for 30 days after inclusion. In

addition, patients were followed up for 30 days after discharge to obtain information about

the 30-day unplanned hospital readmission rate. MaxScore per patient was used to perform

the predictive analysis of MEWS.

Statistical analysis

Descriptive characteristics and frequencies were calculated in SPSS version 22.0 (SPSS, Chicago,

IL, USA). Categorical outcome measures are presented as frequencies and percentages.

Continuous variables are summarised by mean and standard deviation since data was

distributed normally. To illustrate the comparison in adverse events between patients who

had a MEWS < 3, versus MEWS ≥ 3 a chi-squared test was used. P-values below 0.05 were

considered significant.

RESULTSPatient characteristics

A total of 1053 patients were included during the 8-week inclusion period. Table 1 shows

patient characteristics. Most patients were admitted to the Acute Medical Unit (n=408,

38.8%), the least to the general surgery ward (n=113, 10.7%). The mean age of patients in

this cohort was 61.1 (SD 17.6).

84

Figure 2 | Protocol adherenceMeasurement and documentation. Horizontal section I representing all MEWS measurements, regardless of score, Horizontal section II representing MEWS ≥ 3, as recalculated by the coordinating researcher.

Figure 3 | Actions undertaken on patients by clinical staff after critical score reachedN= number of MEWS measurements ≥ 3. *Expectative since this high score is expected as a result of the (known) disease process or the treatment.

85

Table 1 | Patient characteristics (N=1053)

WardPatients Number (%) Male (%) Mean age (SD)

MaxScore*** Median [range]

Acute medical Unit 408 (38.8) 220 (53.9) 61.4 (18.9) 1.0 [0-9]Non-critical score* 365 (89.5) 1.0 [0-2]

Critical score** 43 (10.5) 4.0 [3-9]

Internal medicine 120 (11.4) 60 (50.0) 66.4 (16.8) 2.0 [0-8]Non-critical score* 80 (66.7) 2.0 [0-2]

Critical score** 40 (33.3) 3.5 [3-8]

General surgery 113 (10.7) 69 (61.1) 65.2 (14.5) 2.0 [0-8]Non-critical score* 70 (61.9) 1.0 [0-2]

Critical score** 43 (38.1) 4.0 [3-8]

Vascular/urology/nephrology 140 (13.3) 92 (65.7) 60.1 (14.8) 1.0 [0-6]Non-critical score* 119 (85.0) 1.0 [0-2]

Critical score** 21 (15.0) 3.0 [3-6]

Trauma surgery 151 (14.3) 72 (47.7) 53.8 (18.6) 1.0 [0-5]Non-critical score* 122 (80.2) 1.0 [0-2]

Critical score** 29 (19.2) 3.0 [3-5]

Pulmonary diseases 121 (11.5) 56 (46.3) 61.6 (14.4) 1.0 [0-6]Non-critical score* 97 (80.2) 1.0 [0-2]

Critical score** 24 (19.8) 3.5 [3-6]

Total cohort 1053 (100.0) 569 (54.0) 61.1 (17.6) 1.0 [0-9]Non-critical score* 853 (81.0) 450 (52.8) 60.5 (17.4) 1.0 [0-2]

Critical score** 200 (19.0) 119 (59.5) 63.8 (18.0) 3.0 [3-9]

* MEWS < 3**MEWS ≥ 3***MaxScore: Highest reached MEWS for patients who were in hospital for multiple days.

Table 2 | Patient Outcomes

MEWS < 3 n=853 (81%)

MEWS ≥ 3 n = 200 (19% Significance

Odds Ratio (95% CI)

Composite endpoint reached (%) 16 (1.9) 25 (12.5) p < 0.001¹ 7.5 (3.9 - 14.3)

ICU-admissions 11 (1.3) 14 (7.0) p < 0.001² 5.8 (2.6 - 12.9)

In-hospital mortality 7 (0.8) 12 (6.0) p < 0.001² 7.7 (3.0 - 19.9)

Resuscitation 0 (0.0) 1 (0.5) p = 0.190² -

Readmission (%) 91 (10.8) 35 (18.6) p < 0.05¹ 1.9 (1.2 - 2.9)

Length of Stay (SD) 6.09 (6.9) 15.7 (15.7) p < 0.001³ -

RIT-call (%) - 21 (10.5) - -

1: Pearson Chi-squared 2: Fisher’s exact test 3: Independent samples t-test

86

Measuring and documentation

There were 4041 patient days where vital parameter measurements could have taken place

according to protocol. Fig. 2 displays a flowchart of the measurement and documentation.

368 potential measurement moments were missed because these patients were not present on

the ward during the time of assessment or because they were in palliative care. This resulted in

a total of 3673 morning round measurements in 1053 patients. Of these 3673 vital parameter

measurements, 3270 were explicitly documented in nurses’ charts, resulting in a protocol

adherence of 89.0%. The investigator recalculated all MEWS using the vital parameters

measurements in the charts. The determined MEWS were referred to as recalculated MEWS.

We observed a correct calculation in 2600/3673 (70.8%) of the scores in nurses’ charts, 670

(18.2%) scores were calculated incorrectly. The recalculated MEWS were < 3 in 3316 (90.3%)

and were ≥ 3 in 357 (9.7%) measurements.

Actions performed by clinical staff

In 257 (72.0%) instances in which MEWS ≥ 3 the investigator inquired clinical staff what

action they undertook. Fig. 3 shows the actions undertaken by hospital staff. In 10 (3.5%)

cases no actions could be found in charts and no staff members could answer the questions.

Of the remaining 247 cases a doctor was contacted 169 (68.4%) times and 78 (31.6%) times

no doctor was contacted. The categorised actions performed are displayed in S1 Table and

S2 Table. Of the 169 times a doctor was contacted the doctor intervened 70 (41%) times.

The main reason for not intervening was that clinical staff did not feel the urge to perform an

action since they judged the situation as not alarming enough.

Patient outcomes

The vital parameters to calculate a MEWS were measured in 1053 patients. Two-hundred

patients (19.0%) had a critical score during their hospital stay. The remaining 853 (81%)

patients did not have a critical score. Table 2 shows the relation between a critical MEWS and

patient outcome. Having a critical score was associated with a higher percentage of unplanned

ICU admission [7.0% vs. 1.3%, OR 5.8 (2.6 – 12.9), p < 0.001], and a higher in-hospital

mortality [6.0% vs. 0.8, OR 7.7 (3.0 – 19.9), p < 0.001]. Also, results show that patients

with a critical score had a longer length of stay [15.7 days (SD: 15.7) vs. 6.09 days (SD: 6.9)

p < 0.001] and the 30-day readmission rate was higher [18.6% vs. 10.8%, OR 1.9 (1.2 – 2.9),

p < 0.05] than patients without a critical score. Sensitivity for MEWS related to composite

adverse events was 61%, specificity 83%, positive predicting value 12.5% and the negative

predicting value was 98.1%. MEWS of 3 to 5 show significant more adverse events compared

to MEWS below 3. MEWS above 5 show significant more adverse events than MEWS < 3

(p < 0.001) but compared to MEWS 3-5 no significance was reached (p = 0.196). Fig. 4 shows

patient outcomes compared between different scores.

DISCUSSIONIn this prospective study conducted in a real-life setting, we have demonstrated that adherence

to the MEWS protocol in our hospital was good (89%). However in some cases (18%)

87

the MEWS was calculated incorrectly because values were not added up properly, influencing

the total score. Although, in the majority of the cases the nurse informed the doctor about

the critical score an intervention only occurred in one-third of the cases mostly because

the situation was judged as not alarming. The MEWS of 3 or higher was a strong predictor of

clinical endpoints such as in-hospital mortality, 30-day readmissions, hospital length of stay. In

addition, the negative predictive value of MEWS < 3 in this general hospital population was

98.1% indicating the reliability of this score as a screening tool.

The afferent limb is an important component of a RRS, since an effective clinical response

depends on early recognition of deterioration.8,13 When we implemented the RRS in our hospital

a few years ago the afferent limb was implemented without a clear protocol. Therefore the TTS

did not function properly. We re-trained the clinical staff and a clear protocol was implemented

in 2015. In this protocol nurses were requested to always take a MEWS score in the morning.

The main aim of this study was to analyse protocol adherence after this reimplementation. In

addition, we aimed to analyse the value of the morning MEWS measurement in predicting

clinical outcomes in this general hospital population because this has not been evaluated

in a prospective study in a real-life setting. The results of this study showed a high protocol

adherence with nurses completing MEWS documentation in 89% of the measurements.

However, a percentage (18%) of wrongly documented scores were also seen, likely due to

wrong calculations in adding up separate MEWS parameters. An important finding was that

due to these wrong calculations, a relatively high percentage of critical scores were missed

by nurses. Twenty-eight percent of the critical scores, where a doctor was supposed to be

alarmed, were not recognised by the nurses. Our study has also shown, that doctors were

not contacted in one-third (32%) of the critical scores. When physicians were contacted,

they only undertook an action in 28% of the cases. The main reason for not taking action

was that staff judged the situation as not alarming. These findings are comparable to Jones

Fig 4 | Adverse events compared between MEWS groupsSignificant with MEWS < 3 with a p-level of p < 0.001. OR = Odds ratio

88

et al. (2011) who also found a percentage of 29%.16 Reasons for these findings, as explained

in previous work, are that clinical staff feel the parameter is too rigorous in its cut-offs or

the nursing staff estimate the situation as being under control.17,18 However, previous work

has already demonstrated that changing the critical cut-off to 4 devaluates MEWS as a reliable

screening tool.19

Since creating awareness and emphasising the importance of the MEWS can increase protocol

adherence a secondary aim was to validate the MEWS as a predictor for adverse events in our

own hospital population. We demonstrate prospectively for the first time in a real-life setting

that patients with a MEWS ≥ 3 in one of the morning measurements had an increased risk

for an unplanned adverse event than the patients with a MEWS < 3. No significant increase

was observed for unplanned resuscitations, likely due to the very low incidence of events. To

our knowledge, this is the first study validating this MEWS protocol prospectively in a general

in-hospital population in real-life setting. One other study has prospectively validated the value

of MEWS in predicting adverse events in a European surgical population. This study was also

performed in a real-life setting. Their results are consistent with our findings.20 In addition,

a recent publication in Africa validated the MEWS prospectively in a research setting in low-

resource circumstances.21 They too found that the MEWS was a useful triage tool to identify

patients at the greatest risk of experiencing an adverse event. We also demonstrate for the first

time that MEWS ≥ 3 is associated with a significantly higher readmission rate within 30 days

for a critical score (10.8% vs. 18.6%). Since readmissions are known to increase mortality and

are associated with functional decline, it again underlines the importance of the MEWS as

a screening tool.22-24

MEWS as part of the RRT system, was implemented in many Dutch hospitals to potentially

increase patient safety.25 MEWS is a relatively low-cost and convenient bedside monitoring

tool, however critical scores can lead to a higher workload for clinical staff. This study, however,

again emphasised the clinical importance of recognising patients with a MEWS higher than

3 since these patients are at high risk of developing adverse events. In addition, the negative

predictive value of MEWS < 3 was 98.1 underscoring the importance of MEWS as a screening

tool. Nevertheless, it is worth mentioning only 7% of the patients in our population with

a MEWS ≥ 3 were transferred to the ICU. We do not know how many patients were prevented

from ICU admission by early recognition and prompt treatment on the wards.

The strength of this study is its prospective study design in a real world general hospital sample

in which 3290 MEWS values were analysed. In addition we personally contacted every nurse

who was involved with the MEWS or vital parameter measurements to collect information

daily. This is the largest prospective study conducted so far validating MEWS as a screening

tool in a general in-hospital (medical and surgical) population.26,27 The study was conducted

in a single-centre which uses one specific MEWS protocol. Therefore, results might not be

generalised to hospitals using another EWS protocol. Also, since our aim was to determine

clinical relevance of MEWS in daily practice, a real-life hospital situation was studied. As

89

a result, the determined MEWS and not the completeness of the vital parameter set was taken

into account. This could possibly under- or overestimate the relation between MEWS and

patient outcomes.

CONCLUSIONIn this prospective study performed in a real-life setting we demonstrated that adherence to

the MEWS protocol in our hospital is good (89%). A morning Modified Early Warning Score

of 3 or higher was a strong predictor of clinical endpoints such as in-hospital mortality, 30-day

readmissions, hospital length of stay. In addition, the negative predictive value of MEWS < 3 in

this general hospital population was 98.1% indicating the reliability of this score as a screening

tool. Therefore, it is important to keep emphasising the clinical relevance of the MEWS among

clinical staff.

CONFLICTS OF INTERESTSNone

ACKNOWLEDGEMENTSOur thanks to all the nurses and the doctors of the units for their help during the study. Special

thanks to Edwin Pompe (manager care) and Sascha Spoor (emergency nurse) for their support.

SUPPORTING INFORMATION CAPTIONSS1 Table. Actions undertaken on patients by clinical staff after critical score reached.

*Since it is part of disease/treatment or patient is familiar with abnormalities

S2 Table. Categorisation of actions of clinical staff.

*Since it is part of disease/treatment or patient is familiar with abnormalities

S1 Dataset. Vital parameters and measured MEWS.

S2 Dataset. Categorization actions undertaken at MEWS ≥ 3 by clinical staff.

S3 Dataset. Patient outcomes.

90

REFERENCES1. Ludikhuize J, Smorenburg SM, de Rooij SE, de Jonge E. Identification of deteriorating patients on

general wards; measurement of vital parameters and potential effectiveness of the Modified Early Warning Score. J Crit Care 2012; 2012/02/22:424 e427-413.

2. Kim WY, Shin YJ, Lee JM, et al. Modified Early Warning Score Changes Prior to Cardiac Arrest in General Wards. PLoS One. 2015;10(6):e0130523.

3. Hillman K, Parr M, Flabouris A, Bishop G, Stewart A. Redefining in-hospital resuscitation: the concept of the medical emergency team. Resuscitation. 2001;48(2):105-110.

4. Mathukia C, Fan W, Vadyak K, Biege C, Krishnamurthy M. Modified Early Warning System improves patient safety and clinical outcomes in an academic community hospital. J Community Hosp Intern Med Perspect. 2015;5(2):26716.

5. Alam N, Hobbelink EL, van Tienhoven AJ, van de Ven PM, Jansma EP, Nanayakkara PW. The impact of the use of the Early Warning Score (EWS) on patient outcomes: a systematic review. Resuscitation. 2014;85(5):587-594.

6. Moon A, Cosgrove JF, Lea D, Fairs A, Cressey DM. An eight year audit before and after the introduction of modified early warning score (MEWS) charts, of patients admitted to a tertiary referral intensive care unit after CPR. Resuscitation. 2011;82(2):150-154.

7. Kyriacos U, Jelsma J, Jordan S. Monitoring vital signs using early warning scoring systems: a review of the literature. J Nurs Manag. 2011;19(3):311-330.

8. Ludikhuize J, Borgert M, Binnekade J, Subbe C, Dongelmans D, Goossens A. Standardized measurement of the Modified Early Warning Score results in enhanced implementation of a Rapid Response System: a quasi-experimental study. Resuscitation. 2014;85(5):676-682.

9. Morgan R, Williams F, Wright M. An early warning scoring system for detecting developing critical illness. Clin Intensive Care. 1997;8(2):100.

10. Gao H, McDonnell A, Harrison DA, et al. Systematic review and evaluation of physiological track and trigger warning systems for identifying at-risk patients on the ward. Intensive Care Med. 2007;33(4):667-679.

11. Subbe CP, Gao H, Harrison DA. Reproducibility of physiological track-and-trigger warning systems for identifying at-risk patients on the ward. Intensive Care Med. 2007;33(4):619-624.

12. DeVita MA, Bellomo R, Hillman K, et al. Findings of the first consensus conference on medical emergency teams*. Critical care medicine. 2006;34(9):2463-2478.

13. Kolic I, Crane S, McCartney S, Perkins Z, Taylor A. Factors affecting response to national early warning score (NEWS). Resuscitation. 2015;90:85-90.

14. Smith GB, Prytherch DR, Schmidt PE, Featherstone PI. Review and performance evaluation of aggregate weighted ‘track and trigger’ systems. Resuscitation. 2008;77(2):170-179.

15. Subbe C, Kruger M, Rutherford P, Gemmel L. Validation of a modified Early Warning Score in medical admissions. Qjm. 2001;94(10):521-526.

16. Jones S, Mullally M, Ingleby S, Buist M, Bailey M, Eddleston JM. Bedside electronic capture of clinical observations and automated clinical alerts to improve compliance with an Early Warning Score protocol. Crit Care Resusc. 2011;13(2):83-88.

17. Shearer B, Marshall S, Buist MD, et al. What stops hospital clinical staff from following protocols? An analysis of the incidence and factors behind the failure of bedside clinical staff to activate the rapid response system in a multi-campus Australian metropolitan healthcare service. Bmj Quality & Safety. 2012;21(7):569-575.

18. Davies O, DeVita MA, Ayinla R, Perez X. Barriers to activation of the rapid response system. Resuscitation. 2014;85(11):1557-1561.

91

19. van Rooijen CR, de Ruijter W, van Dam B. Evaluation of the threshold value for the Early Warning Score on general wards. Neth J Med. 2013;71(1):38-43.

20. Smith T, Den Hartog D, Moerman T, Patka P, Van Lieshout EM, Schep NW. Accuracy of an expanded early warning score for patients in general and trauma surgery wards. Br J Surg. 2012;99(2):192-197.

21. Kruisselbrink R, Kwizera A, Crowther M, et al. Modified Early Warning Score (MEWS) Identifies Critical Illness among Ward Patients in a Resource Restricted Setting in Kampala, Uganda: A Prospective Observational Study. PLoS One. 2016;11(3):e0151408.

22. Greysen SR, Stijacic Cenzer I, Auerbach AD, Covinsky KE. Functional impairment and hospital readmission in medicare seniors. JAMA Intern Med. 2015;175(4):559-565.

23. Zanocchi M, Maero B, Martinelli E, et al. Early re-hospitalization of elderly people discharged from a geriatric ward. Aging Clin Exp Res. 2006;18(1):63-69.

24. Balla U, Malnick S, Schattner A. Early readmissions to the department of medicine as a screening tool for monitoring quality of care problems. Medicine (Baltimore). 2008;87(5):294-300.

25. Ludikhuize J, Hamming A, de Jonge E, Fikkers BG. Rapid response systems in The Netherlands. Jt Comm J Qual Patient Saf. 2011;37(3):138-144, 197.

26. Armagan E, Yilmaz Y, Olmez OF, Simsek G, Gul CB. Predictive value of the modified Early Warning Score in a Turkish emergency department. Eur J Emerg Med. 2008;15(6):338-340.

27. Ho le O, Li H, Shahidah N, Koh ZX, Sultana P, Hock Ong ME. Poor performance of the modified early warning score for predicting mortality in critically ill patients presenting to an emergency department. World J Emerg Med. 2013;4(4):273-278.

THE USE OF QUALITY INDICATORS TO ASSESS PATIENT SAFETY

CHAPTER 6

EXPLORING THE PREVENTABLE CAUSES OF UNPLANNED READMISSIONS USING ROOT CAUSE ANALYSIS: COORDINATION OF CARE IS THE WEAKEST LINK

Louise van Galen & Kristien Fluitman | Hanneke Merten | Saskia M. Rombach | Mikkel Brabrand | Tim Cooksley | Christian H. Nickel | Christian P. Subbe |

Mark H. Kramer | Prabath W. Nanayakkara |

*On behalf of the safer@home consortium

Eur J Intern Med 2016;30:18-24

‘The things that count aren’t always the things that can be counted’ Kiki Lombarts

94

ABSTRACTImportance

Unplanned readmissions within 30 days are a common phenomenon in everyday practice and

lead to increasing costs. Although many studies aiming to analyse the probable causes leading

to unplanned readmissions have been performed, an in depth-study analysing the human

(healthcare worker)-, organizational-, technical-, disease- and patient-related causes leading

to readmission is still missing.

Objective

The primary objective of this study was to identify human-, organizational-, technical-,

disease- and patient-related causes which contribute to acute readmission within 30 days

after discharge using a Root-Cause Analysis Tool called PRISMA-medical.

The secondary objective was to evaluate how many of these readmissions were deemed

potentially preventable, and to assess which factors contributed to these preventable

readmissions in comparison to non-preventable readmissions.

Design

Cross-sectional retrospective record study.

Setting

An academic medical centre in Amsterdam, the Netherlands.

Participants

Fifty patients aged 18 years and older discharged from an internal medicine department and

acutely readmitted within 30 days after discharge.

Main Outcome Measures

Root causes of preventable and unpreventable readmissions.

Results

Most root causes for readmission were disease-related (46%), followed by human (healthcare

worker)- (33%) and patient- (15%) related root causes. Half of the readmissions studied were

considered to be potentially preventable. Preventable readmissions predominantly had human-

related (coordination) failures.

Conclusion and Relevance

Our study suggests that improving human-related (coordinating) factors contributing to

a readmission can potentially decrease the number of preventable readmissions.

95

Keywords

Patient readmission, root cause analysis, quality improvement

Highlights

1. PRISMA-analysis is a suitable method to perform in-depth analysis of readmissions.

2. This study provides a structured analysis of root causes for unplanned readmissions.

3. Half of the readmissions studied were considered to be potentially preventable.

4. Healthcare worker coordination failures were mostly responsible.

96

INTRODUCTIONUnexpected hospital readmissions within 30 days after an index admission are highly prevalent

and costly. The proportion of patients readmitted within 30 days varies widely and is estimated

to be between 7 and 24% across borders.1-4 In some countries readmission rates are used as

a quality and safety criteria to rate and reimburse units.5,6 In 2016 the Dutch government will

introduce readmission rates as an official quality indicator in hospitals.7 To identify the patients

at high risk of readmission, attempts have been made to develop a prediction score based

on clinical variables.8,9 However, these scores have not been validated to predict readmissions

in different populations worldwide and show low to moderate discriminatory power in

predicting readmissions.10-12

More importantly, little is known about the preventability of the unexpected readmissions. It

would seem logical that hospitals only have to accept ramifications for preventable readmissions.

Reported preventable readmission rates vary from 5 % to 79 %.13 However, current literature

has not been able to define clear risk factors that predict a preventable readmission, which

may be targeted to reduce the rate of unexpected readmissions and improve the quality of

care.14 In addition, preventability has not yet been defined uniformly.15

The problem is that readmissions appear to be multi-causal. They are potentially related to

multimorbidity and recent studies have found a progressive increase in readmission risk as

the degree of functional impairment increases.3,16,17 Some research groups have investigated

factors contributing to preventable readmissions in more detail and have classified causes

according to not only patient but also system and social causes.6,18,19 These causes, however,

are often non-modifiable and beyond the reach of implementing process improvements in

hospitals. Currently, no studies have been published on other potentially relevant factors such

as the healthcare worker-, organizational-, technical- and patient-related causes that could

contribute to acute readmissions within 30 days.

A useful tool to analyse these types of root causes is the PRISMA-tool (Prevention and Recovery

Information System For Monitoring and Analysis). The main goal of the PRISMA method is to

build a quantitative database of incidents and process deviations, from which conclusions may

be drawn to suggest optimal countermeasures. This method has been accepted by the World

Alliance for Patient Safety of the World Health Organization.20-22

Insight into these characteristics and potential preventability of acute readmissions may be of

help to ultimately reduce the number of readmissions and the costs attached. Understanding

factors contributing to (preventable) readmissions would help physicians increase the safety

surrounding discharge for patients and their caregivers.

In this retrospective record review study, the main aim was to identify the organizational-,

technical-, healthcare worker- and patient-related causes that contribute to readmissions

97

using “PRISMA-medical” analysis. The secondary aim was to evaluate how many of these

readmissions were deemed potentially preventable, and to assess what factors contributed to

these preventable readmissions in comparison to non-preventable readmissions.

SUBJECTS AND METHODSStudy Design

This retrospective, cross-sectional record review study included readmissions at the VU University

Medical Centre in Amsterdam, the Netherlands. The VUmc is an academic medical centre with

approximately 3400 admissions per annum to its Acute Medical Unit (AMU) and a hospital

wide readmission rate of around 10%. Previous studies have shown that 50 PRISMA-analyses

are valid and sufficient basis for a reliable causal-profile.[23] In the current study, we therefore

decided that 50 records would be included for a review to explore the causes and potential

preventability of unplanned readmissions. All readmission records of the year 2013 meeting

the inclusion criteria were selected. To get the most recent results, the reviewers started with

the last record from December 2013 and included each consecutive record backwards until

the number of 50 records was reached, with the last record coming from June 2013.

The following criteria were used for inclusion into the study sample: patients aged 18 years

and older; initial discharge from the internal medicine ward (which included the following

specialties: general internal medicine, nephrology, oncology, geriatrics, haematology,

pulmonary medicine); admission and readmission through the emergency department in 2013

and readmission to any department/ward, regardless of the medical specialty within 30 days

after initial discharge. All records not meeting the above mentioned inclusion criteria were

excluded from the study.

If a patient was readmitted more than once during the study period only the first readmission

was analysed. The local Medical Ethics Committee approved this study.

Assessment

Doctor’s charts, nurse’s charts and electronic patient files including all test results were

available for analysis. For each individual readmission information on patient characteristics

(such as age, co-morbidity23, living situation) and circumstances under which initial discharge

and readmission took place (such as length of stay and specialty for admission) was collected

according to a standardized chart abstraction form (appendix 1). Two medically and PRISMA-

trained investigators (LG, KF) reviewed each case separately and filled out these chart

abstraction forms. Subsequently consensus was reached.

PRISMA Analysis

In order to perform PRISMA-analysis on all readmissions, the above mentioned data collection

forms were used. These consisted of a free text description of the circumstances contributing

to the readmission and identification of direct, indirect and root causes. This information

98

was used to construct a causal tree. At the top of each tree the readmission was placed as

an unforeseen adverse event. The direct causes of the readmissions were noted hereunder.

Under each of the direct causes the indirect causes were stated. By constantly asking ‘why’ an

event and each subsequent event had taken place relevant indirect causes were revealed. This

continued until no more objective information was available to reveal an underlying cause.

The last noted indirect cause was labelled as root cause and was located at the bottom of

the causal tree (Figure 1).

Root causes were classified as technical-, organizational-, human- (healthcare worker) and

patient-related factors according to the Eindhoven Classification Model, see Table 1.20,21

Disease-related causes were added as fifth category to this model. We anticipated that

progression of disease would be identified as root cause in many readmissions without any

technical-, organizational-, healthcare worker- or patient-related factors contributing to

the readmission. Table 1 shows the subcategories of PRISMA-root causes with case examples

from our study. Finally, both reviewers studied the cases independently and extensively. They

concluded whether the readmission related to its index admission and if it was potentially

preventable or not. The readmission was judged to be related to the index admission if both

admissions were based on the same medical issue or if cause of the readmission originated

during index admission. A readmission was judged as potentially preventable if it could have

been reasonably foreseen by discharging physician and could reasonably have been prevented

by any action undertaken by hospital staff or the patient. If consensus could not be reached

satisfactorily, a third independent party was consulted. These cases were then re-analysed and

discussed with a senior physician (PN) and a psychologist with a special interest in PRISMA-

analysis (HM). The mean time to assess a case by a single reviewer was 72 minutes (range

25-180, SD=71), reaching consensus thereafter took up approximately 30 minutes per case.

Statistical Analysis

Descriptive characteristics and frequencies were calculated in SPSS version 22.0. Categorical

outcome measures are presented as frequencies and percentages. Continuous variables are

summarized by median and interquartile ranges since none of them were normally distributed.

To illustrate the comparison in patient characteristics between preventable and non-preventable

readmissions we used the Mann-Whitney U test for continuous and ordinal variables. Pearson’s

chi-square test and Fisher’s exact were used for dichotomous and categorical data. P-values

below 0,05 were considered significant.

RESULTSPatient characteristics

For 46 out of the 50 reviewed cases, consensus on the root-cause analysis and preventability

was reached by the two reviewers. Independent observers were consulted to assess the root

causes and preventability in 4 cases, after which agreement was reached. During the study

period, six cases had more than one readmission within 30 days. Table 2 shows baseline

99

Figure 1 | Two examples of causal trees IA: index admission, DRF: disease-related factor, HRC: human-related coordination, PRF: patient-related factor, H-ex: human-related external.

characteristics of the study population. The mortality one year after index admission was 40%

(n=20). Most readmissions (n=28, 56%) occurred more than one week after discharge, about

one third occurred within 72 hours after discharge (n=17, 34%). In 13 (26%) readmissions

no discharge letter was provided. For 35 (70%) of the readmissions the admitting specialty

for the index admission was the same as for the readmission. The majority (n=45, 90%) lived

independently before the index admission. A large proportion of the population (n=36, 72%)

used five or more drugs at the time of the index admission.

Preventability

Reviewers judged more than half (n=26, 52%) of readmissions as potentially preventable.

All preventable readmissions were found to be related to the index admission. However,

16 (32%) readmissions that were related were considered not preventable. There were no

significant differences in patient characteristics between the (potentially) preventable versus

non-preventable readmission-groups although this should be interpreted with caution because

of the small sample size (Table 2). An example of a related and preventable human-related

caused readmission within one week after discharge concerned a patient who was readmitted

with malaise and self-neglect. During the index admission both patient and family members

had repeatedly expressed their concerns about the discharge. They felt the patient was unable

100

Table 1 | Description of categories of the Eindhoven Classification Model: PRISMA-medical Version.20,21

Main category Subcategory Code Description Example (if available)

Technical External T-ex Technical failures beyond the control of the organization

Design TD Failures to poor design of equipment etc.

Construction TC Correct design inappropriately constructed or placed

Materials TM Material defects not classified under TD or TC

Organizational External O-ex Failures at an organizational level beyond the control and responsibility of the investigating team

Transfer of knowledge

OK Failure resulting from inadequate measures to train or supervise new or inexperienced staff

Protocols OP Failures relating to the quality or availability of appropriate protocols

Management priorities

OM Internal management decisions which reduce focus on patient safety when faced with conflicting priorities

Culture OC Failure due to attitude and approach of the treating organization.

Human External H-ex Human failures beyond the control of the organization/department

Nursing home did not monitor fluid restriction

Knowledge-based behaviour

HKK Failure of an individual to apply their knowledge to a new clinical situation

Not adjusting tramadol dosage to poor kidney function

Qualifications HRQ An inappropriately trained individual performing the clinical task

Co-ordination HRC A lack of task co-ordination within the healthcare team.

Inadequate medication handover to the general practitioner

Verification HRV Failure to correctly check and assess the situation before performing interventions

Urine culture not verified before patient was discharged

Intervention HRI Failure resulting from faulty task planning or performance

Monitoring HRM Failure to monitor the patient’s progress or condition

Response to oral antibiotics not monitored after switching from intravenous antibiotics

101

to deal with her own medication and to take care of herself. Hospital staff did report this in

the medical records, but they did not take extra discharge precautions or arrange extra care

at home. An example of a readmission judged as non-related and non-preventable is the case

of a patient initially being admitted for a community-acquired pneumonia. Two weeks later

the patient was readmitted with angioedema after the start of a new antihypertensive drug

prescribed by the general practitioner, which had nothing to do with the initial admission.

Root causes

After PRISMA-analysis, a total of 100 root causes were identified. Twenty-one (42%) of

the readmissions had a single root cause, 15 (30%) had two root causes, 8 (16%) three root

causes, 5 (10%) four root causes and one readmission (2%) had five root causes. The mean

number of root causes per readmission was 2.0 (SD = 1.1). Figure 1 shows two examples of

causal trees. Table 1 presents the subcategories of PRISMA-root causes with case examples

from our study.

The results in Figure 2 show the distribution over the root causes plotted against preventability. It

illustrates which root causes were identified for preventable and non-preventable readmissions.

Regardless of preventability, most root causes were disease-related (n=46, 46%), followed by

human- (n=33, 33%) and patient- (n=15, 15%) related root causes. Human-related causes

were causes which were healthcare worker related. Disease-related causes (DRF) consisted of

natural progression of disease, which means that they could not be influenced by the physician

or patient. This is, for example, evident in the case of a patient being readmitted due to

symptoms caused by brain metastases from melanoma. Patient-related causes (PRF) originated

from failures related to patient characteristics or conditions which were beyond the control

of hospital staff, for example the refusal of further diagnostics or treatment by the patient.

Finally, 6% of the root causes were unclassifiable. We did not identify any technical (T) or

organizational (O) causes. Organizational causes would include failures that are related to

Table 1 | (continued)

Main category Subcategory Code Description Example (if available)

Skills-based HSS Failure in performance of highly developed skills

Patient Patient-related PRF Failures related to patient characteristics or conditions, which are beyond the control of staff and influence clinical progress

Non-compliance in taking prescribed medication

Disease-related DRF Failures related to the natural progress of disease which are beyond control of patient, its carers and staff

Subjective dyspnoea without disease- or patient-related substrate

102

Table 2 | Baseline characteristics stratified by preventability

Patient CharacteristicsTotal (N=50)

Considered Preventable (N=26)

Considered Non-preventable (N=24) P-value*

Age – median(range) 71.0(32-97) 75.0(41-97) 70.0(32-90) 0,0771

Gender (%(N))

male

female

38.0(19) 62.0(31)

34.6(9) 65.4(17)

41.7(10) 58.3(14)

0,6082

Deceased (%(N))

< 6 months after IA*

6-12 months after IA

> 12 months after IA

Unknown

26.0(13)

14.0(7) 4.0(2)

56.0(28)

26.9(7) 15,4(4) 7,7(2)

50.0(13)

25.0(6) 12,5(3) 0.0(0)

62.5(15)

0,4622

Length of stay IA (days) - median(range) 6.0(2-25) 7.5(2-24) 5.0(2-25) 0,1461

Length of stay RA (days)- median(range) 6.0(2-45) 6.5(2-44) 5.0(2-45) 0,2581

Length between IA – RA (%(N))

< 24 h

24 - 72h

72 h-1 week

More than 1 week

2.0(1)

32.0(16)

10.0(5) 56.0(28)

3.8(1) 26.9(7) 11.5(3) 57.7(15)

0.0(0) 37.5(9) 8.3(2) 54.1(13)

0.9041

Medical Speciality IA (%(N)

Geriatrics

Haematology General internal Nephrology

Oncology

Pulmonary medicine

2.0(1) 2.0(1) 54.0(27)

26.0(13) 14.0(7)

2.0(1)

0.0(0) 0.0(0) 61.5(16) 26.9(7) 11.5(3) 0.0(0)

4.2(1) 4.2(1) 45.8(11) 25.0(6) 16.7(4) 4.2(1)

0,6753

Speciality IA same as RA (%(N))

yes no

70.0(35) 30.0(15)

76.9(20) 23.1(6)

62.5(15) 37.5 (9)

0,2662

Living situation (%(N))

independent dependent

90.0(45) 10.0(5)

88.5(23) 11.5(3)

91.7(22) 8.3(2)

1,03

Living with partner (%(N))

yes

no

Unknown

56.0(28) 42.0(21)

2.0(1)

53.8(14) 42.3(11)

3.8(1)

58.3(14) 41.7(10)

0.0(0)

0.8692

Home care prior to IA (%(N))

None

Formal

Informal

Not applicable

Unknown

32.0(16) 18.0(9) 38.0(19)

10.0(5)

2.0(1)

30.8(8) 30.8(8) 23.1(6) 11.5(3)

3.8(1)

33.3(8) 4.2(1) 54.2(13) 8.3(2)

0.0(0)

1,0003

KATZ*24-score IA – median (range) 0.0(0-6) 1.0(0-6) 0.0(0-6) 0,0941

SNAQ*25-score IA – median (range) 0.5 (0-6) 0.0(0-6) 1.0(0-3) 0,7811

CCI*23 IA- median (range) 6.0 (1-14) 6.0(2-11) 6.5(1-14) 0,6241

103

Table 2 | (continued)

Patient CharacteristicsTotal (N=50)

Considered Preventable (N=26)

Considered Non-preventable (N=24) P-value*

Polypharmacy* (%(N))

yes

no

72.0(36) 28.0(14)

80.8(21) 19.2(5)

62.5(15) 37.5(9)

0,1512

Fall in last 6 months (%(N))

yes

no

Unknown

28.0(14) 70.0(35)

2.0(1)

34.6(9) 65.4(17)

0.0(0)

20.8(5) 75.0(18)

4.2(1)

0,3192

Use of mobility device (%(N))

yes

no

36.0(18) 64.0(32)

38.5(10) 61.5(16)

33.3(8) 66.7(16)

0,7062

Complications during IA (%(N))

yes

no

18.0(9) 82.0(41)

19.2(5) 80.8(21)

16.7(4) 83.3(20)

1.0003

Location of discharge (%(N))

Home

Other institution

84.0(42)

16.0(8)

80.8(21) 19.2(5)

87.5(21) 12.5(3)

0,7043

Living situation after discharge IA dependentindependent (%(N))

yes no

6.0(3) 94.0(47)

7.7(2) 92.3(24)

4.2(1) 95.8(23)

1.0003

Extra care after IA (%(N))

yes

no

22.0(11) 78.0(39)

26.9(7) 73.1(19)

16.7(4) 83.3(20)

0,7223

Outpatient follow-up (%(N))

yes

no

68.0(34) 32.0(16)

61.5(16) 38.5(10)

75.0(18) 25.0(6)

0,3082

Discharge letter send to GP (%(N)

yes

no

74.0(37)

26.0(13)

76.9(20)

23.1(6)

70.8(17)

29.2(7)

0,6242

*IA: Index Admission, RA: Readmission, Polypharmacy: the concomitant use of five or more drugs, CCI: Charlson Comorbidity Index corrected by age[24], SNAQ: assessment of nutritional status[32]:, KATZ: score tool for assessing a patient’s ability to perform activities of daily living [31]**1: Mann-Whitney U, 2: Pearson chi-squared, 3: Fisher’s exact test.

quality and availability of protocols within the department. A technical cause could originate

from failures resulting from poor design of equipment such as software.

Of 100 root causes, 69 were identified in patients with readmissions regarded as preventable, in

contrast to 31 that were identified in a non-preventable readmission. Results reveal that a non-

preventable readmission was never caused by a human-related factor; these were only found in

readmissions regarded as preventable. Nearly half (46.3%) of the root causes in readmissions

104

identified as preventable consisted of human-related factors. Further subcategorized, these

preventable human-related factors originated mostly from coordination (13.0%) and monitoring

(11.6%) failures. The first represent failures in task coordination within a healthcare team, i.e.

no clear medication handover towards the home situation. The latter comprises incompetence

in monitoring a process or patient status. This is exemplified in our study by a patient admitted

with urosepsis where the response to oral antibiotics was not monitored after switching from

intravenous to oral antibiotics. The patient was readmitted a few days after discharge with

high fever, and additional diagnostics at readmission revealed that bacteria found in the urine

culture were not sensitive to the oral antibiotics given. In addition, patient-related causes

explained 17.4% of potentially preventable readmissions. External factors were also present

(7.2%). These are causes originating in areas beyond the control of the hospital. For instance,

causes resulting from failures by a general practitioner or within a nursing home (Figure 1).

Non-preventable readmissions were mostly caused by disease related factors (71.0%).

DISCUSSIONThe purpose of the present retrospective record review study was to identify organizational-,

technical-, human-, disease and patient-related causes which contributed to readmissions.

The most important finding from this study is that half of the readmissions were considered

potentially preventable. The second major finding was that human (healthcare worker)-

related root causes were exclusively found in preventable readmissions. This was mostly due

Figure 2 | Root causes readmissions

105

to human-related coordination failures. Non-preventable readmissions were mostly the result

of disease-related factors.

To our knowledge, the current study is the first on causes and preventability of unplanned

readmissions performed in the Netherlands. However, using large databases, efforts have been

made globally to identify those patients at a higher risk of readmission and to develop risk

models.8,9,12,19 These risk models attempt to predict chances of readmission based on several

patient characteristics (for example the Charlson Comorbidity Index) which can be derived

from big hospital data systems. The problem with these models is that they have not yet been

validated in different populations. In addition, as highlighted by Van Walraven (2015) there is

no risk model incorporating aspects of preventability in predicting readmissions. These aspects

are of particular interest, as they could be the key to improvement measures in preventing

readmissions and facilitating a safer discharge home.10,11,26

Many studies have tried to assess the proportion of preventable readmissions and patient

characteristics related to these readmissions. In a review of 34 studies Van Walraven et al.

(2011) found that the percentage of preventable readmissions varied from 5% to 79%, with

a median of 27.1%.13 Our results of 52% preventable readmissions are in agreement with

results found in more recent studies where percentages are around 45%.6,27,28

As mentioned above, the majority of studies performed focused mainly on patient and disease

characteristics associated with preventable readmissions such as age and comorbidity.3,5,29

However, as already stated by Joynt (2012), there is a growing body of evidence that suggests

that primary drivers of variability in readmission rates reach beyond patient characteristics

since they are multi-factorial.30 Our results are consistent with recent studies which suggest

that in order to predict potentially preventable readmissions one should look into patient- and

human-related factors.6,28 Feigenbaum et al. (2012) suggested that potentially preventable

readmissions were due to patient- and human-related factors such as suboptimal medical

treatment and a lack of coordination in the discharge process.27 Bianco et al. (2012) suggested

that preventable readmissions are often associated with diagnostic and therapeutic errors, but

also with patient-related factors such as non-adherence.31 In contrast to our study, they have

not subdivided root causes into specific areas for improvement. Furthermore, despite the fact

that some previous studies did have independent reviewers, they do not elaborate precisely on

how they interpret and evaluate the reasons for readmission as being potentially preventable.

The outcomes of this study are of interest from a policy perspective. As mentioned before,

readmission rate will be included as an official quality indicator for Dutch hospitals in 2016.7

However, this careful in-depth analysis of 50 readmissions in our study showed that half of

the readmissions is non-preventable and the majority of their causes are either patient- or

disease-related, which mostly lay outside the influence of hospital personal. Therefore, it is

questionable to what extent the general readmission rate can be reliably linked to the quality

provided by the hospital and the hospital care professionals. Our study shows that careful

106

consideration is needed to judge whether a readmission may have been preventable, and

therefor of interest from a policy perspective. A large scale multicentre study analysing

readmissions could be a first step to show whether the results of our study also apply to other

hospitals. The PRISMA-analysis would be a useful tool to use, because it systematically addresses

the causes that may have contributed to the readmission. It is able to provide the reviewer with

the information needed to judge whether a readmission is potentially preventable. Analysing

these causes implicitly provides valuable information for improvement.

Furthermore, the strength of our study lies in case by case comprehensive investigation,

performed by two independent researchers to find the root causes of readmissions using

the full extent of clinical data available for each case. To our knowledge, this is the first study

that explore the underlying causes of readmissions and determine their eventual preventability.

This retrospective nature of this study limited reviewers to patient records which meant that

they had to rely solely on information that was written down in the patient records when

investigating root causes of readmissions. In addition, no technical or organizational factors

could be discovered to be instrumental in readmissions. As illustrated by Van der Schaaf et al.

these often entail latent causes. An organizational factor could for instance be a shortage of

hospital staff.21 However, such causes are not typically written down in patient charts and were

therefore inaccessible to the reviewers. This also applies to cultural factors such as hospital-wide

accepted faulty behaviour (e.g. the hospital-wide acceptance of sending discharge letters to

the general practitioner with substantial delay). This is unfortunate since these organizational

factors could provide very suitable grounds for improvements. Finally, our sample size was

relatively limited. However, when using PRISMA-tool a population size of 50 is thought to offer

a sufficient variability to perform a reliable root cause analysis.32

CONCLUSIONDespite its exploratory nature, this study offers insight into the healthcare worker- and patient-

related causes of readmissions. The relevance of human-related factors is clearly supported

by the fact that human-related causes are exclusively found in preventable readmissions.

The most significant finding of this study was that the most prevalent human-related cause of

preventable readmission was the lack of coordination within the healthcare system. The most

frequent findings were inadequate assessment of a patient’s home situation and the lack of

a proper information handover to the general practitioner, the patient and its carers at home.

This flaw in communication between caregivers is also highlighted by the large number of

missing discharge letters (n=13, 26%). In our opinion these results could function as starting

points for improvement. Perhaps a more standardized method of discharge could help reduce

the amount of potentially preventable readmissions. This should include requirements of

a clear handover of care to the primary physician and a clear assessment of extra care needed

at home during index admission.

To study the possible usefulness of these interventions, prospective study designs are needed

where more relevant and reliable information on factors that may have contributed to

107

the readmission can be collected. Perspectives of patients their informal carers and involved

healthcare professionals can also provide valuable additional information. This will provide

more detailed information on communication breakdowns and gaps in information transfer.

In addition, causal factors responsible for readmissions in different healthcare systems can be

diverse. Shortly a prospective pan-European study named ‘ the CURIOS@-study (CaptUring

Readmission InternatiOnally to prevent readmission by safer@home consortium) will start

recruiting patients with the aim of investigating the predictability and preventability of

readmissions (clinicaltrials.gov: NCT02621723).

ACKNOWLEDGEMENTSAll members of safer@home consortium: M. Brabrand, T. Cooksley, K.S. Fluitman, L.S. van

Galen, R. Kidney, J. Kellett, H. Merten, P.W. Nanayakkara, C.H. Nickel, J. Soong, C.P. Subbe,

L. Vaughan, I. Weichert

SUPPORTING INFORMATION Appendix 1. Acute Readmissions: Data collection sheet chart review.

108

REFERENCES1. Blunt I, Bardsley M, Grove A, Clarke A. Classifying emergency 30-day readmissions in England using

routine hospital data 2004-2010: what is the scope for reduction? Emergency medicine journal : EMJ. 2015;32(1):44-50.

2. Graham H, Livesley B. Can readmissions to a geriatric medical unit be prevented? Lancet. 1983;1(8321):404-406.

3. Donze J, Lipsitz S, Bates DW, Schnipper JL. Causes and patterns of readmissions in patients with common comorbidities: retrospective cohort study. BMJ (Clinical research ed.). 2013;347:f7171.

4. Shimizu E, Glaspy K, Witt MD, et al. Readmissions at a public safety net hospital. PloS one. 2014;9(3):e91244.

5. Balla U, Malnick S, Schattner A. Early readmissions to the department of medicine as a screening tool for monitoring quality of care problems. Medicine. 2008;87(5):294-300.

6. Yam CH, Wong EL, Chan FW, et al. Avoidable readmission in Hong Kong--system, clinician, patient or social factor? BMC health services research. 2010;10:311.

7. Inspectorate for Public Health: Quality indicators hospitals 2016. Utrecht. http://www.igz.nl/Images/IGZ%20Basisset%20kwaliteitsindicatoren%20ziekenhuizen%202016_tcm294-367407.pdf. Updated September 2015. Accessed December, 2015.

8. Donze J, Aujesky D, Williams D, Schnipper JL. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA internal medicine. 2013;173(8):632-638.

9. van Walraven C, Wong J, Forster AJ. LACE+ index: extension of a validated index to predict early death or urgent readmission after hospital discharge using administrative data. Open medicine : a peer-reviewed, independent, open-access journal. 2012;6(3):e80-90.

10. Cotter PE, Bhalla VK, Wallis SJ, Biram RW. Predicting readmissions: poor performance of the LACE index in an older UK population. Age and ageing. 2012;41(6):784-789.

11. Cooksley T, Nanayakkara PW, Nickel CH, et al. Readmissions of medical patients: an external validation of two existing prediction scores. QJM : monthly journal of the Association of Physicians. 2015.

12. Spiva L, Hand M, VanBrackle L, McVay F. Validation of a Predictive Model to Identify Patients at High Risk for Hospital Readmission. Journal for healthcare quality : official publication of the National Association for Healthcare Quality. 2015.

13. van Walraven C, Bennett C, Jennings A, Austin PC, Forster AJ. Proportion of hospital readmissions deemed avoidable: a systematic review. CMAJ : Canadian Medical Association journal = journal de l’Association medicale canadienne. 2011;183(7):E391-402.

14. Vest JR, Gamm LD, Oxford BA, Gonzalez MI, Slawson KM. Determinants of preventable readmissions in the United States: a systematic review. Implementation science : IS. 2010;5:88.

15. Maurer PP, Ballmer PE. Hospital readmissions--are they predictable and avoidable? Swiss medical weekly. 2004;134(41-42):606-611.

16. Kwok T, Lau E, Woo J, et al. Hospital readmission among older medical patients in Hong Kong. Journal of the Royal College of Physicians of London. 1999;33(2):153-156.

17. Greysen SR, Stijacic Cenzer I, Auerbach AD, Covinsky KE. Functional impairment and hospital readmission in medicare seniors. JAMA internal medicine. 2015;175(4):559-565.

18. Lindquist LA, Baker DW. Understanding preventable hospital readmissions: masqueraders, markers, and true causal factors. Journal of hospital medicine. 2011;6(2):51-53.

19. Barnett ML, Hsu J, McWilliams JM. Patient Characteristics and Differences in Hospital Readmission Rates. JAMA internal medicine. 2015.

109

20. van Vuuren W SC, Schaaf TW. The development of an incident analysis tool for the medical fiels Eindhoven website. [http://alexandria.tue.nl/repository/books/493452.pdf]. 1997. Accessed March, 2015.

21. Team helps hospital avoid readmission penalties. Hospital case management : the monthly update on hospital-based care planning and critical paths. 2013;21(4):50-51.

22. Antibiotic therapy for acute appendicitis in adults. Fewer immediate complications than with surgery, but more subsequent failures. Prescrire international. 2014;23(150):158-160.

23. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of chronic diseases. 1987;40(5):373-383.

24. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. STUDIES OF ILLNESS IN THE AGED. THE INDEX OF ADL: A STANDARDIZED MEASURE OF BIOLOGICAL AND PSYCHOSOCIAL FUNCTION. Jama. 1963;185:914-919.

25. Kruizenga HM, de Jonge P, Seidell JC, et al. Are malnourished patients complex patients? Health status and care complexity of malnourished patients detected by the Short Nutritional Assessment Questionnaire (SNAQ). European journal of internal medicine. 2006;17(3):189-194.

26. van Walraven C. The Utility of Unplanned Early Hospital Readmissions as a Health Care Quality Indicator. JAMA internal medicine. 2015:1-2.

27. Feigenbaum P, Neuwirth E, Trowbridge L, et al. Factors contributing to all-cause 30-day readmissions: a structured case series across 18 hospitals. Medical care. 2012;50(7):599-605.

28. Cakir B, Gammon G. Evaluating readmission rates: how can we improve? Southern medical journal. 2010;103(11):1079-1083.

29. van Walraven C, Dhalla IA, Bell C, et al. Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. CMAJ : Canadian Medical Association journal = journal de l’Association medicale canadienne. 2010;182(6):551-557.

30. Joynt KE, Jha AK. Thirty-day readmissions--truth and consequences. The New England journal of medicine. 2012;366(15):1366-1369.

31. Bianco A, Mole A, Nobile CG, Di Giuseppe G, Pileggi C, Angelillo IF. Hospital readmission prevalence and analysis of those potentially avoidable in southern Italy. PloS one. 2012;7(11):e48263.

32. van Wagtendonk I, Smits M, Merten H, Heetveld MJ, Wagner C. Nature, causes and consequences of unintended events in surgical units. The British journal of surgery. 2010;97(11):1730-1740.

CHAPTER 7

HOSPITAL READMISSIONS: A RELIABLE QUALITY INDICATOR?

Louise S. van Galen | Prabath W.B. Nanayakkara

Ned Tijdschr Geneeskd 2016;160:A9885

‘Culture eats policy for breakfast‘ Peter Drucker

112

ABSTRACTThe percentage of readmissions within 30 days after discharge is an official quality indicator

for Dutch hospitals in 2016. In this commentary the authors argue why readmissions cannot

be regarded as a reliable way of assessing quality of healthcare in a hospital. To date, policy

makers have been struggling with its precise definition and the indicator has not been properly

formulated yet. It does not distinguish between planned and unplanned readmissions and

does not take into account the ‘preventability’. Therefore the authors believe that the indicator

in its current form might falsely interpret the quality of care of a hospital and it is questionable

to use readmissions as a quality indicator.

113

MAIN TEXT (TRANSLATED FROM DUTCH)In the 2015 cooperation agreement between the Dutch Inspectorate of Healthcare and Dutch

hospitals it was decided that the percentage of unplanned readmissions within 30 days after

discharge of a clinical admission in the same hospital was added as an official quality indicator.1

Readmission, together with already existing quality indicators: unexpected long length of stay

and in-hospital mortality, are regarded as a ‘major adverse events’ and these deem to indicate

the negative results of clinical healthcare. It is known that in an older population mortality

increases if these patients are readmitted frequently.2 For several years, healthcare policy

makers in the USA and the UK have already used the percentage of readmissions as a marker

of quality and hospital safety. A ‘high’ readmission rate often results in financial penalties in

these countries.3,4 In the Netherlands there is hardly any literature concerning readmissions

and policy makers struggle with the realization of a reliable indicator.

Internationally, there seems to be a lack of insight in the predictability and preventability of

readmissions. As a reaction to this problem extensive research has been performed using

retrospective databases to create prediction models which could potentially indicate the risk of

readmission. In these models demographics and severity of disease characteristics are used to

assess if there is a significant difference in the patients who are, and who are not readmitted.

The question rises if this research and composed models truly add to increasing the quality

of care. Knowing a patient with heart-failure will have a higher chance of getting readmitted

does not tell us what the direct cause of this readmission entails, and can therefore not be

used to improve the handling of these patients.

Current literature has shown that none of the developed prediction models are validated

adequately. These models are not yet capable to predict the chance of readmission in an average

hospital population.7 Also, in daily practice, these ‘static’ predictors, are often not useful as

efficient targets aiming to prevent readmission. Recently published work using a PRISMA-

analysis has shown that causes of potentially preventable readmissions are mostly human-

related coordination and communication failures.8 An example is a patient who is readmitted

due to a medication-related complication which occurred because no clear handover was

available for the general physician, another example from this work is a readmission in which

the home situation of the patient was not assessed thoroughly enough during index admission.

In the elaborate international literature, often no distinction is made between preventable

and non-preventable readmissions when describing potential predictors of readmission. Many

readmissions are the cause of the natural course of disease, for example a readmission because

of an epileptic insult caused by brain metastasis of a melanoma.9,10 This seems important

since preventable readmissions significantly increase mortality within six months as compared

to non-preventable readmission. 3 Available research, however, has not yet defined clear risk

114

factors that could predict or prevent such readmissions. Therefore, policy makers still have

minimal tools to reduce preventable readmissions.

Given the before mentioned one must critically ask if ‘readmission’ is a reliable quality indicator

in its current form. The indicator does not distinguish between predictable and non-predictable

readmissions, and more importantly, no clear definition has been formulated yet. This might

result in systemic distortion. If a hospital has a relatively high percentage of non-preventable

readmissions, this does not automatically imply that their quality of care is not up to standards.

This hospital might strive for admissions as short as possible in patients who are in a terminal

stage of their disease. This could improve their quality of life since these patients are able to

spend as much time as home as possible with their loved ones, and therefore does not indicate

improper quality of care but rather the opposite. In addition, some healthcare institutions

struggle with the balance between high burden on available beds and need for readmissions.

In the current health system doctors are under constant pressure to discharge patient rapidly.

Also, the increase of acute medical units in the Netherlands, which increases patient flow and

decrease hospital length of stay might unintentionally increase readmission rate.

In the Netherlands empirical research concerning readmissions that could be prevented in our

healthcare system is lacking. In addition, patients are also increasingly seen as active partners

in healthcare: it seems logical to ask patients and their carers to participate in the judgement of

quality of care in hospitals. In early 2016 CURIOS@HOLLAND, the first prospective readmission

project will commence in several Dutch hospitals. The aim of this study is to assess risk factors

and causes that could potentially predict preventable readmissions by using the perspectives

of the most important participants in the healthcare chain.

CONCLUSIONThe percentage of readmissions within 30 days after discharge in the same hospital is not

a reliable quality indicator to measure quality of care and patient safety of a hospital. Current

research has mainly focused on retrospective analysis of databases to derive predictors,

however readmissions do not seem to be ‘static’ and are not easily framed into figures. In

order to improve quality of care in hospitals problem-orientated interventions are needed

which mainly focus on unplanned potentially preventable readmissions. A quality indicator

should not use the total amount of readmissions, but should solely address the ones which

are preventable. This could provide us with a more honest and goal-directed way to assess

the quality in hospitals. However, the question remains if it wouldn’t be more beneficial to

drop this indicator in total. In our vision, in every hospital, regardless the preventability, it

should be natural to investigate the causes for each individual readmission that has returned

back to hospital unintentionally within 30 days after discharge.

115

REFERENCES1. Basisset Kwaliteitsindicatoren 2016. Utrecht: Inspectie voor de Gezondheidszorg, 2015. Accessed

November, 2015.

2. Zanocchi M, Maero B, Martinelli E, Cerrato F, Corsinovi L, Gonella M, et al. Early re-hospitalization of elderly people discharged from a geriatric ward. Aging Clin Exp Res 2006;18: 63-9. Medline

3. Balla U, Malnick S, Schattner A. Early readmissions to the department of medicine as a screening tool for monitoring quality of care problems. Medicine (Baltimore) 2008;87: 294-300. Medline

4. Greysen SR, Stijacic Cenzer I, Auerbach AD, Covinsky KE. Functional impairment and hospital readmission in Medicare seniors. JAMA Intern Med 2015;175:559-65. Medline

5. van der Ven MJ, Schoon Y, Olde Rikkert MG. Ongeplande heropnames bij kwetsbare ouderen: Retrospectieve analyse van opnames in een academisch ziekenhuis. Ned Tijdschr Geneeskd 2015;159:A9211. Medline

6. Donze J, Aujesky D, Williams D, Schnipper JL. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA Intern Med 2013;173:632-8. Medline

7. Cooksley T, Nanayakkara PW, Nickel CH, Subbe CP, Kellett J, Kidney R, et al. Readmissions of medical patients: an external validation of two existing prediction scores. QJM 2015 Jul 10. Medline

8. Fluitman K, van Galen L, Merten H, Rombach S, Brabrand M, Cooksley M, et al. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Int Med IN PRESS

9. Halfon P, Eggli Y, Pretre-Rohrbach I, Meylan D, Marazzi A, Burnand B. Validation of the potentially avoidable hospital readmission rate as a routine indicator of the quality of hospital care. Med Care 2006;44:972-81. Medline

10. Bianco A, Mole A, Nobile CG, Di Giuseppe G, Pileggi C, Angelillo IF. Hospital readmission prevalence and analysis of those potentially avoidable in southern Italy. PLoS One 2012;7:e48263. Medline

CHAPTER 8

PHYSICIAN CONSENSUS ON PREVENTABILITY AND PREDICTABILITY OF READMISSIONS BASED ON

STANDARD CASE SCENARIOS

Louise S. van Galen | Tim Cooksley | Hanneke Merten | Mikkel Brabrand | Caroline B. Terwee | Christian H. Nickel | Christian P. Subbe | Rachel M. Kidney |

John Soong | Louella Vaughan | Immo Weichert | Mark H. Kramer | Prabath W. Nanayakkara |

*On behalf of the safer@home consortium

Neth J Med 2016;74(10):434-42

‘Don’t get too comfortable too change’ Graham Dixon

118

ABSTRACTBackground

Policy makers struggle with unplanned readmissions as a quality indicator since integrating

preventability in such indicators is difficult. Most studies on the preventability of readmissions

questioned physicians whether they consider a given readmission to be preventable, from

which conclusions on factors predicting preventable readmissions were derived. There is no

literature on the interobserver agreement of physician judgement.

Aim

To assess the degree of agreement among physicians regarding predictability and preventability

of medical readmissions.

Design

An online survey based on eight real-life case scenarios was distributed to European physicians.

Methods

Physicians were requested to rate from the first four (index admission) scenarios whether they

expected these patients to be readmitted within 30 days (the predictability). The remaining

four cases, describing a readmission, were used to assess the preventability. The main outcome

was the degree of agreement among physicians determined using intra class correlation

coefficient (ICC).

Results

526 European medical physicians completed the survey. Most physicians had internal medicine

as primary specialism. The median years of clinical experience was 11 year. ICC for predictability

of readmission was 0.67 (moderate to good) and ICC for preventability of readmission was

0.13 (poor).

Conclusion

There was moderate to good agreement among physicians on the predictability of readmissions

while agreement on preventability was poor. This study indicates that assessing preventability of

readmissions based solely on the judgement of physicians is far from perfect. Current literature

on the preventability of readmissions and conclusions derived on the basis of physician opinion

should be interpreted with caution.

Keywords

Patient safety, Quality improvement, Readmissions

119

INTRODUCTIONHospital readmissions within 30 days are of interest to many policy makers internationally.1

They are used as a quality and safety indicator with financial penalties levied in many countries

including the United States and United Kingdom.2 The main problem in using readmissions as

a quality indicator is that the preventability of these readmissions is not properly defined and

integrated in this indicator making it difficult to use as a genuine measure of quality of care.3,4

By not distinguishing between preventable and non-preventable readmissions this indicator

might therefore result in distorted evaluation of hospital care. Furthermore, there is increasing

evidence that the causes of mostly medical readmissions are often multifactorial and usually

the result of natural disease progression, underlying comorbidity or socio-environmental

factors beyond the control of hospital and not solely caused by inadequate hospital care.5-9

The use of readmissions as a quality indicator necessitates that they reflect poor care, are

preventable and that a consensus definition for these two aspects is agreed.

Previous research has not yet been able to determine uniform factors related to preventable

readmissions.10 To date, consensus definition of preventability has not been established.

Many studies use the opinion of physicians as the gold standard to determine if readmissions

are preventable, and derive factors that would predict preventable readmissions from these

findings.11-13 However, to our knowledge, no study has yet been performed to examine

the interobserver reliability of the physicians’ judgement on preventability.

Therefore, we performed an international study to assess if there is any consensus between

physicians regarding the predictability and preventability of medical readmissions.

MATERIAL AND METHODSThis study is an initiative of the safer@home consortium, an international group founded in

2013 consisting of 13 acute medical physicians, emergency physicians and epidemiologists

from Europe that focus on readmissions and safer discharge processes.

During the 3-month study period (1 September to 1 December 2015) a survey on eight

cases based on common clinical scenarios (see appendix 1) was distributed to physicians

throughout Europe.

Survey

The survey consisted of eight case-based medical scenarios (Table 1 shows summary of case

vignettes). The scenarios were generated using a Delphi-type methodology, whereby multiple

scenarios were generated and then represented to the safer@home consortium in two rounds.

In the first round underlying assumptions and information leading to different judgements was

explored using current readmission literature. This round took place in a face-to-face half-yearly

consortium meeting with all 13 members. After this seven clinically active medical physicians

in the group were asked to provide examples from their daily work in order to compose cases.

120

In the second round, these cases were discussed in a conference call during which the cases

that would be representable for all countries were selected through consensus. In addition we

assessed if, in our ‘expert’-opinion, the cases could potentially be used to fulfil the purpose

of our research question. Subsequently, a pilot was performed on a small group of physicians

from all countries to ensure cases were understandable and varied sufficiently. Final case

selection ensured that: a) the cases would be representative of patients requiring unselected

medical admission in northern Europe; b) the scenarios covered the range of factors suggested

by the literature to impact on readmissions; c) cases were not traceable to real-life patients.

The online survey consisted of two parts: 1) Physicians were asked about their opinion on

predictability of medical readmissions; from four cases describing an index admission, physicians

were asked to rate the chance of readmission within 30 days. 2) Physicians were asked to

assess the preventability of four described medical readmission cases. From the physicians’

assessment of predictability and preventability, the degree of consensus could be derived.

For both parts of the survey a five point Likert Scale was used as an answering model ((part

1: Definitely not predictable (1) – Definitely predictable (5); part 2: Definitely not preventable

(1) - Definitely preventable (5)).14

Data on the country and primary specialty of the responding physician filling out the survey

and the number of years of clinical experience were collected in order to explore agreement

within these subgroups. The survey was anonymized to ensure the researchers could not trace

which physician filled out which survey. Finally, general comments concerning readmission

could be made after completing the survey.

Distribution

The survey was distributed among physicians throughout Europe, they all worked solely in

a medical specialty and not in any surgical specialty. Invitations were sent to the members

of the Society for Acute Medicine the UK, the Dutch Acute Medicine Society, the Danish

Society for Emergency Medicine, physicians from Switzerland and Ireland using a common

web-based platform SurveyMonkey®. In order to calculate an accurate response rate, each

physician communicated the number of requests sent to one research member (LG), who was

responsible for data processing and statistical analysis. The ethics committee of VU University

Medical Centre, Amsterdam approved the study. No funding was received for this study.

Statistics

Descriptive characteristics and frequencies were calculated in SPSS version 22.0. Ratings of

physicians are presented as frequencies and percentages. Using the intraclass correlation

(ICC, a reliability coefficient) we assessed the agreement among physicians regarding

the predictability and subsequently, the preventability of the assessed medical readmissions.

This coefficient (ICC) is used to assess the agreement of ratings made by multiple observers

(in our study ‘physicians’) measuring the same outcome (in our study ‘the predictability and

121

preventability of readmissions both based on four-real-life readmission scenarios’). The ICC is

a ratio ranging in value between 0 (representing no agreement) and 1 (implying agreement).

Calculating the variance components we constructed the ICC formulas from which the ICC

could be calculated. For dependent variables we used the outcome ‘Likert scores’ and for

random factors ‘physicians’ and ‘case numbers (1.1, 1.2, 1.3, 1.4 and 2.1, 2.2, 2.3, 2.4)’ were

used. The variance among cases (case numbers1.1, 1.2, 1.3, 1.4 and 2.1, 2.2, 2.3, 2.4) were

analysed separately, among physicians, and the random error were calculated in SPSS using

the VARCOMP procedure. From the variance components we calculated the ICC for absolute

agreement as the variance among cases divided by the total variance of the cases, physicians

and random error.15

RESULTSPhysician characteristics

During the three-month study period (1 September to 1 December 2015) the survey was

distributed to physicians in Europe. In total 526 medical physicians filled out the survey.

The overall response rate was 24.2%. Seventy-seven (14.6%) physicians did not complete

all the questions in the survey. Table 2 shows physician characteristics. Dutch physicians were

the largest group of respondents (46.2%), followed by Danish (25.1%) and physicians from

the United Kingdom (23.6%). Most physicians had internal medicine (33.3%) as their primary

specialty followed by acute medicine (24.5%) and geriatrics (12.5%). The median years of

clinical experience was 10.75 (interquartile range: 5-20).

Agreement on predictability of readmission

For the first part of the survey physicians were asked if they could predict a readmission based

on the four case descriptions of medical index admissions. Responses are shown in figure 1.

The results show that there was substantial variation in the degree of predictability between

the physicians’ judgements in all four cases. The cases were assessed with different degrees

of predictability. To illustrate, for case number 1.2, about half of the physicians assessed

the likelihood of readmission as ‘definitely not’ (score 1), while in case number 2.1 over 60%

of the physicians predicted that the patient will definitely be readmitted (score 5). The ICC for

agreement of predictability was 0.67 (Var(Casenumber) 1,444, Var(Observer) 0.054, Var(error)

0.649) which indicates a moderate to strong interobserver agreement between the raters

(physicians). These findings suggest that the surveyed doctors had a moderate to good degree

of agreement about the patients that were prone to come back, they predicted the same

patients as having a higher chance of a readmission occurring.

Agreement on preventability of readmission

In the second part of the survey the respondents were asked to rate the preventability of

four medical readmission cases. The results in figure 2 show the distribution of answers by

the physicians. It shows that the physicians rated the cases differently, there was a wide variety

in assessment. In all four cases no clear majority seemed to rate the same readmissions with

122

Table 1 | Summary of case vignettes

A. Predictability: Please assess on a scale from 1-5 if you think the following four admissions are followed by a readmission within 30 days: Definitely not (1) - Definitely (5).

Demographics Presenting complaint Diagnosis Investigations Management

Other information

Case 1.1

83 year old female

Collapse Atrial Fibrillation, Hypertension,

Urinary tract infection

Raised inflammatory markers, Positive urine culture

Antibiotics, Aspirin

Cardioverts with sepsis treatment

Case 1.2

20 year old female

Headache Migraine CT brain: normal, Lumbar-puncture: normal

Intra-venous fluids, Paracetamol,

Non-steroidal anti-inflammatory

Case 1.3

60 year old female

Dyspnoea

Productive cough

Infective exacerbation of COPD

Sputum culture negative

Oxygen, Bronchodilators, Steroids, Antibiotics

Has home nebulizers

Case 1.4

94 year old female

Dyspnoea Pneumonia Persistently raised inflammatory markers two days before discharge

Antibiotics Chest pain, dyspnoea, vomiting prior to discharge

B. Preventability: Please assess on a scale from 1 to 5 if you think the following four readmissions within 30 days are: Definitely not preventable (1) - Definitely preventable (5).

DemographicsPresenting complaint Diagnosis Investigation Management

Readmission diagnosis

Case 2.1

63 year old lady

Fever Gemcitabine induced fever

None of note Supportive treatment

Neutropenic sepsis 10 days later

Case 2.2

40 year old male

Ascites Childs B cirrhosis

Alcohol Dependency

None of note Abdominal paracentesis,

Diuretics, vitamins, lactulose,

Alcohol support declined

Upper Gastro-intestinal bleed 3 weeks later

Case 2.3

55 year old male

Chest pain Anterior ST-elevation Myocardial infarction

Angiogram, Echo-cardiogram with moderate left ventricular dysfunction

Angioplasty of the left anterior descending artery, Secondary prevention

Pulmonary oedema 3 weeks later

Case 2.4

32 year old female

Loin pain Pyelonephritis, Hydronephrosis due to ureter stenosis

Ultra-sound abdomen

Urine culture

Intra-venous antibiotics as outpatient

Pyelonephritis one month later

123

similar scores. These findings were also reflected in the ICC for this part of the survey. The ICC

was calculated at 0.13 (Var(Casenumber) 0,194, Var(Observer) 0.168, Var(error) 1.076), which

implies poor agreement. Doctors do not seem to agree on the preventability of readmissions.

However, one must note that the variance among case numbers was relatively low which

may indicate that the cases assessed were not sufficiently distinct enough to obtain a high

reliability coefficient.

Subgroup analysis

To assess if there was any difference in agreement between subgroups of physicians

we subdivided the doctors into years of clinical experience. They were grouped based on

clinical experience up to 5 years (n=151, 28,7%), from 5-15 years (n=208, 39,5%), and 15

years and higher (n=167, 31,7%). Results suggest that medical physicians with less clinical

experience had a trend towards greater agreement than those with more clinical experience

as to the likelihood of readmission but these differences are minimal (ICC 0.70, 0.69, 0.63,

respectively). Physicians with more clinical experience seemed to have more agreement

about the preventability of a readmission compared with those with less clinical experience

(ICC 0.08, 0.01, 0.19, respectively).

Table 2 | Physician characteristics

Country Percentage 100% (n=526) Primary specialty

Percentage 100% (n=526)

The Netherlands 46,2 (243) Internal Medicine 33,3 (175)

Denmark 25,1 (132) Acute/Emergency Medicine 24,5 (129)

United Kingdom 23,6 (124) Geriatrics 12,5 (66)

Switzerland 1,9 (10) Other 5,3 (28)

Other 3,2 (17) Nephrology 4,2 (22)

Intensive Care 3,6 (19)

Endocrinology 3,2 (17)

Gastroenterology 2,9 (15)

Pulmonary medicine 2,3 (12)

Haematology 1,7 (9)

Medical Oncology 1,5 (8)

Rheumatology 1,3 (7)

Hepatology 0,8 (4)

Cardiology 0,4 (2)

124

Figure 1 | Part I: The predictability of readmissionSpread of Likert-Scores (Definitely not predictable (1) – Definitely predictable (5)) given per case (1,2,3,4) by 526 physicians (in percentage)

Figure 2 | Part II: The preventability of readmissionSpread of Likert-Scores (Definitely not preventable (1) – Definitely preventable (5)) given per case (5,6,7,8) by 526 physicians (in percentage)

125

DISCUSSIONIn this survey among 526 European physicians, there was moderate agreement as to

the predictability of medical readmissions but poor agreement about their preventability. These

results suggest that doctors agree on the patients who have a higher risk of being readmitted,

but the physicians differ on how preventable these readmissions are. To our knowledge,

the current study is the first to investigate the interobserver reliability on the evaluation of

unplanned readmissions in such a large group of observers.

Unplanned readmissions are a complex phenomenon, which are influenced not only by medical

factors but also by a range of social and political issues.5,16,17 Readmission risk is difficult to

define and is less predictable than mortality.18 Nevertheless, there are a number of risk factors

which are recognized as increasing the risk of readmission and multiple predictive scoring

systems based on these factors have been designed.5,6,19

Although the risk factors for predominantly medical readmission are increasingly well

recognized, the dynamic of how they interact and whether they can be influenced remains

controversial.20 The poor consensus among physicians found in our study as to whether

the readmissions were preventable underlines this issue. A US study of 17 hospitalists reviewing

300 consecutive readmissions also found wide variation in their scoring of preventability,

however comparability might be limited since these were real-life readmissions.21 We can

concur their findings of interobserver variability in a (European) setting.

The above findings illustrate the problem faced by policy makers trying to integrate preventability

in the readmission indicator since doctors, who are supposed to be experts in the field, cannot

even agree on the readmissions that are potentially preventable. Current literature, however,

often uses the opinion of one or more physicians as the gold standard to get insight into

preventability and draw conclusions on factors predicting preventability. The results in this

study, however, demonstrate that the assumptions derived from these studies might lead to

misperception since physicians do not share similar ideas on the potential preventability of

readmissions.22-24 Hence, it can be questioned whether conclusions drawn from these studies

might not provide reliable conclusions to create an appropriate quality indicator.

Readmitted medical patients are a heterogeneous group; there is a wide variation in the age,

comorbidities and social support of these patients. It remains unclear as to whether the factors

which drive unplanned readmission, including medical, social, cultural and environmental, are

modifiable.20 This is reflected by an increasing body of evidence that suggests readmissions do

not always reflect poor care and preventability of these readmissions is poorly defined.3,4,21,25

More research studying ‘the preventability’ in a structured manner might help to improve

the difficult task in creating a reliable indicator.

We used adapted real-life case scenarios in our study, which may be a limitation. This was also

reflected in the comments section, where physicians mentioned they were missing information

126

that would allow them to thoroughly assess the case, for example more details on the patients’

social situation. It would however be difficult to incorporate all the potentially relevant social

and environmental factors into scenarios particularly in a pan-European study where there

is a wide variety of political and health policies that influence readmissions. Furthermore,

in calculating the ICC for the preventability part of the survey one could suggest that there

was little variation in the preventability of the cases. This may reflect either that there was

insufficient variation with regards to preventability within the scenarios, potentially caused

by balancing between uniformity in the cases in a way they could be representable for all

countries participating in the study and enough variation in the cases in order to create

different opinions per case. It may also reflect an uncertainty among physicians regarding

what comprises a preventable admission.

On a final note, our respondents were of high seniority with a median of 11 years of clinical

experience. If clinicians with this level of experience cannot agree on the predictability of

readmission, is it wise to use it as a marker of quality of care?

CONCLUSIONThis study demonstrates that there is moderate agreement among experienced medical

physicians about the predictability of readmissions but poor agreement about their

preventability. Therefore, the conclusions derived from earlier studies on preventability, on

the basis of physician consensus as the gold standard, are questionable. Hence, a good way of

defining and integrating preventability into this quality indicator remains elusive.

ACKNOWLEDGEMENTSA full list of membership of the safer@home Consortium is as follows: M. Braband, T. Cooksley,

L. van Galen, H. Haak, R. Kidney, J. Kellet, H. Merten, P. Nanayakkara, C. Nickel, J. Soong,

C. Subbe, L. Vaughan, I. Weichert.

CONFLICT OF INTEREST STATEMENTThe authors have nothing to disclose.

127

REFERENCES1. Fischer C, Lingsma HF, Marang-van de Mheen PJ, Kringos DS, Klazinga NS, Steyerberg EW. Is

the readmission rate a valid quality indicator? A review of the evidence. PLoS One. 2014;9(11):e112282.

2. Inspectie van de volksgezondheid: Basisset Kwailiteitsindicatoren 2016. Utrecht. http://www.igz.nl/Images/IGZ%20Basisset%20kwaliteitsindicatoren%20ziekenhuizen%202016_tcm294-367407.pdf. Laatste update september 2015. Accessed July, 2016.

3. van Galen LS, Nanayakkara PW. [Hospital readmissions: A reliable quality indicator?]. Ned Tijdschr Geneeskd. 2015;160:A9885.

4. van Walraven C. The Utility of Unplanned Early Hospital Readmissions as a Health Care Quality Indicator. JAMA Intern Med. 2015;175(11):1812-1814.

5. Donze J, Aujesky D, Williams D, Schnipper JL. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA Intern Med. 2013;173(8):632-638.

6. Cooksley T, Merten H, Kellett J, et al. PRISMA Analysis of 30 Day Readmissions to a Tertiary Cancer Hospital. Acute Med. 2015;14(2):53-56.

7. Rico F, Liu Y, Martinez DA, Huang S, Zayas-Castro JL, Fabri PJ. Preventable Readmission Risk Factors for Patients With Chronic Conditions. J Healthc Qual. 2015.

8. Vest JR, Gamm LD, Oxford BA, Gonzalez MI, Slawson KM. Determinants of preventable readmissions in the United States: a systematic review. Implement Sci. 2010;5:88.

9. Anderson MA, Helms LB, Hanson KS, DeVilder NW. Unplanned hospital readmissions: a home care perspective. Nurs Res. 1999;48(6):299-307.

10. Jackson AH, Fireman E, Feigenbaum P, Neuwirth E, Kipnis P, Bellows J. Manual and automated methods for identifying potentially preventable readmissions: a comparison in a large healthcare system. BMC Med Inform Decis Mak. 2014;14:28.

11. van Walraven C, Jennings A, Forster AJ. A meta-analysis of hospital 30-day avoidable readmission rates. J Eval Clin Pract. 2012;18(6):1211-1218.

12. Cakir B, Gammon G. Evaluating readmission rates: how can we improve? South Med J. 2010;103(11):1079-1083.

13. Meisenberg BR, Hahn E, Binner M, et al. ReCAP: Insights Into the Potential Preventability of Oncology Readmissions. J Oncol Pract. 2016;12(2):153-154.

14. Elaine IE, Seaman CA. Likert scales and data analyses. Quality Progress 40.7 2007: 64-65.

15. de Vet HC, Terwee CB, Knol DL, Bouter LM. When to use agreement versus reliability measures. J Clin Epidemiol. 2006;59(10):1033-1039.

16. Billings J, Blunt I, Steventon A, Georghiou T, Lewis G, Bardsley M. Development of a predictive model to identify inpatients at risk of re-admission within 30 days of discharge (PARR-30). BMJ Open. 2012;2(4).

17. Fluitman KS, van Galen LS, Merten H, et al. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Intern Med. 2016.

18. Kansagara D, Englander H, Salanitro A, et al. Risk prediction models for hospital readmission: a systematic review. Jama. 2011;306(15):1688-1698.

19. van Walraven C, Bennett C, Jennings A, Austin PC, Forster AJ. Proportion of hospital readmissions deemed avoidable: a systematic review. CMAJ. 2011;183(7):E391-402.

20. Brown EG, Burgess D, Li CS, Canter RJ, Bold RJ. Hospital readmissions: necessary evil or preventable target for quality improvement. Ann Surg. 2014;260(4):583-589; discussion 589-591.

21. Koekkoek D, Bayley KB, Brown A, Rustvold DL. Hospitalists assess the causes of early hospital readmissions. J Hosp Med. 2011;6(7):383-388.

22. Bianco A, Mole A, Nobile CG, Di Giuseppe G, Pileggi C, Angelillo IF. Hospital readmission prevalence and analysis of those potentially avoidable in southern Italy. PLoS One. 2012;7(11):e48263.

128

23. Shimizu E, Glaspy K, Witt MD, et al. Readmissions at a public safety net hospital. PLoS One. 2014;9(3):e91244.

24. Donze J, Lipsitz S, Bates DW, Schnipper JL. Causes and patterns of readmissions in patients with common comorbidities: retrospective cohort study. Bmj. 2013;347:f7171.

25. Cooksley T, Nanayakkara PW, Nickel CH, et al. Readmissions of medical patients: an external validation of two existing prediction scores. QJM. 2016;109(4):245-248.

CHAPTER 9

PATIENTS’ AND PROVIDERS’ PERCEPTIONS OF THE PREVENTABILITY OF HOSPITAL READMISSION:

A PROSPECTIVE, OBSERVATIONAL STUDY IN 4 EUROPEAN COUNTRIES

“CAPTURING READMISSIONS INTERNATIONALLY TO PREVENT READMISSION BY THE SAFER@HOME

GROUP (CURIOS@)”

Louise S. van Galen | Mikkel Brabrand | Tim Cooksley | Peter M. van de Ven | Hanneke Merten | Ralph K. So | Loes van Hooff | Harm Haak | Rachel M. Kidney |

Christian H. Nickel | John Soong | Immo Weichert | Mark H. Kramer | Christian P. Subbe | Prabath W.B. Nanayakkara |

*On behalf of the safer@home consortium

BMJ Qual Saf. 2017; In press

‘Patient safety is a dissatisfier, safe care is normal, non-safe care makes unsatisfied’ Jan Klein

132

ABSTRACT Background

Unscheduled readmissions are increasingly being used as a quality indicator. In comparison to

extensive US data, no large-scale European data are available.

Objectives

To investigate opinions of readmitted patients, their carers, nurses and physicians on

predictability and preventability and using majority opinion to determine contributing factors

that could potentially foresee (preventable) readmissions.

Methods

This prospective observational study included medical patients readmitted unscheduled within

30 days in 15 European hospitals. Readmitted patients, carers and treating professionals

were surveyed to assess the discharge process and predictability and preventability of

the readmission. Variables associated with risk of readmission were also collected. Cohen’s

Kappa measured pairwise agreement of considering readmission as predictable/preventable

by patients, carers and professionals. Subsequently, multivariable logistic regression identified

risk factors associated with predictability and preventability.

Results

In this cohort (1398 readmissions), the majority deemed 27.8% readmissions potentially

predictable and 14.4% potentially preventable. Consensus on predictability and preventability

was poor, especially between patients and professionals (kappa’s ranged from 0.105-

0.173). The interviewed selected different factors potentially associated with predictability

and preventability. When a patient reported that he was ready for discharge during index

admission, readmission was deemed less likely by majority (predictability OR 0.55;CI 0.40-

0.75, preventability OR 0.35;CI 0.24-0.49).

Conclusions

There is no consensus between readmitted patients, their carers, treating nurses and physicians

about predictability and preventability of readmissions, nor associated risk factors. The patient

reporting not feeling ready for discharge was strongly associated with preventability and

predictability. Therefore, healthcare workers should question patient’s readiness to go home

timely before discharge.

Keywords

Readmission, Healthcare quality indicator, Patient involvement, Patient

discharge, Communication

133

INTRODUCTIONUnplanned readmissions within 30 days are perceived as a major adverse event after hospital

discharge. Current readmission rates vary from 10-30% internationally and the related costs

are increasing.1 Internationally, readmission rates are often used to audit and reimburse units

and used as a quality of care monitoring tool.2,3 In the Netherlands, readmission rates were

introduced as an official quality indicator in 2016.4

Due to an aging population and healthcare policy changes, Emergency Department (ED)

admissions have risen in the last few years.5,6 As a result, hospital beds are becoming scarce

and physicians are under constant pressure to discharge patients rapidly.

The problem with the use of readmissions as a quality indicator is that not all readmissions

are preventable and causes for readmission might find their origin in natural progression or

unavoidable recurrence of underlying diseases.7,8 In addition, despite many efforts to reduce

readmissions such as the Hospital Readmission Reduction Plan and telephonic follow-up

interventions, it is still questionable whether these interventions truly result in lower

readmission rates.5-8 Moreover, since preventability has not been defined uniformly, it remains

uncertain whether this quality indicator, in its current form, is reliable.9,10 Several studies have

used the opinion of physicians as the gold standard to determine whether readmissions are

preventable, and have defined factors that would predict preventability from these findings.11-13

However, recent work has shown that even among doctors there is poor consensus about

the preventability of a readmission.14

Although many attempts have been made to create readmission prediction models, there

are still no widely utilized and internationally validated tools predicting the chance of

readmission.15-17 Most risk models were developed in the USA and Canada, and due to

differences in healthcare systems and case-mix they probably may not be suitable to be used

in European populations.18-20 In addition, they do not measure preventability. A recent analysis

demonstrated that causes of potentially preventable readmissions are mostly human-related

coordination and communication failures.21 A few studies have been performed investigating

opinions of patients and health-care workers on the preventability of readmissions and

discharge planning.22,23 Yet, most of these studies had small sample sizes, and questioned

patients retrospectively after discharge.

Therefore, we performed the first prospective observational study of 1398 unscheduled

medical readmissions to 15 centres in 4 European countries aiming to investigate:

(1) the opinions of readmitted patients, their carers, nurses and physicians on the predictability

and preventability of the readmissions;

134

(2) the contributing factors that could potentially predict (preventable) readmissions using

the majority opinion as the gold standard.

METHODSSites and Participants

This research project ‘CURIOS@’ (CaptUring Readmissions InternatiOnally to prevent

readmissions by safer@home consortium) took place in academic and non-academic centres

using the network of the safer@home (‘Scoring Acute admissions For Estimating Readmission’)

consortium. The study was an initiative of the collaboration founded in 2014 consisting of 12

acute medical physicians, emergency physicians and epidemiologists from Europe which focus

on readmissions and safer discharge processes. Fifteen centres participated (nine centres in

the Netherlands, three in the United Kingdom, one in Ireland, and two in Denmark). The data

collected was derived from two clinical episodes: readmission (RA) and index admission (IA)

between 1 January 2016 and 1 November 2016. Data collection took place on the unit of

readmission at any time within 72 hours after the readmission. Ethics Committees of all

participating sites individually approved the study, after primary approval was obtained by

the coordinating centre (VUmc, Amsterdam).

Eligible patients were those who had unscheduled admissions through ED, AMU or other

clinical areas, were aged above 18 years and readmitted to hospital following a previous

inpatient episode to any clinical specialty for a minimum of one night in the previous 30 days.

If a patient was readmitted more than once, only first readmission was included. In order to

be included, readmissions had to be to a medical (cardiology, geriatrics, gastroenterology,

haematology, internal medicine, nephrology, neurology, oncology, pulmonary medicine,

rheumatology) ward. Excluded were patients: readmitted electively for procedures, surgery or

chemotherapy; with an IA for psychiatry or gynaecology; who stayed shorter than one night

during IA or RA; transferred to the ward from an initial admission to the ICU; admitted to

another institution in their IA. All patients gave written informed consent.

Data Collection

After instructions and training from the coordinating investigator (LG), centres were allowed to

participate, aiming to include at least 50 patients consecutively. Investigating site researchers

were deemed competent once they followed adequate training and enough time for them

was allocated to perform interviews, data collection and root cause analysis. They were all

medically trained but not involved in patientcare during inclusion period. After obtaining

written informed consent, site researchers surveyed readmitted patients: the questionnaire

consisted of seven questions about their readiness for discharge during IA and predictability

and preventability of their readmission (Supplement 1). This questionnaire was constructed

using available literature and after reaching consensus among safer@home group members.

In addition, we ensured it consisted of questions that could easily be asked by physicians

and answered by patients in daily practice, and are reproducible to be used in other settings.

135

The questionnaires were tested in a pilot group before agreeing on a final amended version.

Subsequently, after obtaining patient’s permission, a carer (defined as a person providing

unpaid intensive and long-term care because of a personal relationship) if available, was

approached in person or by telephone, to answer two questions about predictability and

preventability of the readmission. Lastly, a doctor and a nurse responsible for the patient during

readmission were interviewed. To reduce bias, all interviews were performed separately by site

researcher. This researcher also answered the same two questions after assessing answers

given by the interviewees. In addition, data-items pre-identified through literature as being

potentially predictive of a (preventable) readmission were collected.15,16 These variables (i.e.

Charlson Comorbidity Index24 and Clinical Frailty Scale25) were extracted during readmission,

directly from patient, using clinical notes of IA and discharge communication. The dataset

contained no patient identifiable variables.

PRISMA-analysis

To assess predictability and preventability, we asked all interviewed about the reasons for

readmission. Subsequently, site researcher qualified these into one or more root causes

categorized in disease-, patient-, healthcare worker-, organizational- or other causes, originally

identified by PRISMA-analysis supplying us with more information about probable root causes

for the readmission. PRISMA-analysis has previously shown to provide objective and structured

insight into causes for adverse events, by composing root causal trees for adverse events.21,26

This method has been accepted by the World Alliance for Patient Safety (see Supplement 2 for

more detailed description).27

Measurement of Predictability and Preventability

Since a gold standard defining predictability and preventability is not available, after reaching

consensus in our half-yearly consortium meetings a new variable was composed. It was

decided that if a majority (50 percent or more) of interviewed groups (patients, carers, doctors,

nurses, researchers), assessed the readmission as predictable or preventable (options yes, no,

unknown), the readmission was decided as predictable or preventable. For example, if a carer

was not available, if 2 out of 4 remaining interviewed assessed readmission as preventable, it

was deemed preventable. If all 5 interviewees answered the questions, 3 ‘yes’ answers were

needed for this conclusion. In a separate analysis we used the answers given by the members

of the 5 interviewed groups separately and regarded option ‘unknown’ as missing.

Statistical Analysis

Statistical analysis was performed in SPSS version 22.0. Categorical variables are summarized

as frequencies and percentages. Continuous variables are summarized by mean and standard

deviation (SD) in case of a normal distribution or median and ranges otherwise.

Cohen’s kappa (κ) was used to measure agreement of predictability and preventability

assessments (yes, no, unknown) separately for each pair of five interviewed groups. McNemar

136

Bowker test was performed to assess whether different pairs of interviewed groups varied in

proportions of ‘yes’, ‘no’ and ‘unknown’ answers.

Logistic regression analyses were used to find variables that were associated with assessments

of preventability and predictability of readmissions. For all categorical variables where

‘unknown’ or ‘don’t know’ was an option, these were considered as “missing”. However,

these constituted less than 10% of the total answers. For predictability and preventability

assessed by the individual groups, the option ‘don’t know’ was considered as missing in

multivariable models, regardless of its percentage. (Table 1 and 2). Separate models were built

for each interviewed group and for majority judgement and separate for preventability and

predictability. Only variables with a two-sided P ≤ .10 in univariable logistic regression analyses

were included in a multivariable logistic regression analyses where backward elimination

was used to find a minimum set of variables that were independently associated with

predictability and preventability. To account for differences between countries, country was

included as a predictor in all models. The magnitude of the association between predictors and

outcome was quantified using Odds Ratios (OR) together with their 95% confidence interval.

The discriminative ability was quantified by means of the Area under the Receiver Operator

Curve (ROC) curve.

RESULTSPatient and hospital and readmission characteristics

During the study period 1961 patients were eligible of which 1398 patients participated,

resulting in an inclusion rate of 71.3%. The reasons for exclusion were: patient was already

sent home on day of intended inclusion, 41.0% (231 of 563); unwilling to participate, 20.4%

(115 of 563); being too ill, 19.2% (108 of 563); language barrier, 8.6%, (48 of 563); patient

deceased at RA, 2.1% (12 of 563); and other reasons i.e. being in quarantine, 8.7% (49 of

563). The median age was 70 (range 18-96), the median number of included patients per

hospital was 71 (range 48-226), other characteristics are given in Table 1.

The majority of interviewed deemed 27.8% (390 of 1398) of the readmissions as a potentially

predictable, and 14.4% (202 of 1398) potentially preventable. The assessment per interviewed

is listed in Table 2.

Consensus on readmission (κ)

Table 3 shows Cohen’s kappa for the two questions put forward to all individuals

interviewed: 1. ‘Do you feel the current readmission was expected?’, and 2. ‘Do you feel

the current readmission was preventable?’ (yes, no, don’t know). For predictability, none of

the kappa’s were satisfactory, they were all below κ = 0.7. The poorest consensus was found

between patient and physician, and patient and nurse, with a rate of κ = 0.173 and κ =

0.153, respectively. The highest kappa was found for physician and researcher (κ = 0.607).

The consensus on preventability of readmission was also unsatisfactory, the highest score

137

in this analysis was κ = 0.473, measuring agreement between researcher and physician.

The poorest scores in this group, κ = 0.105 and κ = 0.135, were found comparing patient to

physician, and nurse, respectively. McNemar Bowker test was P < .05 for all, indicating that

proportions of respondents answering ‘yes’, ‘no’ and ‘don’t know’ differed between all pairs of

interviewed groups.

Contributing factors to assessing the predictability and preventability of readmission

Using multivariate models, factors potentially contributing to predictability and preventability

of readmission were identified according to the opinion of the majority. Subsequently, models

were composed per interviewed group. The variables used in the models (Table 1) associated

with assessing readmission as potentially predictable and preventable, were different for every

group interviewed (Table 4 and 5). Using the opinion of the majority as the gold standard

(Table 4), factors significantly associated with a higher predictability of readmission included:

Having a non-elective IA (OR 2.55; 95% CI 1.59-4.08), having more than five admissions in

the year before IA (OR 2.69; 95% CI 1.72-4.20), lower age (OR 0.98; 95% CI 0.97-0.99),

higher CFS (OR 1.29; 95% CI 1.18-1.42), and a higher CCI (OR 1.08; 95% CI 1.02-1.15). When

a patient reported having felt ready at discharge during IA, the readmission was deemed less

likely (OR 0.55; 95% CI 0.40-0.75). Having a follow-up planned (OR 0.52; 95% CI 0.35-0.78),

and feeling ready for discharge (OR 0.35; 95% CI 0.24-0.49) were significantly associated with

a readmissions being deemed less preventable. As is illustrated in table 5, with exception of

the physician, all interviewed considered readmissions more predictable and preventable when

a patient reported not feeling ready for discharge. Table 4 and 5 demonstrate the discriminative

ability of the twelve models by means of the AUC varied from moderate to good (0.65–0.74).

Root causes

All interviewed were asked to qualify reasons for readmission into one or more of five available

root causes. Each readmission could have more than one root cause per interviewed and

the maximum causes per patient was 25. In total 6895 root causes were identified, with

a mean of 5.3 per patient (SD 1.9). Most root causes were disease-related 67.9% (4686 of

6895), followed by healthcare worker-related 17.5% (1208 of 6895), patient-related causes

10.9% (749 of 6895), organizational 2.2% (153 of 6895) and non-classifiable 1.4% (99 of

6895). In univariate analysis, when the composed variable for majority opinion was used,

disease-related root causes were negatively associated with a readmissions being considered

predictable [83.7% vs. 71.8%] (OR 0.5; 95% CI 0.37-0.66) and preventable [87.6% vs.

38.0%] (OR 0.09; 95% CI 0.06-0.12), healthcare worker-related root causes were positively

associated with readmissions being more predictable [11.9% vs. 20.6%] (OR 1.92; 95%

CI 1.39-2.65) and preventable [5.6% vs. 66.3%] (OR 33.43; 95% CI 22.48-49.727).

138

Tab

le 1

| ‘P

atie

nt c

hara

cter

istic

s’ab

Ch

arac

teri

stic

Tota

l (n

=13

98)

Pred

icta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Pred

icta

ble

vs

Pred

icta

ble

Prev

enta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Prev

enta

ble

vs

Prev

enta

ble

No

n-P

red

icta

ble

(n=

1008

)Pr

edic

tab

le

(n=

390)

No

n-P

reve

nta

ble

(n=

1196

)Pr

even

tab

le

(n=

202)

Cou

ntry

, N

o. (

%)

Th

e N

ethe

rlan

ds

911

(65.

2)67

9 (6

7.4)

232

(59.

5)P

< 0

.001

785

(65.

6)12

6 (6

2.4)

P <

0.0

01

D

enm

ark

262

(18.

7)19

1 (1

8.9)

71 (

18.2

)24

2 (2

0.2)

20 (

9.9)

U

nite

d K

ingd

om a

nd Ir

elan

d22

5 (1

6.1)

225

(16.

1)87

(22

.3)

169

(14.

1)56

(27

.7)

Pati

ent

age,

med

ian

(ran

ge),

y

70 (

18-9

6)70

(18

-96)

69 (

18-9

6)P

= 0

.03

70 (

18-9

6)70

.5 (

18-9

6)P

= 0

.62

Mal

e se

x, N

o. (

%)

737

(52.

7)53

8 (5

3.4)

199

(51.

0)P

= 0

.43

627

(52.

4)11

0 (5

4.5)

P =

0.5

9

Part

nere

d/M

arri

ed,

No.

(%

)82

8 (5

9.2)

61.3

(61

.1)

53.

8 (4

8.0)

P =

0.0

1173

1 (6

1.1)

97 (

48.0

)P

< 0

.001

IA t

ypec ,

Non

-ele

ctiv

e, N

o. (

%)

1195

(85

.2)

830

(82.

3)36

1 (9

2.6)

P <

0.0

0183

0 (8

2.3)

361

(92.

6)P

= 0

.21

LOS

IAc ,

med

ian

(ran

ge),

d4

(0-1

91)

4 (0

-136

)4

(0-1

91)

P <

0.0

014

(1-1

91)

4 (1

-136

)P

= 0

.94

Leng

th b

etw

een

IA a

nd R

Ac ,

med

ian

(ran

ge),

d9

(0-3

0)9

(0-3

0)10

(0-

30)

P =

0.0

329

(0-3

0)7

(0-3

0)P

= 0

.007

Clin

ical

Fra

ilty

Scal

e, m

ean

(SD

)25

4.4

(1.6

)4.

2 (1

.6)

4.9

(1.7

)P

< 0

.001

4.4

(1.6

)4.

5 (1

.6)

P =

0.3

3

Cha

rlso

n C

omor

bidi

ty In

dex,

m

edia

n (r

ange

)24

2 (0

-11)

4 (0

-136

)4

(0-1

91)

P <

0.0

012

(0-1

1)2.

5 (0

-11)

P =

0.6

8

HO

SPIT

AL

Scor

e16,d,

mea

n (S

D)

4.4

(2.1

)4.

2 (2

.0)

5.0

(2.2

)P

< 0

.001

4.4

(2.1

)4.

6 (2

.2)

P =

0.0

8

Num

ber

of m

edic

ine

at

disc

harg

e (c

hron

ic a

nd

non-

chro

nic

use)

, N

o. (

%)

0-

425

7 (1

8.4)

201

(20.

0)56

(14

.4)

P =

0.0

0122

2 (1

8.6)

35 (

17.3

)P

= 0

.062

5-

8 39

5 (2

8.3)

299

(29.

7)96

(24

.7)

342

(28.

7)53

(26

.2)

>

874

3 (5

3.3)

506

(50.

3)23

7 (6

0.9)

629

(52.

7)11

4 (5

6.4)

DN

R po

licy

at d

isch

arge

, D

o re

susc

itat

e, N

o. (

%)

961

(73.

1)71

2 (7

4.9)

249

(68.

8)P

= 0

.022

821

(73.

1)14

0 (7

3.3)

P =

0.9

6

ICU

/CC

U/H

DU

adm

issi

on d

urin

g IA

, Ye

s, N

o (%

)11

5 (8

.2)

94 (

9.4)

21 (

5.4)

P =

0.0

1510

2 (8

.5)

13 (

6.5)

P =

0.3

2

139

Tab

le 1

| (c

ontin

ued)

Ch

arac

teri

stic

Tota

l (n

=13

98)

Pred

icta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Pred

icta

ble

vs

Pred

icta

ble

Prev

enta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Prev

enta

ble

vs

Prev

enta

ble

No

n-P

red

icta

ble

(n=

1008

)Pr

edic

tab

le

(n=

390)

No

n-P

reve

nta

ble

(n=

1196

)Pr

even

tab

le

(n=

202)

Cog

niti

ve im

pair

men

t (o

ffici

al

diag

nosi

s)e ,

Yes

, N

o. (

%)

41 (

3.0)

23 (

2.3)

18 (

4.7)

P =

0.0

230

(2.

5)11

(5.

6)P

= 0

.02

Falle

n at

leas

t on

ce in

last

hal

f ye

ar,

Yes,

No.

(%

)38

0 (2

8.9)

279

(29.

6)10

1 (2

7.3)

P =

0.4

230

9 (2

7.6)

71 (

36.2

)P

= 0

.014

Tota

l adm

issi

ons

year

pri

or t

o IA

, el

ecti

ve a

nd n

on-e

lect

ive,

N

o. (

%)

0

507

(36.

3)40

4 (4

0.1)

103

(26.

4)P

< 0

.001

440

(36.

8)67

(33

.2)

P =

0.6

1

1-

574

5 (5

3.3)

534

(53.

0)21

1 (5

4.1)

632

(52.

8)11

3 (5

5.9)

>

514

6 (1

0.4)

70 (

6.9)

76 (

19.5

) 12

4 (1

0.4)

22 (

10.9

)

Dis

char

ge le

tter

issu

ed a

t ti

me

RA,

Yes,

No.

(%

)10

45 (

75.8

)75

9 (7

6.5)

286

(74.

1)

P =

0.3

576

.9 (

905)

140

(69.

7)P

= 0

.027

Follo

w-u

p pl

anne

d at

dis

char

ge

IA,

Yes,

No.

(%

)11

34 (

84.1

)82

4 (8

5.0)

310

(81.

6)P

= 0

.12

992

(86.

1)14

2 (7

2.1)

P <

0.0

01

Has

pat

ient

bee

n se

en b

y a

doct

or b

etw

een

IA a

nd R

A?,

Ye

s, N

o. (

%)

867

(66.

8)61

5 (6

6.1)

252

(68.

9)P

= 0

.34

735

(66.

6)13

2 (6

8.4)

P =

0.6

2

Was

a s

umm

ary

prov

ided

at

disc

harg

e? Y

es,

No.

(%

)99

4 (7

2.6)

720

(78.

5)27

4 (7

9.0)

P =

0.8

686

0 (

80.2

)13

4 (6

9.8)

P =

0.0

01

Did

you

(th

e pa

tien

t) s

ugge

st

doct

ors

to s

tay

long

er a

t IA

? Ye

s, N

o. (

%)

131

(9.4

)74

(7.

4)57

(14

.9)

P <

0.0

0110

3 (8

.7)

28 (

13.9

)P

< 0

.001

140

Tab

le 1

| (c

ontin

ued)

Ch

arac

teri

stic

Tota

l (n

=13

98)

Pred

icta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Pred

icta

ble

vs

Pred

icta

ble

Prev

enta

ble

maj

ori

ty

Tota

l (n

=13

98)

P V

alu

e fo

r m

ajo

rity

no

n-

Prev

enta

ble

vs

Prev

enta

ble

No

n-P

red

icta

ble

(n=

1008

)Pr

edic

tab

le

(n=

390)

No

n-P

reve

nta

ble

(n=

1196

)Pr

even

tab

le

(n=

202)

Did

you

(th

e pa

tien

t) f

eel b

ette

r at

dis

char

ge (

IA)?

Yes

, N

o. (

%)

950

(69.

3)70

7 (7

1.9)

243

(63.

3)P

= 0

.002

839

(71.

9)11

1 (5

5.5)

P <

0.0

01

Did

you

(th

e pa

tien

t) f

eel r

eady

at

dis

char

ge (

IA)?

Yes

, N

o. (

%)

1014

(75

.1)

772

(78.

9)24

2 (6

3.9)

P <

0.0

0190

0 (7

7.7)

114

(57.

3)

P =

0.0

02

a Mis

sing

dat

a in

clud

e th

e fo

llow

ing:

did

you

fee

l bet

ter

for

disc

harg

e (n

=3)

, did

you

fee

l rea

dy f

or d

isch

arge

(n=

1), w

as a

sum

mar

y pr

ovid

ed a

t di

scha

rge?

( n=

12).

b Va

riab

les

whe

re ‘

unkn

own’

or

‘don

’t k

now

’ w

as a

nsw

ered

, th

ese

wer

e re

gard

ed a

s m

issi

ng s

ince

the

y w

ere

belo

w 1

0% (

see

met

hods

): N

umbe

r of

med

icin

e at

di

scha

rge

(n=

3), D

NR

polic

y at

dis

char

ge (n

=84

), IC

U/C

CU

/HD

U a

dmis

sion

dur

ing

IA (n

=4)

, Cog

niti

ve im

pair

men

t (o

ffici

al d

iagn

osis

) (n=

16),

Fal

len

at le

ast

once

in

last

hal

f ye

ar (n

=84

), D

isch

arge

lett

er is

sued

at

tim

e RA

(n=

20),

Fol

low

-up

plan

ned

at d

isch

arge

IA (n

=49

), H

as p

atie

nt b

een

seen

by

a do

ctor

bet

wee

n IA

and

RA

(n

=10

1), W

as a

sum

mar

y pr

ovid

ed a

t di

scha

rge

(n=

122)

wer

e an

swer

ed w

ith

‘unk

now

n’ b

y th

e ex

ecut

ive

rese

arch

er. D

id y

ou s

ugge

st d

octo

rs t

o st

ay lo

nger

at

IA?

(n=

9),

Did

you

fee

l bet

ter

at d

isch

arge

? (n

=28

), D

id y

ou f

eel r

eady

at

disc

harg

e? (

n=40

) w

as a

nsw

ered

wit

h ‘d

on’t

kno

w’

by t

he p

atie

nt s

urve

yed.

c

IA:

Inde

x A

dmis

sion

, RA

: Re

adm

issi

on,

LOSI

A:

Leng

th o

f st

ay I

A.

d Fo

r P

(pro

cedu

re)

1 po

int

was

giv

en t

o al

l no

n-m

edic

al i

ndex

adm

issi

ons,

0 t

o m

edic

al I

As,

m

issi

ng d

ata

was

con

side

red

as n

orm

al. e O

ffici

al d

iagn

osis

of

cogn

itiv

e im

pair

men

t in

med

ical

his

tory

, but

ass

esse

d as

men

tally

com

pete

nt t

o an

swer

the

que

stio

ns

(i.e.

usi

ng M

MSE

>20

)

141

Table 2 | ‘Predictability and preventability according to the interviewed’ ab

Patients: Predictable, No. (%) Patients: Preventable, No. (%)No 1037 (74.4) No 875 (62.9)

Yes 325 (23.3) Yes 337 (24.2)

Don’t know 31 (2.2) Don’t know 180 (12.9)

Physicians: Predictable, No. (%) Physicians: Preventable, No. (%)No 844 (64.7) No 1000 (76.6)

Yes 412 (31.6) Yes 222 (17.0)

Don’t know 49 (3.8) Don’t know 83 (6.4)

Nurses: Predictable, No. (%) Nurses: Preventable, No. (%)No 741 (55.4) No 856 (64.2)

Yes 427 (31.9) Yes 250 (18.2)

Don’t know 170 (12.7) Don’t know 237 (17.7)

Carers: Predictable, No. (%) Carers: Preventable, No. (%)No 566 (61.3) No 556 (59.3)

Yes 359 (38.3) Yes 298 (31.8)

Don’t know 13 (1.4) Don’t know 83 (8.9)

Researchers: Predictable, No. (%) Researchers: Preventable, No. (%)No 841 (60.3) No 988 (70.8)

Yes 508 (36.4) Yes 281 (20.1)

Don’t know 46 (3.3) Don’t know 126 (9.0)

Majority: Predictable, No. (%) Majority: Preventable, No. (%)No 1008 (72.1) No 1196 (85.6)

Yes 390 (27.8) Yes 202 (14.4)

a Questions: 1. Do you feel the current readmission was expected?’ (yes, no, don’t know), and 2. ‘Do you feel current readmission was preventable?’ (yes, no, don’t know)b Missing data include the following: patient predictable (n=5), patient preventable (n=6), physician predictable (n=93), physician preventable (n=93), carer predictable (n=460), carer preventable (n=461), nurse predictable (n=60), nurse preventable (n=62), researcher predictable (n=3), researcher preventable (n=3).

142

Table 3 | ‘Consensus on readmission’

Kappa Preventability (κ)

Kappa Predictability (κ)

- Patient Physician Nurse Carer Researcher

Patient - 0.105 0.135 0.360 0.174

Physician 0.173 - 0.273 0.225 0.473

Nurse 0.153 0.338 - 0.194 0.356

Carer 0.289 0.230 0.243 - 0.312

Researcher 0.231 0.607 0.408 0.380 -

Table 4 | ‘Factors positively associated with predictability and preventability of readmissions assessed by the majority’a

Predictability Final model (n=1240), AUROC: 0.71 Odds Ratio (95% CI) P value

Patient age, y 0.98 (0.97-0.99) P < 0.001

IAb type, Non-elective 2.55 (1.59-4.08) P <0.001

Clinical Frailty Scale25 1.29 (1.18-1.42) P < 0.001

Charlson Comorbidity Index24 1.08 (1.02-1.15) P = 0.013

Length between IA and RAb, d 1.02 (1.00-1.03) P = 0.053

Total admissions year prior to IA, elective and non-elective Overall P < 0.001

0

1-5 1.12 (0.82-1.52) P = 0.47

> 5 2.69 (1.72-4.20) P < 0.001

Did you (the patient) suggest to doctors to stay longer at IA? Yes

1.54 (0.99-2.40) P = 0.055

Did you (the patient) feel ready at discharge (IA)? Yes 0.55 (0.40-0.75) P < 0.001

Preventability Final model (n=1155), AUROC: 0.68 Odds Ratio (95% CI) P value

Discharge letter issued at time of readmission, Yes 0.68 (0. 47-1.00) P = 0.051

Follow-up planned at discharge IA, Yes 0.52 (0.35-0.78) P = 0.002

Did you (the patient) feel ready at discharge (IA)? Yes 0.35 (0.24-0.49) P < 0.001

a Corrected for country for all. b IA: Index Admission, RA: Readmission.

143

Tab

le 5

| ‘F

acto

rs p

ositi

vely

ass

ocia

ted

with

rea

dmis

sion

bei

ng a

sses

sed

as p

redi

ctab

le a

nd p

reve

ntab

le p

er in

terv

iew

ed g

roup

’ ab

Fin

al m

od

els

Pati

ent

Pred

icta

ble

(n

=12

40)

Pati

ent

Prev

enta

ble

(n

=10

67)

Phys

icia

nPr

edic

tab

le

(n=

1216

)

Phys

icia

n

Prev

enta

ble

(n=

978)

Nu

rse

Pred

icta

ble

(n=

945)

Nu

rse

Prev

enta

ble

(n=

961)

Car

er

Pred

icta

ble

(n=

860)

Car

er

Prev

enta

ble

(n=

816)

Res

earc

her

Pr

edic

tab

le(n

=11

71)

Res

earc

her

Pr

even

tab

le(n

=10

11)

AU

ROC

Fin

al m

odel

0.72

0.74

0.71

0.65

0.74

0.70

0.68

0.71

0.71

0.68

Ch

arac

teri

stic

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

e

Od

ds

Rat

io

(95%

CI)

, P-

valu

ePa

tien

t ag

e, y

2.34

(1

.42-

4.02

),

P <

0.0

01

--

1.01

(1

.00-

1.02

), P

= 0

.058

0.98

(0

.96-

0.99

), P

< 0

.001

--

--

-

Mal

e se

x-

--

--

--

--

-

Part

nere

d/M

arri

ed-

--

0.67

(0

.48-

0.94

), P

= 0

.02

-0.

53

(0.3

8-0.

75),

P <

0.0

01

-0.

58

( 0.

42-0

.81)

, P

= 0

.001

-0.

73

(0.5

3-1.

01),

P =

0.0

54

IA t

ype,

Non

-ele

ctiv

e2.

34

(1.4

2-4.

02),

P <

0.0

01

-2.

21

(1.4

4-3.

38),

P <

0.0

01

-2.

03

(1.2

5-3.

23),

P =

0.0

04

1.81

(1

.03-

3.18

), P

< 0

.039

1.63

(1

.06-

2.51

), P

= 0

.026

-1.

95

(1.2

9-2.

93),

P =

0.0

01

-

Leng

th o

f st

ay IA

, d

--

--

--

--

--

Leng

th b

etw

een

IA

and

RA,

d-

--

--

--

0.96

(0

.94-

0.98

), P

< 0

.001

0.96

(0

.94-

0.98

), P

< 0

.001

Clin

ical

Fra

ilty

Scal

e25

--

1.31

(1

.20-

1.43

), P

< 0

.001

-1.

38

(0.9

6-1.

98),

P =

0.0

87

-1.

18

(1.0

7-1.

30),

P =

0.0

01

-1.

30

(1.1

9-1.

42),

P <

0.0

01

Cha

rlso

n C

omor

bidi

ty

Inde

x24

--

1.07

(1

.01-

1.13

), P

= 0

.029

--

0.92

(0

.85-

0.99

), P

= 0

.035

--

1.09

(1

.03-

1.16

), P

= 0

.004

0.92

(0

.86-

1.00

), P

= 0

.036

HO

SPIT

AL

Scor

e16-

--

--

--

--

1.12

(1

.03-

1.21

2),

P =

0.0

07

144

Tab

le 5

| (c

ontin

ued)

Fin

al m

od

els

Pati

ent

Pred

icta

ble

(n

=12

40)

Pati

ent

Prev

enta

ble

(n

=10

67)

Phys

icia

nPr

edic

tab

le

(n=

1216

)

Phys

icia

n

Prev

enta

ble

(n=

978)

Nu

rse

Pred

icta

ble

(n=

945)

Nu

rse

Prev

enta

ble

(n=

961)

Car

er

Pred

icta

ble

(n=

860)

Car

er

Prev

enta

ble

(n=

816)

Res

earc

her

Pr

edic

tab

le(n

=11

71)

Res

earc

her

Pr

even

tab

le(n

=10

11)

Num

ber

of m

edic

ine

at

disc

harg

e (c

hron

ic a

nd

non-

chro

nic

use)

--

--

P =

0.0

82-

P =

0.0

01P

= 0

.094

--

0-4

--

--

--

-

5-8

--

--

1.70

(1

.06-

2.72

), P

= 0

.028

-0.

74

(0.4

8-1.

15),

P =

0.1

78

1.00

(0

.63-

1.59

), P

= 0

.97

--

Mor

e th

an 8

--

--

1.55

(0

.98-

2.45

), P

= 0

.059

-1.

45

(0.9

6-2.

19),

P =

0.7

5

1.41

(0

.93-

2.15

), P

= 0

.11

--

DN

R po

licy

at

disc

harg

e, D

o re

susc

itat

e

-1.

44

(1.0

0-2.

07),

P =

0.0

48

--

0.6

2

(0.4

2-0.

92),

P =

0.0

18

--

--

-

ICU

/CC

U/H

DU

ad

mis

sion

dur

ing

IA,

Yes

--

--

--

--

--

Cog

niti

ve im

pair

men

t (o

ffici

al d

iagn

osis

), Y

es-

--

--

--

-2.

03

(0.8

9-4.

63),

P =

0.0

91

Falle

n at

leas

t on

ce in

la

st h

alf

year

, Ye

s-

--

--

1.37

(0

.95-

1.95

), P

= 0

.089

--

--

145

Tab

le 5

| (c

ontin

ued)

Fin

al m

od

els

Pati

ent

Pred

icta

ble

(n

=12

40)

Pati

ent

Prev

enta

ble

(n

=10

67)

Phys

icia

nPr

edic

tab

le

(n=

1216

)

Phys

icia

n

Prev

enta

ble

(n=

978)

Nu

rse

Pred

icta

ble

(n=

945)

Nu

rse

Prev

enta

ble

(n=

961)

Car

er

Pred

icta

ble

(n=

860)

Car

er

Prev

enta

ble

(n=

816)

Res

earc

her

Pr

edic

tab

le(n

=11

71)

Res

earc

her

Pr

even

tab

le(n

=10

11)

Tota

l adm

issi

ons

year

pr

ior

to IA

, el

ecti

ve

and

non-

elec

tive

--

P <

0.0

01

-P

= 0

.002

P =

0.0

96P

= 0

.090

-P

< 0

.001

-

0-

--

--

1-5

--

1.24

(0

.93-

1.67

), P

= 0

.149

-1.

45

(1.0

3-2.

04),

P =

0.0

31

1.39

(0

.94-

2.04

), P

= 0

.096

0.98

(0

.71-

1.34

), P

= 0

.876

1.40

(1

.04-

1.88

), P

= 0

.026

-

> 5

--

2.97

(1

.92-

4.59

), P

< 0

.001

-2.

37

(1.4

4-3.

88),

P =

0.0

01

1.79

(1

.03-

3.11

), P

= 0

.039

1.68

(1

.00-

2.80

), P

= 0

.049

-3.

40

(2.1

7-5.

34),

P <

0.0

01

-

Dis

char

ge le

tter

issu

ed

at t

ime

RA,

Yes

--

--

--

--

--

Follo

w-u

p pl

anne

d at

di

scha

rge

IA,

Yes

--

-0.

62

(0.4

1-0.

94),

P =

0.0

26

-0.

56

(95%

CI

0.37

-0.8

5),

P =

0.0

06

-0.

66

(95%

CI

0.43

-1.0

1),

P =

0.0

57

-0.

42

(95%

CI

0.22

-0.6

1),

P <

0.0

01

Has

pat

ient

bee

n se

en

by a

doc

tor

betw

een

IA a

nd R

A?

Yes

2.08

(1

.47-

2.93

),

P <

0.0

01

--

--

--

--

-

Was

a s

umm

ary

prov

ided

at

disc

harg

e?

Yes

--

-0.

67

(0.4

6-0.

98),

P =

0.0

4

1.38

(0

.96-

1.98

), P

= 0

.087

--

--

-

Did

you

(th

e pa

tien

t)

sugg

est

doct

ors

to s

tay

long

er a

t IA

? Ye

s

-2.

18

(1.3

5-3.

53),

P =

0.0

02

1.56

(0

.99-

2.46

), P

= 0

.056

--

--

--

-

Did

you

(th

e pa

tien

t)

feel

bet

ter

at d

isch

arge

(IA

)? Y

es

0.68

(0

.47-

0.98

),

P =

0.0

39

0.69

(0

.48-

0.98

), P

= 0

.039

-0.

62

(0.4

4-0.

87),

P =

0.0

06

--

--

--

146

Tab

le 5

| (c

ontin

ued)

Fin

al m

od

els

Pati

ent

Pred

icta

ble

(n

=12

40)

Pati

ent

Prev

enta

ble

(n

=10

67)

Phys

icia

nPr

edic

tab

le

(n=

1216

)

Phys

icia

n

Prev

enta

ble

(n=

978)

Nu

rse

Pred

icta

ble

(n=

945)

Nu

rse

Prev

enta

ble

(n=

961)

Car

er

Pred

icta

ble

(n=

860)

Car

er

Prev

enta

ble

(n=

816)

Res

earc

her

Pr

edic

tab

le(n

=11

71)

Res

earc

her

Pr

even

tab

le(n

=10

11)

Did

you

(th

e pa

tien

t)

feel

rea

dy a

t di

scha

rge

(IA)?

Yes

0.29

(0

.20-

0.42

),

P <

0.0

01

0.22

(0

.15-

0.32

), P

< 0

.001

0.71

(0

.52-

0.97

), P

= 0

.033

- 0

.46

(0

.33-

0.65

), P

< 0

.001

0.47

(0

.33-

0.68

), P

< 0

.001

0.42

(0

.30-

0.60

), P

< 0

.001

0.24

(0

.17-

0.34

), P

< 0

.001

0.48

(0

.36-

0.65

), P

< 0

.001

0.46

(0

.33-

0.65

) P

< 0

.001

a C

orre

cted

for

cou

ntry

for

all

b IA

: In

dex

Adm

issi

on,

RA:

Read

mis

sion

.

147

DISCUSSIONThis prospective European multicentre study of 1398 unscheduled medical readmissions

revealed that there was poor consensus on predictability and preventability among readmitted

patients, their carers, nurses, physicians. Especially, there was little consensus between patients

and their physicians. In addition, factors that could potentially contribute to (preventable)

readmission were notably different according to every group interviewed. Not feeling ready

for discharge was strongly associated with predictability and preventability when opinion

of the majority interviewed was used as the gold standard. This was also underlined when

opinions of the interviewees were taken separately. In addition, according to the opinion of

the majority, 27.8% of the readmissions were deemed predictable and 14.4% preventable.

Readmissions thought to have been caused by healthcare worker failures were more often

deemed preventable.

Our study is the largest multicentre study performed to date, and the first investigating

European readmissions prospectively. Previous large scale studies were mostly performed in

the USA and due to the differences in healthcare systems, results of these studies are probably

not applicable in a European setting.20 In addition, our study is one of the first studies in Europe

where opinions of most important stakeholders in the care-chain were taken into account.

Most previous studies were small retrospective studies, or did not involve all stakeholders.7,21,28

The percentages of readmissions which were deemed preventable are in line with previous

research, but their numbers do distinctly differ per interviewed group.11 An important finding

is the discrepancy in assessment between all interviewed, especially between patient and their

treating physician, the latter assessing less readmissions as likely preventable. This implies that

these healthcare professionals do not agree with their patients about the predictability and

preventability and associated factors.

Although, multiple risk models have been composed trying to create readmission prediction

models, most models do not perform satisfactorily in a European population.20 This was also

demonstrated in our cohort, 53.8% had a low risk of readmission according to the HOSPITAL-

score (Supplement 3).16 In addition, many variables in these models are not modifiable, and

therefore not suitable as interventions to improve the healthcare chain.18-20

According to our results, if a patient reported not feeling ready for discharge this was a risk

factor associated with a higher chance of (preventable) readmission. Auerbach et al. (2016)

also reported this by stating, that a proportion of US readmissions may be prevented with

better attention to patients’ readiness for discharge.23 In addition, we inquired all interviewed

for root causes that could better qualify the lack of readiness. Readmissions were more often

deemed preventable if attributed to healthcare worker-related causes, which are probably

modifiable and therefore a potential focus for improvement. Commencing improvement by

simply asking patient at the bedside whether he feels ready before discharge may be the first

step in understanding each other’s perspectives and could make other prediction models less

148

relevant. A recent study has underlined this by showing that early preparation for discharge

resulted in significant reductions in patients reporting they were unaware of problems to watch

out for post-discharge, and patients who did not understand their recovery plan.29 In addition,

Greysen et al. (2017) reported most readmitted patients understood their post-discharge

plan but were not explicitly asked about anticipated difficulties carrying out the aspects of

this plan.30

The interviewed groups did not agree on predictability, let alone on preventability. A high

readmission rate should be a signal for a hospital to further look into the causes, but looking

solely at the rate and penalizing the hospitals, without correcting for case-mix and looking

more closely into the type of care provided on the work floor is questionable. Our work

demonstrates that defining a gold standard integrating preventability in a quality indicator

is difficult.

The limitations of this study must be acknowledged. We collected data from four different

countries in 15 hospitals all using different healthcare systems, which may make results less

generalizable. However, European healthcare systems are more alike when compared to US

system and we corrected for country in our analysis. Another limitation is asking opinions

during readmission about the index admission which could potentially lead to (recall)bias and

subjectivity. The majority opinion was used as the gold standard to decide on preventability

and predictability which may be questionable. Therefore, we also reported on perspectives per

interviewed group separately. 8

CONCLUSIONThe European acute healthcare chain is under increasing pressure, potentially resulting in more

unscheduled readmissions. This international multicentre study performed in 2016 is the first

to prospectively assess this problem in Europe. The 1398 readmitted patients, their carers,

treating nurses and their physicians do not agree on the predictability and preventability

of readmissions, let alone associated risk factors. This raises the question of the validity of

readmissions as a quality monitor. Extensive research has been performed on risk models, but

healthcare professionals simply asking the patient whether they are feeling ready for discharge

may be a key question to target in preventing unnecessary readmissions.

ACKNOWLEDGEMENTSThe authors have nothing to disclose. No funding was received for this study.

A full list of membership of the safer@home consortium is as follows: Mikkel Braband, Tim

Cooksley, Louise S. van Galen, Harm R. Haak, Rachel M. Kidney, John Kellet, Hanneke Merten,

Prabath W. Nanayakkara, Christian. H. Nickel, John Soong, Christian P. Subbe, Immo Weichert.

Collaborators (alphabetically ordered): Lisette Ackermans1, MD, Alex B. Arntzenius14, MD,

PhD, Dennis Barten7, MD, Maarten H. van der Bie14, MD, Tom Boeije15, MD, Ella Chaudhuri16,

149

MD, Marieke Diepenbroek-Meekes17, Eva Durinck6, Bsc, Jelle A. van Erven1, BSc, MD, Tabitha

A. Graaff-de Kooter6, MD, Frits Holleman18, MD, PhD, Job Huussen17, MD, W. Jansen6, Jonas

Junghans Jensen19, BSc, John Kellet20, MB, FRCPI, Tessa Knol21, BSc, Irene Lee13, MD, Darleen A.

Leung13, MD, Bas van Lieshout18, BSc, Arend-Jan Meinders21, MD, Nieke E. Mullaart-Jansen15,

MD, Sietske van Nassau8, BSc, Jonas Larsen Pedersen19, BSc, Rune Overgaard Jensen19, BSc,

Anouschka Pronk21, MD, Mette Rahbek Kristensen2, Amy M. Ridge9, MD, Timo Roeleveld14,

MD, Marije C. Schipper18, BSc, Daisy Vedder15, MD, Joris van der Vorst18, BSc, MD, Joyce

Wachelder8, MD

1Department of Internal Medicine, VU University Medical Centre, Amsterdam, The Netherlands 2Department of Emergency Medicine, Hospital of South West Jutland, Denmark3Department of Acute Medicine, University Hospital of South Manchester, Manchester, United Kingdom4Department of Epidemiology and Biostatistics, VU University Medical Centre Amsterdam,

Amsterdam, The Netherlands5Department of Public and Occupational Health, EMGO Institute for Health and Care Research,

VU University Medical Centre, Amsterdam, The Netherlands 6Department of Quality, Safety & Innovation, Albert Schweitzer Hospital,

Dordrecht, The Netherlands 7Department of Emergency Medicine, VieCuri Hospital, Venlo, The Netherlands 8aDepartment of Internal Medicine, Maxima Medisch Centre, Eindhoven/Veldhoven, The Netherlands8bDepartment of Internal Medicine, Division of General Internal Medicine, Maastricht University

Medical Centre+. Maastricht University. 9Department of Acute Medicine, St. James Hospital, Dublin, Ireland10Department of Emergency Medicine, University Hospital Basel, Switzerland11Imperial College London, United Kingdom12Department of Acute Medicine, The Ipswich Hospital NHS Trust, Ipswich, United Kingdom13Department of Acute Medicine, Ysbyty Gwynedd Hospital, Wales, United Kingdom14Department of Internal Medicine, Spaarne Gasthuis, Hoofddorp, The Netherlands 15Department of Emergency Medicine, Westfries Gasthuis, Hoorn, The Netherlands 16Department of Acute Medicine, North Bristol NHS Trust, Bristol, United Kingdom17Department of Internal Medicine, Slingeland Hospital, Doetinchem, The Netherlands18Department of Internal Medicine, Academic Medical Centre, Amsterdam, The Netherlands19Department of Emergency Medicine, Odense University Hospital, Denmark20Thunder Bay Regional Health Sciences Centre, Thunder Bay, Ontario, Canada 21Department of Internal Medicine, Antonius Hospital, Nieuwegein, The Netherlands

SUPPLEMENTAL MATERIAL

Supplemental Material 1 ‘Questionnaire’

Supplemental Material 2 ‘PRISMA model’

Supplement Material 3 ‘Patient characteristics (additional)’

150

REFERENCES1. Greysen SR, Stijacic Cenzer I, Auerbach AD, Covinsky KE. Functional impairment and hospital

readmission in Medicare seniors. JAMA Intern Med. 2015;175(4):559-565.

2. Blunt I, Bardsley M, Grove A, Clarke A. Classifying emergency 30-day readmissions in England using routine hospital data 2004-2010: what is the scope for reduction? Emerg Med J. 2015;32(1):44-50.

3. Balla U, Malnick S, Schattner A. Early readmissions to the department of medicine as a screening tool for monitoring quality of care problems. Medicine (Baltimore). 2008;87(5):294-300.

4. Basisset Kwaliteitsindicatoren 2014. Utrecht: Inspectie voor de gezondheidszorg, 2015. Accessed January, 2016.

5. Pines JM, Hilton JA, Weber EJ, et al. International perspectives on emergency department crowding. Acad Emerg Med. 2011;18(12):1358-1370.

6. VUmc Netwerk Acute Zorg. Ontwikkelingen aanbod acute patiënten SEH’s RsHsQQ, 2014 en 2015 ROAZ-regio VUmc. https://www.vumc.nl/afdelingen-themas/719134/8513847/8514115/20150929_RapportAcuteZorgQ11.pdf. Published September 2015. Accessed August, 2016.

7. Bianco A, Mole A, Nobile CG, Di Giuseppe G, Pileggi C, Angelillo IF. Hospital readmission prevalence and analysis of those potentially avoidable in southern Italy. PLoS One. 2012;7(11):e48263.

8. Halfon P, Eggli Y, Pretre-Rohrbach I, Meylan D, Marazzi A, Burnand B. Validation of the potentially avoidable hospital readmission rate as a routine indicator of the quality of hospital care. Med Care. 2006;44(11):972-981.

9. van Galen LS, Nanayakkara PW. [Hospital readmissions: A reliable quality indicator?]. Ned Tijdschr Geneeskd. 2016;160:A9885.

10. van Walraven C. The Utility of Unplanned Early Hospital Readmissions as a Health Care Quality Indicator. JAMA Intern Med. 2015;175(11):1812-1814.

11. van Walraven C, Jennings A, Forster AJ. A meta-analysis of hospital 30-day avoidable readmission rates. J Eval Clin Pract. 2012;18(6):1211-1218.

12. Cakir B, Gammon G. Evaluating readmission rates: how can we improve? Southern medical journal. 2010;103(11):1079-1083.

13. Meisenberg BR, Hahn E, Binner M, et al. ReCAP: Insights Into the Potential Preventability of Oncology Readmissions. J Oncol Pract. 2016;12(2):153-154.

14. L.S. van Galen TC, H. Merten et al.*On behalf of the safer@home consortium. (10-2016) Physician Consensus on preventability and predictability of Readmissions based on standard case scenarios. Running title: Physician consensus on readmissions. Neth J Med. 2016 Dec;74(10):434-442.

15. van Walraven C, Dhalla IA, Bell C, et al. Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. CMAJ. 2010;182(6):551-557.

16. Donze J, Aujesky D, Williams D, Schnipper JL. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA Intern Med. 2013;173(8):632-638.

17. Donze JD, Williams MV, Robinson EJ, et al. International Validity of the HOSPITAL Score to Predict 30-Day Potentially Avoidable Hospital Readmissions. JAMA Intern Med. 2016;176(4):496-502.

18. Kansagara D, Englander H, Salanitro A, et al. Risk prediction models for hospital readmission: a systematic review. JAMA. 2011;306(15):1688-1698.

19. Zhou H, Della PR, Roberts P, Goh L, Dhaliwal SS. Utility of models to predict 28-day or 30-day unplanned hospital readmissions: an updated systematic review. BMJ Open. 2016;6(6):e011060.

20. Cooksley T, Nanayakkara PW, Nickel CH, et al. Readmissions of medical patients: an external validation of two existing prediction scores. QJM. 2016;109(4):245-248.

151

21. Fluitman KS, van Galen LS, Merten H, et al. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Intern Med. 2016;30:18-24.

22. Mixon AS, Goggins K, Bell SP, et al. Preparedness for hospital discharge and prediction of readmission. J Hosp Med. 2016;11(9):603-609.

23. Auerbach AD, Kripalani S, Vasilevskis EE, et al. Preventability and Causes of Readmissions in a National Cohort of General Medicine Patients. JAMA Intern Med. 2016;176(4):484-493.

24. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383.

25. Rockwood K, Song X, MacKnight C, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ. 2005;173(5):489-495.

26. van Wagtendonk I, Smits M, Merten H, Heetveld MJ, Wagner C. Nature, causes and consequences of unintended events in surgical units. Br J Surg. 2010;97(11):1730-1740.

27. Vuuren Wv, Shea CE, Schaaf TWvd. The development of an incident analysis tool for the medical fields. Eindhoven University of Technology.

28. Magdelijns FJ, Schepers L, Pijpers E, Stehouwer CD, Stassen PM. Unplanned readmissions in younger and older adult patients: the role of healthcare-related adverse events. Eur J Med Res. 2016;21(1):35.

29. Harrison JD, Greysen RS, Jacolbia R, Nguyen A, Auerbach AD. Not ready, not set...discharge: Patient-reported barriers to discharge readiness at an academic medical center. J Hosp Med. 2016;11(9):610-614.

30. Greysen SR, Harrison JD, Kripalani S, et al. Understanding patient-centred readmission factors: a multi-site, mixed-methods study. BMJ Qual Saf. 2016.

CHAPTER 10

HOSPITAL STANDARDIZED MORTALITY RATIO: A RELIABLE INDICATOR OF QUALITY OF CARE?

Jelle A. van Erven | Louise S. van Galen | Asselina A. Hettinga-Roest | Ellen P. Claessens | Jan C. Roos | Mark H. Kramer | Prabath W. Nanayakkara

Submitted

‘They die one by one’ Carol Haraden

154

ABSTRACTObjective

This study investigates (1) whether the HSMR (Hospital Standardized Mortality

Ratio) model under- or overestimates disease severity and (2) the completeness of

the data collected by administrators to calculate HSMR in a cohort of deceased patients with

the diagnosis pneumonia.

Design, Participants and Main Outcome Measures

In this retrospective cohort study PSI scores, abbMEDS scores and associated mortality

probabilities were obtained from 32 deceased pneumonia-patients over the year 2014 in

the VU University Medical Centre. These were compared to mortality probabilities of the Central

Bureau for Statistics (CBS) calculated for every patient using the HSMR model. Clinical charts

were examined to extract relevant comorbidities to determine the reliability of data sent to

the national registration of hospital care.

Results

Risk categories determined by using the PSI and the abbMEDS were significantly higher

compared to the risk categories according to HSMR (p = 0.001, respectively p = 0.000).

The mean difference between the number of comorbidities in our registration and the coders’

registration was 1.97 (p = 0.00). The mean difference was 0.97 (p = 0.000) for the number

of comorbidities of influence on the Charlson Comorbidity Index (CCI) and 1.25 (p=0.001) for

the calculated CCI.

Conclusion

The results of this study suggest that the mortality probabilities as calculated by the CBS

are an underestimation of the risk of dying for each patient. Our study also showed that

the registration of data sent to the CBS underestimated the actual comorbidities of the patient,

and could possibly influence the HSMR.

Key words

Data registration, disease severity, HSMR, quality indicator, patient outcomes

155

INTRODUCTIONSince March 2014, Dutch hospitals are compulsory to be transparent about their mortality

rates.1 To be able to compare the quality of hospital care using their mortality rate, these

rates have to be standardized in order to correct for the differences in the case-mix.2 This

standardized ratio is represented in the ‘Hospital Standardized Mortality Ratio’ and is the ratio

of the observed to the expected deaths, derived from data from the LBZ (LBZ: National

registration of hospital care).3 The expected deaths are calculated with the use of a statistical

model that corrects for certain factors such as age, socio-economic status and comorbidity.3

In 2014, this model contained standardized mortality ratios of 50 diagnosis groups (SMRs),

which account for 80% of in-hospital death. This has been extended to 157 diagnosis groups

(SMRs) in 2015.

Over the year 2014, the VUmc (VU Medical Centre) had a relatively high HSMR, in part caused

by a high SMR for the diagnosis group ‘pneumonia’. The SMR of a diagnosis group can be

used to investigate the cause of unexpected high mortality in a hospital more specifically

than by solely using the HSMR.4 For this reason a commission of independent external

investigators in VUmc were asked to investigate this high SMR. The aim was to investigate

whether preventable/avoidable factors contributed to these deaths. Their report showed no

avoidable causes of death in this cohort. These findings suggest that the cause of the high

SMR for pneumonia is probably due to other (unknown) factors. It could, for example, be

caused by insufficient registration of comorbidities or wide variations in the disease severity.

In the clinical setting, physicians and nurses use several different scoring systems to determine

the severity and to predict the mortality of pneumonia using patient characteristics such as

age, blood urea and respiratory rate. Two of the best-validated and most used scoring systems

are the PSI (Pneumonia Severity Index) and the abbMEDS (Abbreviated Mortality in Emergency

Department Sepsis).5,6

The HSMR is calculated by the CBS (Central bureau of statistics) and the data used for

this calculation is registered by DHD (Dutch Hospital Data) within the context of the LBZ.

The Medical Administration Office of each hospital provides the information that is used.

The HSMR is, among other covariates, derived from the primary diagnosis and the Charlson

Comorbidity Index (CCI),7which are obtained from patients’ charts and documented by coders.

This underlines the importance of a complete administration, as deficient or faulty data might

directly influence the HSMR. Van der Laan et al. (2013) showed that the effect of registering

10 percent more comorbidities could result in a decrease of 5 points of the HSMR.8 Although

the administration of data has improved significantly since the implementation of the HSMR

as an indicator of quality of care, there still might be inconsistencies in the (comorbidity) data

extracted by coders and registered by DHD, when compared with the actual data extracted by

doctors from the patients’ charts.8,9

Therefore, the main aim of this study is to examine whether the HSMR model under- or

overestimates the disease severity of pneumonia patients when compared to routinely used

156

clinical severity scores. Our secondary aim is to investigate the completeness of the data sent

to DHD to calculate the HSMR.

METHODSIn 2014, 32 deceased patients were registered in the ‘pneumonia’-group at the VUmc. In order

to obtain PSI scores and abbMEDS scores for these patients, patients’ charts were examined for

information needed to calculate these scores from which corresponding mortality probabilities

could be calculated. Missing information was considered as not contributing to the score.

The HSMR is calculated by logistic regression using below mentioned covariates with data

provided by hospital coders. With this information, regression coefficients for these covariates

are estimated and are used to calculate mortality probabilities for each individual admission.3

The results of the calculations is sent to each hospital in the annual HSMR report.

The HSMR is calculated using the following covariates3

– Age at admission

– Sex

– SES (socio-economic status) of the postal area of the patient’s address. The SES

classification per postal code is compiled by the Netherlands Institute for Social

Research (SCP)

– Severity of main diagnosis. Instead of CCS diagnosis subgroups (Clinical

Classifications Software: a tool to cluster patient diagnoses into a manageable

number of clinically meaningful categories, based on the International Classification

of Diseases. The CCS makes little distinction in regard to disease severity when

categorizing diagnosis codes), a classification of severity of the main diagnosis in

terms of mortality rates is used, as suggested by Van den Bosch et al. (2011)10

– Urgency of admission (elective, acute)

– Comorbidity (17 comorbidity groups of the Charlson comorbidity index7)

– Source of admission (home, nursing home or other institution, hospital)

– Year of discharge

– Month of admission

In order to compare the mortality probabilities derived from the PSI scores and abbMEDS scores

(which correspond with ordinal risk categories) and the mortality probabilities calculated by

the CBS (which can be considered a continuous variable), new categories needed to be formed

for the latter. It was decided to form three sets of categories from the CBS data, one for each

of the scores. Table 1 shows the risk categories and corresponding mortality probabilities of

the two scoring systems. The consensus was that the best way to establish limits for new

categories, was by using the median between each of the mortality probabilities, as those

are the mean of that risk category. As can be seen in table 1, the lowest risk categories

of the PSI predict a risk of 0.1% and of 0.6%. The median between these risks is 0.35,

therefore, the limits of the PSI categories used are 0 – 0.35; 0.35 – 0.75; 0.75 – 5.2; 5.2 – 18.1;

18.1 – 100 and the limits for the abbMEDS are 0 – 11.55; 11.55 – 32.85; 32.85 – 100.

157

The newly formed categories of the CBS calculated mortality probabilities were compared to

the categories of the PSI scores and abbMEDS scores. A Wilcoxon Sign-Rank Test was used for

statistical analysis to test for conformity.

To investigate whether data sent to DHD significantly differed from what is found in patients’

charts, data was gathered on the total amount of comorbidities that were present in charts,

which of these were directly of influence on the CCI (excluding the comorbidities that are not

in the Charlson comorbidity index) and finally the estimated CCI by hospitals itself. The coders

in VUmc primarily look at the discharge letter and only broaden their scope when they presume

this to be insufficient. In this study one researcher (JVE) thoroughly checked every patient’s

chart which included the discharge letter. If there was any uncertainty concerning a possible

comorbidity or diagnosis, a second researcher (PN) was consulted and consensus was reached.

The data that the CBS used was obtained from the Medical Administration Office. A paired

t-test was used to analyse the difference between our registration and the coder’s registration.

For all analyses, a two-tailed p-value of less than 0.05 was considered statistically significant.

RESULTSTable 2 gives an overview of the patient characteristics of our population. 10 patients had

a cause of death other than respiratory failure or sepsis.

Mortality probabilities

Table 3 illustrates the dispersion of mortality probabilities calculated by the CBS using the HSMR

model and those of the two clinical scoring systems. It can be seen that for the majority of

patients the estimated risk of dying within 28-30 days is much higher according to the clinical

scoring systems than the estimated risk of dying as calculated using the SMR-model. Especially

the abbMEDS assesses the risk to be significantly higher than the CBS does.

Descriptive statistics of conformity were performed and this showed that for the PSI 18 patients

were in a higher risk category than according to the CBS (SMR), 3 were in a lower category

and 11 were in the same category. When looking at the abbMEDS, all patients were either in

the same risk category (10) or in a higher risk category (22) compared to SMR.

Table 1 | Risk categories and corresponding mortality probabilities of the scoring systems.

PSI5 abbMEDS6,11

Low risk I 0.1% Low risk 3.6%

Low risk II 0.6% Intermediate risk 19.5%

Low risk III 0.9% High risk 46.2%

Medium risk 9.5%

High risk 26.7%

The mortality probabilities of the risk categories for the abbMEDS score are derived from a study by Roest et al. (2015)11

158

Table 2 | Demographic and clinical characteristics of the deceased patients in the ‘pneumonia’-group over the year 2014

Characteristics

Deceased patients (n=32)

Number Percentage

Demographic factor Age > 65 year Female sex

Nursing home resident

25

12

6

78.13

37.5

18.75

Admissions > 2 last 12 months ICU admissions last 12 months

Unexpected long admission*

16

14

6

50

43.75

18.75

Cause of death

Respiratory failure

Sepsis

Myocardial infarction

Heart failure

Other

14

12

5

4

1

43.75

37.5

15.63

12.5

3.13

Other clinical characteristics Immunocompromised§ Do not resuscitate

Polypharmacy±

Limited mobility&

Delirium

Malnutrition$

10

26

28

24

8

12

31.25

81.25

87.5

75

25

37.5

* = An admission minimally 50 percent longer than expected for a specific primary diagnosis. The calculation of the expected length of admission takes into account the age of the patient, primary diagnosis and any possible interventions.§ = Immunodeficiency by the use of immuno-suppressive drugs, by neutro- or leukopenia or other causes.± = The chronic use of ≥ 5 medications.& = Patient uses devices for mobility or was bedridden.$ = Patient has a SNAQ (Short Nutritional Assessment Questionnaire) of ≥ 2 or when the patient was described as cachexic.

Further analysis showed a significant increase in assigned risk categories for the PSI (p<0,001)

and for the abbMEDS (p<0,001) compared to the SMR. This indicates that risks of dying of

these patients according to clinical scoring systems were significantly higher than the risks of

dying according to SMR calculated by the CBS.

Registration of data

Figure 1 shows the number of comorbidities, the number of comorbidities influencing the CCI

and the calculated CCI itself from our own registration and those same outcome measures

which medical coders registered. For each of the outcome measures the mean of our registered

number is higher than the mean of what the coders registered.

159

As table 4 shows, the mean difference between the number of comorbidities in our registration

and the coder’s registration is 1,97. The mean difference between our registration and

the coders’ registration for the number of comorbidities of influence of the CCI and of the CCI

itself are 0,97 and 1,25 respectively. All of these results are statistically significant.

Table 3 | Mortality probabilities calculated by the CBS and those derived from the scoring systems

Patient nr. Mortality probabilities CBS (SMR) PSI abbMEDS

1 0.87% 0.60% 3.60%

2 6.29% 26.70% 19.50%

3 14.70% 26.70% 46.20%

4 6.43% 9.50% 46.20%

5 10.01% 9.50% 19.50%

6 3.64% 26.70% 19.50%

7 18.02% 26.70% 19.50%

8 15.33% 26.70% 46.20%

9 10.62% 9.50% 19.50%

10 12.19% 26.70% 19.50%

11 8.76% 9.50% 3.60%

12 18.05% 26.70% 46.20%

13 18.60% 9.50% 19.50%

14 3.45% 0.90% 19.50%

15 5.39% 26.70% 46.20%

16 5.76% 26.70% 19.50%

17 11.02% 26.70% 19.50%

18 9.34% 26.70% 46.20%

19 6.68% 26.70% 19.50%

20 7.74% 9.50% 19.50%

21 5.92% 9.50% 19.50%

22 6.18% 9.50% 19.50%

23 24.59% 26.70% 19.50%

24 10.54% 26.70% 19.50%

25 13.70% 9.50% 19.50%

26 5.01% 9.50% 19.50%

27 1.54% 9.50% 19.50%

28 2.89% 26.70% 3.60%

29 6.32% 26.70% 46.20%

30 12.87% 9.50% 19.50%

31 13.37% 26.70% 46.20%

32 21.24% 9.50% 19.50%

Green represents higher probability than calculated by the CBS (using the SMR-model), red represents a lower probability.

160

Figure 1

Comorbiditi

es study

Comorbiditi

es coders

Comorbiditi

es Charlson stu

dy

Comorbiditi

es Charlson co

ders

Charlson-sc

orestu

dy

Charlson-sc

oreco

ders0

2

4

6

8

10

Study vs. Coder's registration

Coun

t

Figure 1 | Compared means and SD of the outcome measures extracted during our study versus what coders registered and the CBS used to calculate the HSMR.

Table 4 | Statistical analysis of the outcome measures

Mean SD* SE* 95% CI p-value

Comorbidities study – comorbidities CBS

1.96875 2.53345 .44785 1.05535-

2.88215

.000

Comorbidities Charlson study – Comorbidities Charlson CBS

0.96875 1.37921 .24381 .47149 -1.46601

.000

Charlson index study – Charlson index CBS

1.25000 1.95101 .34489 .54658 -1.95342

.001

Outcome measures in the table are the number of comorbidities and the number of comorbidities influencing the Charlson Comorbidity Index and the Charlson Comorbidity index.* SD: Standard deviation, SE: Standard error of the mean.

DISCUSSIONThe findings in this paper indicate that (1) the risk of dying for an individual patient is

higher when calculated using validated clinical scoring systems than the risk of dying

calculated with the HSMR model. (2) The total number of comorbidities and the number of

comorbidities influencing the CCI is higher according to our registration than according to

the coders’ registration.

The results in this study further support the suggestion that has been made by Pleizier et al.

that the SMR for more diagnosis groups besides cerebrovascular diseases will also decrease

when adjusted for the severity of disease.12 They concluded that within the SMR group

‘cerebrovascular diseases’ there is no distinction between ‘stroke’, ‘cerebral haemorrhage’

161

and ‘subarachnoid haemorrhage’ while their mortality rates differ greatly.12 The mortality rates

were 18, 43 and 35 percent respectively, and when these differences were not taken into

account, the influence on the SMR could be considerable.13 They recalculated the SMR for

‘cerebrovascular diseases’ after correcting for the above mentioned sub-diagnoses and found

that this gave a reduction from 119 (95%-CI 105-133) to 102 (95%-CI 91-115).12 Beside this

diagnosis group, this possibly is also true for other SMR groups such as the ‘pneumonia-group’.

A subdivision for ‘cerebrovascular diseases’ was easily made by just looking at the mortality

rates for several sub diagnoses within that group. This however, is a lot harder for a diagnosis

group such as ‘pneumonia’, where there are no known distinct sub diagnoses. To make

a subdivision for ‘pneumonia’, two different scoring systems were used that indicate severity

of disease. The best way to prove that a subdivision by each of these scores has a direct

effect on the SMR, is by adjusting the SMR model in the same way Pleizier et al. did.12 They

incorporated a division in risk categories into the logistic regression model, just like the other

covariates. In our study, it was decided to compare the mortality probabilities of the validated

scores to the mortality probabilities calculated by the CBS with the use of the SMR-model.

The results show that for the large majority of patients the expected mortality within 28-30

days is much higher according to the two scoring systems than to the score calculated by

the CBS. Probably partly caused by underscoring the number of co-morbidity, but also a lack

of proper adjustment for the severity of the disease pneumonia in the individual patient.

These two scoring systems are widely used in clinical settings when dealing with pneumonia

patients and have been validated.6,13 They are specifically designed to assess the severity of

pneumonia/sepsis and should therefore be taken seriously as predictors of death. This suggests

that the mortality probabilities according to the HSMR model of CBS are an underestimation

of the real risk of dying for each patient. Naturally, estimating disease severity with the use of

nine variables results in a simplification of reality. In addition, it is known that university medical

centres predominantly provide tertiary care for a case-mix of patients with a higher severity of

disease than peripheral hospitals. Therefore, they might falsely have a ‘higher’ HSMR.

Our results indicate that the mortality probabilities calculated by the PSI and the abbMEDS

are higher than what the CBS calculated. It could be argued that the steps between the risk

categories of these scoring systems are fairly big. Therefore, when a patient is placed in

the highest risk category of for example the abbMEDS, their risk of dying could be even

higher than 46.2 percent. However, table 2 does compare categorical variables (the mortality

probabilities calculated by the scoring systems) with a continuous variable (the mortality

probabilities calculated by the CBS), which implies that these risks will almost always differ

from the risks as calculated by the CBS.

The secondary aim of this study was to assess the registration of comorbidities from the patients’

charts by medical coders. An unanticipated finding was that the source of admission was in

every case ‘home’. It seemed as though no distinction was made between ‘home’ and ‘nursing

home’. Nevertheless, table 1 shows that 6 out of 32 patients were admitted from a nursing

162

home. This could potentially have an impact on the HSMR as a whole, however, this influence

is probably rather small. It must be acknowledged that the source of admission is not primarily

registered by coders, but they are responsible for checking this registration.

Van der Laan et al. already underlined the influence of the registered number of comorbidities

on the HSMR.8 With this in mind, an average difference of two registered comorbidities seems

significant enough to be of influence of the HSMR. For a comorbidity to be of influence of

the HSMR it needs to add to the CCI, so to make the previous assumption more likely, the CCI

of every patient was also taken in consideration. It was found that there was a statistically

significant difference of 1.25 points, between the calculated CCI based on our registration and

the calculated CCI based on the coders’ registration. This strongly suggests that the apparent

insufficiently registered number of comorbidities does directly influence the HSMR. As stated

earlier, coders are dependent on proper documentation by others, including doctors. They

primarily look at the discharge letter and surgery reports, and are not expected to go through

the entire patient chart, mainly since this would be too time-extensive. This lack of time

might be one of the causes of the apparent under registration of comorbidities. One other

cause explaining the under registration is that according to coding protocol, sometimes an

ICD-10 (International Classification of Diseases) code which has less impact on the HSMR

than the actual diagnosis has to be selected. Although the precise impact cannot be judged

by the results of this study, these findings do raise the question whether the HSMR is reliable

enough to estimate what it is supposed to do or to be published for everyone to see.

The limitations of this work must be acknowledged. The self-formed categories composed

to compare categorical and continuous variables are merely based on what was thought to

be the most logical way to do this. Hence, a note of caution is due here when interpreting

these results. Also, in this study no control group was investigated. This withholds us

the opportunity to compare the mortality probabilities of the living patients with the deceased

patients’ and therefore we weren’t able to investigate if the severity of disease was greater in

the deceased group.

CONCLUSIONHospitals are compulsory to publish their HSMRs, which gives patients and healthcare

institutions the opportunity to judge and compare hospitals on the basis of this number.

However, we demonstrated that differences in case-mix and the incompleteness of the data

used to calculate the HSMR could negatively influence the HSMR. Although it seems quite

logical to look at the number of deaths in each hospital as an indicator of quality of care, there

are numerous pitfalls hidden in using the HSMR as a quality indicator. Therefore, HSMR should

always be interpreted with caution and openly publishing HSMRs may have unfair negative

consequences for some hospitals.

163

REFERENCES1. Hospital Standardised Mortality Ratio, de Praktijk Index, www.hsmr.nl. Accessed January 3, 2017.

2. Understanding HSMRs. A Toolkit on Hospital Standised Mortality Ratios. http://www.drfoster.com/wp-content/uploads/2014/09/HSMR_Toolkit_Version_9_July_2014.pdf. Published 2014. Accessed December, 2016.

3. van der Laan JdB, A., van den Akker-Ploemacher J, Penning C, Pijpers F. HSMR 2014 methodological report, november 2015. http://www.hsmr.nl/wp-content/uploads/2016/01/2015hsmrmethodologicalreport2014.pdf. Accessed November, 2016.

4. Jarman B, Pieter D, van der Veen AA, et al. The hospital standardised mortality ratio: a powerful tool for Dutch hospitals to assess their quality of care? Qual Saf Health Care. 2010;19(1):9-13.

5. Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med. 1997;336(4):243-250.

6. Vorwerk C, Loryman B, Coats TJ, et al. Prediction of mortality in adult emergency department patients with sepsis. Emerg Med J. 2009;26(4):254-258.

7. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383.

8. Central bureau of statistics (Van der Laan J). Quality of the Dutch Medical Registration (LMR) for the calculation of the Hospital Standardised Mortality Ratio. ISSN 2013; 1572-0314.

9. Tol J, Broekman M, Brauers M, van Gulik T, Busch OR, Gouma DJ. [Reliability of the registration of data on complex patients: effects on the hospital standardised mortality ratio (HSMR) in the Netherlands]. Ned Tijdschr Geneeskd. 2012;156(49):A4918.

10. van den Bosch WF, Spreeuwenberg P, Wagner C. [Hospital standardised mortality ratio (HSMR): adjustment for severity of primary diagnosis can be improved]. Ned Tijdschr Geneeskd. 2011;155(27):A3299.

11. Roest AA, Tegtmeier J, Heyligen JJ, et al. Risk stratification by abbMEDS and CURB-65 in relation to treatment and clinical disposition of the septic patient at the emergency department: a cohort study. BMC Emerg Med. 2015;15:29.

12. Pleizier CM, Geerlings W, Pieter D, Boiten J. Patientmix influences HSMR. Medisch Contact 2010. 36: 1777-1779.

13. Bots ML, Jager-Geurts H, Berger-van Sijl M. Risk of dying after first hospital admission for a cerebrovascular accident in the Netherlands. Cardiovascular disease in the Netherlands. The Hague, 2006.

CHAPTER 11

SUMMARY OF MAIN RESULTS, GENERAL DISCUSSION AND FUTURE PERSPECTIVES

‘We cannot solve our problems with the same thinking we used when we created them’ Albert Einstein

167

This final chapter summarizes and discusses the findings of the work presented in this thesis. In

addition, it presents some future perspectives on patient safety in the acute healthcare chain.

The main aim of this thesis was to obtain insight and provide potential improvement strategies

in safety issues within the healthcare chain using novel ways to investigate these matters.

The acute healthcare chain is increasingly under pressure due to changing demography

(ageing population) and healthcare policy changes. This has led to an increased work

burden for healthcare workers and has negatively affected communication and coordination

within the chain resulting in poor handovers and inadequate teamwork.1-4 It is important

to address these issues since they can eventually result in serious adverse events such as

unplanned readmission, ICU admission, and death potentially jeopardizing the patient safety

and adversely affecting the clinical outcome. In every ten patients, one is still harmed while

receiving hospital care.5 To enhance patient safety, we must change the mindset and culture in

hospitals. In order to do so, creating awareness for patient safety issues in healthcare workers

is needed. In addition, the healthcare workers should be stimulated to take an active role in

these matters with the aim of increasing their intrinsic motivation so that they can actually

contribute to improving the patient safety on the workfloor.6 This thesis aimed to focus on this

awareness by addressing bottlenecks within the acute healthcare chain by working on research

questions originating from daily issues on the floor. The ideas for these projects originated on

the workfloor using healthcare worker point of views aiming to concentrate on ‘real’ systems

rather than ‘ideal’ systems so that our results would be applicable in the real-life settings. In

addition, we hypothesized that the recognition and tackling of these well-known (clinical)

problems would attract the attention of the healthcare workers and thereby help change their

mindset.7 Therefore, the research presented in this thesis was performed on the frontline using

healthcare worker perspectives. In addition, since the patient is in the centre of the healthcare

chain, and the only one who sees and experiences all aspects of it, we deemed it valuable

to incorporate his/her opinion when formulating research and improvement strategies. We

hypothesized that using the perspectives of all stakeholders in the acute healthcare chain to

assess daily safety issues faced in the field would create new insights, awareness and potential

strategies to improve patient safety.

SUMMARY OF MAIN FINDINGSIn chapter 2 we provided a systematic literature overview describing the effects of an

acute medical unit on patient outcomes.8 In addition, we reviewed the current situation in

the Netherlands and made recommendations that could be used when implementing an AMU.

The review demonstrated that current literature proves the AMU to be an effective model to

improve patient flow in the acute care chain. However, the current situation in the Netherlands

showed that existing AMUs were started up individually and are therefore heterogeneously

organized. In order to optimize the effectiveness and compare the quality of care in our AMUs,

a clear national guideline is needed providing a gold standard, especially since the results of

this paper demonstrated that more and more AMUs will be developed in the coming years in

the Netherlands. In addition, we found that to date, very few studies have been performed

168

on the effectivity of the AMUs in the Netherlands. More evidence based research might prove

useful to learn from each other.

In chapter 3, we performed a feasibility study using Patient Reported Outcome Measures

(PROMs) to measure the perceived quality of care and feeling of safety in an AAU.9 The results

of this study showed that the measurement of generic PROMS in the acute admission unit,

with a heterogeneous patient population, is feasible. In general, most patients were positive

about their feeling of safety during their stay on the AAU and regarding the contribution

of the AAU to their recovery. Patients were positive about the questionnaire since they felt

the hospital really wanted to know about their wellbeing, and their perspectives on received

quality and safety of care. They also reported that the follow-up telephone call gave them

the opportunity to share their experiences and reflect on their recent admission. Healthcare

workers benefited from this study too since the PROMs data collection and analysis could

be performed easily making it possible to report the results to the nurses and physicians on

the ward rapidly. This feedback motivated them to keep up their good work and improve

patient care. However, since this was the first (small) study done assessing PROMs in an acute

admission unit future research warrants larger sample sizes to investigate the use of PROMs to

compare hospitals, establish reference values and determine ideal follow-up times.

In chapter 4 and 5 we assessed whether deteriorating patients on clinical wards were

recognized timely by healthcare workers and use of screening instruments such as early

warning scores could predict and prevent serious adverse events in these patients.

PRISMA-medical is an analysis tool which investigates the root causes of adverse events and

incidents in healthcare.10,11 It has shown to be an effective starting point for improvement

strategies in patient safety. In chapter 4 we identified healthcare worker-, organizational-,

technical-, disease- and patient-related causes that contributed to 50 acute unplanned ICU

admissions.12 We found that almost half of the root causes were human- (healthcare worker)

related, predominantly including human monitoring and intervention failures, indicating flaws

in monitoring the patients progress or condition and faulty task planning and performance.

The other half of the root causes were disease-related, comprising the root causes related to

the natural progression of the disease, as could be expected in this severely ill population with

a high mortality rate. We also investigated the recognition of these deteriorating patients on

the clinical wards and found that only in 1% of the vital sets an explicit correct ‘MEWS’ was

reported, although in 43% of the measurements patients had a critical score. This withholds

clinical implications, since it seems essential that the hospital staff in the chain recognize these

critical scores so that the deteriorating patients are detected and treated on time. A proper

implementation of the early warning score described in chapter 5 could prove beneficial. This

study was performed before adequate reimplementation of this protocol, the study described

in chapter 5 was performed afterwards.

In chapter 5 we determined protocol adherence and predictive value of a rapid response

system, with a main focus on the effectuation of the afferent limb, assessed by the use of

169

MEWS on the wards.13 This study exposed that protocol adherence to the MEWS protocol was

good (89%), however one-third of the critical scores were calculated wrongly. Results of this

study indicate that a critical MEWS at 8 a.m. in the morning was a strong predictor of adverse

events associated with poor patient safety: in-hospital mortality, hospital length of stay, cardiac

arrests, ICU-admissions and unplanned 30-day readmissions. This indicates the reliability of

a once a day MEWS measurement as a screening tool: ‘A MEWS in the morning, a very

good warning’.

In chapter 6-10 we evaluated the clinical use and reliability of quality indicators to assess

patient safety. The two quality indicators chosen (unplanned readmissions and Standardized

Hospital Mortality Ratio) are regarded as ‘major adverse events’ and are thought to indicate

the quality of care in Dutch hospitals.14 A great deal of resistance exists in healthcare workers

who find these measures unconnected to their actual work on the floor, too time consuming

and bureaucratic.1,14,15

PRISMA-analysis was used in chapter 6 to create a root causal profile of 50 unplanned

readmissions.16 After composing the root causal trees the two researchers judged if

the readmission could have been prevented. After analysis, half of the readmissions were

deemed potentially preventable. Non-preventable readmissions were mostly the result of

disease-related factors. Healthcare worker-related root causes were solely found in preventable

readmissions, mostly due to human related coordination failures representing failures in task

coordination within a healthcare team, for example no clear medication handover towards

the home situation. This study underlines the importance of communication and coordination

within the chain to potentially prevent adverse events.

In the commentary in chapter 7 we elaborate our perspectives regarding readmissions

and argue why readmission as a quality indicator in its current form cannot be regarded

as a reliable way of assessing quality of care.17 The indicator which is presently used does

not integrate the aspect of preventability and overgeneralizes readmissions not correcting for

hospitals’ case-mix. It remains a struggle for policy makers to compose a reliable indicator since

readmissions are multi-causal events. In addition, the reasons for readmissions are often not

solely related to the course of events in hospital during index admission but to the bottlenecks

in the whole acute care chain.

In chapter 8 we performed an international study to assess the degree of agreement among

physicians regarding predictability and preventability of medical readmissions using a survey

of eight real-life scenarios.18

The results from the 526 participating physicians illustrated that there was moderate to

good agreement among physicians on the predictability of readmissions while agreement on

preventability was poor. This study therefore questions the conclusions derived from current

literature on the basis of physician opinion and advices to interpret these with caution.

170

Chapter 9 describes a multicentre study which included 1398 patients in Europe. During this

study we made an inventory of the opinions of readmitted patients, their carers, nurses, and

physicians regarding the predictability and preventability of readmissions and subsequently used

their opinions to find which factors were associated with their judgements on predictability and

preventability. We demonstrated that (1) consensus among the interviewed on predictability

and preventability was poor (2) factors associated with deemed predictability and preventability

also distinctly differed. However, if a patient reported not feeling ready for discharge, this

was significantly associated with a higher predictability and preventability of readmissions by

the majority. This study has implications for the practice since it emphasizes the importance

of communication during the discharge process. Healthcare workers should question patients

timely about their readiness to go home before discharge in order to potentially reduce

(preventable) readmissions.

In chapter 10 of this thesis we compared the HSMR to clinical disease severity scores in

a cohort of deceased patients with the diagnosis pneumonia. Furthermore, we assessed

the completeness of the data collected by administrators to calculate the HSMR. Results

reveal that risk categories calculated using clinical disease severity scores for pneumonia were

significantly higher than the mortality risk calculated by the HSMR. In addition, we found

the completeness of the data sent to the Central Bureau for Statistics underestimated the actual

comorbidities, and could possibly influence HSMR. Therefore, HSMR should be interpreted

with caution and openly publishing HSMRs may have unfair negative consequences for

some hospitals.

GENERAL DISCUSSION: WHAT CAN WE LEARN FROM THESE RESULTS This thesis has assessed patient safety from a clinical point of view. We aimed to provide an

insight in the whole healthcare chain and have investigated the bottlenecks in the places

where care is transferred from one link to the next in the chain.

System approach

Patient safety is defined as: ‘preventing errors and adverse effects to patients associated with

health care’.5 Results of this thesis have revealed that most adverse events have multifactorial

causes and an easy remedy to prevent them does not exist. The hospital is a high risk

environment and adverse events or incidents occur if when individual barrier weaknesses align

and the system as a whole fails, as was illustrated by the Swiss Cheese model of James Reason

(Figure 1).19,20 This causes “a trajectory of accident opportunity”, in which the hazard passes

through all of the holes in all of the defences, leading to a failure. The complex and risky

environment in healthcare in which no clear cause-effects exists, pleads for a system approach

to improve patient safety. Root cause analysis tools such as the PRISMA-analysis (used in

this thesis) have been found to be useful for such an approach since it aims to investigate

whole system failures of adverse events comprising patient safety by simply asking the ‘why’

question.10,11

171

Soft skills and behaviour

Another lesson learned in this thesis is that even though healthcare is rapidly evolving in this

golden age of technology and digital evolution, this does not automatically guarantee safer

environments. Our work has shown that most adverse events are still the result of human-

(healthcare worker) related failures, namely poor communication and coordination within

the chain, which are often caused by overloaded medical teams and a culture of silence in

hospitals. As the pressure on the healthcare system is not expected to decrease in the coming years, and the ongoing fragmentation in care will result in more handovers, more awareness

for these ‘soft skills’ and behaviour is warranted. Clear handover methods are required to

transfer information adequately and completely.3,21,22 In addition, to change the culture in

the wards all healthcare workers within the chain should feel that they can have a say about

the course of events of a patient’s journey. The major feature of mistakes is that they repeat

themselves and in order to prevent repetition all people involved should be able to ‘stop-the-

chain’ and call out the (potential) harm on the lurk.23

Beneficial safety implementations

Since the first report of EMGO-NIVEL in 2004, patient safety has been an important subject

of attention in the Dutch healthcare system. Despite the fact that not all interventions and

improvement strategies to enhance patient safety have been found useful in clinical practice

and resistance exists among healthcare workers, this thesis has shown that an adequately

implemented monitoring instrument can actually predict and possibly prevent adverse

events.14,22,24 One of our PRISMA-studies (chapter 4) revealed that monitoring patients

adequately is essential to prevent unplanned ICU-admission. Nevertheless, to implement

Track and Trigger Systems sufficiently, proper training of the medical staff is required, as was

Figure 1 | The Swiss cheese model by James Reason19,20

172

illustrated in chapter 4. Nurses and physicians working with these systems must be aware

of the effectiveness of these tools in preventing adverse events so that the adherence to

the guidelines can be improved. Another focal point to improve patient safety is the stimulation

of multidisciplinary work in hospitals to create a more efficient patient flow. The research in

this thesis has shown that the implementation of measures such as acute medical units, which

are designed to improve multidisciplinary teamwork and effectuate rapid diagnostics and

therapy, are an effective and safe method to streamline the healthcare chain. Also, patients on

these units were positive about these wards and felt safe on them.

The reliability of quality indicators

Despite positive effects of the increased attention to safety, this thesis has also exposed not

all of these safety implementations are thought to be beneficial by the healthcare workers in

the field and therefore, lack the support of these essential stakeholders. Due to the increased

focus on safe hospital environments by the government, the Inspectorate of Health and

accreditation bodies (i.e. JCI25), clinicians feel under constant pressure to perform perfectly

and feel they are held personally responsible for the adverse events. This results in a reluctance

to talk openly about potential improvement strategies to prevent these incidents in the future,

and does not set the right example for other healthcare workers (in training) around them.

In addition, they feel that the quality indicators they ought to report on, which are supposed

to measure quality and safety in hospitals, do not actually do so.1,15,26,27 Many clinicians feel

the quality indicators do not focus on clinical actions undertaken during their daily work and

are actually far from real patient outcomes since they often do not integrate the ‘end-users’

during the formulation of these indicators. This thesis has underlined this by demonstrating

that the current quality indicators ‘readmission rate’ and ‘HSMR’ in its current form are

probably unreliable since they do not integrate clinical severity of disease of an individual

patient and often do not correct for case-mix. These quality indicators tend to overgeneralize

and shift the attention from the patient to ‘the system’. This has a negative effect as with

this approach an individual assessment of a patient’s journey becomes less important. All

patients have a unique story which should be addressed in its own way - one size does not

fit all. The research performed on the quality indicators presented in this thesis was mostly

qualitative and small-scale, with the focus on these individual stories, since adverse events and

unsafe situations are fortunately still rare. Therefore, we plead for a cautionary interpretation

of the quality reports on these indicators which are currently open to public.

The added value of stakeholder perspectives

Studies in this thesis have also taught us that integrating the perspectives of patient and

healthcare worker can contribute to new insights. As was hypothesized, their experiences

were found to be valuable resources for identifying their needs and evaluating patient

safety outcomes.5 During our work we gathered the opinions of patients who had and had

not suffered harm, to not solely focus on the caveats, but to also gain knowledge about

the protective factors.1 The added value of incorporating patients’ perspectives is exemplified

by the CURIOS@-study in which we asked the opinions of 1398 patients and their direct

173

carers about their readmission. It has been shown that the extensive amount of research on

readmissions performed to date has mainly focused on measurable ‘risk factors’ while the only

issue which was found to be significantly relevant for the patient was whether he or she was

feeling ready for discharge. This shifts back attention to our social skills and to the soft side of

medicine. Communication in the medical world still seems to be the key, but do we ever really

ask all the relevant questions on the bedside?

Methodologic considerations

Performing research in the acute healthcare chain is not straightforward. During the course of

this trajectory we have faced some methodologic challenges. As we dealt almost exclusively

with a heterogeneous hospital population in our studies and we investigated safety procedures

in only one centre for some of our studies, this might make results less generalizable.

Thinking of a proper study design to answer our research questions was not always easy since

the research performed in this thesis aimed to capture new perspectives on patient safety

and not many studies with validated methodologies on this theme have been performed.

To overcome this, we sought help of the many experts at an early stage to assist us with

drafting an appropriate research protocol. In addition, the recruitment of patients for research

in an acute setting was sometimes challenging. Patients who have just been admitted are

often thrown out by new information after having spoken to multiple therapists. Yet another

person (researcher) at their bedside requesting to participate in a study can sometimes just be

too overwhelming, and therefore these patients might not be willing to grant consent. Also,

admitted patients can be severely ill or cognitively impaired and not suitable for inclusion.

Excluding these groups may result in selection bias. Another consideration when interpreting

of our work is the fact that part of our conclusions are based on retrospective data analysis

of the patient records. This means researchers had to rely solely on information that was

written down in the patient records when investigating causes for adverse events. This might

have underestimated the failures caused by cultural or organizational behaviours which are

often not presented in written charts. Furthermore, the limitations in our multicentre studies

should be noted. The physicians answering our survey of adapted real-life cases reflected that

they were missing information that would allow them to assess the case more thoroughly,

such as the patients’ social situation. A major limitation of the multicentre readmission study

performed was the potentially (recall)bias and subjectivity. In addition, a newly composed

variable ‘majority opinion’ was used as the gold standard to decide on preventability and

predictability. One might argue that this is questionable.

FUTURE PERSPECTIVESStart with facing the facts, valuable leadership from bottom-up

In order to enhance patient safety we must start with the end-users in the field. The hospital

remains an environment in which ‘humans’ work, and unfortunately, humans make errors.

Nevertheless, these humans almost always do work with best intentions. Yet, they sometimes

do not get the best results, mainly because the system they work in often does not allow them

174

to do so.28 The Berwick-committee has underlined this in their report ‘crossing the quality’

chasm which was a reaction to the report “To err is human”. In this report they stated that

simply trying harder does not work: a redesign of the healthcare system is needed.29,30 This

report focused on chief executives in healthcare to motivate them to incorporate new best

practices. This implementation is not done overnight however, it takes an average of 17 years

to broadly spread best practices in healthcare. These findings are especially important for

hospital directors, who carry the final responsibility for a proper safety system in hospitals and

should disseminate it adequately. It is known that poor leadership is one of the most significant

factors contributing to the occurrence of incidents and adverse events in healthcare.31

Therefore, these medical leaders should know and understand the field to design the most

suitable safety system and to address current bottlenecks. A way to do this is by making

the leaders ‘walk the talk’, an initiative by dr. Kaplan, an internist and the CEO of the Virginia

Medical Centre. Directors should literally walk through the hospital to see its daily functioning,

in order to show concern and be visible on the work floor. This initiative has proven to be

effective in the University Medical Centre of Utrecht.1 The concept could bridge the hierarchic

gap between the medical staff on the work floor and the hospital directors.23 In addition,

the dialogue created is the start of a bottom-up approach in contrast with the top-down

approach most healthcare workers presently endure. Conversing on a micro level will create

new insights into current obstacles and potential improvement strategies.1

Adapt approach to safety, from safety 1 to safety 2

Safety implementations have improved patient safety in Dutch hospitals.32 However, due to

increased attention to this subject, patients and healthcare workers actually feel less safe –

this is called the safety paradox.33 Even though we have improved in patient safety matters,

we are not ‘done’ and cannot sit still, as was underlined by the work in this thesis. Firstly, it

Figure 2 | Our perspectives on improvement strategies in patient safety

175

must be emphasized that the improvement of patient safety is an ongoing cycle, which needs

constant attention and evaluation (Figure 3).5 Secondly, the coming years will probably be

even more challenging considering patients will only become older and more frail, technology

more advanced and working together more complicated.32 Considering increasing demands

and growing system complexity, we must therefore adjust our approach to patient safety.34

The current approach to patient safety, labelled safety-1, is established on the ‘find and fix’

model. This causes a linear model of cause and effect instead of the preferred system approach

we have discussed earlier. It is essential to learn from the far more frequent cases where

things go right and develop ways to support, augment and encourage these. In order to

find out what does work, a shift to a proactive safety-2 management could prove helpful in

the future.1,7,34 It concentrates on how everyday performance usually succeeds rather than

on why it occasionally fails, and actively strives to improve the former rather than simply

preventing the latter. In this way we will prevent degrading the resources and procedures

needed to make things go right.7,34

Face the hierarchic culture, from extrinsic to intrinsic motivation

An integrated, positive and transparent culture on wards is an essential factor to improve

patient safety. The culture on hospital wards is still mostly created by the medical specialists

working on these wards. Even though hierarchical gaps have closed over the past decades,

medical specialists are still at the top of the medical hierarchy and are seen as central role

players. Yet, specialists often do not feel they are part of the hospitals organization since

a mismatch/gap between hospital directors and medical specialists exits.1 In most specialisms

a culture of silence and not addressing colleagues exists. This prevents medical staff from

learning from each other and speaking openly about their doubts and (almost) failures. When

an incident does occur and everyone starts talking about this, the potentially responsible

Figure 3 | Patient safety cycle, World Health Organization 20035

176

healthcare worker may feel that he is the only one who is doing something wrong, resulting

in a ‘blame and shame’ culture. A culture change providing attitude changes towards patient

safety in professionals is needed to increase self-reflection and shift the attention on patient

safety from extrinsic to intrinsic motivation. Intrinsic motivation entails activities which are

undertaken because of internal motivation and provide immediate effect after the action

itself whereas extrinsic motivation makes people undertake actions to get a reward or to

dodge a penalty from an external party without being interested in the immediate potential

effect. An organization-wide reduction of adverse events can be achieved by high levels of

ward’s shared values, beliefs and behaviours plus an individual’s perception of contributing

to the culture.6 Soft skills such as communication and team training for non-technical skills

within the healthcare chain are key focal points. To improve these behavioural skills and to

change the culture, education should be provided from the early onset of a medical career.

Unfortunately, the time and resources reserved for such trainings is increasingly under pressure

in the last decades because of budget cuts. The tendency is to choose efficiency and high

production over investment in quality for the future.

E-health: Match supply and demand. Which way to go?

Due to new technologies more care can be delivered at home in the future. E-health is the use

of technology to support or improve a patient’s health and the healthcare system. It offers

patients the ability to direct and self-manage their own health. The use of e-consultation,

virtual monitoring systems and therapy at home (i.e. chemotherapy) might be beneficial

to relieve the pressure in hospitals and to make patients more aware of their own health

situation. A massive amount of novel technologies is released daily which state they could

potentially enhance the quality of care. A problem with these new inventions is that they are

often created by an external company with commercial intentions and look good on paper,

but it is known that they have not all been designed using insights from the frontline. It is

necessary for healthcare institutions to work together to match these inventions to the actual

needs on the floor. In addition, since these strategies might decrease the hours in hospital,

insurance companies and hospitals should think of novel ways to pay for this care. Finally, we

must not forget that even though these new flashy methods seem beneficial, they do not

substitute the need for proper communicational skills which are still essential for safe care.

Also, less modern inventions such as a consultation on the telephone and proper discharge

letters should not be left out in seeking for alternative care methods.35

Rely on patient’s journeys: Talk with them, not about them

Research has indicated that better patient experiences are associated with better clinical

outcomes, less healthcare utilization and better patient safety within hospitals.36 However,

current patient experience has shown little association with hospital quality management

strategies.37 To match supply and demand, more patient perspectives could be helpful. As

was highlighted in this thesis, patients are often willing to participate, however, barriers do

still exist. The traditional hierarchy between patient and doctors still makes some patients

feel subordinate to their clinician. This is an important barrier to their involvement in error

177

reduction.38 In addition, patients might feel they are being labelled ‘difficult’ and therefore

choose to take a more passive role in protecting their own personal safety. To enhance their

active role, frameworks have been developed by error victims to describe general courses of

action by which patients can contribute to their safety and increase resilience. These include:

1) informing the management plan by sharing information with clinicians and asking questions

about treatment decisions, 2) monitoring and ensuring safe delivery of treatment for example

by self-administration of medication and 3) informing systems improvement for example by

providing feedback on care quality.39

Another helpful tool that could be constructive to investigate patient experiences is

the usage of ‘patient tracers’, which are originally used by consultants and auditors from

the JCI (Joint Commission International).1,25 In this method the travelled path of a patient is

followed to stepwise investigate and reconstruct what would happen in an ideal situation,

who would carry responsibility and what would be needed (structure) to provide optimal

quality of care. The principal ‘tell me, show me, and take a good look’ is used to evaluate

the course through the healthcare chain. A tracer looks into the whole system and the process

(the healthcare chain) in contrast to individual acting and enhances the ‘safety 2’ approach.7

In the Netherlands, the nationwide network for acute care has recently reported and tracked

11 patient journeys in the acute chain. An executive from an insurance company and a patient

(representative) have successfully worked together on this assessment, and have found areas for

potential improvement.40

Personal reflection

On a final note, I would like to elucidate my personal reflection on the results of this thesis.

During the course of this PhD trajectory I have learned a lot about research practices, patient

safety and scientific writing. In my opinion, the thesis presented assesses safety issues from

a new perspective, addressing it in a more practical way as it was assessed by ‘insiders’. Some

might argue that results from our studies are logical findings, and causes for adverse events

are widely known by all healthcare workers. However, we noticed that our new and ‘fresh’

research strategies in combination with the quick feedback of results to healthcare workers

created an unexpected Hawthorne effect.42 This effect, which is also referred to as the observer

effect, is a type of reactivity in which individuals modify an aspect of their behaviour in response

of their awareness of being observed (participating in our research). During the research

performed, awareness was created in healthcare workers to actually actively think and reflect

with us about patient safety matters and the effectiveness of current safety implementations

and quality indicators. The interest in this research field was also underlined by the great

number of young medical students and trainees that were motivated to help us to perform our

studies. The attention and time they spent on participating patients was greatly appreciated,

and during their work they not only learned research basics but also gained knowledge and

awareness of the relevance of patient safety. Hopefully, this thesis will function as a starting

point to improve patient safety in the overloaded acute healthcare chain by initiating the shift

of extrinsic motivation to the required intrinsic motivation in the healthcare system.

178

REFERENCES1. Dillman R, Wagner C, Schellekens W, Klein J, Jager M, van der Grinten T. Patientveiligheid. Utrecht:

de Tijdstroom uitgeverij. 2016

2. Arora VM, Farnan JM. Inpatient Service Change: Safety or Selection? JAMA. 2016;316(21):2193-2194.

3. Greenstein EA, Arora VM, Staisiunas PG, Banerjee SS, Farnan JM. Characterising physician listening behaviour during hospitalist handoffs using the HEAR checklist. BMJ Qual Saf. 2013;22(3):203-209.

4. Pannick S, Beveridge I, Wachter RM, Sevdalis N. Improving the quality and safety of care on the medical ward: A review and synthesis of the evidence base. Eur J Intern Med. 2014;25(10):874-887.

5. World Health Organization. Patient Safety. http://www.euro.who.int/en/health-topics/Health-systems/patient-safety/patient-safety. Accessed January 3, 2017.

6. Mardon RE, Khanna K, Sorra J, Dyer N, Famolaro T. Exploring relationships between hospital patient safety culture and adverse events. J Patient Saf. 2010;6(4):226-232.

7. Braithwaite J, Wears RL, Hollnagel E. Resilient health care: turning patient safety on its head. Int J Qual Health Care. 2015;27(5):418-420.

8. van Galen LS, Lammers EM, Schoonmade LJ, Alam N, Kramer MH, Nanayakkara PW. Acute medical units: The way to go? A literature review. Eur J Intern Med. 2016.

9. Galen LS, der Schors W, Damen NL, Kramer M, Wagner C, Nanayakkara P. Measurement of generic patient reported outcome measures (PROMs) in an acute admission unit: A feasibility study. Acute medicine. 2016;15(1):13-19.

10. van Vuuren W SC, van der Schaaf TW. The Development of an Incident Analysis Tool For the Medical Field. Eindhoven University of Technology: Eindhoven. 1997.

11. van der Schaaf TW HM. PRISM-Medical. A Brief Description. Eindhoven University of Technology, Faculty of Technology Management, Patient Safety Systems: Eindhoven. 2005.

12. van Galen LS, Struik PW, Driesen BE, et al. Delayed Recognition of Deterioration of Patients in General Wards Is Mostly Caused by Human Related Monitoring Failures: A Root Cause Analysis of Unplanned ICU Admissions. PLoS One. 2016;11(8):e0161393.

13. van Galen LS, Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. PLoS One. 2016;11(8):e0160811.

14. Basisset Kwaliteitsindicatoren 2014. Utrecht: Inspectie voor de gezondheidszorg, 2015. Accessed January, 2016.

15. Poortmans J. Onderzoek - Zorgkwaliteit in ziekenhuizen. Driemaaldaags een pijnschaal. De Groene Amsterdammer. 2016(49).

16. Fluitman KS, van Galen LS, Merten H, et al. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Intern Med. 2016;30:18-24.

17. van Galen LS, Nanayakkara PW. [Hospital readmissions: A reliable quality indicator?]. Ned Tijdschr Geneeskd. 2016;160:A9885.

18. van Galen LS, Cooksley T, Merten H, et al. Physician consensus on preventability and predictability of readmissions based on standard case scenarios. Neth J Med. 2016;74(10):434-442.

19. Reason J. Human error: models and management. BMJ. 2000;320(7237):768-770.

20. Reason J. James Reason: patient safety, human error, and Swiss cheese. Interview by Karolina Peltomaa and Duncan Neuhauser. Qual Manag Health Care. 2012;21(1):59-63.

21. O’Leary KJ, Buck R, Fligiel HM, et al. Structured interdisciplinary rounds in a medical teaching unit: improving patient safety. Arch Intern Med. 2011;171(7):678-684.

179

22. Veiligheidsmanagementsysteem. Communicatie tussen hulpverleners volgens het SBAR-proces. VMS Veiligheidsprogramma. 02-2009.

23. Interview with Dr. Gary S. Kaplan: Determined Steps to Transformation. http://ethix.org/2011/01/11/dr-gary-s-kaplan-determined-steps-to-transformation. Published 2011. Accessed February 16, 2017.

24. Langelaan M, de Bruijne M, Baines R, et al. NIVEL Monitor Zorggerelateerde Schade 2011/2012. Dossieronderzoek in Nederlandse ziekenhuizen. http://www.nivel.nl/sites/default/files/bestanden/monitor_zorggerelateerde_schade_2011_2012.pdf. Published 2013, Accessed December 28, 2016.

25. The Joint Comission. https://www.jointcommission.org/performance_measurement.aspx. Accessed February 2, 2017.

26. Ikkersheim D, van der Avoort J. KPMG Plexus. Onderzoek kosten kwaliteitsmetingen Nederlandse Vereniging van Ziekenhuizen (NVZ). https://www.nvz-ziekenhuizen.nl/_library/31906 Published 2015. Accessed January 10, 2017.

27. Botje D, Ten Asbroek G, Plochg T, et al. Are performance indicators used for hospital quality management: a qualitative interview study amongst health professionals and quality managers in The Netherlands. BMC Health Serv Res. 2016;16(1):574.

28. Holusha J. W. Edwards Deming, Expert on Business Management, Dies at 93. http://www.nytimes.com/1993/12/21/obituaries/w-edwards-deming-expert-on-business-management-dies-at-93.html?pagewanted=all. Accessed February 16, 2017.

29. Berwick DM. A user’s manual for the IOM’s ‘Quality Chasm’ report. Health Aff (Millwood). 2002;21(3):80-90.

30. Institute of Medicine. To err is human: building a safer health system. http://www.nationalacademies.org/hmd/~/media/Files/Report%20Files/1999/To-Err-is-Human/To%20Err%20is%20Human%201999%20%20report%20brief.pdf. Published November 1999. Accessed December 28, 2016.

31. Leisikov I. Patientveiligheid, de rol van de bestuurder (PhD thesis TU Delft). 2010. Accessed February 15, 2017.

32. Veiligheid Nederlandse ziekenhuizen aanzienlijk verbeterd. NFU. http://www.nfu.nl/actueel/veiligheid-nederlandse-ziekenhuizen-aanzienlijk-verbeterd. Published 2013. Accessed February 17, 2017.

33. Boutelier H. Boom Lemma Uitgevers. De veiligheidsupopie. Hedendaags onbehagen en verlangen rond misdaad en straf. 2005.

34. Hollnagel E, Wears RL, Braithwaite J. From Safety-I to Safety-II: A White Paper. https://www.england.nhs.uk/signuptosafety/wp-content/uploads/sites/16/2015/10/safety-1-safety-2-whte-papr.pdf. Published 2015, Accessed January 10, 2017.

35. Car J, Sheikh A. Telephone consultations. BMJ. 2003;326(7396):966-969.

36. Anhang Price R, Elliott MN, Zaslavsky AM, et al. Examining the role of patient experience surveys in measuring health care quality. Med Care Res Rev. 2014;71(5):522-554.

37. Groene O, Arah OA, Klazinga NS, et al. Patient experience shows little relationship with hospital quality management strategies. PloS one. 2015;10(7):e0131805.

38. Doherty C, Stavropoulou C. Patients’ willingness and ability to participate actively in the reduction of clinical errors: a systematic literature review. Soc Sci Med. 2012;75(2):257-263.

39. Peat M, Entwistle V, Hall J, Birks Y, Golder S. Scoping review and approach to appraisal of interventions intended to involve patients in patient safety. J Health Serv Res Policy. 2010;15 Suppl 1:17-25.

40. Landelijk Netwerk Acute Zorg. Uitkomsten bespreking patient journeys. Samenvattend overzicht. Published December 2016. Accessed February 17, 2017.

CHAPTER 12

DUTCH SUMMARY – NEDERLANDSE SAMENVATTING PATIËNTVEILIGHEID IN DE ACUTE ZORGKETEN:

“IS HET THUIS VEILIGER?”

‘The art of medicine consists of amusing the patient while nature cures the disease’ Voltaire

183

PATIËNTVEILIGHEID IN DE ZORG Dit proefschrift gaat over het inzichtelijk maken en verbeteren van patiëntveiligheid

in Nederlandse en Europese ziekenhuizen. Het onderzoek in dit proefschrift beschrijft

nieuwe inzichten en potentiele verbeterstrategieën ten behoeve van veiligheidskwesties in

de acute zorgketen.

Door toenemende vergrijzing , de marktwerking en aanhoudende bezuinigingen in de zorg

neemt de werkdruk op ziekenhuizen en zorgverleners toe. Ondanks een afname van

spoedeisende hulp presentaties in de laatste jaren, is er een toename van het aantal opnames.

Hierdoor zijn ziekenhuizen vaak vol en loopt de patiëntenstroom op de spoedeisende hulp

dagelijks vast. Met name de toename van stroom van oudere patiënten zorgt voor opstopping

omdat deze meer en vaak langere toegepaste zorg vergt door hun multimorbiditeit, polyfarmacie

en atypische presentatie van klachten. Huisartsen trachten acute patiëntenstromen indien

mogelijk af te wenden uit het ziekenhuis door patiënt zelf te behandelen, op te nemen in

een eerstelijns verblijf of te verwijzen voor een poliklinisch consult, zij functioneren als een

poortwachter. Helaas zijn door de toename van sociale hulpvragen van met name oudere

patiënten (‘het gaat niet meer thuis’) in combinatie met het terugschroeven van budget voor

verpleegtehuizen ook de huisartsen in de zorgketen vaak overbelast.

Communicatie in en coördinatie binnen de zorgketen zijn essentieel voor de stroomlijning

ervan, echter komen deze factoren door de huidige belasting op het zorgstelsel in het geding.

Hierdoor komt patiëntveiligheid potentieel in gevaar en kunnen ‘Serious Adverse Events’

(ernstige ongewenste voorvallen) zoals ongeplande heropname, onverwachte intensive care

opname of in het ergste geval overlijden het gevolg zijn. Patiëntveiligheid wordt gedefinieerd

als: het vermijden, voorkomen, en verminderen van ongewenste uitkomsten of lichamelijk

letsel ten gevolge van het zorgproces. Ondanks toegenomen aandacht voor en projecten ter

bevordering van patiëntveiligheid komen deze ongewenste uitkomsten nog steeds te vaak

voor. Zo krijgt 1 op elke 10 patiënten te maken met ongewenste schade tijdens zijn traject

door de zorgketen.

Om de patiëntveiligheid te bevorderen is er een cultuurverandering nodig in ziekenhuizen.

Zorgverleners en bestuurders dienen zich bewust te zijn van hun rol in het systeem, en

verantwoordelijkheid te nemen ten behoeve van de veiligheid van de patiënt. Door de grote

administratieve last van kwaliteitsindicatoren van externe partijen hebben zorgverleners het

gevoel dat het daadwerkelijke probleem op de vloer niet goed in kaart wordt gebracht, dan

wel gemeten. De motivatie voor het goed uitvoeren van zorg lijkt nu vooral ‘het aankruisen

van extern opgelegde hokjes’ in plaats van intrinsiek willen dat de patiënt daadwerkelijk in een

veiligere omgeving zorg krijgt.

De belangrijkste speler binnen patiëntveiligheid blijft uiteraard de patiënt. Te vaak echter,

wordt er in onderzoek en hervorming van de zorg, niet de opinie gevraagd van deze essentiële

stakeholder. De patiënt blijft de enige persoon die het traject van begin tot einde in diepgang

184

meemaakt, en hij zal dan ook waardevolle inzichten hebben ten behoeve van de verbetering

van zorg.

In dit proefschrift zijn de inzichten gebruikt van de stakeholders die op de frontlinie betrokken

zijn bij (bijna) incidenten in het ziekenhuis, met een grote focus op de patiënt en zijn

zorgverleners. Onze hypothese was dat het gebruik van deze perspectieven zou leiden tot

nieuwe inzichten, verbeterstrategieën, en tevens meer bewustwording onder zorgverleners

zou creëren.

DE START VAN DE ACUTE ZORGKETEN: NIEUWE MANIEREN VAN ZORG ORGANISEREN In het eerste deel van dit proefschrift bestudeerden wij nieuwe manieren om het eerste deel van

de acute zorgketen voor patiënten effectiever en veiliger te maken. In de afgelopen decennia

zijn er verschuivingen geweest in het originele model van opname via de spoedeisende hulp

naar (gewone) klinische afdelingen. Acute opname afdelingen (AOA), bestaan in Nederland

sinds 2000. Het zijn afdelingen die bemand en uitgerust zijn om patiënten op te nemen met

een acuut medische aandoening voor een snelle multidisciplinaire en medisch specialistische

evaluatie. De zorg op deze afdeling is voor een bepaalde periode (meestal tussen 24 en 72

uur) waarna patiënten naar huis worden ontslagen, dan wel naar gewone klinische afdelingen

worden overgeplaatst. In hoofdstuk 2 beschreven wij in een systematisch literatuuroverzicht

de positieve effecten van het implementeren van een AOA op patiënt uitkomsten zoals

opnameduur en patiënttevredenheid. Ook gaven wij de huidige stand van zaken in Nederland

weer en deden wij een aanbeveling voor een nationale richtlijn. In hoofdstuk 3 werden

patiënten gedurende hun opname op de AOA middels gevalideerde vragenlijsten (PROMS:

Patient Reported Outcome Measures) gevraagd naar hun ervaringen en gevoel van veiligheid.

Patiënten op deze afdeling voelden zich veilig en de resultaten van het onderzoek konden

direct teruggekoppeld worden naar de verpleegkundigen en artsen op de werkvloer, wat hen

motiveerden deze goed gewaardeerde zorg te blijven leveren.

VERMIJDEN VAN ONGEWENSTE VOORVALLEN OP DE KLINISCHE AFDELINGEN Ondanks het feit dat niet alle geïmplementeerde interventies en strategieën voor het verbeteren

van patiëntveiligheid gebruiksvriendelijk en bijdragend worden bevonden in de klinische

praktijk, toont dit proefschrift dat een adequaat ingevoerd instrument daadwerkelijk kan

bijdragen aan het voorspellen en mogelijk voorkomen van ongewenste uitkomsten. In

hoofdstuk 4 identificeerden wij factoren die bijdroegen aan 50 ongeplande Intensive Care

opnames. Middels een PRISMA-analyse werd gevonden dat voornamelijk menselijk gerelateerde

monitoringsfouten een oorzaak waren voor deze ongewenste uitkomsten. Doordat patiënten

op de klinische afdelingen niet adequaat werden bewaakt, was er vaak te laat herkenning

van hun achteruitgang, waardoor overplaatsing naar de Intensive Care noodzakelijk was.

In hoofdstuk 5 toonden wij het effect aan van een adequaat geïmplementeerde ‘Early

185

Warning Score’, een score die is ontwikkeld om deze patiënten tijdig op te sporen om verdere

achteruitgang te voorkomen. Deze tool zorgt dat er bij een afwijkende score tijdig een arts

wordt gewaarschuwd, waardoor eerder de juiste zorg wordt geleverd. Een afwijkende score

was in de heterogene ziekenhuispopulatie een goede voorspeller van ongewenste uitkomsten,

wat de betrouwbaarheid van dit instrument benadrukt. 89% van de meldingen werd volgens

protocol uitgevoerd, desalniettemin werd nog steeds een aanzienlijk deel van kritieke scores

gemist, hier is nog ruimte voor verbetering.

HET GEBRUIK VAN KWALITEITSINDICATOREN OM PATIËNTVEILIGHEID TE METEN EN WAARBORGEN In Nederland zijn er momenteel ongeveer 3400 kwaliteitsindicatoren in de gezondheidszorg.

Er is veel weerstand onder zorgverleners over deze indicatoren: ze staan niet in verbinding met

het echte werk op de vloer, zijn tijdsintensief, en te administratief. Het percentage heropnames

binnen 30 dagen na een eerdere klinische opname in hetzelfde ziekenhuis is samen met een

onverwacht lange opnameduur en overlijden in het ziekenhuis, één van de ‘major adverse

events’ die gelden als indicator voor een negatieve uitkomst van klinische ziekenhuiszorg in

Nederland. In de Verenigde Staten en het Verenigd Koninkrijk is het percentage heropnames al

enige jaren één van de indicatoren voor de kwaliteit en veiligheid van ziekenhuizen. Een hoog

percentage heropnames kan in deze landen resulteren in een geldboete.

Dit proefschrift trekt het huidige gebruik van kwaliteitsindicatoren ‘heropnames’ en ‘overlijden

in het ziekenhuis’ in twijfel, en beargumenteert dat de gepresenteerde cijfers van deze

indicatoren vaak geen betrouwbare afspiegeling zijn van de daadwerkelijke veiligheid en

kwaliteit van geleverde zorg in ziekenhuizen.

Middels een PRISMA-analyse in hoofdstuk 6 maakten wij van 50 heropnames een oorzaken-

risicoprofiel. Dit onderzoek toonde aan dat vermijdbare heropnames met name veroorzaakt

worden door menselijke coördinatie-fouten, bijvoorbeeld geen goede medicatieoverdracht naar

de thuissituatie. Deze studie benadrukte het belang van adequate en efficiënte communicatie

en coördinatie in de zorgketen om zo mogelijk vermijdbare heropnames te voorkomen. Het

opiniestuk in hoofdstuk 7 becommentarieert waarom de heropname in zijn huidige vorm geen

betrouwbare indicator is van kwaliteit van zorg. Het waarheidsgetrouw vormgeven van deze

indicator is een opgave, want heropnames hebben vaak meerdere oorzaken, binnen en buiten

(de macht van) het ziekenhuis. De huidige indicator houdt geen rekening met vermijdbaarheid,

en generaliseert ziekenhuizen door niet te corrigeren voor hun ziektezwaarte. In hoofdstuk 8

hebben wij gekeken of medisch specialisten (in opleiding) het met elkaar eens zijn over de

vermijdbaarheid van heropnames. Dit is gedaan omdat in de huidige literatuur conclusies

met name berust op de opinie van een klein aantal artsen. Dit onderzoek liet zien dat de

526 artsen die deelnamen het niet eens waren over de vermijdbaarheid van de heropname.

Er was wel consensus over de patiënten die een hogere kans hadden om heropgenomen te

worden. Dit onderzoek geeft aan dat huidige literatuur met een kritisch oog bekeken moet

worden en dat vermijdbaarheid opnemen in de kwaliteitsindicator ‘heropname’ een lastige

186

taak is. Relatief de meeste onderzoekuren in dit proefschrift werd verricht in onderzoek dat

wordt gepresenteerd in hoofdstuk 9. In dit onderzoek vroegen wij 1398 patiënten in 15

Europese ziekenhuizen gedurende hun heropname naar de reden van hun heropname. Ook

vroegen wij een mantelzorger, verpleegkundige en arts van het behandelteam naar hun

perspectieven betreffende dit ongewenste incident. Het onderzoek concludeerde dat met name

de communicatie tussen arts en patiënt tijdens ontslag erg belangrijk was. Als patiënten zich

niet klaar voelden voor ontslag gedurende hun eerste opname, was de kans significant groter

dat zij terugkwamen en dat dit vermijdbaar was dan als zij wel klaar waren. Dit onderzoek

plaats een kritische noot bij de veelvuldige heropname-predictiemodellen. Deze modellen

nemen statische factoren mee om een heropname te voorspellen, dit onderzoek toonde aan

dat met name ‘zachte factoren’ zoals communicatie en empathie belangrijk zijn.

De ‘Hospital Standardized Mortality Ratio’, is een ratio die de werkelijke sterfte in een

ziekenhuis afzet tegen de sterfte die op basis van de patiëntkenmerken wordt verwacht. Het

fungeert als indicator voor potentieel vermijdbare sterfte. In hoofdstuk 10 zetten wij deze

openbaar gepubliceerde cijfers in een ander perspectief. De studie, uitgevoerd in 32 patiënten

overleden aan een longontsteking, demonstreerde dat de risico scoren die klinisch gebruikt

worden om ziekte-ernst van de patiënten met longontsteking te beoordelen significant hogere

sterftecijfers gaven dan de sterftecijfers die werden berekend door de HSMR. Daarnaast

onthulde deze studie dat de volledigheid van gegevens die naar het Centraal Bureau voor

Statistiek gaan de werkelijke comorbiditeit onderschatte. Deze bevindingen kunnen de HSMR

negatief beïnvloeden, en de vraag is dan ook of deze openbaar gepubliceerde cijfers een

werkelijke weergave zijn van de kwaliteit van zorg.

WAT KUNNEN WIJ MET DEZE RESULTATEN? Dit proefschrift heeft vanuit een klinisch perspectief de knelpunten in de acute zorgketen

benaderd. De resultaten hebben ons het volgende geleerd:

→ Het onderzoeken van patiëntveiligheid vereist een systeembenadering in plaats

van een persoon- of ‘schakel’-benadering. Een ongewenste uitkomst vindt vaak

pas plaats als een aaneenschakeling van fouten in hetzelfde traject, in dit geval

de keten, plaatsvindt.

→ De meeste ongewenste uitkomsten vinden nog steeds plaats door foutief menselijk

handelen, met name inadequate en inefficiënte coördinatie en communicatie in

de zorgketen zijn belangrijke factoren voor incidenten. Veilige en gestructureerde

overdracht is nodig om kwaliteit van zorg te waarborgen.

→ Iedereen in de zorgketen moet zich kunnen uitspreken over een eventuele ‘bijna’

fout en de mogelijke consequenties. Dit betekent dat de hiërarchische structuur

plaats moet maken voor gelijkwaardig hoor en wederhoor onder zorgverleners en

ondersteunende functies in de zorgketen.

→ Om veiligheidsstrategieën in een zorgsysteem goed te implementeren is

goede training en besef van het belang van deze instrumenten in de dagelijkse

praktijk nodig.

187

→ Huidige kwaliteitsindicatoren kunnen een goed middel zijn om verder intern

onderzoek te verrichten als cijfers afwijken van de norm. Zij moeten echter niet

dienen als openbaar vergelijkingsmiddel, met name omdat de meeste indicatoren

geen betrouwbare weerspiegeling van de daadwerkelijke kwaliteit en veiligheid is.

→ Het meenemen van de perspectieven van alle stakeholders in de zorgketen

zorgt voor nieuwe inzichten en potentiele verbeterslagen ten behoeve van

patiëntveiligheid. Daarnaast creëert het betrekken van zorgverleners awareness en

intrinsieke motivatie voor manieren om veiligheid daadwerkelijk te verbeteren.

→ De ‘zachte’ kanten (zoals communicatie en empathie) van menselijk handen zijn

vaak lastig te meten, deze eigenschappen zijn echter vaak doorslaggevend voor

veilige en patiëntvriendelijke zorg.

HET BEVORDEREN VAN PATIËNTVEILIGHEID IN DE ACUTE ZORGKETEN IN DE TOEKOMST In de afsluitende woorden van dit proefschrift doen wij een aanbeveling voor het bevorderen

van patiëntveiligheid in de toekomst. Enkele kernpunten:

→ De medische leiders (bestuurders, managers) in het ziekenhuis moeten de feiten

onder ogen zien, zij zijn immers eindverantwoordelijk in een ziekenhuis. Dit betekent

dat zij de werkvloer op moeten om in gesprek te gaan met hun werknemers en

patiënten om zo te kunnen zien waar verbetering nodig is. Ziekenhuizen moeten

een bottom-up benadering kiezen, in plaats van een top-down.

→ Door toenemende huidige focus op patiëntveiligheid is er sprake van een

veiligheidsparadox: wij voelen ons minder veilig. Daarom zou veiligheid meer

bestudeerd moeten worden vanuit de momenten dat het wel goed gaat, in plaats

van die ene keer dat het (bijna) mis ging. Dit zorgt voor een positievere benadering

van het probleem en omarmt beschermende factoren.

→ Door de ouderwetse hiërarchische structuur in ziekenhuizen wordt cultuur op

afdelingen vaak nog bepaalt door medisch specialisten. Hierdoor worden mensen

vaak niet aangesproken op normen, waarden en gedrag die patiëntveiligheid

potentieel in gevaar kunnen brengen. Communicatie binnen het team en niet-

technische skills moeten getraind worden in teamverband om deze ouderwetse

cultuur te doorbreken.

→ Zorgverleners moeten intrinsiek gemotiveerd zijn om bij te dragen aan

patiëntveiligheid zodat zij meteen het effect van hun handeling kunnen zien. Dit in

tegenstelling tot de nu vaak aanwezige extrinsiek motivatie: mogelijke beloningen

en boetes door overkoepelende instanties die bepaalde indicatoren behartigen.

→ Technologische innovaties om zorg meer buiten het ziekenhuis te laten

plaatsvinden zullen in de komende jaren zijn opmars maken. Deze zien er op

papier vaak veelbelovend uit, echter zijn ze vaak niet ontworpen met inzichten van

de zorgverleners op de vloer. Daarnaast kunnen deze E-health mogelijkheden niet

de zachte kanten, zoals communicatie en empathie overnemen. Het gebruik zal in

de praktijk door de eindgebruikers getest en gevalideerd moeten worden.

188

→ Positieve patiëntervaringen leiden tot betere uitkomsten. Om kwaliteit

daadwerkelijk te meten, moeten wij de input van patiënten gebruiken om in te zien

wat zij belangrijk vinden. Een manier om dit te doen is bijvoorbeeld door middel

van ‘patiënt tracers’ waarin samen met de patiënt het gehele pad (patient journey)

van ervaringen door de zorgketen wordt doorlopen.

APPENDIX

LIST OF PUBLICATIONS LIST OF SCIENTIFIC PRESENTATIONS

AUTHOR AFFILIATIONS WORD OF THANKS

CURRICULUM VITAE/BIOGRAPHY

193

LIST OF PUBLICATIONS 2015

– Cooksley T, Nanayakkara PW, Nickel CH, Subbe CP, Kellett J, Kidney R, van Galen LS et al. (7-2015) Readmissions of medical patients: an external validation of two existing prediction scores. QJM

– Arntzenius AB, van Galen LS. (11-2015) Budesonide related adrenal insufficiency. BMJ Case reports

2016 – Fluitman KS & van Galen LS, Merten H, Cooksley T, Nanayakkara PW, Kramer MH et al. (1-2016)

Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Eur J Intern Med.

– van Galen LS, Nanayakkara PW. (1-2016) Hospital readmissions: A reliable quality indicator? Ned Tijdschr Geneeskd.

– van Galen LS, van der Schors W, Damen NL, Kramer MH, Wagner C, Nanayakkara PW. (3-2016) Measurement of generic patient reported outcome measures (PROMS) in an acute admission unit: A feasibility study. Acute Med.

– van Galen LS & Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. (8-2016) A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. PLoS One

– van Galen LS & Struik PW, Driesen BE, Merten H, Ludikhuize J, van der Spoel JI, Kramer MH, Nanayakkara PW. (8-2016) Delayed recognition of deterioration of patients in general wards is mostly caused by human related monitoring failures: A root cause analysis of unplanned ICU admissions. PLoS One

– van Galen LS, Cooksley T, Merten H, Brabrand M, Terwee CB, Nickel CH et al. (10-2016) Physician Consensus on preventability and predictability of Readmissions based on standard case scenarios. Neth J med.

– van Galen LS, Lammers EM, Schoonmade LJ, Alam N, Kramer MH, Nanayakkara PW (11-2016) Acute Medical Units: The way to go? A literature review. Eur J Intern Med.

2017 – van Galen LS & Wachelder J. (4-2017) Cruising through the journey without getting drowned:

The saga of a PhD student in the Netherlands. Acute Med.

– Kidney R, Sexton E, van Galen LS, Silke B, Nanayakkara P, Kellet J. (4-2017) Hospital Readmissions - Independent Predictors of 30-day Readmissions derived from a 10 year Database. Acute Med.

– van Galen LS, Brabrand M, Cooksley T, van de Ven PM, Merten H, So RK et al. (5-2017) Patients’ and providers’ perceptions of the preventability of hospital readmission: a prospective, observational study in 4 European countries. BMJ Qual Saf.

LIST OF PUBLICATIONS

194

LIST OF SCIENTIFIC PRESENTATIONS – Cooksley T, Nanayakkara P, Nickel CH, Subbe CP, van Galen LS, Kellett J, Kidney RM, Merten H,

Vaughan L, Brabrand M. *On behalf of the safer@home consortium. Predicting readmissions – Existing scoring models of little use to acute physicians? Poster presentation at the Society of Acute Medicine Annual Spring Conference, 2015, Bristol, United Kingdom.

– van Galen LS, Subbe CP, Brabrand M, Cooksley T, Kellett J, Kidney RM, Merten H, Nickel CH, Vaughan L, Nanayakkara P.*On behalf of the safer@home consortium. Study proposal: CURIOS@. Poster presentation at the Society of Acute Medicine Annual Spring Conference, 2015, Bristol, United Kingdom.

– van Galen LS & Nanayakkara PWB. Study protocol CURIOS@: CaptUring Readmissions InternatiOnally to prevent readmissions by safer@home consortium. Oral presentation at the 4th annual Dutch Acute Medicine congress, 2015, Rotterdam, the Netherlands.

– van Galen LS & Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. Oral presentation at the 5th annual Dutch Acute Medicine congress, 2016, Enschede, the Netherlands.

– van Galen LS, Cooksley T, Merten H, et al. *On behalf of the safer@home consortium. Physician consensus on preventability and predictability of readmissions based on standard case scenarios. Oral and poster presentation at Society of Acute Medicine Annual autumn Conference, 2016, Edinburgh, Schotland.

– van Galen LS & Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. Poster presentation at the Society of Acute Medicine Annual autumn Conference, 2016, Edinburgh, Schotland.

– van Galen LS & Fluitman KS, Merten H, et al. *On behalf of the safer@home consortium. Exploring the preventable causes of unplanned readmissions using root cause analysis: Coordination of care is the weakest link. Poster presentation at Society of Acute Medicine Annual autumn Conference, 2016, Edinburgh, Schotland.* Prize for best poster.

– van Galen LS, Brabrand M, Cooksley T, van der Ven PM, Merten H, So, RK, van Hoof L, Haak HR, Kidney RM, Nickel CH, Soong J, Weichert I, Kramer MH, Subbe CP, Nanayakkara PWB. *On behalf of the safer@home consortium. Preliminary results from a prospective European study on predictability and preventability of readmissions using patient/healthcare worker perspectives (CURIOS@). Oral presentation at the Society of Acute Medicine Annual autumn Conference, 2016, Edinburgh, Schotland.

– van Galen LS & Dijkstra CC, Ludikhuize J, Kramer MH, Nanayakkara PW. A Protocolised Once a Day Modified Early Warning Score (MEWS) Measurement Is an Appropriate Screening Tool for Major Adverse Events in a General Hospital Population. Selected as ‘paper of the month’ by the director of the Swiss Safety Foundation, October 2016.

LIST OF SCIENTIFIC PRESENTATIONS

195

AUTHOR AFFILIATIONSNadia Alam, Section acute medicine, Department of Internal Medicine, VU Medical Centre,

Amsterdam, the Netherlands

Mikkel Brabrand, Department of Emergency Medicine, Hospital of South West Jutland,

Esbjerg, Denmark

Nikki L. Damen, NIVEL Netherlands Institute for health services research, Utrecht,

The Netherlands

Ellen P. Claessens, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Tim Cooksley, Department of Acute Medicine, University Hospital of South Manchester,

Manchester, United KingdomCasper C. Dijkstra, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Babiche E. Driesen, Section acute medicine, Department of Emergency Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Jelle A. van Erven, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Kristien Fluitman, Department of Internal Medicine, VU Medical Centre, Amsterdam,

the Netherlands; Sahlgrenska Academy, Institute of Medicine, Gothenburg University,

Gothenburg, Switzerland.

Louise S. van Galen, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Harm Haak, Department of Internal Medicine, Maxima Medisch Centre, Eindhoven/Veldhoven,

The Netherlands; Department of Internal Medicine, Division of General Internal Medicine,

Maastricht University Medical Centre Maastricht University, Maastricht, The Netherlands

Asselina A. Hettinga-Roest, Section acute medicine, Department of Internal Medicine, VU

Medical Centre, Amsterdam, the Netherlands

Loes van Hooff, Department of Emergency Medicine, VieCuri Hospital, Venlo, The Netherlands

Rachel M. Kidney, St. James Hospital, Dublin, Ireland

Mark H. Kramer, Department of Internal Medicine, VU Medical Centre, Amsterdam,

the Netherlands

Eline M. Lammers, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Jeroen Ludikhuize, Department of Anaesthesiology, Academic Medical Centre, Amsterdam,

the Netherlands

Hanneke Merten, Department of Public and Occupational Health, EMGO Institute for Health

and Care Research, VU University Medical Centre, Amsterdam, The Netherlands

Prabath W. Nanayakkara, Section acute medicine, Department of Internal Medicine, VU

Medical Centre, Amsterdam, the Netherlands

Christian H. Nickel, Department of Emergency Medicine, University Hospital Basel,

Basel, Switzerland

Saskia M. Rombach, Department of Internal Medicine, VU University Medical Centre,

Amsterdam, The Netherlands

AUTHOR AFFILIATIONS

196

Jan C. Roos, Bureau of medical affairs, VU University Medical Centre, Amsterdam,

the Netherlands

Linda J. Schoonmade, Medical library, VU University, Amsterdam, the Netherlands

Wouter van der Schors, NIVEL Netherlands Institute for health services research, Utrecht,

The Netherlands

John Soong, Imperial College London, United Kingdom

Ralph K. So, Department of Quality, Safety & Innovation, Albert Schweitzer Hospital,

Dordrecht, The Netherlands

Johannes I. van der Spoel, Department of Intensive Care, VU University Medical Centre,

Amsterdam, The Netherlands

Patricia W. Struik, Section acute medicine, Department of Internal Medicine, VU Medical

Centre, Amsterdam, the Netherlands

Christian P. Subbe, Ysbyty Gwynedd Hospital, Wales, United Kingdom

Caroline B. Terwee, Department of Public and Occupational Health, EMGO Institute for

Health and Care Research, VU University Medical Centre, Amsterdam, The Netherlands

Louella Vaughan, Department of Medicine and Therapeutics, Chelsea and Westminster

Hospital, London, United Kingdom

Peter M. van de Ven, Department of Epidemiology and Biostatistics, VU University Medical

Centre Amsterdam, Amsterdam, The Netherlands

Cordula Wagner, VU University Medical Centre, Amsterdam, The Netherlands; NIVEL

Netherlands Institute for health services research, Utrecht, The Netherlands; Department of

Public and Occupational Health, EMGO Institute for Health and Care Research, VU University

Medical Centre, Amsterdam, The Netherlands

Immo Weichert, The Ipswich Hospital NHS Trust, Ipswich, United Kingdom

198

WORD OF THANKS

WORD OF THANKS

199

ALEXARNTZENIUSALINAOFFRINGAAMYVANGALENANNAEMANUEL

ANNEFLEURKOOPENANNELIESBROUWER

ANNELIESKABALTANOUSCHKAPRONK

BARBARABOSBASVANLIESHOUT

BOBOSCASPERDIJKSTRACHANTALWIEPJES

CHRISSUBBECHRISTELDEBLOKCHRISTIANNICKEL

CONNECTCARCORDULAWAGNER

DAISYVEDDERDANIELVANRAALTEDAVIDBRINKMANDEMUSONTWERPT

DENISEHOOGEVEENDENNISBARTEN

DIDIBRINKEDINKLEGEMATE

DUTCHACUTEMEDICINEEDMÉESCHRIJVERELINELAMMERSELLEMIJNKUIPER

ERIKVANBOMMELEVADURINCK

FRITSHOLLEMANHANNEKEMERTEN

ICARVUIRISDENIE

JANINESTOLWIJKJANKLEIN

JELLETICHELAARJELLEVANERVEN

JEROENLUDIKHUIZEJESBOS

JESKEVANDIEMENJESSICABEKEMA

JOERITIJDINKJONNESIKKENS

JORINDEKUIPERSJORISVANDERVORSTJORNWOERDEMAN

JOSIPCAR

JOYCEWACHELDERKARINKAASJAGER

KEWAJAMJAARNULACHTKEWAJENNES

KOENVANBEERSKRISTIENFLUITMAN

LENNARTTONNEIJCKLIEKEGLAUDEMANSLISETTEACKERMANS

LKCMEDICINELYNNVANDIJK

MAARTENVANDERBIEMAARTJEKLAVER

MAMAMARCELMUSKIET

MARIEKEDIEPENBROEKMARIEKEVANNOORTWIJK

MARIJESCHIPPERMARKHOLLANDMARKKRAMERMARTINECARIS

MCDONALDFAMILYMICHAELREUMERMANMICHIELVANAGTMAEL

MIKKELBRABRANDNADIAALAMNANALORNA

NICKDEJONGENIENKENOTA

PAPAPHILIPBOS

PRABATHNANAYAKKARARACHELKIDNEY

RISHINANNANPANDAYRSGVRIENDINNETJES

SAFERATHOMECONSORTIUMSECTIEFARMACOTHERAPIE

SHANEVANGALENSIETSKEVANNASSAU

SOCIETYACUTEMEDICINESTIENEKEDOORNWEERD

TABITHADEGRAAFFTIMCOOKSLEYTIMSCHUTTETOMBOEIJE

VANGALENFAMILIEWESSELFUIJKSCHOT

WILMAJANSENWOUTERVANDERSCHORS

YVOSMULDERS

200

CURRICULUM VITAE / BIOGRAPHYLouise Sandra van Galen was born on 27th of June

1989 in Hengelo (OV), the Netherlands. She is half-

Irish. Louise finished her secondary school in Harderwijk

(RSG ‘t Slingerbos), where she graduated cum laude in

2007. In secondary school she also received her diploma

‘International Baccalaureate English A2 higher level’. She

moved to Amsterdam at the age of 18 to start her medical

training at the VU University Medical Centre. During her

study she worked for the BIS Foundation in Leiden as

a transplantation employee and went abroad to Aruba for

her clinical internship in surgery. Louise received her medical

degree in the summer of 2014. Subsequently, she worked

as a junior doctor in internal medicine at the Westfries Gasthuis in Hoorn for eight months

after which she started her PhD on patient safety in the acute healthcare chain at the VUmc

under supervision of dr. Prabath Nanayakkara and professor Mark Kramer. The research

projects in her PhD attempt to work on research questions originating from daily clinical issues

on the workfloor.  She was the coordinating researcher for the first European prospective

readmission study ‘CURIOS@’ (“CaptUring Readmission data InternatiOnally by Safer@home

group”). This is a research project that originated from the safer@home consortium, an

international group founded in 2013 consisting of 13 acute medical physicians, emergency

physicians and epidemiologists from Europe that focus on readmissions and safer discharge

processes. Results of her PhD research are presented in this thesis. During her PhD, Louise

also worked as a staff member for the section pharmacotherapy and developed and taught

pharmacotherapy education to medical students. Beside her medical duties, Louise has several

functions in voluntary work such as ambassador for the Hubrecht Institute in Utrecht, as

advisor in the Amsterdam city centre panel, and as a board member of the Medical Business

Education Committee. After her PhD Louise hopes to become a specialist in Acute Internal

Medicine, and to prolong her research in the acute healthcare chain field. In September 2017

Louise will start a post-doc fellowship at Nanyang University in Singapore under supervision of

dr. Josip Car in his health services outcomes research group.

Louise lives in Amsterdam with her partner, Philip Bos.

CURRICULUM VITAE / BIOGRAPHY

201

SPONSORS