practice nurse involvement in the management of adults

SYSTEMATIC REVIEW

©2016 U

Practice nurse involvement in the management ofadults with type 2 diabetes mellitus attending ageneral practice: results from a systematic review

Deborah Parker,1,2,4 Diane Maresco-Pennisi,1,2 Karen Clifton,1,2 Ridhwan Shams3 and Jacqui Young1

1School of Nursing, Midwifery and Social Work, The University of Queensland, 2Australian Centre for Evidence Based Community Care:

A Collaborating Centre of the Joanna Briggs Institute, The University of Queensland, 3School of Medicine, University of Queensland, St Lucia,

and 4Centre for Applied Nursing Research, Western Sydney University and the Ingham Applied Medical Research Institute, Liverpool, Australia

AB S T R AC T

Aim: Using themethodology of the Joanna Briggs Institute, a systematic review of current research was performed todetermine if the addition of management by nurses had been more effective in improving clinical outcomes ofpatients with type 2 diabetes attending a general practice compared with standard care.

Methods: A three-step literature search was conducted for suitable English studies with quantitative clinicaloutcomes that had been published from January 1990 to May 2014. Randomised controlled trials (RCTs) wereparticularly sought after; however, other research designs were considered. Articles were assessed by twoindependent reviewers for methodological validity, prior to inclusion in the review, using standardised criticalappraisal instruments from the Joanna Briggs Institute. When possible, quantitative data were pooled in statisticalmeta-analysis.

Results: Seven studies were of suitable quality and relevance for the review: these included three randomisedcontrol trials; two cluster- RCTs; a cluster, nonrandomised, controlled before-after study; and a cluster observationalcohort study. These studies yield evidence that nurse management in addition to standard general practitioner careleads to modest improvements in blood pressure and total cholesterol levels in adults with type 2 diabetes attendinga general practice.

Conclusion: Meta-analysis identified modest, significant improvements amongst participants in nurse manage-ment interventions (NMIs) in the following clinical outcomes: mean SBP, mean DBP and mean total cholesterol. Themajority of outcomes studied did not show any advantage to adding NMIs to general practitioner care. Two studiesreported significant improvements of participants with poor control in mean haemoglobin A1c. An RCT thatinvestigates the effect of NMIs on patients, with poor control in regard to clinical outcomes and cost effectiveness,is recommended.

Key words: diabetes mellitus type 2, general practice, meta-analysis, nurse management, practice nurse

Int J Evid Based Healthc 2016; 14:41–52.

Background

I n the year 1996, diabetes was first listed as an Aus-

tralian national health priority.1 Now 5% of Australian

city-dwellers livewith this disease, and in remoteareas, it is

estimated that one out of every 11 people has diabetes.2

Correspondence: Deborah Parker, Centre for Applied Nursing

Research, Locked bag 7103, Liverpool BC. NSW 1871, Australia.

E-mail: [email protected]

DOI: 10.1097/XEB.0000000000000080

International Journal of Evidence-Based Healthcare � 2016 University

niversity of Adelaide, Joanna Briggs Institute. U

Type 2diabetes accounts for 85–90%of all diabetes cases,

and eachweek in Australia approximately 1160 new cases

of type 2 diabetes are identified.1,3

The increase rate in type 2 diabetes occurrences has

been a worrying trend, as it often leads to debilitating

symptoms and an early death. It is associated with

circulatory complications, chronic kidney disease, eye

complications and a significant reduction in quality of

life.4 In 2010, 5.4% (7750/143 519) of Australian deaths

were the result of diabetes or causes related to diabetes.2

of Adelaide, Joanna Briggs Institute 41

nauthorized reproduction of this article is prohibited.

mailto:[email protected]

D Parker et al.

©2016 U

In 2008–2009, almost $1507 million was spent on

diabetes in Australia. Of this, 43% was on hospital-

admitted patients, 24% on out-of-hospital medical serv-

ices and 33% on blood glucose lowering medicines.5

Diabetes management is a drain on one’s personal

finances as well. On average, the direct individual

medical costs of someone with diabetes are two to three

times above the average.3

As with other chronic diseases, the management of

diabetes is complex, time consuming and challenging.

Studies have shown that approximately 50% of patients

with chronic diseases do not receive suitable manage-

ment.6,7 Such findings have led to calls for a paradigm

shift in how healthcare is delivered to patients who have

a chronic disease. Ongoing research continues to search

for methods to manage chronic disease that are cost-

effective, accessible and widely available.8

Evidence has identified nurses to be well suited to

deliver education on health and illness prevention to

patients with chronic diseases. Nurses employed in

general practice (i.e. practice nurses) are particularly well

placed to implement behaviour change strategies and to

monitor patient compliance. Data from international

studies confirm that nurses working in general practice

effectively manage the complex needs of patients with

chronic diseases.9 A systematic review by Laurant et al.9

concluded that interventions that replaced doctors in

primary care with properly trained nurses resulted in

comparable processes of care, utilisation of resources,

costs and patient health outcomes.

In the United Kingdom, nurses have long partici-

pated in the management of chronic diseases.10 In

Australia, a recent national survey of Australian practice

nurses with a response rate of 32% (n¼ 701/2161)

verifies that the growing role of the Australian practice

nurse includes chronic disease management.11 The

proportions of practice nurses that reported that they

undertook education, assessment and management on

a weekly or daily basis for specific chronic diseases were

as follows: asthma (46%), arthritis (29%), cardiovascular

disease (76%), diabetes (59%) and wound management

(93%).11 Approximately half (52%) reported that they

frequently prepared chronic disease management

plans.11 It has been suggested by Jolly12 that up to

70% of work performed by doctors in Australian general

practice could actually be performedwell by nurses. The

number of Australian practices employing a nurse has

been climbing over the last decade; in 2007, 58% of

general practitioners (GPs) employed a nurse and in

2012, this figure rose to 63%.11 In 2012, there were

10 693 nurses employed at a GP; nearly 3000 more than

those employed in GPs in 2007.11

42 International Journal of Evidence-Based


Some believe that adults with type 2 diabetes would

benefit by having nurses involved with their manage-

ment, and several studies have shown that nurses work-

ing with doctors in general practice significantly

improved outcomes in patients with type 2 diabe-

tes.13–17 Although improvements in the patients’ glycae-

mic control have been reported, Renders et al.18 have

called for researchers to focus more broadly on the

clinical outcomes in interventions to improve the man-

agement of diabetes in general practice, outpatient and

community settings.

AimsThe aim of this systematic review was to determine if the

additional management by nonspecialist nurses was

more effective in improving clinical outcomes of patients

with type 2 diabetes in general practice than standard

GP care alone. This review is an update of a 2012

review.19 Studies with interventions performed by a

nurse practitioner, a diabetic nurse educator or a study

nurse were excluded from the review to retain focus on

nonspecialist nurses, often referred to as practice nurses,

employed in general practice.

Search strategyA search strategy was designed to identify peer-

reviewed scientific articles as well as other types of

research publications, for example, doctoral theses or

government reports. The search was limited to publi-

cations written in English and was undertaken in three

steps. First, an initial limited search of Medline and

CINAHL involved an analysis of the title, abstract and

index terms of relevant articles to identify relevant

search terms. Next, identified search terms were used

to search relevant databases, and, lastly, the reference

lists of all identified articles were manually searched for

additional studies.

The search strategy was designed to identify both

published and unpublished literature. The first published

search was limited to English language articles published

from January 1990 to December 2011. After publication

of the source report, the search was rerun in July 2014, to

update the review for this article, using the same search

strategy. This 2014 search resulted in two new articles,

published after 2012, being included in the review.

Therefore, the final search for this article was performed

in 2014, and, in total, the literature for this article was

searched from 1990–2014.

Initial keywords used were as follows: diabetes, dia-

betes mellitus, type 2 diabetes, nurse-led, nurse-directed,

practice nurse, primary care, primary healthcare, general

practice, family practice and GP.

Healthcare � 2016 University of Adelaide, Joanna Briggs Institute


SYSTEMATIC REVIEW

©2016 U

Twenty-four databases for published literature were

searched; these included but were not limited

to CINAHL, EMBASE, Medline, PubMed and Scirus.

Fourteen databases containing unpublished literature

were searched that included but were not limited to the

following: American Nurses Association, Australian

Digital Theses Program, Google Scholar System for

Information on Grey Literature in Europe. A full list of

the databases used in the search can be found in the

source report.19

The following specific search strategy was applied:

nurse-led OR nurse-run OR nurse-directed OR ‘nurse

managed’ OR ‘nurse care manage�’ OR ‘nurse coordi-

nated’ OR ‘shared care’ AND ‘general practice’ OR ‘family

practice’ OR ‘primary care’ OR ‘primary healthcare’ AND

‘diabet�’ AND ‘rct’ OR ‘random allocation’ OR ‘random-

ised control trial’ OR ‘comparative stud�’ OR ‘interrupted

time series’ OR ‘clinical trial’ OR ‘prospective stud�’ OR‘study design’ OR ‘evaluation research’ OR ‘controlled

stud�’ OR ‘cohort’ OR ‘case control’ OR ‘interrupted time

series’ OR ‘experimental stud�’.

Inclusion criteriaStudies of nurse management interventions (NMIs)

involving adult participants (aged above 18 years) diag-

nosed with diabetes type 2, attending a general/family

practice, were reviewed. Only studies that involved a

nurse who was both employed by the practice and

played a lead role in the intervention care were included.

Studies using a randomised controlled trial (RCT) or a

cluster RCT (cRCT) design were particularly sought after;

however, other research designs (i.e. controlled clinical

trials, interrupted time series, controlled before and after

designs and observational studies) were considered as

well. Studies were selected that reported on changes in

one ormore clinical outcomes for example, SBP, DBP, BMI,

Table 1. Results of JBI-MAStARI critical appraisal of i

JBI-MAStARI critical appraisal checklist for randomised/nonr

Reference

Blackberry et al.21 cRCT

Cleveringa et al. 200826 cRCT

Gabbay et al. 200628 RCT

Meulepas et al.27 cCBA

Taylor et al. 200320 RCT

Taylor et al. 200529 RCT

JBI-MAStARI critical appraisal checklist for comparable cohort/case-con

Reference CC

Juul et al.25

cCBA, cluster controlled before after; cOC, cluster observational cohort; cRCT, cluster Runclear if the criterion was met; Y, yes the criterion was met.



cholesterol levels, haemoglobin A1c (HbA1c) and fasting

blood glucose.

Methods of the reviewStudy appraisalA Joanna Briggs Institute (JBI) approach was used for the

review. Studies were assessed by two independent

reviewers for methodological quality using one of two

standardised critical appraisal instruments, which are

part of the JBI Meta-analysis of Statistical Assessment

and Review Instrument (JBI-MAStARI). The JBI-MAStARI

tools offer a list of criteria and reviewers assess whether

each criterion was met (yes, no or unclear). The studies

were appraised for aspects such as randomisation/allo-

cation, blinding, measure methods, group comparison

and statistical analysis. The reviewer determined if the

study had met the criterion with a yes, no or unclear. To

be included in the review, a study had to meet at least

five out of ten criteria (‘yes’ by both reviewers) for

experimental studies or nine for case control, cohort

and descriptive studies. A third reviewer was used to

resolve any disagreements. The results of this assess-

ment can be viewed in Table 1; question numbers in the

table correspond with those on the published tool.

Specific details about the interventions, study popu-

lations, methods and outcomes of significance to the

review question and specific objectives were extracted

from articles using the standardised data extraction tool

from JBI-MAStARI.

Meta-analysisWhen possible, quantitative data from the studies were

pooled in statistical meta-analysis using RevMan 5

(Cochrane Collaboration). Some studies did not report/

provide standard deviations and/or 12 month mean

data25 and were excluded from the meta-analysis. For

ncluded studies

andomised controlled trials

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10

Y N Y N U Y Y Y Y Y

Y U U Y U N Y Y Y Y

Y U U U U Y Y Y Y Y

N U N N U Y Y Y Y Y

U U U N Y Y Y Y Y Y

Y U U Y U Y Y Y Y Y

trol studies

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9

U Y Y Y Y Y Y Y Y

CT; N, no the criterion was not met; N/A, not applicable; Q, question; U, it was



D Parker et al.

©2016 U

each forest plot, experimental condition was used to

denote the NMI group and control to denote the usual

care/GP intervention group. RevMan 5 (Cochrane Col-

laboration) was used to create forest plots of the meta-

analysis results. All quantitative data were entered twice.

Wherever appropriate, relative risks and/or odds ratios

and their associated 95% confidence interval were cal-

culated for analysis of categorical data. For continuous

data that had been collected using the same scale,

weighted mean differences and standard deviations

were calculated; for data collected using different scales,

standardised mean differences were calculated. Signifi-

cant heterogeneity was assumed at P< 0.1 using the chi-

square test for homogeneity. As the interventions were

complex and varied, heterogeneity was anticipated in

the meta-analysis, and a random effects meta-analysis

was utilised.

Four studies used a cluster design in which groups of

individuals are organised to receive either intervention

or control treatment rather than individuals being organ-

ised to receive either intervention or control treatment.

In all four included cluster studies, the GP clinic was the

unit of allocation. For the Blackberry et al.21 and Clever-

inga et al.26 studies, the research team randomised the

GP clinics to receive either intervention or control,

whereas the Meulepas et al.27 and Juul et al.25 studies

collected data from GP clinics that were providing inter-

vention or control treatment for their patients for non-

random reasons independent of the research study.

There are advantages to using a cluster designs, and it

is a valid approach; it is also acceptable to combine cRCTs

with RCTs in ameta-analysis,22 as long as the effect of the

intervention on a cluster design and individual design

are considered comparable,23 as they are in this case.

However, as participants within a cluster have a

tendency to be more alike than they are to participants

in other clusters, this design requires special treatment.

We adhered to recommendations to adjust data from

included cluster studies, when needed, to avoid unit-of-

Table 2. Assessment of bias in included cluster studi

Recruitedpatients beforeclusters random-

isedBaseline wabalanced

Blackberry et al.21 cRCT Y Y

Cleveringaet al. 200826

cRCT U Y

Meulepaset al.27

cCBA N Y

Juul et al.25 cOC N/A N/A

cCBA, cluster controlled before after; cOC, cluster observational cohort; cRCT, cluster R



analysis errors;22 and to assess cluster studies for cluster-

specific bias,23 as shown in Table 2.

Blackberry et al.21 and Meulepas et al.27 each adjusted

for clustering errors in their analysis as well as reporting

raw data. Cleveringa et al.26 mention adjusting for clus-

tering in the power analysis; however, it was unclear if

they adjusted for clustering when calculating their

results. Blackberry et al.21 reported an intracluster cor-

relation coefficient (ICC) value, a measure of the amount

of clustering in a sample, for each outcome. Meulepas

et al.27 adjusted for clustering using a published ICC

(0.05) that was appropriate for use in studies where the

allocation unit is a GP clinic.24 We used this same

published ICC (0.05) to adjust the Cleveringa et al.26 data

for clustering or to adjust the raw dichotomous data

from Blackberry et al.21 or Meulepas et al.27 We adjusted

the data for clustering, using amethod recommended by

Cochrane, that for continuous variables involved using

the ICC to reduce the sample size resulting in the

calculation of effective sample size and for dichotomous

variables using the ICC to reduce both the number of

events and sample size.22 Juul et al.25 report that they

adjusted for clustering in the analysis of their data; they

had estimated differences in mean proportions by

binomial regressions with identity link, systematic

effect of type of practices and random difference

between practices.

ResultsThe updated search led to the inclusion of two more

articles, which brought our total yield to seven peer-

reviewed research articles that were methodically

assessed for inclusion (Fig. 1), resulting in seven final

articles that originated from Australia, the United States,

Canada, the Netherlands and Denmark. The main reason

for related studies to be excluded was that the nurse was

a diabetes educator, an advanced practice nurse or nurse

practitioner or that the nurse had an unclear or minor

contribution within a multidisciplinary intervention.

es

sData was col-lected from all

clusters

Analysisadjusted

for cluster-ing

Comparablewith individ-ual trials

Y Y Y

Y Y Y

N Y Y

N Y Y

CT.



Records identified throughdatabase searching

(n = 804)

Records after duplicates removed(n = 805)

Records screened(n = 806)

Full-text articles assessedfor eligibility

(n = 49)

Studies included inquantitative synthesis

(meta-analysis)(n = 7)

Full-text articles excluded,with reasons

(n = 42)

Records excluded(n = 757)

Additional records identifiedthrough other sources

(n = 2)

Iden

tifi

cati

onS

cree

ning

Elig

ibili

tyIn

clud

ed

Figure 1. Study selection process.

SYSTEMATIC REVIEW

©2016 U

Individual reasons for exclusion are reported in the

source review.19

The interventionsNMIs were varied and included both individual and

group education for participants on lifestyle, medication

adjustment and treatment decisions based on algor-

ithms (Table 3). In Blackberry et al.21, this education

occurred over the telephone. In the Taylor et al. study,20

practice nurses titrated medications for diabetes, cho-

lesterol and hypertension following algorithms based on

national guidelines. Two studies used a computer

decision support system.26,27 Outcomes reported in

the seven reviewed studies included one or more of

the following: HbA1c, fasting blood glucose, lipids [total

cholesterol (TC), LDL], BMI, SBP or DBP (Table 3).

Main clinical outcomesGlycated haemoglobin A1cAll seven included studies assessed HbA1c. Meta-

analysis performed on the final mean HbA1c values of

four of these studies21,26–28 did not identify any

advantage to participants who had received the NMIs



as compared with standard GP care (Fig. 2). The results of

Taylor et al.29, which were not included in the meta-

analysis because the standard deviation data were not

reported, described an advantage to participants receiv-

ing the NMI at 4 months that approached significance

(P¼ 0.10). The cluster observational cohort study per-

formed by Juul et al.25 found comparable mean HbA1c

values when comparing practices with well-imple-

mented nurse-led diabetes consultations to practices

that did not have a nurse. The study of Taylor et al.20

was excluded from the meta-analysis as it was limited to

participants with long-standing diabetes, one or more

major medical comorbid conditions and an HbA1c

greater than 10% at baseline. However, they did observe

a significant decrease in HbA1c mean in the participants

receiving NMI as compared with controls at 1 year

(P¼ 0.01, effect size 0.37).20

Several studies assessed the number of participants

who reached an HbA1c target. Meta-analysis combining

data from two such studies, one using an HbA1c target of

less than 7%26 and the other a target of less than 7.5%,27

showed that NMIs did not impact the proportion

of participants that had reached the HbA1c target at



Table 3. Included studies

Studies Method/length Participants Interventions Usual care

Blackberry et al.21

Victoria,Australia

Cluster RCT;follow-up: 18months

Sites¼ 59 GP clinics; partici-pants had diabetes type 2attending a GP; interventionn¼ 236 control n¼ 237

Practice nurses fromintervention prac-tices received 2 daysof training in a tele-phone coachingprogramme, whichaimed to deliver 8telephone and oneface-to-face coachingepisodes per patient

Usual care by their GP,which may haveincluded referral to dia-betes educators, dieti-tians and diabetessocialists as part of thestandard diabetes careof that practice

Cleveringaet al.26 theNetherlands

Cluster RCT;follow-up: 1 year

Sites¼ 55 GP clinics participantshad diabetes type 2 attend-ing a GP; interventionn¼ 1699 control n¼ 1692

A 1 hr consultationwith nurse who useda computeriseddecision support sys-tem (CDSS) withdiagnostic and treat-ment algorithms anda recall system

Usual care was providedby the GP or a nurseunder direction of theGP

Gabbay et al.28

Hershey, Penn-sylvania, USA

RCT; follow-up: 1year

Sites¼ 2 GP clinics participantshad diabetes (95% had type2) attending a GP clinic; inter-vention n¼ 150 controln¼ 182

Nurse implementedalgorithms undersupervision of GP.Also included werebehavioural goal set-ting, individualisedcare plans, self-man-agement education,follow-up phonecalls and referrals

Usual care by their GP,which might includereferrals to other prac-titioners

Juul et al.25

DenmarkCluster observa-

tional cohortstudy; follow-up:cholesterol: 450days; HbA1c: 180days

Sites¼ 193 GP clinics; partici-pants had diabetes type 2attending a GP; n¼ 13 117

Practices with wellimplemented nurse-led type 2 diabetesconsultations

Practices with either nonurse employed or noimplemented nurse-ledconsultations

Taylor et al.20

USARCT; follow-up: 1

yearSites¼ 1 GP clinic; participants

had long-standing diabetes(type 2¼ 97% usual care,93% intervention) with oneor more major medicalcomorbid conditions and anHbA1c greater than 10% atbaseline attending a GP;intervention n¼ 61 controln¼ 66

Individual consultationwith nurse to reviewmedical, lifestyle andpsychosocial statusand develop self-management plan.Group classes eachweek for 4 weeksusing a workbookand discussion for-mat. Telephone fol-low-up to reviewpatient goals, medi-cation use, symp-toms, glucosemonitoring, BPmonitoring and self-management activi-ties. Nurses followedalgorithms based onnational guidelinesto titrate medi-cations for diabetes,cholesterol and hy-pertension

Usual care with GP includ-ing diabetes pamphlets,Medic Alert pamphletand instructions tomaintain contact withGP and attend generaldiabetes classes at alocal medical centre.Participants told after 1year they would receivea workbook and meetwith nurse care man-ager or attend a group

D Parker et al.

46 International Journal of Evidence-Based Healthcare � 2016 University of Adelaide, Joanna Briggs Institute

©2016 University of Adelaide, Joanna Briggs Institute. Unauthorized reproduction of this article is prohibited.

Table 3. (Continued)

Studies Method/length Participants Interventions Usual care

Taylor et al.29

CanadaRCT; follow-up: 4

monthsSites¼ 1 GP clinic participants

had type 2 diabetes attend-ing a GP; intervention n¼ 20control n¼ 19

Four or five visits bynurse at home, 1visit from a dietician(nurse present) and1 visit from exercisephysiologist. Initialvisit including assess-ment of knowledgeand education pro-vided based on thisassessment. Focus ofsubsequent meetingswas self-manage-ment. Case confer-ences with usualphysician for medi-cation adjustmentand advice

Usual care by GP,included one quarterlyscheduled visit

Meulepas et al.27

the NetherlandsCluster controlled

before-afterstudy/1 year

GP clinics¼ 51; patients withtype 2 diabetes attending aprimary care clinic; interven-tion n¼ 431 control n¼ 469

Nurse provided infor-mation on lifestylemanagement for riskfactors and adjustedmedication at quar-terly visits. Also usedthe diabetes supportservice (DSS), apatient recall regis-ter. The before meandata were compar-able for HbA1c(intervention, 7.3; SD1.2 and control, 7.2;SD 1.1); for SBP(intervention, 149;SD 22.9 and control,150; SD 23.3); forDBP (intervention,84;SD 12.1 and con-trol, 84; SD 12.5); fortotal cholesterol(intervention, 5.0; SD1.0 and control, 5.1;SD 0.9). The inter-vention and controlgroups each had82% smokers, amean BMI of 29 kg/m2, a mean HDLratio of 4.3 and com-parable mean trigly-cerides (intervention,1.8; SD 1.2 and con-trol, 1.9; SD 1.2)

Care by GP and DSS

cCBA, cluster controlled before after; cOC, cluster observational cohort; cRCT, cluster RCT; GP, general practice; HbA1c, glycated haemoglobin.

SYSTEMATIC REVIEW

©2016 U

12 months (P¼ 0.19). At the commencement of the

Taylor et al.20 study, 43% of participants in the NMI

group had reached an HbA1c target of less than 7.5

as compared with just 25% in the control group; this

difference was significant (P< 0.03). When assessing the

International Journal of Evidence-Based Healthcare � 2016 University of Adelaide, Joanna Briggs Institute 47

niversity of Adelaide, Joanna Briggs Institute. Unauthorized reproduction of this article is pr

reduction of participants with high HbA1c, the Meulepas

et al.27 study found that the NMI group had a significant

improvement in reducing the proportion of participants

with high HbA1c (�8.5%) over the standard care group

(OR of 0.5; 95% CI 0.3 to 0.9). In their cluster observational

ohibited.

HbA1c, mean

SBP Mean

DBP Mean

Total Cholesterol Mean

Study or Subgroup

Blackberry 2013

Cleveringa 2008

Gabbay 2006

Meulepas 2008

Total (95% CI)

Heterogeneity: Tau2 = 4.27; Chi2 = 7.25, df = 3 (P = 0.06); I2 = 59%

Test for overall effect: Z = 3.24 (P = 0.001)

Mean

133

143

129

153

SD

14

20

18

22.5

Total

189

421

150

233

993

Mean

136

147

138

155

SD

16

20.8

19

24.3

Total

186

419

182

253

1040

Weight

27.6%

29.5%

21.9%

21.0%

100.0%

IV, Random, 95% CI

-3.00 [-6.04, 0.04]

-4.00 [-6.76, -1.24]

-9.00 [-12.99, -5.01]

-2.00 [-6.16, 2.16]

-4.40 [-7.06, -1.74]

Experimental Control Mean Difference Mean Difference

IV, Random, 95% CI

-100 -50 0 50 100

Favours [experimental] Favours [control]

Study or Subgroup

Blackberry 2013

Cleveringa 2008

Gabbay 2006

Meulepas 2008

Total (95% CI)



Mean

76

80

72

83

SD

9

11

9

11.1

Total

188

421

150

233

992

Mean

77

82

78

86

SD

11

10.6

10

12.1

Total

186

419

182

253

1040

Weight

24.2%

27.6%

24.2%

24.1%

100.0%

IV, Random, 95% CI

-1.00 [-3.04, 1.04]

-2.00 [-3.46, -0.54]

-6.00 [-8.05, -3.95]

-3.00 [-5.06, -0.94]

-2.96 [-4.97, -0.96]


IV, Random, 95% CI

-100 -50 0 50 100


Study or Subgroup

Blackberry 2013

Cleveringa 2008

Meulepas 2008

Total (95% CI)



Mean

4.2

4.6

4.7

SD

0.95

0.9

0.9

Total

200

421

233

854

Mean

4.28

4.8

4.8

SD

1.05

1.1

1.1

Total

200

419

253

872

Weight

23.3%

48.5%

28.2%

100.0%

IV, Random, 95% CI

-0.08 [-0.28, 0.12]

-0.20 [-0.34, -0.06]

-0.10 [-0.28, 0.08]

-0.14 [-0.24, -0.05]


IV, Random, 95% CI

-2 -1 0 1 2


Study or Subgroup

Blackberry 2013

Cleveringa 2008

Gabbay 2006

Meulepas 2008

Total (95% CI)



Mean

7.85

6.9

7.45

7.1

SD

1.24

1.1

1.4

0.9

Total

221

421

150

233

1025

Mean

7.91

6.9

7.4

7.3

SD

1.42

1

1.8

1.2

Total

219

419

182

253

1073

Weight

16.6%

46.3%

8.9%

28.2%

100.0%

IV, Random, 95% CI

-0.06 [-0.31, 0.19]

0.00 [-0.14, 0.14]

0.05 [-0.29, 0.39]

-0.20 [-0.39, -0.01]

-0.06 [-0.17, 0.04]


IV, Random, 95% CI

-2 -1 0 1 2Favours [experimental] Favours [control]

(a)

(b)

(c)

(d)

Figure 2. When compared to usual care, nurse-led interventions did not improve HbA1c but were more effective at 1 year inreducing mean SBP, DBP and total cholesterol.

D Parker et al.

48 International Journal of Evidence-Based Healthcare � 2016 University of Adelaide, Joanna Briggs Institute

©2016 University of Adelaide, Joanna Briggs Institute. Unauthorized reproduction of this article is prohibited.

SYSTEMATIC REVIEW

©2016 U

cohort study, Juul et al.25 found that compared with

practices with no nurses, the mean proportion of

patients with an HbA1c at least 8.0% for practices with

a nurse employed was lower; this held true both in

practices where nurse consultations for diabetes patients

were well implemented (3.7% reduction in HbA1c; 95%

CI �6.7 to �0.6) and in practices where there were

nurses but no organised nurse consultations (3.2%

reduction; 95% CI �7.5; 1.0), but this result was compli-

cated by the fact that statistically more patients in the

nurse-employed group had undergone HbA1c testing.

Systolic blood pressureSix of the included studies assessed participant SBP. The

data of four studies that had reported mean SBP data on

a total of 4983 participants21,26–28 were combined in

meta-analysis and a modest, significant decrease in final

mean SBP was identified in the NMI group as compared

with the GP standard care group [MD¼�4.40 (95% CI

�7.06, �1.74), Z¼ 3.24, P¼ 0.001] (Fig. 2). Taylor et al.29

found nomean SBP difference between intervention and

control at 4 months (P¼ 0.17). Taylor et al.20, who looked

at the effect of NMIs on participants with poor HbA1c

control at baseline, reported deterioration of SBP in both

groups with the intervention group deteriorating less

than GP standard care at 1 year (effect size of 0.28). The

data of two studies that assessed the portion of partici-

pants meeting an SBP target of 140 or 150mmHg26,27

were combined for meta-analysis; resulting in an insig-

nificant result of OR¼ 0.75 (95% CI 0.51, 1.09), P¼ 0.13.

Taylor et al.20 observed that more participants in the NMI

group, as compared with the standard care group, had

met a SBP target of 130mmHg or less, yielding a P value

approaching significance (P¼ 0.06).

Diastolic blood pressureSix studies assessed participant DBP with most reporting

a modest advantage to participants that had received an

NMI. The mean DBP data of four studies and a total of

4982 participants21,26–28 were combined in meta-

analysis, and a significant decrease in mean DBP in

the NMI group as compared with usual GP care was

observed [MD¼�2.96 (95% CI –4.97, –0.96), Z¼ 2.90,

P¼ 0.004] (Fig. 2). Taylor et al.29 also reported a signifi-

cantly greater reduction of mean DBP at 4 months in the

NMI group as compared with controls (P¼ 0.04). When

studying the effect of NMIs on participants with poorly

controlled HbA1c at baseline, Taylor et al.20 found no

advantage to being allocated to the NMI for either mean

DBP or the proportion reaching a DBP target of

85mmHg or less. Meulepas et al.27, in contrast, observed

that a significantly, slightly larger portion of participants



achieved a DBP target of less than 85mmHg in the NMI

group than the control group, OR of 1.4 (1.0–2.1).

Total cholesterolTC was assessed in all of the included studies with the

exception of the Gabbay study.28 Meta-analysis was

performed combining the data of 4676 participants from

three studies,21,26,27 and a statistically significant, modest

improvement in mean TC was observed in participants

receiving NMIs as compared with those receiving stand-

ard care [MD¼�0.14 (95% CI �0.24, �0.05), Z¼ 2.98,

P� 0.003] (Fig. 2). Taylor et al.29 did not observe a

difference between intervention and control at 4 months

for mean TC (P¼ 0.97). When studying participants with

poor HbA1c control at baseline, however, Taylor et al.20

observed a significantly greater improvement of mean

TC amongst participants allocated to NMIs as compared

with those in the control group (P¼ 0.01, effect size of

0.18). The Juul et al.25 cluster observational cohort study

did not find a difference in the mean TC values of

participants originating from practices with well-imple-

mented nurse-led diabetes consultations as compared

with practices that either did not employ a nurse or

employed a nurse but did not have organised nurse

consultations for diabetes patients.

Four studies reported on the portion of participants

who met a TC target at the commencement of the

study.20,26,21,25 A meta-analysis was performed combin-

ing the data of two studies26,21 and 4276 participants

that measured the participants’ ability to meet a TC

target of 4.5mmol/l or less,26 or less than 5mmol/l27

(Fig. 2). The proportion of participants that reached the

target at 12 months amongst those receiving NMIs were

not significantly larger than those who had received

standard care OR¼ 0.87 [(95% CI 0.70, 1.09) Z¼ 1.21,

P¼ 0.22]. Taylor et al.20 found no significant difference

between NMI and standard GP care in participants with

poor HbA1c control at baseline using a TC goal of less

than 5.17mmol/l. Likewise, the Juul et al.25 cluster obser-

vational cohort study did not find a significant difference

comparing practices with or without a well-implemented

nurse plan in regards to participants reaching a TC goal

of at least 5mmol/l.

Low density lipoproteinFive of the included studies assessed the effect of NMIs

on LDL. One RCT24 and two cRCTs21,26,28 that had

reported the mean final LDL cholesterol levels of 4123

participants were suitable for meta-analysis (Fig. 2).

When combined, LDL cholesterol levels in the interven-

tion group were comparable to those of controls

[MD¼�0.07 (95% CI �0.16, 0.02), Z¼ 1.55, P¼ 0.12]



D Parker et al.

©2016 U

(Fig. 2). Taylor et al.29 did not observe a difference

between intervention and control in mean LDL at 4

months (P¼ 0.98). The study by Taylor et al.20, looking

at patients with poor HbA1c control at baseline,

observed a significantly greater improvement in mean

LDL (P¼ 0.02, effect size 0.330) amongst those receiving

NMIs as compared with controls but no difference was

observed between NMI and control groups, when

measuring the proportion of participants reaching a

target LDL.

HDL cholesterol, triglycerides, fasting bloodglucose, BMI, weight, smokersNo advantage was observed for the participants receiv-

ing NMIs over GP standard care for the following out-

comes: BMI, HDL cholesterol, smoking, triglycerides,

mean weight. Meta-analysis was performed, when

possible, which resulted in nonsignificant outcomes

(Table 4). Taylor et al.20 found no significant difference

between NMI and standard GP care in participants with

poor HbA1c control at baseline with regards to HDL

cholesterol, LDL cholesterol, triglycerides or BMI. For

fasting blood glucose, however, there was greater

improvement in the NMI group than in the standard

care group that approached significance (P¼ 0.07).20

Combined outcomeOf note, the study by Cleveringa et al.26 calculated the

mean 10-year-UKPDS coronary heart disease (CHD) risk

(%) of participants in NMI and control groups and found

that those who had received NMIs had significantly

slightly less CHD risk; OR of 1.4, 95% CI (0.3–2.6)

P� 0.05. The following participant data had been used

to calculate this odds ratio: date of onset, age, duration of

diabetes, sex, ethinicity, smoking, A1C, SBP, TC and HDL

cholesterol. Cleveringa also measured the proportion of

participants in control and NMI groups that met four

treatment targets (A1C, systolic, TC and LDL cholesterol

targets) yielding a result which significantly favoured the

Table 4. Medical outcomes that did not improve wit

Outcome Studies

Mean LDL cholesterol Blackberry et al.21; Cleveringa et

Mean triglycerides Blackberry et al.21; Meulepas et a

Mean weight Blackberry et al.21; Gabbay et al.2

BMI target Blackberry et al.21; Meulepas et a

HbA1c target Cleveringa et al.26; Meulepas et a

SBP target Cleveringa et al.26; Meulepas et a

Smoking target Blackberry et al.21; Meulepas et a

Total cholesterol target Cleveringa et al.26; Meulepas et a

Smoking target Blackberry et al.21; Meulepas et a

HbA1c, glycated haemoglobin.



NMI group with an OR of 1.6 (95% CI, 1.3–2.1).26 It is

unclear if the authors accounted for the effect of cluster-

ing in these analyses.

DiscussionMeta-analysis identified modest, significant improve-

ments amongst participants in NMIs in the following

clinical outcomes: mean SBP, DBP and TC. When looking

specifically at participants with poor control, two studies

reported improvements in mean HbA1c that were sub-

stantial.20,27 Meta-analysis performed on the majority of

outcomes showed no difference between control and

NMI groups, namely, mean HbA1c, mean triglycerides,

mean weight, mean HDL cholesterol, mean LDL choles-

terol, or the proportion of participants reaching a goal/

target for HbA1c, SBP, DBP, TC, BMI or smoking.

HbA1cThe clinical outcome HbA1c is of particular importance

because of the strong, well-documented correlation

between HbA1c and type 2 diabetes complications.30

The HbA1c analysis, however, did not show a significant

advantage for participants who had received NMIs.

Cleveringa et al.26 rationalised that their negative HbA1c

result had occurred ‘because mean A1C at baseline was

almost at the treatment target, there was little room for

improvement’. The mean HbA1c target for Cleveringa

was 7% and participants’ HbA1c at baseline was certainly

nearly at this target (intervention 7.1%, control 7.0%).26

In contrast, the Taylor et al.20 study, by design,

involved participants with poor HbA1c control at base-

line (intervention mean HbA1c, 9.5%; control mean

HbA1c, 9.5%), and significant results were achieved. In

the study’s conclusion, a significant difference was

observed when comparing the 1.14 mean reduction in

mean HbA1c of the intervention group to the 0.35 mean

reduction of those receiving standard care.20 For the

cluster, controlled, before–after nonrandomised study

by Meulepas et al.27, the authors found that ‘the patient-

h NMIs

Z Value P Value

al.26; Meulepas et al.27 1.55 0.12

l.27 1.46 0.148 0.02 0.98

l.27 1.12 0.26

l.27 1.32 0.19

l.27 1.51 0.13

l.27 1.39 0.22

l.27 1.21 0.16

l.27 1.39 0.98



SYSTEMATIC REVIEW

©2016 U

oriented interventions seem to especially have an

effect on poorly controlled patients’. Subanalysis of

participants exhibiting poor control at baseline (an

HbA1c greater than 8.5%), in this study, revealed that

in the study’s conclusion at 1 year, the number of

participants with an HbA1c greater than 8.5% had halved

from 13 to 6 in the intervention group, a significant

difference to the reduction observed in the control

group (16 to 14).27

LimitationsThere were notable limitations to the review. Our meta-

analysis combined studies with a mixture of interven-

tion designs, which affected the outcome. The Taylor

et al.20 study had a small sample size (intervention

n¼ 61, control n¼ 66), and the study by Taylor et al.29

was hampered by both a very small sample size (inter-

vention, n¼ 20; control, n¼ 19) as well as a very short

intervention period (4 months). The authors of Taylor

et al.29 acknowledged that, in all likeliness, their study

had lacked the power to reach significance. Two weak-

nesses of the Meulepas et al.27 study need to be con-

sidered.27 Because of funding constraints, not all

practices could start the intervention treatment at the

same time; the first cohort formed the intervention

group, and the practices on the waiting list formed

the control group.27 This nonrandom allocation of

participants is very likely to introduce bias. In addition,

all participating practices of the Meulepas et al.27 study

had already participated in Phase 1 of the study in which

participants in the intervention arm received 12months

of logistic call support given the name, the diabetes

support service (DSS).27 Phase 2 incorporated an NMI in

half of the practices and used DSS in both intervention

and control, leaving the influence of the NMI in isolation

unclear.27 In Phase 1, DSS had not significantly improved

HbA1c control over standard care whereas in Phase 2,

significant improvements were observed with DSS

together with NMI compared with DSS alone.27 It is

recommended that controlled before–after studies

should only be included in a systematic review if they

have at least two intervention sites and two control sites,

the timing period for the control and intervention

groups were the same, and the intervention and control

groups were comparable on key characteristics;23 the

Meulepas et al.27 study meets all of these criteria. We

performed a sensitivity analysis that involved running

the meta-analyses without the Meulepas et al.27 data.

The results of running the meta-analyses without the

Meulepas et al.27 data did not differ in any substantial

way to the meta-analyses that had included the

Meulepas et al.27 data (data not shown).



Nurse management correlates with moremanagementTwo studies reported on several management outcomes

of interest. The practice nurse of the Gabbay et al.28 study

was given authority to refer participants receiving the

intervention to a dietician and/or a diabetes educator.

Compared with the GP, significantly more referrals were

made by the nurse who referred 70% of participants in

the NMI group to a certified diabetes educator and 53%

to a dietician; during this same time, the GP had referred

3% to a certified diabetes educator and 3% to a dieti-

cian.28 Gabbay et al.28 also found that participants who

had received NMIs had experienced significantly better

process measures/screening for the following: ophthal-

mological exam, foot exam, microalbuminuria screening,

pneumonia vaccination, dietician visit, diabetes nurse

educator visit and smoking cessation counselling. In the

cluster observational cohort study by Juul et al.25, partici-

pants in general practices with well-implemented nurse-

led type 2 diabetes consultations were more likely (mean

difference of 6%, 95% CI 1.5–11.4) to have had their

HbA1c measured during the 180 days of observation

than participants from practices that did not have a

nurse employed.

ConclusionAs practice nurses are often afforded more time in their

consultations, they may be suitable to collaborate in the

care of diabetes patients, especially in remote areas of

Australia where there are shortages of GPs and diabetes

specialists required for treatment and management of

diabetes.2 A review of the literature and subsequent

meta-analysis indicates that practice nurses employed

in a GP, participating in the management of adults with

diabetes type 2, yielded only modest improvements in a

minority of clinical outcomes. In an RCT specifically for

diabetes patients with poor control at baseline, however,

Taylor et al.20 observed significantly more robust

reduction of mean HbA1c, SBP and DBP with an inter-

vention that allowed nurses to titrate diabetes medi-

cations. Also of note, Gabbay et al.28 identified a

correlation between nurse management and signifi-

cantly improved diabetes-related screening such as

increased foot exams. Well-designed RCTs are needed,

which measure practice nurse management of diabetes

type 2 patients with poor control on clinical outcomes

that include HbA1c, SBP, DBP, TC as well as outcomes

related to screenings, such as foot ulcers.

AcknowledgementsMr. Ridhwan Shams was supported by a University of

Queensland Summer Research Scholarship.



D Parker et al.

©2016 U

Conflicts of interestThere are no conflicts of interest.

References1. National Health Priority Action, Council. National service

improvement framework for diabetes. Canberra: Australian

Government Department of Health and Ageing; 2006.

2. Alliance NRH. Diabetes in rural Australia: fact sheet 21

Canberra: Diabetes Australia; 2011.

3. Diabetes Australia. Diabetes Australia’s plan for a better life

for people with diabetes. Canberra: Diabetes Australia; 2004.4. Australian Institute of Health and Welfare. Australia’s health

2014. Canberra: Australian Institute of Health and Welfare;

2014; 141-143.

5. Australian Institute of Health and Welfare. Diabetes expen-

diture in Australia 2008–09. Canberra: Australian Institute

of Health and Welfare; 2013.

6. NSW Chronic Care Program. Strengthening general prac-

titioner involvement in chronic care: review and recom-

mendations Sydney: NSW Department of Health; 2005.

7. Oldroyd J, Proudfoot J, Infante FA, et al. Providing health-

care for people with chronic illness: the views of Australian

GPs. Med J Aust 2003; 179: 30–3.

8. Wagner EH. Chronic disease management: what will it take

to improve care for chronic illness? Eff Clin Pract 1998; 1:

2–4.

9. Laurant M, Reeves D, Hermens R, Braspenning J, Grol R,

Sibbald B. Substitution of doctors by nurses in primary care.

Cochrane Database Syst Rev 2005; 2: CD001271.

10. Zwar N, Harris M, Griffiths R, et al. A systematic review of

chronic disease management. Sydney: Research Centre for

Primary Healthcare and Equity, School of Public Health and

Community Medicine, UNSW; 2006.

11. Alliance AML. 2012 General practice nurse National Survey

Report.Australian Medicare Local Alliance; 2012.

12. Jolly R. Practice nursing in Australia. Report No.: 10, 2007–

08.Parliament of Australia, Department of Parliamentary

Services; 2007.

13. Davidson MB. Effect of nurse-directed diabetes care in a

minority population. Diabetes Care 2003; 26: 2281–7.

14. Harris DI, Cracknell P. Improving diabetes care in general

practice using a nurse-led, GP supported clinic: a cohort

study. Pract Diabetes Int 2005; 22: 295–301.

15. Vrijhoef HJM, Diederiks JPM, Spreeuwenberg C, Wolffen-

buttel BHR. Substitution model with central role for nurse

specialist is justified in the care for stable type 2 diabetic

outpatients. J Adv Nurs 2001; 36: 546–55.

16. Woodward A, Wallymahmed M, Wilding J, Gill G. Improved

glycaemic control an unintended benefit of a nurse-led

cardiovascular risk reduction clinic. Diabetic Med 2005; 22:

1272–4.

17. Woodward V. Commentary on Halcomb E.J., Patterson E. &

Davidson P.M. (2006) Evolution of practice nursing in

Australia. Journal of Advanced Nursing 55(3), 376–390.

J Adv Nurs 2006; 55: 388–90.



18. Renders CM, Valk GD, Griffin SJ, Wagner EH, van Eijk JT,

Assendelft WJJ. Interventions to improve the management

of diabetes in primary care, outpatient, and community

settings: a systematic review. Diabetes Care 2001; 24:

1821–33.

19. Parker D, Clifton K, Shams R, Young J. The effectiveness of

nurse-led care in general practice on clinical outcomes

in adults with type 2 diabetes. JBI Lib Syst Rev 2012; 10:

2514–58.

20. Taylor C, Miller N, Reilly K, et al. Evaluation of a nurse-

care management system to improve outcomes in

patients with complicated diabetes. Diabetes Care 2003;

26: 1058–63.

21. Blackberry ID, Furler JS, Best JD, et al. Effectiveness of

general practice based, practice nurse led telephone

coaching on glycaemic control of type 2 diabetes: The

Patient Engagement and Coaching for Health (PEACH)

pragmatic cluster randomised controlled trial. BMJ 2013;

347: f5272.

22. McKenzie J, Ryan R, Di Tanna GL. Cochrane Consumers and

Communication Review Group: Cluster randomised con-

trolled trials 2014; http://cccrg.cochrane.org (March 2014;

accessed May 2015).

23. Higgins JPT, Green SE (Eds): Cochrane handbook for system-

atic reviews of interventions. Version 5.1.0. The Cochrane

Collaboration; 2011. (Updated March 2011). www.cochrane-

handbook.org.

24. Campbell MK, Mollison J, Grimshaw JM. Cluster trials in

implementation research: estimation of intracluster corre-

lation coefficients and sample size. Stat Med 2001; 20:

391–9.

25. Juul L, Maindal HT, Frydenberg M, Kristensen JK, Sandbaek

A. Quality of type 2 diabetes management in general

practice is associated with involvement of general practice

nurses. Prim Care Diabetes 2012; 6: 221–8.

26. Cleveringa FGW, Gorter KJ, van den Donk M, Rutten GEHM.

Combined task delegation, computerized decision support,

and feedback improve cardiovascular risk for type 2 dia-

betic patients: a cluster randomized trial in primary care.

Diabetes Care 2008; 31: 2273–5.

27. Meulepas MA, Braspenning JCC, de Grauw WJ, Lucas AEM,

Wijkel D, Grol R. Patient-oriented intervention in addition to

centrally organised checkups improves diabetic patient

outcome in primary care. Qual Saf Healthcare 2008; 17:

324–8.

28. Gabbay RA, Lendel I, Saleem TM, et al. Nurse case manage-

ment improves blood pressure, emotional distress and

diabetes complication screening. Diabetes Res Clin Pract

2006; 71: 28–35.

29. Taylor KI, Oberle KM, Crutcher RA, Norton PG. Promoting

health in type 2 diabetes: nurse–physician collaboration in

primary care. Biol Res Nurs 2005; 6: 207–15.

30. Leal J, Gray AM, Clarke PM. Development of life-expectancy

tables for people with type 2 diabetes. Eur Heart J 2009; 30:

834–9.



http://www.cochrane-handbook.org/

http://www.cochrane-handbook.org/

practice nurse involvement in the management of adults

Documents