table of contents-five key articles for sit cliniciansfigure 1. the syndrome of chronic critical...

FIVE KEY ARTICLES FOR CLINICIANS IN THE SIT CLINICAL TRIAL

1) CONCISE CLINICAL REVIEW: CHRONIC CRITICAL ILLNESS Nelson JE et al. Am J Respir Crit Care Med. 2010; 182:446-454

2) COMMUNICATION ABOUT CHRONIC CRITICAL ILLNESS Nelson JE et al. Arch Intern Med. 2007; 167(22):2509-2515

3) ONE YEAR TRAJECTORIES OF CARE AND RESOURCE UTILIZATION FOR RECIPIENTS OF PROLONGED MECHANICAL VENTILATION Unroe M et al. Ann Intern Med. 2010;153:167-175

4) BRAIN DYSFUNCTION: ANOTHER BURDEN FOR THE CHRONICALLY CRITICALLY ILL Nelson JE et al. Arch Intern Med. 2006;166:1993-1999

5) A PROGNOSTIC MODEL FOR 1-YEAR MORTALITY IN PATIENTS REQUIRING PROLONGED MECHANICAL VENTILATION Carson SS et al. Crit Care Med. 2008;36(7):2061-2069

Concise Clinical Review

Chronic Critical IllnessJudith E. Nelson1, Christopher E. Cox2, Aluko A. Hope1,3, and Shannon S. Carson4

1Division of Pulmonary, Critical Care, and Sleep Medicine, Department of Medicine, and Hertzberg Palliative Care Institute, Mount Sinai School ofMedicine, New York, New York; 2Division of Pulmonary and Critical Care Medicine, Department of Medicine, Duke University School of Medicine,Durham, North Carolina; 3Department of Geriatrics and Palliative Medicine, Mount Sinai School of Medicine, New York, New York; 4Division ofPulmonary and Critical Care Medicine, Department of Medicine, University of North Carolina, Chapel Hill, School of Medicine, Chapel Hill, NorthCarolina

Although advances in intensive care have enabled more patients tosurvive an acute critical illness, they also have created a large andgrowing population of chronically critically ill patients with pro-longed dependence on mechanical ventilation and other intensivecare therapies. Chronic critical illness is a devastating condition:mortality exceeds that for most malignancies, and functional de-pendence persists for most survivors. Costs of treating the chroni-cally critically ill in the United States already exceed $20 billion andare increasing. In this article, we describe the constellation of clinicalfeatures that characterize chronic critical illness. We discuss theoutcomesof this condition including ventilator liberation,mortality,and physical and cognitive function, noting that comparisonsamong cohorts are complicated by variation in defining criteriaand care settings. We also address burdens for families of thechronically critically ill and the difficulties they face in decision-making about continuation of intensive therapies. Epidemiologyand resource utilization issues are reviewed to highlight the impactof chronic critical illness on our health care system. Finally, wesummarize the best available evidence formanaging chronic criticalillness, including ventilator weaning, nutritional support, rehabili-tation, and palliative care, and emphasize the importance of effortsto prevent the transition from acute to chronic critical illness. Assteps forward for thefield,we suggest a specificdefinitionof chroniccritical illness, advocate for the creation of a research networkencompassing a broad range of venues for care, and highlight areasfor future study of the comparative effectiveness of differenttreatment venues and approaches.

Keywords: respirator, artificial; critical illness; chronic disease; respira-tory care units

Although advances in intensive care have enabled more pa-tients to survive an acute critical illness, they have also createda large and growing population of patients with prolongeddependence on mechanical ventilation and other intensive caretherapies. The term ‘‘chronically critically ill’’ was coined forthis group by Girard and Raffin in a 1985 article that asked in itstitle, ‘‘to save or let die’’? (1). They focused on patients whosurvived an initial episode of critical illness but remaineddependent on intensive care, neither dying in the acute period

of intensive care unit (ICU) treatment nor recovering. Esti-mates indicate that there are more than 100,000 such patients inthe United States at any point in time, and increasing numbersin some other countries. Chronic critical illness is a devastatingcondition for patients and their families and, at a cost exceeding$20 billion each year, for the U.S. health care system as a whole.Incidence and expenditures are rising as more older adults, whoalready comprise the majority of chronically critically ill pa-tients, receive aggressive medical and surgical treatments.

In this article, we review clinical features and outcomes ofthe chronic critical illness syndrome, its impact on the healthcare system, and the challenges it presents for treatment anddecision-making. Data are drawn from a search of the Medlinedatabase from 1975 to March 2010 for all English languagearticles using the terms ‘‘chronic critical illness’’ or ‘‘chronicallycritically ill,’’ ‘‘prolonged critical illness,’’ ‘‘prolonged mechan-ical ventilation,’’ ‘‘prolonged ventilator weaning,’’ ‘‘post-ICU,’’‘‘long-term acute care facility,’’ ‘‘respiratory care unit,’’ and‘‘tracheostomy’’ or ‘‘tracheotomy’’; we also hand-searchedreference lists and author files.

CLINICAL FEATURES OF THE CHRONIC CRITICALILLNESS SYNDROME

The hallmark of chronic critical illness is respiratory failurerequiring prolonged dependence on mechanical ventilation.Although the term ‘‘prolonged mechanical ventilation’’ has beenused in the literature to describe periods of ventilator depen-dence ranging from 2 days to 4 weeks (2–5), this period is usuallymeasured in weeks for the chronically critically ill. Besidesprolonged ventilator dependence, evidence suggests that chroniccritical illness is a syndrome comprising additional characteristicfeatures. These include profound weakness attributed to myop-athy, neuropathy, and alterations of body composition includingloss of lean body mass, increased adiposity, and anasarca (6);distinctive neuroendocrine changes including loss of pulsatilesecretion of anterior pituitary hormones, contributing to lowtarget organ hormone levels and impaired anabolism (7, 8);increased vulnerability to infection, often with multiresistantmicrobial organisms (9, 10); brain dysfunction manifesting ascoma or delirium that is protracted or permanent (11); and skinbreakdown associated with nutritional deficiencies, edema, in-continence, and prolonged immobility (12). Patient reportsdocument significant distress from symptoms including pain,thirst, dyspnea, depression, and anxiety, and from inability tocommunicate during endotracheal intubation (13). This constel-lation of features, summarized in Figure 1, serves as a frameworkfor the clinical definition of chronic critical illness. Some of thesefeatures (e.g., brain dysfunction, symptom distress) may bepresent during acute critical illness (or other conditions), buttheir prolonged duration and intensity in the chronic phase of

(Received in original form February 2, 2010; accepted in final form March 6, 2010)

Supported by an Academic Career Leadership Award from the National Instituteon Aging (K07-AG034234; J.E.N.) and by a Mentored Patient-oriented ResearchCareer Development Award from the National Heart, Lung, and Blood Institute(K23 HL081048; C.E.C.).

Correspondence and requests for reprints should be addressed to Judith E.Nelson, M.D., J.D., Box 1232, Mount Sinai School of Medicine, 1 Gustave LevyPlace, New York, NY 10029. E-mail: [email protected]

Am J Respir Crit Care Med Vol 182. pp 446–454, 2010Originally Published in Press as DOI: 10.1164/rccm.201002-0210CI on May 6, 2010Internet address: www.atsjournals.org

critical illness are distinctive. Other features (e.g., changes inbody composition and neuroendocrine patterns) have been de-scribed only in the chronic phase. Chronic critical illness isuniquely characterized by the presence of these features asa clinical constellation in association with prolonged dependenceon mechanical ventilation.

Between 5 and 10% of patients who require mechanicalventilation for acute conditions develop chronic critical illness(14–16). Patients from any type of medical or surgical ICU can beaffected. On the basis of data from statewide databases, the mean(SD) age for adult patients is 65 (15) years (17, 18); for those inspecialized weaning facilities, it is in the eighth decade (17, 19).Patients are evenly divided according to sex, and comorbiditiesare common (11, 17–19). Patients with trauma as an admittingdiagnosis are usually younger, more likely male, and have fewercomorbidities (18). More than one-third of chronically criticallyill patients receive care in teaching hospitals (18).

OUTCOMES

Generalization of outcomes from published reports is compli-cated by variation in study populations described as chronicallycritically ill, in definitions of outcomes of interest, and in post–acute care practices that affect hospital use. A threshold periodof mechanical ventilation, ranging from 2 to 30 days, has beenused to define the majority of cohorts for longitudinal studies(2–5). To limit the heterogeneity of these cohorts and therebyimprove comparability of outcomes across different studies,a consensus conference established a formal definition forprolonged mechanical ventilation: at least 21 consecutive dayson the ventilator for more than 6 hours/day (2). Other cohortstudies have identified chronically critically ill patients byelective placement of a tracheotomy to facilitate prolongedventilation and weaning efforts (20–23). Referral for tracheot-omy reflects the clinician’s judgment that the patient willneither wean nor die in the immediate future and thus providesa point of demarcation between acute and chronic critical illnessthat is both clinically meaningful and practical. In population-based studies, International Classification of Diseases (ICD)coding and Diagnosis-Related Groups (DRGs) have been used,respectively, to identify patients requiring specified thresholdperiods of mechanical ventilation and those receiving tracheot-

omy for failure to wean from the ventilator (17, 18). One studyvalidated an algorithm defining ‘‘prolonged mechanical venti-lation’’ on the basis of coding as either (1) ICD-9-96.72(mechanical ventilation . 96 h) plus an ICU length of stay ofat least 21 days or (2) classification in DRG 541/542 (tracheot-omy for a condition other than head, neck, or face disease)(sensitivity, 97.6%; specificity, 96.4%) (17). To address theimpact of cohort definition on outcome results, another studycompared outcomes of cohorts from the same hospital that weredefined either by placement of tracheotomy after at least 96hours of mechanical ventilation or by ventilation for at least 21days (24). In that study, patients identified by tracheotomy hadhigher rates of 1-year survival than those ventilated for at least21 days (52 vs. 42%), although both definitions captured a groupof patients with high resource use and poor clinical outcomes.

Ventilator Liberation

Between 30 and 53% of chronically critically ill patients areliberated from mechanical ventilation (defined as dischargedalive and breathing without assistance) in the acute care hospital(14, 25). Average time to ventilator liberation varies with severityand type of illness or injury, but typically ranges from 16 to 37days after intubation for respiratory failure (14, 17, 24, 25). Mostpatients who fail to achieve ventilator independence within 60days do not do so later (26, 27). Better outcomes are reported forsome specialized weaning units, but they often select patientswith higher potential for ventilator liberation and rehabilitation(10, 26, 28). Reimbursement incentives discourage some weaningfacilities from admitting patients who have severe irreversiblepulmonary processes, require hemodialysis, or have profoundneurologic injuries; outcomes in such facilities may be morefavorable because of this admission bias.

Mortality

Although patients who remain ventilator dependent are athigher risk of death, successful weaning does not ensure long-term survival as most patients with chronic critical illness haveunderlying comorbid conditions, residual organ dysfunction,and intercurrent complications (Table 1). Acute care hospitalmortality for unselected patients is generally reported in therange of 20–49% (11, 17, 24, 25). One-year mortality acrossstudy populations is 48–68% with little change over the past 20

Figure 1. The syndrome of chronic critical illness. Mostchronically critically ill patients are older adults who haveunderlying comorbid conditions and develop sepsis andother acute comorbidities with treatment for acute med-ical, surgical, neurologic, or cardiac critical illness. Beyondprolonged ventilator dependence, which is its hallmark,increasing evidence indicates that chronic critical illness isa syndrome encompassing other characteristic clinicalfeatures and affecting multiple systems and organs.

Concise Clinical Review 447

years (5, 10, 14, 24, 25, 28). Compared with patients requiringshort-term ventilation, the risk of death in chronically criticallyill patients remains particularly high between 60 and 100 daysafter initiation of mechanical ventilation (24).

Functional and Cognitive Impairments

Nearly all patients with chronic critical illness leave the hospitalwith profound impairments of physical function, cognitivestatus, or both, and most therefore require institutional care(11, 14, 29–31). Hospital readmission rates during the year afterhospital discharge exceed 40% (32). Patients discharged toextended care facilities who cannot be sufficiently rehabilitatedfor return to home by 6 months usually remain institutionalizeduntil death (33). Across multiple studies, fewer than 12% ofchronically critically ill patients were alive and independent 1year after their acute illness (11, 28, 31). Long-term survivorswho are able to respond to surveys of health-related quality oflife (HRQOL) typically report better emotional and socialfunction than physical function or symptom experience (14,34). Although these findings are consistent with studies ofHRQOL in survivors of acute critical illness, which indicatethat some patients can adapt emotionally to profound changesin health status (35), interpretation must be tempered by thefact that the majority of chronically critically ill patients do notsurvive 1 year and most survivors lack sufficient cognitivefunction to respond to HRQOL surveys (11).

Family Burdens

Chronic critical illness also imposes heavy burdens on families,who experience high rates of depression and practical andfinancial hardships (36, 37). Evidence from studies of informalcaregivers of patients requiring prolonged mechanical ventila-tion indicates that depressive symptoms are more severe in thisgroup than among caregivers of patients with Alzheimer’sdisease or spinal cord injury and that depression continues formonths after the patient’s discharge from the hospital (37).These caregivers also report a decline in physical health andincrease in ‘‘caregiving overload’’ during the postdischargeperiod (37). In a study of post-ICU caregiver burden, multivari-able regression analysis identified the placement of a tracheot-omy in the ICU as a significant predictor of lifestyle disruptionfor informal caregivers of ICU survivors at 1 year after ICUadmission (38). Another study, focusing on patients who un-derwent tracheotomy after at least 4 days of mechanicalventilation or ventilation for at least 21 days, found that 61%of 1-year survivors still required daily assistance from informal

caregivers, who reported ‘‘a lot’’ or ‘‘severe’’ stress fromcaregiving; 84% of these caregivers had either quit work orsignificantly changed work hours to accommodate the patient’scaregiving needs (39). Several studies have shown that burdensfor families are not limited to those who provide the care athome (36, 37); depressive symptoms, caregiving overload, andphysical deterioration may actually be more severe amongfamilies of patients who are institutionalized than of thosewho return home (37). In addition, families of patients withprotracted critical illness often face significant losses of incomeand/or savings, even if the patient is insured (40).

Mortality Prediction

Evidence suggests that several patient-related factors contributesignificantly to variation in clinical outcomes. Advanced age andresidual organ failures, both common characteristics of chron-ically critically ill patients, are associated with higher mortality,as is poor prior functional status (29, 31, 41). Younger patientswith trauma initiating critical illness have had more favorableoutcomes (21). Neither the APACHE (Acute Physiology andChronic Health Evaluation) system nor other models forpredicting mortality of acutely critically ill patients are validfor chronic critical illness (41, 42). A simple mortality predictionmodel was developed to identify chronically critically ill pa-tients at high risk for 3-month and 1-year mortality (25);a multicenter study to provide external validation of this modelis ongoing. The mortality model does not address prognosis forfunctional or cognitive recovery, which many patients andfamilies consider as seriously as prospects for survival (20).

Deficiencies in Physician–Patient–Family Communication

Studies reveal that these clinical outcomes are poorly under-stood by family decision-makers for chronically critically illpatients and even by physicians. In questionnaire-based in-terviews conducted shortly after tracheotomy, ICU patients andsurrogates reported that key aspects informing the decision toprovide prolonged life support were routinely omitted byphysicians; for example, 80 and 93% of the respondents re-ceived no information about possible functional dependency athospital discharge or about expected 1-year survival, respec-tively (20). Similarly, qualitative interviews of family surrogatesin another study found that three-quarters lacked accurateestimates of survival, functional status, and caregiving needsfor patients requiring prolonged mechanical ventilation; 93%expected that the patient would survive for at least 1 year, andless than 30% anticipated any functional limitation or other

TABLE 1. OUTCOMES OF CHRONIC CRITICAL ILLNESS RELATIVE TO COHORT DEFINITION*

Cohort Definition Ventilation for >14 d Ventilation for >21 dTracheotomy for Prolonged

Mechanical Ventilation

Study (reference) Combes et al. (34) Carson et al. (25) Cox et al.† (24) Cox et al.† (24) Engoren et al. (14)n 347 200 114 267 347Age (yr), median (IQR) or mean 6 SD 63 6 14, 67 6 13‡ 58 (42–69) 66 (47–74) 66 (45–75) 64, 71x

Duration of ventilation, median (IQR)or mean 6 SD

36 6 25, 37 6 28‡ 35 (26–51) 27 (23–36) 16 (10–24) 23–30k

Hospital length of stay, median (IQR) — 51 (36–72) 39 (30–52) 29 (22–38) 28–37k

Died in hospital, % 43 41 31 20 22Discharged home, % — 11 4 7 —Alive at 12 mo, % 32 48 42 52 50

* Studies were selected if they were prospective cohort studies published after 2000, enrolled patients in the acute hospital setting who received at least 14 days ofmechanical ventilation, and reported 12-month survival for comparison.

† Study reporting outcomes of two different cohorts distinguished by definition.‡ Mean 6 SD for ICU survivors and for nonsurvivors, respectively.x Median for hospital survivors and for nonsurvivors, respectively.k Range of medians for hospital survivors who were ventilator dependent, liberated from ventilator but with tracheotomy, or liberated and decannulated; and for

nonsurvivors.

448 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 182 2010

impairment of quality of life (39). These expectations divergeddramatically from those held by the patients’ physicians, whowere less optimistic about chances for 1-year survival (expectedby 44% of physicians) and much less so for functional recovery(expected by 6% of physicians). Both families and physicianssignificantly overestimated the patients’ actual outcomes (39).

IMPACT ON HEALTH CARE SYSTEM

Chronic critical illness is a serious and growing problem for theU.S. health care system and an emerging challenge in othercountries (23, 43). In population-based studies in the UnitedStates, the incidence of respiratory failure requiring mechanicalventilation has been increasing by as much as 5.5% per year (44,45) and will soon increase at an even faster rate as BabyBoomers pass age 60, when the risk of respiratory failure risesdramatically (46). Numbers of the chronically critically ill, whoare mostly older adults, will increase as much or more. Analysisof a state database of hospital discharges showed that theincidence of tracheotomy for prolonged ventilation nearly tri-pled between 1993 and 2002, from 8.3 per 100,000 population to24.2 per 100,000 (18). Another population-based study pro-jected that the number of patients requiring mechanical venti-lation for at least 7 days in the United States will more thandouble from 250,000 in 2000 to more than 600,000 in 2020 (44).Although the chronically critically ill account for fewer than 10%of those receiving mechanical ventilation, they consume between20 and 40% of ICU bed days and other critical care resources (16,18). Because functional limitations are common, even patientswho recover sufficiently to permit discharge from an inpatientfacility typically require paid caregiving as outpatients or familymembers must leave jobs to provide ongoing care. The overallcost to the health care system for the management of chroniccritical illness already exceeds an estimated $20 billion per year(29) and is expected to climb with increases in the incidence ofthis syndrome and in overall expenditures for critical care, whichnearly doubled between 1985 and 2000 and represent 13% of allhospital costs in the United States (47).

Cost-Effectiveness

A cost-effectiveness analysis calculated that providing pro-longed mechanical ventilation to Medicare-eligible patientswith multiple comorbid conditions exceeds $200,000 for eachquality-adjusted life year gained, as compared with patients whohad life-sustaining therapies withdrawn before Day 14 ofmechanical ventilation (29). This analysis demonstrated thatincremental costs associated with prolonged life support weremost sensitive to acute hospital costs (rather than post–acutecare facility costs) and hospital readmissions. Innovations toreduce costs in addition to improving clinical outcomes are

needed. A specialized disease management program failed toreduce the risk of readmission for patients ventilated for 3 daysor more, but it was associated with a decreased length of stayduring rehospitalization from 16 to 11.4 days, yielding anaverage cost saving of more than $50,000 per patient (3).

Venues of Care

Patients with chronic critical illness receive care across a rangeof venues (Table 2). Besides acute care hospitals, these includelong-term acute care facilities, skilled nursing facilities,inpatient rehabilitation facilities, and chronic ventilator facil-ities, with various resources for the complex needs and de-pendencies that characterize the chronically critically ill.Diagnosis-Related Groups (DRGs) covering these patientsare among the most heavily weighted, supporting relativelyhigh reimbursement to acute care hospitals. Nevertheless, highcosts for long-stay outliers are a burden for these hospitals,creating an incentive for transferring chronically critically illpatients to post–acute care facilities for further attempts atweaning and rehabilitation. Such facilities have become profit-able to operate, contributing to rapid expansion of the for-profit long-term acute care (LTAC) industry, which grew ata rate of 12% per year between 1993 and 2003 (48). Medicarepayments to LTACs, which cover more than 80% of theirdischarges, have increased by 15% annually (48). Costs overthe entire episode of illness are lower for chronically criticallyill Medicare beneficiaries who are transferred on mechanicalventilation to LTACs (48), probably because of lower nurse-to-patient ratios and staff salaries, and efficiencies in ventilatorweaning and rehabilitation services. Data are inconclusive,however, on whether patient survival is affected by LTACtransfer (48). Reports by individual LTACs suggest thatpatients are being transferred from acute care hospitals earlierin the course of critical illness, with higher degrees of illnessseverity (49). If these trends of earlier transfer to LTACscontinue, the cost savings associated with care in such facilitiesmay be offset by an increase in rates of subsequent readmis-sions to acute care hospitals. Congress has mandated reform ofpayment to post–acute care facilities; this process is currentlyin the demonstration phase and has expanded to address therole of acute care hospitals in providing care for chronicallycritically ill patients (50).

TREATMENT CHALLENGES

Even as attention focuses increasingly on chronic critical illnessfrom the perspectives of epidemiologic and health services,empirical research to define effective methods of treatmentremains scant. Most data on specific therapeutic approachesderive from descriptive studies conducted in single centers,

TABLE 2. MAIN VENUES FOR CARE OF THE CHRONICALLY CRITICALLY ILL*

Nursing Intensity/Cost Patient Acuity Specialized Rehabilitation Approach

Acute care hospitalsIntensive care units 1 1 1 1 1 1 1Critical care stepdown units 1 1 1 1 1Specialized weaning units 1 1 1 1 1 1 1Medical–surgical ward 1 1 1Long-term acute care (LTAC)or long-term care hospitals (LTCH)

1 1 1 1 1 1 1

Skilled nursing facilities 1 1 1 1

* Depending on geographic location and available resources within and outside the acute hospital setting, most chronically critically ill patients in the United States arecared for in acute care hospitals, long-term acute care facilities or, less commonly, skilled nursing facilities. Relative levels of nursing intensity/cost, patient acuity, andavailability of specialized rehabilitation are noted for the typical facility in these venue categories, although there is variation within as well as across categories.


leaving clinicians to rely mainly on their own experience andextrapolation of evidence from studies of acutely critically illpatients, which may lack external validity in this setting.

Ventilator Weaning

As shown for ICU patients in prospective, randomized, con-trolled trials, a before–after study showed the effectiveness ofa protocol implemented by respiratory therapists for weaningpatients with tracheotomy from prolonged mechanical ventila-tion in a long-term acute care facility (51). Median time toventilator liberation was 17 days during an 18-month periodafter the protocol was implemented, compared with 29 days ina historical control group. Therapists used a rapid shallowbreathing index (RSBI) of no more than 80 as an ‘‘accelerationstep’’ to advance patients to spontaneous breathing trials. Asubsequent prospective observational study found that a higherthreshold RSBI of 100 (as used in patients with shorter termventilation) accelerated weaning without significantly loweringthe specificity of the index (52). In many LTACs and othervenues, weaning protocols are successfully managed by non-physicians (19, 53). Typically, trials of pressure support at a level(10–15 cm H2O) that is approximately half of full ventilatorsupport are followed by spontaneous breathing trials, usinga ‘‘trach collar’’ or ‘‘T-piece approach’’ for progressively in-creasing periods (2, 43). Standardized criteria can be used toassess readiness for final removal of a tracheotomy tube(‘‘decannulation’’) after ventilator liberation (54).

Nutritional Support

Beyond ventilator weaning, the syndrome of chronic criticalillness calls for a broader, multidisciplinary, therapeutic ap-proach that addresses all major clinical features (Figure 2).Strategies have been suggested, but not empirically tested, toaddress the kwashiorkor-like malnourished state (6). Rationalgoals of nutritional support reflect a balancing of potentialbenefits of providing metabolic substrates to minimize furtherloss of lean body mass against adverse consequences of over-feeding and other risks. For the patient with a functional

gastrointestinal tract, enteral feeding is recommended as first-line therapy, and observational evidence favors placement ofa percutaneous endoscopic gastrostomy or jejunostomy fornutritional support to exceed 30 days (55). Stress hyperglyce-mia, originating during acute critical illness, typically persistsand requires insulinization.

Functional and Cognitive Recovery

Integration of physical therapy in a comprehensive rehabilita-tive model for care is supported by a consensus of expertopinion (2) and by evidence emerging from the setting of acutecritical illness, which indicates that early mobilization maymitigate development, severity, and/or duration of post-ICUmuscle weakness (56). Initiation of this approach during thechronic phase of critical illness, when the patient is alreadycachectic, profoundly weak, and debilitated, has not yet beenspecifically studied. Biochemical evidence supports treatmentwith calcitriol and pamidronate to attenuate accelerated boneloss (22), but clinical implications for recovery of strength orfunction remain unclear. Although some data suggest thatsedation and analgesia can be reduced after tracheotomy (57),it is not known whether this will decrease the prevalence orprolonged duration of brain dysfunction. Extrapolation fromthe acute ICU setting suggests that, because of their deliriogenicpotential (58, 59), benzodiazepines should be avoided aspossible. Haloperidol and newer atypical antipsychotics suchas ziprasidone have been recommended for control of agitationor delirium in the ICU, but data are limited regarding theirefficacy in reducing delirium, especially the hypoactive form(60).

Preventing Complications

Attempts to prevent and treat infectious and other complica-tions, which cause morbidity, mortality, and persistent ventila-tory insufficiency (61), require assiduous efforts. Patients facea ‘‘triple threat’’ of risk of infection, the most commoncomplication (10): barrier breaches, such as intravenous cathe-terization and skin breakdown; exposure to virulent and re-

Figure 2. Comprehensive care for thechronically critically ill. Comprehensive carefor the chronically critically ill includes mul-tiple components, as illustrated here anddiscussed more fully in text, with five keygoals: ventilator liberation, nutritional sup-port, cognitive and functional recovery, pre-vention of complications, and attention topalliative needs. Given the unique and com-plex challenges, a dedicated interdisciplin-ary team of professionals may be bestequipped to provide this care.


sistant pathogens in ICU and post–ICU care environments; andpostulated ‘‘immune exhaustion’’ from recent critical illness andcomorbidities (9). Processes of care should be systematized tomaximize use of essential preventive measures such as hand-washing, isolation, removal of unnecessary indwelling catheters,restriction of antibiotic use, and best practices for maintainingskin integrity (12). Source identification and control shouldfocus first on possible line sepsis, pneumonia, and Clostridiumdifficile colitis, which account for the majority of infections (9).

Care Models

Effective and efficient care can be provided to the chronicallycritically ill outside of the ICU, either in specialized in-hospitalunits or free-standing facilities, with lower levels of nursingintensity, technology, and ancillary care (62). In these venues,nurses have played a key role in structuring and managing careby an interdisciplinary team that is dedicated to the special needsof this resource-intensive patient group (12, 62). A ‘‘mobile’’team led by advanced practice nurses using a protocol-basedapproach improved outcomes and reduced costs for ICU pa-tients requiring mechanical ventilation for more than 3 days(63).

Tracheotomy Timing and Other Issues

Although placement of a tracheotomy for patients with pro-longed weaning failure is a clinical marker of the transitionbetween the acute and chronic phases of critical illness, there isdebate about whether earlier placement of a tracheotomy canreduce ventilator days and therefore reduce the likelihood ofcomplications that can lead to chronic critical illness (64).Multicenter trials designed to help resolve this debate areongoing, while the average time from initiation of mechanicalventilation to tracheotomy placement is decreasing in clinicalpractice (18, 65). It is likely that other evidence-based practicesthat reduce ventilator days for patients in the acute ICU settingwill help to decrease the incidence of chronic critical illnesswhen applied systematically during the acute phase. Efficientliberation from mechanical ventilation requires organized ICUmanagement practices, preferably directed by certified intensiv-ists in a ‘‘closed’’ ICU model (66, 67). Weaning protocols shouldinclude daily spontaneous breathing trials (68), daily lighteningof continuous sedatives (69), and avoidance of long-actingsedatives (70), as all of these strategies shorten duration ofmechanical ventilation and ICU length of stay. Systematicapproaches to prevention of ventilator-associated pneumoniaand central line–associated bloodstream infections are sup-ported by existing evidence (71).

Palliative Care

Palliative care is an essential component of comprehensivetreatment for all chronically critically ill patients, includingthose receiving life-prolonging therapies. This care includessensitive, effective, proactive, and ongoing communication withpatients and families about prognosis, achievable goals oftreatment, and alternatives to continuation of critical care(20). Ideally, goals should be defined by the patient’s prefer-ences for treatment and views of acceptable function andquality of life, but most chronically critically ill patients areunable to participate directly in discussions or decision-makingand few have designated a surrogate decision-maker or pre-pared another advance directive (72). A prospective studyshowed that treatment and decision-making in chronic criticalillness generally proceeded without direct input from thepatient (most lacked capacity when critical illness becamechronic and 85% lacked an advance directive expressing

treatment preferences) (72). Limitation (withholding or with-drawal) of life-supporting therapies (mechanical ventilation,renal replacement therapy, artificial nutrition, intravenoushydration, or vasopressors) was rare—fewer than 1 in 5 (39 of203) patients—and late in the course (median, 39 d fromhospital admission), when the patient was near death. Inanother study, one-third of family surrogates denied any rolein deciding to continue mechanical ventilation for a prolongedperiod, stating their understanding that this decision was madeexclusively by the physician (39).

To ensure meaningful participation by patients and familieswho wish to share in decision-making, clinicians should engagethem in a mutual exchange, providing relevant medical in-formation in terms that are clear and understandable toa layperson, while investigating the patient’s values and goals.A new model based on four simple measures may be useful toestimate 1-year survival of patients requiring prolonged me-chanical ventilation (25). The use of objective mortality pre-diction models for guiding discussions of prognosis and goals ofcare is controversial. The SUPPORT (Study to UnderstandPrognoses and Preferences for Outcomes and Risks of Treat-ments) study, which tested communication of model-derivedprognostic information from physicians to patients througha research nurse intermediary, did not favorably alter physicianbehavior or clinical outcomes for seriously ill patients hospital-ized with acute illness (73). It is possible, however, that a simplermodel designed to identify chronically critically ill patients whoare at the greatest risk for mortality with a high degree ofspecificity will allow clinicians to be more confident in discus-sing poor prognoses directly. Patients, families, and evenclinicians may fail to appreciate the ongoing risks of death orsevere disability when the patient has just survived the acutephase of critical illness (20, 39). A new brochure about chroniccritical illness is available as an adjunct to direct cliniciancounseling for education of patients and families (74); the valueof printed informational materials for this purpose has beenshown in randomized, controlled trials (75, 76). In a randomized,controlled, multicenter trial that is newly funded by theNational Institutes of Health, this brochure will be given tofamilies of chronically critically ill patients in both the controland intervention groups; in addition, a ‘‘Supportive InformationTeam’’ including a palliative care physician and nurse, whichthe ICU attending physician will have the option to join, willconduct proactive meetings with families in the interventiongroup.

Interdisciplinary support addressing families’ emotional,spiritual, and practical needs is helpful as a framework fordiscussions and decision-making about continuation of intensivecare therapies when critical illness enters a chronic phase.Palliative care consultants are increasingly available to helpwith communication challenges and provide other support forpatients and families, as well as to optimize symptom controland transitional planning. Early integration of palliative carewith treatments for cure or longer life, ideally from admission tothe ICU and through the chronic phase, is recommended.

CONCLUSION

In their 1985 article, Girard and Raffin asked whether weshould attempt to save the chronically critically ill or let themdie (1). This question remains immediately relevant, but thepresent state of the evidence, 25 years later, does not yetsupport a definite response. Various factors have made itdifficult for research on chronic critical illness to progress morequickly. The successes of acute critical care in achieving short-term survival have partly obscured the scope and severity of the


problem of chronic critical illness, delaying the emergence ofthis area as an important focus of scientific investigation.Another barrier is the diversity of venues in which care iscurrently provided to the chronically critically ill, compoundingthe problem of generalization from one setting to another whilealso increasing the challenge of recruiting large cohorts forresearch. To move forward at a pace that matches the increasingincidence of this condition and the magnitude of its impact, thefield needs consensus on a definition. We suggest that place-ment of a tracheotomy after at least 10 days of mechanicalventilation be used to define the onset of chronic critical illnessbecause this definition incorporates the clinician’s judgmentthat the patient is not expected to die or to wean from theventilator in the immediate future. Although it may not benecessary or possible for all studies of chronic critical illness toconform exactly to this definition (e.g., studies using adminis-trative data as currently classified) (17), a common definitionwould be beneficial for interventional studies in which patientsare prospectively enrolled. We also suggest that federal fundsbe dedicated to support the development of a large researchnetwork encompassing a broad range of venues for care of thechronically critically ill. Pathobiology and pathophysiology ofchronic as distinct from acute critical illness deserve furtherscientific investigation. We also need well-designed trials testingapproaches to the many clinical challenges, from managementof prolonged mechanical ventilation to nutritional support totreatment of delirium, symptom distress, and physical weakness.Regarding appropriate venues for care, current evidence lagsfar behind trends in practice. Comparative effectiveness re-search that includes detailed economic analyses should beconducted to compare the cost-effectiveness of transferringchronically critically ill patients from acute hospitals to special-ized facilities (77). Whereas existing evidence for optimalmanagement strategies remains limited, data on long-termoutcomes are available and clear. These data should not beignored during discussions with patients and their families aboutappropriate goals of continuing intensive care therapies whencritical illness becomes chronic.

Author Disclosure: J.E.N. received more than $100,001 from the NationalInstitutes of Health (NIH) in sponsored grants (K02 Independent ScientistResearch Career Development Award, K07 Academic Career Leadership Award,and R21 and R01 research awards), more than $100,001 from the AmericanCancer Society as a research grant, and $10,001–$50,000 from the Departmentof Veterans Affairs as a consultant for the ICU-palliative care quality improvementinitiative in Veterans Integrated Service Network 3. C.E.C. received more than$100,001 from the NIH in sponsored grants (K23 Career Development Award).A.A.H. does not have a financial relationship with a commercial entity that has aninterest in the subject of this manuscript. S.S.C. received $1,000–$4,999 fromPassy-Muir Co. and $1,000–$4,999 from Asthmatx Co. in consultancy oradvisory board fees, and more than $100,000 from the NIH in institutionalresearch grants.

References

1. Girard K, Raffin TA. The chronically critically ill: to save or let die?Respir Care 1985;30:339–347.

2. MacIntyre NR, Epstein SK, Carson S, Scheinhorn D, Christopher K,Muldoon S. Management of patients requiring prolonged mechanicalventilation: report of a NAMDRC consensus conference. Chest 2005;128:3937–3954.

3. Daly BJ, Douglas SL, Kelley CG, O’Toole E, Montenegro H. Trial ofa disease management program to reduce hospital readmissions of thechronically critically ill. Chest 2005;128:507–517.

4. Chelluri L, Im KA, Belle SH, Schulz R, Rotondi AJ, Donahoe MP, SirioCA, Mendelsohn AB, Pinsky MR. Long-term mortality and quality oflife after prolonged mechanical ventilation. Crit Care Med 2004;32:61–69.

5. Gracey DR, Naessens JM, Krishan I, Marsh HM. Hospital and post-hospital survival in patients mechanically ventilated for more than 29days. Chest 1992;101:211–214.

6. Hollander JM, Mechanick JI. Nutrition support and the chronic criticalillness syndrome. Nutr Clin Pract 2006;21:587–604.

7. Van den Berghe G, de Zegher F, Veldhuis JD, Wouters P, Awouters M,Verbruggen W, Schetz M, Verwaest C, Lauwers P, Bouillon R, et al.The somatotropic axis in critical illness: effect of continuous growthhormone (GH)-releasing hormone and GH-releasing peptide-2 in-fusion. J Clin Endocrinol Metab 1997;82:590–599.

8. Van den Berghe G, de Zegher F, Veldhuis JD, Wouters P, Gouwy S,Stockman W, Weekers F, Schetz M, Lauwers P, Bouillon R, et al.Thyrotrophin and prolactin release in prolonged critical illness:dynamics of spontaneous secretion and effects of growth hormone-secretagogues. Clin Endocrinol (Oxf) 1997;47:599–612.

9. Kalb TH, Lorin S. Infection in the chronically critically ill: unique riskprofile in a newly defined population. Crit Care Clin 2002;18:529–552.

10. Scheinhorn DJ, Hassenpflug MS, Votto JJ, Chao DC, Epstein SK, DoigGS, Knight EB, Petrak RA. Post-ICU mechanical ventilation at 23long-term care hospitals: a multicenter outcomes study. Chest 2007;131:85–93.

11. Nelson JE, Tandon N, Mercado AF, Camhi SL, Ely EW, Morrison RS.Brain dysfunction: another burden for the chronically critically ill.Arch Intern Med 2006;166:1993–1999.

12. Carasa M, Polycarpe M. Caring for the chronically critically ill patient:establishing a wound-healing program in a respiratory care unit. Am JSurg 2004;188:18–21.

13. Nelson JE, Meier DE, Litke A, Natale DA, Siegel RE, Morrison RS.The symptom burden of chronic critical illness. Crit Care Med 2004;32:1527–1534.

14. Engoren M, Arslanian-Engoren C, Fenn-Buderer N. Hospital and long-term outcome after tracheostomy for respiratory failure. Chest 2004;125:220–227.

15. Seneff MG, Zimmerman JE, Knaus WA, Wagner DP, Draper EA.Predicting the duration of mechanical ventilation: the importance ofdisease and patient characteristics. Chest 1996;110:469–479.

16. Wagner DP. Economics of prolonged mechanical ventilation. Am RevRespir Dis 1989;140:S14–S18.

17. Kahn JM, Carson SS, Angus DC, Linde-Zwirble WT, Iwashyna TJ.Development and validation of an algorithm for identifying pro-longed mechanical ventilation in administrative data. Health ServOutcomes Res Methodol 2009;9:117–132.

18. Cox CE, Carson SS, Holmes GM, Howard A, Carey TS. Increase intracheostomy for prolonged mechanical ventilation in North Caro-lina, 1993–2002. Crit Care Med 2004;32:2219–2226.

19. Scheinhorn DJ, Hassenpflug MS, Votto JJ, Chao DC, Epstein SK,Doig GS, Knight EB, Petrak RA. Ventilator-dependent survivors ofcatastrophic illness transferred to 23 long-term care hospitals forweaning from prolonged mechanical ventilation. Chest 2007;131:76–84.

20. Nelson JE, Mercado AF, Camhi SL, Tandon N, Wallenstein S, AugustGI, Morrison RS. Communication about chronic critical illness. ArchIntern Med 2007;167:2509–2515.

21. Engoren M, Arslanian-Engoren C. Hospital and long-term outcome oftrauma patients with tracheostomy for respiratory failure. Am Surg2005;71:123–127.

22. Nierman DM, Mechanick JI. Biochemical response to treatment of bonehyperresorption in chronically critically ill patients. Chest 2000;118:761–766.

23. Estenssoro E, Reina R, Canales HS, Saenz MG, Gonzalez FE, ApreaMM, Laffaire E, Gola V, Dubin A. The distinct clinical profile ofchronically critically ill patients: a cohort study. Crit Care 2006;10:R89.

24. Cox CE, Carson SS, Lindquist JH, Olsen MK, Govert JA, Chelluri L.Differences in one-year health outcomes and resource utilization bydefinition of prolonged mechanical ventilation: a prospective cohortstudy. Crit Care 2007;11:R9.

25. Carson SS, Garrett J, Hanson LC, Lanier J, Govert J, Brake MC,Landucci DL, Cox CE, Carey TS. A prognostic model for one-yearmortality in patients requiring prolonged mechanical ventilation. CritCare Med 2008;36:2061–2069.

26. Bigatello LM, Stelfox HT, Berra L, Schmidt U, Gettings EM. Outcomeof patients undergoing prolonged mechanical ventilation after criticalillness. Crit Care Med 2007;35:2491–2497.

27. MacIntyre NR, Cook DJ, Ely EW Jr, Epstein SK, Fink JB, Heffner JE,Hess D, Hubmayer RD, Scheinhorn DJ. Evidence-based guidelinesfor weaning and discontinuing ventilatory support: a collective taskforce facilitated by the American College of Chest Physicians; theAmerican Association for Respiratory Care; and the AmericanCollege of Critical Care Medicine. Chest 2001;120:375S–395S.


28. Stoller JK, Xu M, Mascha E, Rice R. Long-term outcomes for patientsdischarged from a long-term hospital-based weaning unit. Chest 2003;124:1892–1899.

29. Cox CE, Carson SS, Govert JA, Chelluri L, Sanders GD. An economicevaluation of prolonged mechanical ventilation. Crit Care Med 2007;35:1918–1927.

30. Douglas SL, Daly BJ, Gordon N, Brennan PF. Survival and quality oflife: short-term versus long-term ventilator patients. Crit Care Med2002;30:2655–2662.

31. Carson SS, Bach PB, Brzozowski L, Leff A. Outcomes after long-termacute care. An analysis of 133 mechanically ventilated patients. Am JRespir Crit Care Med 1999;159:1568–1573.

32. Douglas SL, Daly BJ, Brennan PF, Gordon NH, Uthis P. Hospitalreadmission among long-term ventilator patients. Chest 2001;120:1278–1286.

33. Nasraway SA, Button GJ, Rand WM, Hudson-Jinks T, Gustafson M.Survivors of catastrophic illness: outcome after direct transfer fromintensive care to extended care facilities. Crit Care Med 2000;28:19–25.

34. Combes A, Costa M-A, Trouillet J-L, Baudot J, Mokhtari M, Gibert C,Chastre J. Morbidity, mortality, and quality-of-life outcomes ofpatients requiring >14 days of mechanical ventilation. Crit CareMed 2003;31:1373–1381.

35. Hofhuis JGM, van Stel HF, Schrijvers AJP, Rommes JH, Bakker J,Spronk PE. Conceptual issues specifically related to health-relatedquality of life in critically ill patients. Crit Care 2009;13:118.

36. Van Pelt DC, Milbrandt EB, Qin L, Weissfeld LA, Rotondi AJ, SchulzR, Chelluri L, Angus DC, Pinsky MR. Informal caregiver burdenamong survivors of prolonged mechanical ventilation. Am J RespirCrit Care Med 2007;175:167–173.

37. Douglas SL, Daly BJ. Caregivers of long-term ventilator patients:physical and psychological outcomes. Chest 2003;123:1073–1081.

38. Van Pelt DC, Schulz R, Chelluri L, Pinsky MR. Patient-specific, time-varying predictors of post-ICU informal caregiver burden: thecaregiver outcomes after ICU discharge project. Chest 2010;137:88–94.

39. Cox CE, Martinu T, Sathy SJ, Clay AS, Chia J, Gray AL, Olsen MK,Govert JA, Carson SS, Tulsky JA. Expectations and outcomes ofprolonged mechanical ventilation. Crit Care Med 2009;37:2888–2894.

40. Swoboda SM, Lipsett PA. Impact of a prolonged surgical critical illnesson patients’ families. Am J Crit Care 2002;11:459–466.

41. Dematte D’Amico JE, Donnelly HK, Mutlu GM, Feinglass J, JovanovicBD, Ndukwu IM. Risk assessment for inpatient survival in the long-term acute care setting after prolonged critical illness. Chest 2003;124:1039–1045.

42. Carson SS, Bach PB. Predicting mortality in patients suffering fromprolonged critical illness: an assessment of four severity-of-illnessmeasures. Chest 2001;120:928–933.

43. Schonhofer B, Euteneuer S, Nava S, Suchi S, Kohler D. Survival ofmechanically ventilated patients admitted to a specialised weaningcentre. Intensive Care Med 2002;28:908–916.

44. Zilberberg MD, de Wit M, Pirone JR, Shorr AF. Growth in adultprolonged acute mechanical ventilation: implications for healthcaredelivery. Crit Care Med 2008;36:1451–1455.

45. Carson SS, Cox CE, Holmes GM, Howard A, Carey TS. The changingepidemiology of mechanical ventilation: a population-based study.J Intensive Care Med 2006;21:173–182.

46. Behrendt CE. Acute respiratory failure in the United States: incidenceand 31-day survival. Chest 2000;118:1100–1105.

47. Halpern NA, Pastores SM, Greenstein RJ. Critical care medicine in theUnited States 1985–2000: an analysis of bed numbers, use, and costs.Crit Care Med 2004;32:1254–1259.

48. Medicare Payment Advisory Commission (MEDPAC). Report to theCongress: new approaches in Medicare, Chapter 5: Defining long-termcare hospitals. Washington DC: MEDPAC; 2004. Accessed June 1,2010. Available at www.medpac.gov/publications%5Ccongressional_reports%5CJune04_ch5.pdf

49. Scheinhorn DJ, Chao DC, Stearn-Hassenpflug M, LaBree LD, HeltsleyDJ. Post-ICU mechanical ventilation: treatment of 1,123 patients ata regional weaning center. Chest 1997;111:1654–1659.

50. RTI International. Medicare post-acute care payment reform demon-stration. Accessed June 1, 2010. Available at: http://www.pacdemo.rti.org

51. Scheinhorn DJ, Chao DC, Stearn-Hassenpflug M, Wallace WA. Out-comes in post-ICU mechanical ventilation: a therapist-implementedweaning protocol. Chest 2001;119:236–242.

52. Chao DC, Scheinhorn DJ. Determining the best threshold of rapidshallow breathing index in a therapist-implemented patient-specificweaning protocol. Respir Care 2007;52:159–165.

53. Carasa M, Nespoli G. Nursing the chronically critically ill patient. CritCare Clin 2002;18:493–507.

54. Ceriana P, Carlucci A, Navalesi P, Rampulla C, Delmastro M,Piaggi G, De Mattia E, Nava S. Weaning from tracheotomy inlong-term mechanically ventilated patients: feasibility of a decisionalflowchart and clinical outcome. Intensive Care Med 2003;29:845–848.

55. Dwolatzky T, Berezovski S, Friedmann R, Paz J, Clarfield AM, StessmanJ, Hamburger R, Jaul E, Friedlander Y, Rosin A, et al. A prospectivecomparison of the use of nasogastric and percutaneous endoscopicgastrostomy tubes for long-term enteral feeding in older people. ClinNutr 2001;20:535–540.

56. Schweickert WD, Pohlman MC, Pohlman AS, Nigos C, Pawlik AJ,Esbrook CL, Spears L, Miller M, Franczyk M, Deprizio D, et al. Earlyphysical and occupational therapy in mechanically ventilated, criti-cally ill patients: a randomised controlled trial. Lancet 2009;373:1874–1882.

57. Nieszkowska A, Combes A, Luyt C-E, Ksibi H, Trouillet J-L, Gibert C,Chastre J. Impact of tracheotomy on sedative administration, seda-tion level, and comfort of mechanically ventilated intensive care unitpatients. Crit Care Med 2005;33:2527–2533.

58. Pisani MA, Murphy TE, Araujo KLB, Slattum P, Van Ness PH,Inouye SK. Benzodiazepine and opioid use and the duration ofintensive care unit delirium in an older population. Crit Care Med2009;37:177–183.

59. Pandharipande P, Shintani A, Peterson J, Pun BT, Wilkinson GR, DittusRS, Bernard GR, Ely EW. Lorazepam is an independent risk factorfor transitioning to delirium in intensive care unit patients. Anesthe-siology 2006;104:21–26.

60. Girard TD, Pandharipande PP, Carson SS, Schmidt GA, Wright PE,Canonico AE, Pun BT, Thompson JL, Shintani AK, Meltzer HY,et al. Feasibility, efficacy, and safety of antipsychotics for intensivecare unit delirium: the MIND randomized, placebo-controlled trial.Crit Care Med 2010;38:428–437.

61. Gracey DR. Ventilator care beyond the intensive-care unit. Mayo ClinProc 1995;70:595–597.

62. Rudy EB, Daly BJ, Douglas S, Montenegro HD, Song R, Dyer MA.Patient outcomes for the chronically critically ill: special care unitversus intensive care unit. Nurs Res 1995;44:324–331.

63. Burns SM, Earven S, Fisher C, Lewis R, Merrell P, Schubart JR, TruwitJD, Bleck TP. Implementation of an institutional program to improveclinical and financial outcomes of mechanically ventilated patients:one-year outcomes and lessons learned. Crit Care Med 2003;31:2752–2763.

64. Griffiths J, Barber VS, Morgan L, Young JD. Systematic review andmeta-analysis of studies of the timing of tracheostomy in adultpatients undergoing artificial ventilation. BMJ 2005;330:1243.

65. Freeman BD, Borecki IB, Coopersmith CM, Buchman TG, FreemanBD, Borecki IB, Coopersmith CM, Buchman TG. Relationshipbetween tracheostomy timing and duration of mechanical ventilationin critically ill patients. Crit Care Med 2005;33:2513–2520.

66. Pronovost PJ, Angus DC, Dorman T, Robinson KA, Dremsizov TT,Young TL. Physician staffing patterns and clinical outcomes incritically ill patients: a systematic review. JAMA 2002;288:2151–2162.

67. Carson SS, Stocking C, Podsadecki T, Christenson J, Pohlman A,MacRae S, Jordan J, Humphrey H, Siegler M, Hall J. Effects oforganizational change in the medical intensive care unit of a teachinghospital: a comparison of ‘‘open’’ and ‘‘closed’’ formats. JAMA 1996;276:322–328.

68. Ely EW, Baker AM, Dunagan DP, Burke HL, Smith AC, Kelly PT,Johnson MM, Browder RW, Bowton DL, Haponik EF. Effecton the duration of mechanical ventilation of identifying patientscapable of breathing spontaneously. N Engl J Med 1996;335:1864–1869.

69. Kress JP, Vinayak AG, Levitt J, Schweickert WD, Gehlbach BK,Zimmerman F, Pohlman AS, Hall JB, Kress JP, Vinayak AG,et al. Daily sedative interruption in mechanically ventilatedpatients at risk for coronary artery disease. Crit Care Med 2007;35:365–371.

70. Carson SS, Kress JP, Rodgers JE, Vinayak A, Campbell-Bright S, LevittJ, Bourdet S, Ivanova A, Henderson AG, Pohlman A, et al. Arandomized trial of intermittent lorazepam versus propofol with daily


interruption in mechanically ventilated patients. Crit Care Med 2006;34:1326–1332.

71. Pronovost P, Needham D, Berenholtz S, Sinopoli D, Chu H, Cosgrove S,Sexton B, Hyzy R, Welsh R, Roth G, et al. An intervention todecrease catheter-related bloodstream infections in the ICU. N EnglJ Med 2006;355:2725–2732.

72. Camhi SL, Mercado AF, Morrison RS, Du Q, Platt DM, August GI,Nelson JE. Deciding in the dark: advance directives and continua-tion of treatment in chronic critical illness. Crit Care Med 2009;37:919–925.

73. SUPPORT Principal Investigators. A controlled trial to improve care forseriously ill hospitalized patients. JAMA 1995;274:1591–1598.

74. Society of Critical Care Medicine. Chronic critical illness in adultsrequiring prolonged mechanical ventilation [support brochure].

Accessed June 1, 2010. Available at http://www.myicucare.org/Support_Brochures/Pages/ChronicCriticalIllness.aspx

75. Lautrette A, Darmon M, Megarbane B, Joly LM, Chevret S, Adrie C,Barnoud D, Bleichner G, Bruel C, Choukroun G, et al. A commu-nication strategy and brochure for relatives of patients dying in theICU. N Engl J Med 2007;356:469–478.

76. Azoulay E, Pochard F, Chevret S, Jourdain M, Bornstain C, Wernet A,Cattaneo I, Annane D, Brun F, Bollaert P-E, et al. Impact of a familyinformation leaflet on effectiveness of information provided to familymembers of intensive care unit patients: a multicenter, prospective,randomized, controlled trial. Am J Respir Crit Care Med 2002;165:438–442.

77. Kahn JM. The evolving role of dedicated weaning facilities in criticalcare. Intensive Care Med 2010;36:8–10.


ORIGINAL INVESTIGATION

Communication About Chronic Critical IllnessJudith E. Nelson, MD, JD; Alice F. Mercado, RN, MBA; Sharon L. Camhi, MD; Nidhi Tandon, MD;Sylvan Wallenstein, PhD; Gary I. August, BS; R. Sean Morrison, MD

Background: Despite poor outcomes, life-sustainingtreatments including mechanical ventilation are contin-ued for a large and growing population of patients withchronic critical illness. This may be owing in part to alack of understanding resulting from inadequate com-munication between clinicians and patients and fami-lies. Our objective was to investigate the informationalneeds of patients with chronic critical illness and theirfamilies and the extent to which these needs are met.

Methods: In this prospective observational study con-ducted at 5 adult intensive care units in a large, university-affiliated hospital in New York, New York, 100 patientswith chronic critical illness (within 3-7 days of electivetracheotomy for prolonged mechanical ventilation) or sur-rogates for incapacitated patients were surveyed using an18-item questionnaire addressing communication aboutchronic critical illness. Main outcome measures in-cluded ratings of importance and reports of whether in-formation was received about questionnaire items.

Results: Among 125 consecutive, eligible patients, 100(80%) were enrolled; questionnaire respondents included2 patients and 98 surrogates. For all items, more than 78%of respondents rated the information as important for de-cision making (!98% for 16 of 18 items). Respondents re-ported receiving no information for a mean (SD) of 9.0 (3.3)of 18 items, with 95% of respondents reporting not receiv-ing information for approximately one-quarter of the items.Of the subjects rating the item as important, 77 of 96 (80%)and 69 of 74 (93%) reported receiving no information aboutexpected functional status at hospital discharge and prog-nosis for 1-year survival, respectively.

Conclusions: Many patients and their families may lackimportant information for decision making about con-tinuation of treatment in the chronic phase of critical ill-ness. Strategies for effective communication in this clini-cal context should be investigated and implemented.

Arch Intern Med. 2007;167(22):2509-2515

A DVANCES IN INTENSIVE CAREunit (ICU) care have para-doxically created a grow-ing population of patientswith prolonged depen-

dence on life-sustaining therapies includ-ing mechanical ventilation.1,2 For patientswith “chronic critical illness,”1 prognosis ispoor: the majority are dead within 6 months,and institutionalization with extreme func-tional dependence and severe cognitive im-pairment is typical for survivors.2-4 Costs ofcare are enormous for the health care sys-tem as a whole and for tens of thousandsof individual families.5

Despite these outcomes, many pa-tients and their families choose to con-tinue life-sustaining therapies when criti-cal illness enters a chronic phase. Evidencesuggests that this may be owing in part toa lack of understanding of chronic criti-cal illness resulting from inadequate com-munication between clinicians and pa-tients and families. In the Study toUnderstand Prognoses and Preferences forOutcomes and Risks of Treatments (SUP-

PORT), less than 40% of patients treatedin ICUs for more than 2 weeks reportedhaving a discussion with their physicianabout prognosis or preferences for life-sustaining treatment.6 Almost 50% of thosewho preferred care focused solely on theircomfort even at the expense of shorter lifethought that the treatment they receivedwas contrary to their preference, and ap-proximately 25% did not know the clini-cal team’s approach to their care.6 At a uni-versity-affiliated ICU, 54% of families failedto comprehend the diagnosis, prognosis,or treatment after meeting with a physi-cian.7 Almost 20% of 70 surrogate deci-sion makers for patients expected to staymore than 3 days in medical or surgicalICUs at 2 medical centers reported receiv-ing no prognostic information.8 In ICUfamily conferences, physicians com-monly missed opportunities to explorecomments about patient treatment pref-erences, as would be required for appro-priate clinical decision making.9

We previously used qualitative meth-ods to determine what information is rel-

Author Affiliations: Division ofPulmonary, Critical Care, andSleep Medicine, Department ofMedicine (Drs Nelson, Camhi,and Tandon, Ms Mercado, andMr August), BrookdaleDepartment of Geriatrics andAdult Development, HertzbergPalliative Care Institute,(Drs Nelson and Morrison), andDepartment of CommunityMedicine (Dr Wallenstein),Mount Sinai School ofMedicine, New York, New York.

(REPRINTED) ARCH INTERN MED/ VOL 167 (NO. 22), DEC 10/24, 2007 WWW.ARCHINTERNMED.COM2509

©2007 American Medical Association. All rights reserved. at MT SINAI SCHOOL OF MEDICINE, on September 9, 2010 www.archinternmed.comDownloaded from

http://www.archinternmed.com

evant and important for discussion and decision mak-ing when critical illness becomes chronic,10 a juncturethat we and others have defined by the placement of tra-cheotomy for ICU failure to wean from mechanical ven-tilation.1,4 Families described distress about what they per-ceived to be inadequacies in communication from

clinicians and particularly regretted that such commu-nication typically occurred, if at all, only after pro-longed, unsuccessful treatment of chronic critical ill-ness, when the patient was near death or discharge to askilled nursing facility for permanent custodial care.10 Weundertook the present study to explore further the in-formational needs of patients with chronic critical ill-ness and their families, to evaluate the extent to whichthese needs are met in current practice, and to identifyfactors associated with communication of informationabout chronic critical illness.

METHODS

SETTING AND SUBJECTS

We conducted this study between 2003 and 2005 in our large(1100-bed), tertiary care, university-affiliated hospital in NewYork, New York. Each weekday, our research nurse (A.F.M.)visited the hospital’s 5 adult ICUs (medical, surgical, cardiac,cardiothoracic surgical, and neurosurgical) to identify all pa-tients with chronic critical illness, ie, those undergoing trache-otomy for failure to wean from mechanical ventilation.4,5 Within3 to 7 days after the tracheotomy, we sought research consentfrom the individual (patient or surrogate) who had given con-sent for the tracheotomy. These decision makers (2 patientsand 98 surrogates) and the patients were our study subjects(n=100), to whom the research nurse administered our studyquestionnaire in-person on the same day that research con-sent was obtained. A priori, we excluded patients who had ahistory of tracheotomy (and their surrogates) and subjects withinsufficient English proficiency to participate without transla-tion. Our study received institutional review board approval.

QUESTIONNAIRE DEVELOPMENT

Through a computerized bibliographic literature review and useof qualitative research methods, we previously identified the fol-lowing 6 general domains of information that are considered bypatients, families, and clinicians to be relevant and important fordecision making when critical illness continues into a chronicphase10: (1) nature of the patient’s illness and treatments; (2) prog-nosis for outcomes including ventilator independence, func-tion, and quality of life; (3) impact of treatment on patient expe-rience, including symptom burden; (4) potential complicationsof treatment; (5) expected care needs after hospitalization; and(6) alternatives to continuation of treatment.10 For the presentstudy, we developed items within each of these domains asking(A) whether the information was considered important by the re-spondent (4-point Likert scale: strongly agree, agree, disagree, andstrongly disagree) and (B) whether this information was com-municated by the clinicians (primary or ICU physician, nurse,social worker, or other clinician).

We pretested the instrument with 10 surrogates of patientsreceiving treatment for chronic critical illness at our hospitaland incorporated their feedback in revising the questionnaire.With the main questionnaire, we then distributed a clinical sen-sibility questionnaire addressing face and content validity aswell as clarity, utility, redundancy, and discriminability11 to thephysician and nursing directors (n=10) of our adult ICUs andto 10 staff nurses in our medical ICU, where tracheotomy forfailure to wean from the ventilator is frequently performed. Usingthe sensibility ratings and additional suggestions, we furthermodified the main questionnaire. We then obtained written andoral feedback from a national interdisciplinary panel of 8 phy-sician- and nurse-researchers who are recognized as experts in

Table 1. Characteristics of 100 Patients With ChronicCritical Illnessa

Characteristic Patients

Age, median [range], y 74 [21-100]Male sex, No. (%) 55 (55)Race/ethnicity, No. (%)

White, non-Hispanic 42 (42)Black, non-Hispanic 26 (26)Hispanic 24 (24)Asian 8 (8)

Residence prior to hospital, No. (%)Home 81 (83)Other hospital/nursing home 15 (15)Adult residence 2 (2)

Marital status, No. (%)Married 53 (54)Divorced/separated 11 (11)Widowed 22 (22)Never married 12 (12)Unknown 1 (1)

Religious identification, No. (%)Catholic 35 (35)Protestant 10 (10)Jewish 24 (24)Other/unknown 31 (31)

Insurance coverage, No. (%)Medicare 54 (54)Private 27 (27)Medicaid 19 (19)

Charlson Comorbidity Index, mean [SD] 3.2 [2.7]ICU at time of tracheotomy/study enrollment, No. (%)

Medical ICU 61 (61)Neurosurgical ICU 20 (20)Surgical ICU 10 (10)Cardiothoracic surgical ICU 6 (6)Cardiac care unit 2 (2)Cardiac progressive care unit 1 (1)

Primary ICU admitting diagnosis, No. (%)Pulmonary 32 (32)Cardiovascular 25 (25)Neurological 16 (16)Surgical 14 (14)Other 12 (12)

Length of ICU stay, median [range], d 15 [1-178]Apache II score at study entry, mean [SD] 19.5 [5.4]Cause of prolonged respiratory failure, No. (%)b

Acute lung disease 52 (27)Chronic lung disease 18 (10)Cardiac disease 19 (10)Neurologic disease 36 (19)Postoperative condition 20 (11)Other 45 (24)

Length of hospital stay, median [range], d 51 [15-224]Hospital mortality, No. (%) 26 (26)

Abbreviation: ICU, intensive care unit.aSum is less than 100 for items for which data were unavailable for some

subjects.bSum is greater than 100 because prolonged respiratory failure was

typically multifactorial.




clinician-patient communication. Our final questionnaire con-tained questions about 18 specific topics of information withinthe 6 general domains as well as demographic questions aboutthe respondent and the patient.

QUESTIONNAIRE ADMINISTRATION

The research nurse administered the questionnaire and re-corded all responses. For patients (n=2) who had provided theirown consent to tracheotomy, we assessed cognition at the timeof administration of the questionnaire using the 6-item Orien-tation-Memory-Concentration test12 and the Confusion Assess-ment Method for the ICU.13

DATA ANALYSES

We summarized the clinical sensibility ratings using means (andstandard deviations). We used regression analysis to examineassociations between sociodemographic and health character-istics of patients and of respondents’ sociodemographic char-acteristics with respondents’ ratings of items as important andtheir reports of receiving or lacking information on the topicscovered in the questionnaire. For the latter analyses, we usedas our dependent variable the number of topics on which re-spondents reported receiving no information.

RESULTS

The clinical sensibility of the questionnaire was ratedhighly (on a 1-5 scale): face validity, mean (SD) of 4.6(0.4); content validity, 4.7 (0.7); clarity, 4.3 (0.7); util-ity, 4.2 (0.7); redundancy, 4.0 (1.0); and discriminabil-ity, 4.1 (0.7).

During the study period, 162 patients underwent tra-cheotomy for failure to wean in our institution’s adultICUs, of whom 125 (77%) met eligibility criteria for thisstudy (37 patients were excluded for the following rea-sons: underwent a prior tracheotomy [n=5], lacked En-glish-language proficiency [n=3], lacked a surrogate de-cision maker for a patient without capacity [n=24], ormet other a priori exclusion criteria [n=5]). Among 125consecutive, eligible, decision makers whom we ap-proached for study participation, we enrolled 100, an over-all response rate of 80%. With respect to sex, age, race/ethnicity, ICU site, hospital length of stay and hospitalmortality of the patients, subjects enrolled in the studywere not significantly different from eligible subjects whodid not participate in the research.

Characteristics of our respondents and the patients aregiven in Table 1 and Table 2, respectively. As is truenationally,5,14,15 our patients with chronic critical illnesswere mainly older adults, with a median age of 74 years(interquartile range, 59-80 years) and multiple comorbidillnesses and reasons for prolonged mechanical ventila-tion. Almost all (98 of 100) patients lacked the capacityto participate in the decision to undergo tracheotomy andcontinued critical care treatment. Most respondents werefamily members, among whom the largest group (n=46[47%]) were adult offspring. Our responding group wasracially and ethnically diverse, and two-thirds were col-lege educated.

Respondents confirmed the importance of informa-tion in the domains addressed by our questionnaire. For

16 of 18 items, more than 98% of the respondents agreedthat information about this topic was important for theirdecision making when the critical illness entered a chronicphase, as indicated by the need for tracheotomy (Table3).The remaining items, both of which related to prognosisfor mortality from the illness, were rated as important by89% (86 of 97 respondents [chances of death in the hos-pital]) and 78% (74 of 97 respondents [chances of deathwithin a year after hospital discharge]). Regression analy-ses did not show any significant variation in ratings of theimportance of questionnaire items according to charac-teristics of either the patients or their surrogates.

Almost half of the respondents reported receiving noinformation about the majority of items in our question-naire (Table 3). By these reports, among the 18 items, cli-nicians provided no information for a mean (SD) of 9.0 (3.3)items. All respondents stated that they did not receive in-formation about more than 1 item, and 95% stated that theywere given no information for one-quarter of these items(Figure). As reported by our respondents, most receivedinformation about the need for tracheotomy and its im-mediate implications (eg, impact on speaking and eat-ing). Fewer reported receiving information about short-term consequences of chronic critical illness (eg, distressingsymptoms, complications, and risk of mortality in the hos-pital). For items relating to long-term burdens and effects(eg, expected functional status after hospitalization anddeath within a year), the majority of respondents reportedreceiving no information (Table 3).

One item in the questionnaire asked whether the re-spondent considered it important in deciding about tra-

Table 2. Characteristics of 98 Surrogates Responding toChronic Critical Illness Communication Questionnairea

Characteristic Surrogates, No. (%)

Relationship to subjectSpouse/partner 28 (29)Child 46 (47)Parent 7 (7)Other relative 12 (12)Other 5 (5)

Years known patient"10 2 (2)#10 96 (98)

Male sex 33 (33)Ethnicity

White, non-Hispanic 35 (36)Black, non-Hispanic 28 (29)Hispanic 24 (25)Asian 9 (9)Other 1 (1)

Religious identificationCatholic 39 (41)Protestant 22 (23)Jewish 22 (23)Other 13 (14)

EducationCollege 64 (66)High school graduate 21 (22)No/some high school 12 (12)

aTwo patients as well as these 98 surrogates were our questionnairerespondents.




cheotomy and continuation of critical care treatments “toknow about choices other than continuing mechanicalventilation. (For example, taking the patient off the ven-tilator, even if he cannot breathe on his own, with treat-ment for shortness of breath or other symptoms.)” Of 97respondents who accorded importance to this informa-tion, 80 (83%) reported that they did not receive it. Eightypercent (77 of 96 respondents) also reported receivingno information about the patient’s expected functionalstatus at hospital discharge, while two-thirds (63 of 96respondents) stated they were given no information about“what the patient’s cognitive status (ability to think andunderstand, mental state) is expected to be when he leavesthe hospital.” Of 74 respondents who considered it im-portant to know “the chances that the patient might diewithin a year after he leaves the hospital,” 69 (93%) re-ported they did not receive this information. Length of

stay in the ICU and in the hospital emerged as the onlyfactors that were associated (Spearman correlation co-efficients, 0.21 and 0.23, respectively) with the numberof items on which respondents reported that informa-tion was inadequate: decision makers for patients withlonger lengths of stay reported lacking information on agreater number of topics they considered important.

COMMENT

This study focused on information communicated by cli-nicians when ICU patients and their surrogates face de-cisions about continuing treatment into the chronic phaseof critical illness. Using qualitative methods, we previ-ously identified domains of information that patients,families, and clinicians consider relevant and important

Table 3. Communication About Topics Rated Important by Family Decision Makers for Patients With Chronic Critical Illness

Topic of Information

Respondents, No. (%)

Rating asImportanta

Reporting NoInformationReceivedb

Why mechanical ventilation is needed 100 (100) 3 (3)Why tracheotomy is needed 100 (100) 1 (1)How tracheotomy might affect ability to speak 100 (100) 16 (16)How tracheotomy might affect ability to eat 100 (100) 25 (25)Chances of liberation from mechanical ventilation 100 (100) 40 (40)Time to ventilator liberation 100 (100) 53 (53)Symptoms during continued treatment 100 (100) 44 (44)Complications that might develop 99 (100) 22 (22)Risk of death during hospitalization 87 (89) 40 (46)Benefits and risk of cardiopulmonary resuscitation 97 (99) 55 (57)Discharge site from hospital (eg, home or nursing home) 97 (100) 53 (55)Choices other than continuing mechanical ventilation 98 (98) 81 (83)Expected functional status after hospitalization 97 (99) 78 (80)Quality of life after hospitalization 97 (99) 70 (72)Cognitive status after hospitalization 97 (99) 63 (65)Death within a year after hospital 74 (77) 69 (93)Services that might be needed after discharge 97 (100) 80 (82)Financial burden of the illness for family 93 (98) 70 (75)

aSubjects responded “agree” or “strongly agree” with the statement “It is important for me to know. . .” for each of these topics. Although there were 100respondents overall, not all answered with respect to every topic.

bFor the denominator, we used the number of respondents rating the item as important to know.

50

40

30

20

10

01-3 4-6 7-9 10-12 13-15 16-18

No. of Items

Resp

onde

nts

With

out I

nfor

mat

ion,

%

Figure. Distribution of items in Chronic Critical Illness 18-item questionnaire on which respondents (n=100) reported receiving no information.




for such decision making.10 In this quantitative study, weconducted “real-time” interviews in ICUs of a prospec-tive cohort of decision makers (almost all surrogates) forpatients with chronic critical illness to determine the im-portance of specific topics within those informational do-mains and the extent of communication about these top-ics. We identified a series of items considered importantby the respondents, including information about “choicesother than continuing mechanical ventilation (for ex-ample, removing the ventilator even if the patient can-not breathe on his own, with treatment for shortness ofbreath or other symptoms);” about “pain, shortness ofbreath, and other sources of distress and suffering thatthe patient might experience with continuing mechani-cal ventilation;” about expected functional and cogni-tive status after treatment; and about prospects for sur-vival in the year following hospital discharge. Allrespondents reported that they did not receive informa-tion on important topics, and approximately half of therespondents reported that for the majority of topics, cli-nicians provided no information.

Prior research has documented deficiencies in ICU com-munication and associations with adverse consequences forpatients, families, and professional caregivers.6,7,16,17 Fami-lies in a large teaching hospital’s medical ICU lacked a ba-sic understanding of diagnosis, prognosis, and treat-ments.7 Further research conducted in multiple centersshowed a high prevalence of anxiety and depression amongfamilies, which was associated with factors related to com-munication.18 “Usual care” in a leading US academic medi-cal center consisted primarily of informal communicationof essential information by nurses and trainees, without in-volvement of attending physicians or coordination amongproviders; formal family meetings were deferred until fur-ther critical care treatment was considered futile, leadingto delay in identifying of appropriate care goals and pro-longation of nonbeneficial ICU care.17 A large, interna-tional study found that without direct discussion, clini-cians presumed preferences favoring cardiopulmonaryresuscitation for most mechanically ventilated patients ad-mitted to their ICUs, whereas half of patients with explicitdirectives preferred not to be resuscitated in the event ofarrest.19 Our study is consistent with these reports but spe-cifically addresses communication at the juncture of acuteand chronic critical illness, as indicated by the placementof tracheotomy for failure to achieve ventilator indepen-dence after prolonged ICU treatment. To our knowledge,this is the first study to focus on communication in thiscontext.

In our study, patients and surrogates identified infor-mation that they considered important for decisions aboutcontinuing treatment into the chronic phase of criticalillness. Previous studies also suggest that such informa-tion would influence decision making. For example, Friedet al20 found that most older adults with serious illnesswould refuse life-sustaining treatments if the expectedoutcome were survival with severe functional or cogni-tive impairment; the prospect of such impairment influ-enced treatment preferences even more than the likeli-hood of death, and many subjects stated that they woulddecline treatment if there was even a 50% chance of se-vere functional or cognitive impairment. Several studies

have shown that information about the probability of sur-vival after cardiopulmonary resuscitation significantly low-ers the proportion of individuals who say they would wantthis intervention.21-23 Among 1-year survivors of me-chanical ventilation for at least 2 days, many stated thatknowledge of the intensity of pain and discomfort dur-ing treatment would influence their decision making aboutsuch ventilation, leading them to refuse treatment athigher levels of symptom distress.24 A small group of sur-vivors of a week or more of mechanical ventilation re-ported that future decision making about mechanical ven-tilation would be influenced by their health status andby the resulting financial burdens for their families.25 Thus,although the present study did not directly measure theimpact of perceived inadequacies in communication onactual decision making, our results suggest that effortsto improve communication about chronic critical ill-ness would be valued by patients and families and mighthelp to align treatment decisions with their values andpreferences. In addition, evidence indicates that these ef-forts would also promote other desirable outcomes, in-cluding higher levels of family satisfaction26-28; less anxi-ety, depression, and posttraumatic stress among familymembers29,30; fewer conflicts31,32; and earlier establish-ment of realistic and appropriate care goals and more ef-ficient use of ICU resources.17,33-35 Patients with chroniccritical illness and their families often misinterpret theplacement of tracheotomy as a sign of hope for full re-covery10 and go on to experience grave disappointment,disillusionment, and anger toward caregivers as the re-ality of protracted and typically progressive illness be-comes clear. Even if earlier communication about ex-pected outcomes does not change decision making at theoutset, such communication may help to prepare pa-tients and families for events, discussions, and deci-sions they are likely to face later.

Although we are unable to predict the experience andoutcomes of individual patients with absolute certainty, ex-isting evidence provides a solid foundation for communi-cation about the benefits and burdens of treating chroniccritical illness. Such treatment is burdensome for many pa-tients and families and is expensive for them and for ourhealth care system, while achieving limited clinical ben-efits.2,4,5,36-39 It is known that mortality rates during and soonafter treatment for chronic critical illness are high.2,4,40 Itis alsoknownthat a largeproportionofpatientswithchroniccritical illness remain permanently dependent on life-sustaining therapies.2,4,40 Fewof thesepatients return to func-tional lives in the community; most are extremely depen-dent for their daily activities and require institutionalcare.4,37,39,40 Perhaps of greatest concern to patients and fami-lies,20,41 brain dysfunction including coma and delirium ishighly prevalent and often permanent.4 Skilled clinicianscan communicate this information in a clear, understand-able, and compassionate way. Some patients and familieswill still choose to continue life-prolonging treatments, andthis may at times be appropriate. However, decision mak-ing needs to be appropriately informed—that is, informedto the extent desired by the decision makers and to the de-gree supported by relevant prior research. Our findings in-dicate that communication to support decision makingabout treatment of chronic critical illness can be im-




proved. They also help to focus attention on areas of in-formation that patients and families value and need most.

Several strategies to improve communication duringacute critical illness have been tested with success. Tworigorous trials demonstrated benefit from distribution ofprinted informational materials for ICU families, with29

or without30 a proactive, protocol-directed approach tofamily conferencing. An “intensive communication in-tervention” consisting of a mandatory, interdisciplinarymeeting with ICU families proved superior to “usual care”in a before-after comparison.17 Consultations by ethicsteams when “value-laden treatment conflicts” arose inICU patient care,34 decreased ICU and hospital stays anddays of “nonbeneficial” treatments compared with stan-dard treatment, and patients, families, and clinicians gen-erally rated the intervention as informative and helpful.Appropriate training of clinicians in communication yieldssustained improvements in these essential skills.42-44

Several inferences are possible from our finding thatthe lengths of ICU and hospital stay were associated withthe extent of communication as reported by the respon-dents. It may be that patients or families who wished tocontinue life-prolonging treatments for longer periodswere less receptive to clinician communication or haddifferent informational needs. Alternatively, the clini-cians may have avoided discussions with such patientsand families. Involvement of multiple clinicians over ex-tended periods in the hospital may have blurred lines ofresponsibility for communication or exposed patients andfamilies to contradictory information from different care-givers. Another possibility is that inadequate communi-cation about prognosis and treatment led to delayed imple-mentation of appropriate care plans and prolongation oftherapies with limited clinical benefit.

Our study has limitations. First, since we did not di-rectly observe communications, our findings about the ex-tent of information may not reflect with complete accu-racy the information actually provided by clinicians. Friedet al45 found a low level of agreement between patients andclinicians on whether the clinician had discussed prognos-tic information. Stress, anxiety, and depression18,46 may im-pair families’ comprehension through mechanisms includ-ing denial and distraction and introduce bias in reportingabout the ICU experience. We believe it is valuable, how-ever, to investigate communication from the perspectiveof critically ill patients and their families, who are the fo-cus of our care and who require the information we com-municate because they have the authority and responsi-bility to make decisions about treatment. Second,“prolonged mechanical ventilation” has been defined forother purposes in terms of duration of time on the venti-lator,47 whereas we and others have identified patients withchronic critical illness by the placement of tracheotomy af-ter ICU weaning failure.4,5,48,49 A third limitation is that thisstudy was conducted in a single institution and may notbe fullygeneralizable toothercare settings;NewYork,whereour hospital is located, does not have long-term acute carefacilities, which might affect length of hospital stay (al-though not overall duration of care for chronic critical ill-ness). On the other hand, we recruited subjects from 5 ICUsacross a broad spectrum of case mix and clinical practiceand studied a patient group with characteristics and out-

comes that are similar to those reported from diverse set-tings.2,37-39,50 Other strengths include our use of “real-time” data collection from patients and families who wereactually making treatment decisions, rather than retro-spective interviews of a limited group of survivors or ques-tioning of subjects about hypothetical scenarios; the re-sponse rate, which compares favorably with that achievedin other research conducted with ICU families in the UnitedStates; and the rigorous process we used to develop a validquestionnaire.

CONCLUSIONS

Chronic critical illness is now known to be a devastat-ing condition imposing heavy burdens on patients, fami-lies, professional caregivers, health care systems, and so-ciety as a whole. When critical illness becomes chronic,continuation of intensive care treatment requires ex-traordinary resources but entails significant distress forpatients and families while yielding limited clinical ben-efits. In specific situations, a choice to continue this treat-ment may be reasonable, particularly if the therapeutictrial is time limited, progress is periodically reevalu-ated, and attention is given simultaneously to the pallia-tive needs of the patient and family. It will never be ap-propriate, however, to treat chronic critical illness withouta thorough discussion of benefits and burdens and an ex-ploration of the needs, values, preferences and goals ofthe patient and family as they relate to the proposed treat-ment. This study suggests that many patients with chroniccritical illness and their families are making crucial medi-cal decisions in a “black hole,” without essential infor-mation. At the same time, our findings can help to guideimprovement in communication by identifying topics aswell as patients and families requiring special attention.Together with recent research about the process of ICUcommunication and studies testing the effectiveness ofcommunication strategies in the context of acute criti-cal illness, the present study of informational contentstrengthens the evidence base for clinical practice andfor future interventional research focusing specifically oncommunication about chronic critical illness.

Accepted for Publication: August 5, 2007.Correspondence: Judith E. Nelson, MD, JD, Box 1232,Mount Sinai School of Medicine, One Gustave Levy Place,New York, NY 10029 ([email protected]).Author Contributions: Dr Nelson had full access to allof the data in the study and takes responsibility for theintegrity of the data and the accuracy of the data analy-sis. Study concept and design: Nelson and Morrison. Ac-quisition of data: Nelson, Mercado, and Camhi. Analysisand interpretation of data: Nelson, Camhi, Tandon,Wallenstein, August, and Morrison. Drafting of the manu-script: Nelson. Critical revision of the manuscript for im-portant intellectual content: Nelson, Mercado, Camhi,Tandon, Wallenstein, and Morrison. Statistical analysis:Tandon, Wallenstein, August, and Morrison. Obtainedfunding: Nelson. Administrative, technical, and material sup-port: Nelson, Mercado, Tandon, and August. Study su-pervision: Nelson and Morrison.




Financial Disclosure: None reported.Funding/Support: This study was funded by Clinical Re-search Grant CG-005-N from the American Lung Asso-ciation. Dr Nelson is the recipient of an Independent Sci-entist Research Career Development Award (K02AG024476) from the National Institute on Aging.Role of the Sponsor: The funding agency played no rolein the design or conduct of the study; collection, man-agement, analysis, or interpretation of the data; or prepa-ration, review, or approval of the manuscript.

REFERENCES

1. Nierman DM, Nelson JE, eds. Chronic critical illness. Crit Care Clin. 2002;18(3):461-715.

2. Carson SS. Outcomes of prolonged mechanical ventilation. Curr Opin Crit Care.2006;12(5):405-411.

3. Camhi SL, Nelson JE. Chronic critical illness. In: Vincent JL, ed. Yearbook of In-tensive Care Medicine. New York, NY: Springer; 2007:908-914.

4. Nelson JE, Tandon N, Mercado A, Camhi SL, Ely EW, Morrison RS. Brain dys-function: another burden for the chronically critically ill. Arch Intern Med. 2006;166(18):1993-1999.

5. Carson SS, Bach PB. The epidemiology and costs of chronic critical illness. CritCare Clin. 2002;18(3):461-476.

6. Teno JM, Fisher E, Hamel MB, et al. Decision-making and outcomes of prolongedICU stays in seriously ill patients. J Am Geriatr Soc. 2000;48(5)(suppl):S70-S74.

7. Azoulay E, Chevret S, Leleu G, et al. Half the families of intensive care unit pa-tients experience inadequate communication with physicians. Crit Care Med. 2000;28:3044-3049.

8. LeClaire MM, Oakes JM, Weinert CR. Communication of prognostic informationfor critically ill patients. Chest. 2005;128(3):1728-1735.

9. Curtis JR, Engelberg RA, Wenrich MD, Shannon SE, Treece PD, Rubenfeld GD.Missed opportunities during family conferences about end-of-life care in the in-tensive care unit. Am J Respir Crit Care Med. 2005;171(8):844-849.

10. Nelson JE, Kinjo K, Meier DE, Ahmad K, Morrison RS. When critical illness be-comes chronic: informational needs of patients and families. J Crit Care. 2005;20(1):79-89.

11. Feinstein AR. Clinical sensibility. In: Feinstein AR, ed. Clinimetrics. New Haven,CT: Yale University Press; 1987:141-166.

12. Katzman R, Brown T, Fuld P, Peck A, Schechter R, Schimmel H. Validation of ashort Orientation-Memory-Concentration Test of cognitive impairment. Am JPsychiatry. 1983;140(6):734-739.

13. Ely EW, Inouye SK, Bernard GR, et al. Delirium in mechanically ventilated pa-tients: validity and reliability of the confusion assessment method for the inten-sive care unit (CAM-ICU). JAMA. 2001;286(21):2703-2710.

14. Cox CE, Carson SS, Holmes GM, Howard A, Carey TS. Increase in tracheostomyfor prolonged mechanical ventilation in North Carolina, 1993-2002. Crit Care Med.2004;32(11):2219-2226.

15. Dewar DM, Kurek CJ, Lambrinos J, Cohen IL, Zhong Y. Patterns in costs andoutcomes for patients with prolonged mechanical ventilation undergoing tra-cheostomy: an analysis of discharges under diagnosis-related group 483 in NewYork State from 1992 to 1996. Crit Care Med. 1999;27(12):2640-2647.

16. Nelson JE, Angus DC, Weissfeld L, et al. End-of-life care for the critically ill: anational ICU survey. Crit Care Med. 2006;34(10):2547-2553.

17. Lilly CM, De Meo DL, Sonna LA, et al. An intensive communication interventionfor the critically ill. Am J Med. 2000;109(6):469-475.

18. Pochard F, Azoulay E, Chevret S, et al. Symptoms of anxiety and depression infamily members of intensive care unit patients: ethical hypothesis regarding de-cision-making capacity. Crit Care Med. 2001;29(10):1893-1897.

19. Cook DJ, Guyatt G, Rocker G, et al. Cardiopulmonary resuscitation directives onadmission to intensive-care unit: an international observational study. Lancet.2001;358(9297):1941-1945.

20. Fried TR, Bradley EH, Towle VR, Allore H. Understanding the treatment prefer-ences of seriously ill patients. N Engl J Med. 2002;346(14):1061-1066.

21. Murphy DJ, Burrows D, Santilli S. The influence of the probability of survival onpatients’ preferences regarding cardiopulmonary resuscitation. N Engl J Med.1994;330(8):545-549.

22. Schonwetter RS, Walker RM, Kramer DR, Robinson BE. Resuscitation decisionmaking in the elderly. J Gen Intern Med. 1993;8(6):295-300.

23. Schonwetter RS, Teasdale TA, Taffet G, Robinson BE, Luchi RJ. Educating theelderly: cardiopulmonary resuscitation decisions before and after intervention.J Am Geriatr Soc. 1991;39(4):372-377.

24. Mendelsohn AB, Belle SH, Fischhoff B, Wisniewski SR, Degenholtz H, ChelluriL. How patients feel about prolonged mechanical ventilation 1 year later. Crit CareMed. 2002;30(7):1439-1445.

25. Guentner K, Hoffman LA, Happ MB, et al. Preferences for mechanical ventilationamong survivors of prolonged mechanical ventilation and tracheostomy. Am JCrit Care. 2006;15(1):65-77.

26. McDonagh JR, Elliott TB, Engelberg RA, et al. Family satisfaction with family con-ferences about end-of-life care in the intensive care unit: increased proportionof family speech is associated with increased satisfaction. Crit Care Med. 2004;32(7):1484-1488.

27. Heyland DK, Cook DJ, Rocker GM, et al. Decision-making in the ICU: perspectivesof the substitute decision-maker. Intensive Care Med. 2003;29(1):75-82.

28. Heyland DK, Rocker GM, Dodek PM, et al. Family satisfaction with care in theintensive care unit: results of a multiple center study. Crit Care Med. 2002;30(7):1413-1418.

29. Lautrette A, Darmon M, Megarbane B, et al. A communication strategy and brochurefor relatives of patients dying in the ICU. N Engl J Med. 2007;356(5):469-478.

30. Azoulay E, Pochard F, Chevret S, et al. Impact of a family information leaflet oneffectiveness of information provided to family members of intensive care unitpatients: a multicenter, prospective, randomized, controlled trial. Am J RespirCrit Care Med. 2002;165(4):438-442.

31. Studdert DM, Mello MM, Burns JP, et al. Conflict in the care of patients with pro-longed stay in the ICU: types, sources, and predictors. Intensive Care Med. 2003;29(9):1489-1497.

32. Abbott KH, Sago JG, Breen CM, Abernethy AP, Tulsky JA. Families looking back:one year after discussion of withdrawal or withholding of life-sustaining support.Crit Care Med. 2001;29(1):197-201.

33. Campbell ML, Guzman JA. Impact of a proactive approach to improve end-of-life care in a medical ICU. Chest. 2003;123(1):266-271.

34. Schneiderman LJ, Gilmer T, Teetzel HD, et al. Effect of ethics consultations onnonbeneficial life-sustaining treatments in the intensive care setting: a random-ized controlled trial. JAMA. 2003;290(9):1166-1172.

35. Dowdy MD, Robertson C, Bander JA. A study of proactive ethics consultationfor critically and terminally ill patients with extended lengths of stay. Crit CareMed. 1998;26(2):252-259.

36. Nelson JE, Meier DE, Litke A, Natale DA, Siegel RE, Morrison RS. The symptomburden of chronic critical illness. Crit Care Med. 2004;32(7):1527-1534.

37. Nasraway SA, Button GJ, Rand WM, Hudson-Jinks T, Gustafson M. Survivorsof catastrophic illness: outcome after direct transfer from intensive care to ex-tended care facilities. Crit Care Med. 2000;28(1):19-25.

38. Dasgupta A, Rice R, Mascha E, Litaker D, Stoller JK. Four-year experience witha unit for long-term ventilation (respiratory special care unit) at the ClevelandClinic Foundation. Chest. 1999;116(2):447-455.

39. Carson SS, Bach PB, Brzozowski L, Leff A. Outcomes after long-term acute care:an analysis of 133 mechanically ventilated patients. Am J Respir Crit Care Med.1999;159(5, pt 1):1568-1573.

40. Scheinhorn DJ, Hassenpflug MS, Votto JJ, et al. Post-ICU mechanical ventila-tion at 23 long-term care hospitals: a multicenter outcomes study. Chest. 2007;131(1):85-93.

41. Fried TR, Byers AL, Gallo WT, et al. Prospective study of health status prefer-ences and changes in preferences over time in older adults. Arch Intern Med.2006;166(8):890-895.

42. Back AL, Arnold RM, Baile WF, et al. Efficacy of communication skills trainingfor giving bad news and discussing transitions to palliative care. Arch Intern Med.2007;167(5):453-460.

43. Fallowfield L, Jenkins V, Farewell V, Solis-Trapala I. Enduring impact of com-munication skills training: results of a 12-month follow-up. Br J Cancer. 2003;89(8):1445-1449.

44. Fallowfield L, Lipkin M, Hall A. Teaching senior oncologists communication skills:results from phase I of a comprehensive longitudinal program in the UnitedKingdom. J Clin Oncol. 1998;16(5):1961-1968.

45. Fried TR, Bradley EH, O’Leary J. Prognosis communication in serious illness:perceptions of older patients, caregivers, and clinicians. J Am Geriatr Soc. 2003;51(10):1398-1403.

46. Pochard F, Darmon M, Fassier T, et al. Symptoms of anxiety and depression infamily members of intensive care unit patients before discharge or death: a pro-spective multicenter study. J Crit Care. 2005;20(1):90-96.

47. MacIntyre NR, Epstein SK, Carson S, Scheinhorn D, Christopher K, Muldoon S.Management of patients requiring prolonged mechanical ventilation: report of aNAMDRC consensus conference. Chest. 2005;128(6):3937-3954.

48. Estenssoro E, Reina R, Canales HS, et al. The distinct clinical profile of chroni-cally critically ill patients: a cohort study. Crit Care. 2006;10(3):R89.

49. Nierman DM, Mechanick JI. Biochemical response to treatment of bone hyper-resorption in chronically critically ill patients. Chest. 2000;118(3):761-766.

50. Scheinhorn DJ, Hassenpflug MS, Votto JJ, et al. Ventilator-dependent survivorsof catastrophic illness transferred to 23 long-term care hospitals for weaningfrom prolonged mechanical ventilation. Chest. 2007;131(1):76-84.




One-Year Trajectories of Care and Resource Utilization for Recipientsof Prolonged Mechanical VentilationA Cohort StudyMark Unroe, MD; Jeremy M. Kahn, MD, MSc; Shannon S. Carson, MD; Joseph A. Govert, MD; Tereza Martinu, MD; Shailaja J. Sathy, MD;Alison S. Clay, MD; Jessica Chia, MD; Alice Gray, MD; James A. Tulsky, MD; and Christopher E. Cox, MD, MPH

Background: Growing numbers of critically ill patients receive pro-longed mechanical ventilation. Little is known about the patterns ofcare as patients transition from acute care hospitals to postacutecare facilities or about the associated resource utilization.

Objective: To describe 1-year trajectories of care and resourceutilization for patients receiving prolonged mechanical ventilation.

Design: 1-year prospective cohort study.

Setting: 5 intensive care units at Duke University Medical Center,Durham, North Carolina.

Participants: 126 patients receiving prolonged mechanical ventila-tion (defined as ventilation for !4 days with tracheostomy place-ment or ventilation for !21 days without tracheostomy), as well astheir 126 surrogates and 54 intensive care unit physicians, enrolledconsecutively over 1 year.

Measurements: Patients and surrogates were interviewed in thehospital, as well as 3 and 12 months after discharge, to determinepatient survival, functional status, and facility type and duration ofpostdischarge care. Physicians were interviewed in the hospital toelicit prognoses. Institutional billing records were used to assigncosts for acute care, outpatient care, and interfacility transportation.Medicare claims data were used to assign costs for postacute care.

Results: 103 (82%) hospital survivors had 457 separate transitionsin postdischarge care location (median, 4 transitions [interquartile

range, 3 to 5 transitions]), including 68 patients (67%) who werereadmitted at least once. Patients spent an average of 74% (95%CI, 68% to 80%) of all days alive in a hospital or postacute carefacility or receiving home health care. At 1 year, 11 patients (9%)had a good outcome (alive with no functional dependency), 33(26%) had a fair outcome (alive with moderate dependency), and82 (65%) had a poor outcome (either alive with complete func-tional dependency [4 patients; 21%] or dead [56 patients; 44%]).Patients with poor outcomes were older, had more comorbid con-ditions, and were more frequently discharged to a postacute carefacility than patients with either fair or good outcomes (P ! 0.05for all). The mean cost per patient was $306 135 (SD, $285 467),and total cohort cost was $38.1 million, for an estimated $3.5million per independently functioning survivor at 1 year.

Limitation: The results of this single-center study may not beapplicable to other centers.

Conclusion: Patients receiving prolonged mechanical ventilationhave multiple transitions of care, resulting in substantial health carecosts and persistent, profound disability. The optimism of surrogatedecision makers should be balanced by discussions of these out-comes when considering a course of prolonged life support.

Primary Funding Source: None.

Ann Intern Med. 2010;153:167-175. www.annals.orgFor author affiliations, see end of text.

The 300 000 patients per year who receive life supportin intensive care units (ICUs) for much longer than

average have been labeled as receiving “prolonged mechan-ical ventilation” (1, 2). These patients utilize a dispropor-tionately large amount of health care resources and haverelatively poor long-term outcomes (2–4). Patients reportdiminished quality of life, have important functional andcognitive limitations, require prolonged informal caregiv-ing assistance, and have high 1-year mortality rates (5–7).Despite these outcomes, the number of patients per yearreceiving prolonged mechanical ventilation will probablyexceed 600 000 within a decade, with associated hospitalcosts of $50 billion to $60 billion (2). Decision makers’hope for patient survival coupled with an incomplete un-derstanding of the specific implications of providing pro-longed mechanical ventilation may contribute to the in-creasing incidence (8).

Patients receiving prolonged mechanical ventilationincur disproportionately high acute care costs and utilizepostacute care facilities more frequently than other patientgroups, which makes them of special interest to health carepayers, such as the Centers for Medicare & Medicaid Ser-

vices, and to the postacute care industry (3). Due in part tosubstantial annual costs, the high reimbursement rates forfacility-based care, and the high margins reported by post-acute care facilities, the Deficit Reduction Act of 2005directed the Centers for Medicare & Medicaid Services toreform the current system of postacute care payment (9).

For these reasons, a clearer description of patients’postdischarge paths through different care facilities, the as-sociated burden of functional limitations, and patient-level

See also:

PrintEditors’ Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168Editorial comment. . . . . . . . . . . . . . . . . . . . . . . . . . 204Summary for Patients. . . . . . . . . . . . . . . . . . . . . . . I-56

Web-OnlyAppendixAppendix TablesConversion of graphics into slides

Annals of Internal Medicine Article

© 2010 American College of Physicians 167

factors associated with high resource utilization and pooroutcomes can assist in clinical decision making, institu-tional planning, payment reform, and the design of futureinterventions targeted to these unique patients. Therefore,we aimed to describe the overall trajectories of care locationand costs for a consecutively enrolled cohort of patientsreceiving prolonged mechanical ventilation.

METHODSStudy Design and Participants

We conducted a 1-year, prospective cohort study inthe adult general surgical, trauma, neurologic, cardiotho-racic surgery, cardiac, and medical ICUs at Duke Univer-sity Medical Center, Durham, North Carolina. Partici-pants, including 126 patients and their 126 surrogates and54 primary ICU physicians, were enrolled beginning inApril 2006 through daily screening of ICUs by study staff,with follow-up completed in April 2008. Details aboutrates of enrollment, exclusion, and declining to participate,as well as characteristics of the surrogates and physicians,are provided elsewhere (5).

Patients were eligible for the study if they were age 18years or older and met either of 2 common definitions ofprolonged mechanical ventilation: mechanical ventilationfor 21 or more days with less than 48 hours of unassistedbreathing or 4 or more days of ventilation and placementof a tracheostomy for an expected prolonged requirementfor ventilatory support (10). Exclusion criteria were lack ofan identifiable surrogate; English-language fluency poorenough to require a translator; tracheostomy placement foreither emergency indications or an ear-, nose-, or throat-related diagnosis; or preadmission receipt of a tracheos-

tomy. We defined the surrogate as the person most in-volved in the decision to place a tracheostomy and the onemost likely to provide most of the postdischarge care. En-rolled physicians were self-identified as the primary ICUphysician for each patient.

Data Collection and VariablesWe collected data from medical records, administra-

tive billing records, and participant interviews. Study staffabstracted clinical data from patients’ charts and hospitalelectronic records to record admitting diagnoses and oper-ative procedures, sociodemographic characteristics, Charl-son comorbidity scores (11), acute physiology scores rep-resenting illness severity on the day of tracheostomyplacement (12), mechanical ventilation course, and hospi-tal and ICU lengths of stay.

Patients, surrogates, and physicians were interviewedin person within 48 hours of meeting study eligibility cri-teria. Follow-up interviews were performed with patientsand surrogates either by telephone or in person 3 and 12months later, with 3-month interviews primarily informingvital status and resource utilization calculations. We com-pleted 100% of interviews with surrogates and patients,excluding patients who had died (36 patients [29%] at 3months and 56 patients [44%] at 12 months) or whoshowed clinically significant cognitive impairment (36 pa-tients [29%] at 3 months and 31 patients [25%] at 12months), as defined by a score of less than 20 on theFolstein Mini-Mental State Questionnaire (13). The6-item activities of daily living instrument (14) quantifieddependencies in basic functioning, including bathing,dressing, feeding, transferring from bed to chair, bladderand bowel control, and use of the toilet.

Quality of life was assessed with the EuroQol-5D(EQ-5D) (15, 16), an instrument with evidence of validityin survivors of critical illness and as a surrogate-completedproxy measure. We adjusted survival for quality of life bymultiplying each patient’s 3- and 12-month EQ-5D indexscores (1 " excellent quality of life; #0.1 " worse thandeath) by the total days alive in the 2 periods (0 to 3months and 4 to 12 months) preceding each measurement.We defined poor quality of life as an index score of 0.44 orlower, 2 SDs below the U.S. population average for per-sons aged 55 to 65 years (17). We defined good quality oflife as a score greater than 0.80, the U.S. population aver-age for similar age groups, and fair quality of life as inter-mediate scores (18). Because nearly one third of patientswere too disabled to complete interviews during follow-up,we used surrogate assessments of patients’ quality of lifeand functional status in analyses. Scores on the EQ-5Dwere highly correlated (r " 0.94; P ! 0.001) between sur-rogates and cognitively intact patients. Surrogates and phy-sicians also reported whether they expected patients to sur-vive and to have complete functional independence at 1year, with responses of “strongly agree” or “agree” consid-ered as “high expectations” and “don’t know,” “disagree,”

Context

A growing number of patients opt for prolonged mechani-cal ventilation, often with the understanding that theirchances for recovery are good.

Contribution

In this study, 126 patients in intensive care units who re-quired prolonged mechanical ventilation had high mortal-ity and multiple transitions of care, incurring estimatedcosts of $3.5 million per patient over the following 12months. Only 11 patients were alive with no functionaldependencies at 1 year.

Caution

The care and transitions all took place in a single medicalcenter and region.

Implication

Patients requiring prolonged mechanical ventilation havecomplicated courses and poor outcomes that often con-trast with initial expectations for recovery.

—The Editors

Article Trajectories of Care in Prolonged Mechanical Ventilation

168 3 August 2010 Annals of Internal Medicine Volume 153 • Number 3 www.annals.org

or “strongly disagree” recorded as “low expectations” foreither item.

Data on resource utilization were obtained by re-viewing medical records, administrative billing records,and participant interviews (Appendix, available at www.annals.org). We determined costs for the primary hos-pitalization by using itemized charges from each pa-tient’s administrative billing record and converting theseto costs by using department-specific cost-to-charge ra-tios obtained from the Centers for Medicare & Medic-aid Services hospital cost reports (19). We used partici-pant interviews to record any hospital readmission,postacute care facility admission, or use of home healthservice. Duration of care for each episode, includinglength of ICU stay during readmissions, was verified byreview of medical records whenever possible. We esti-mated costs for postdischarge care episodes by using2006 region-specific mean daily ICU, hospital, andpostacute care costs obtained from Medicare claims.Physician costs were estimated at 17% of hospital costs,as in previous analyses (20). Clinic visits and ambula-tory surgical procedures were recorded from medicalcharts and assigned costs on the basis of relevant Cur-rent Procedural Terminology codes. Air and groundtransportation costs were assigned for interfacility trans-fers on the basis of administrative billing records.

Statistical AnalysisThe primary outcomes of the study were 1-year sur-

vival, functional status, and health care–associated resourceutilization. We also described health outcomes by combin-ing 1-year survival and functional status into 3 simple cat-egories designated to be equally interpretable by patients,surrogates, and physicians: good outcome (alive with nodependencies in activities of daily living), fair outcome(alive with 1 to 5 dependencies in activities of daily living),or poor outcome (either dead or alive with dependencies inall 6 activities of daily living).

We present categorical data as numbers (percent-ages) and continuous data as means (SDs) or medians(interquartile ranges [IQRs]). We examined factors as-sociated with grouped health outcomes (good, fair, orpoor) as appropriate for data distribution by using Pear-son chi-square and Fisher exact tests for categorical vari-ables and analysis of variance tests or Kruskal–Wallistests for continuous variables. Because cost data wereskewed, we used log-transformed values in analyses. Ad-ditional ventilator outcomes data are shown in the Ap-pendix. We used Stata software, version 11 (StataCorp,College Station, Texas) for all analyses and considered aP value less than 0.05 to be significant.

Role of the Funding Source

The Duke University Institutional Review Board ap-proved all study procedures. No external funding was re-ceived for this study.

Table 1. Baseline Characteristics and Hospital DischargeDisposition*

Characteristic Value

Age, yMean (SD) 55 (16)Range 19–85

Female, n (%) 50 (40)

Race or ethnicity, n (%)White 67 (53)Black 48 (38)Native American 7 (5)Asian 2 (2)Hispanic 2 (2)

Place of residence before admission, n (%)Home 124 (98)Nursing facility 2 (2)

Employment status, n (%)Employed full- or part-time 41 (32)Not currently employed, retired, or disabled 85 (68)

Less than high school education, n (%)†Patients 15 (12)Surrogates 9 (8)

Insurance status, n (%)Private 72 (57)Government (Medicare or Medicaid) 34 (27)Self-pay 20 (16)

Median comorbid conditions (IQR) 2 (0–4)

Median dependencies in ADLs (IQR) 0 (0–0)

Primary ICU admission diagnosis, n (%)‡Respiratory failure 29 (23)Neurologic 29 (23)Trauma 27 (21)Postoperative 26 (21)Septic shock 10 (8)Cardiac 5 (4)

Median Acute Physiology Score (IQR)§ICU day 1 19 (15–24)Day of tracheostomy placement 15 (12–19)

Median ventilator days (IQR)Before tracheostomy! 11 (8–17)Total 27 (18–24)

Median length of stay (IQR), dICU 26 (22–42)Hospital 39 (28–57)

Hospital discharge disposition, n (%)Home without paid home health care 6 (5)Home with paid home health care 14 (11)Long-term acute care facility 36 (29)Skilled nursing facility 17 (13)Rehabilitation facility 23 (18)Other hospital 3 (2)Still in acute care hospital at 1 y 1 (1)Inpatient hospice facility 3 (2)Dead 23 (18)

ADL " activity of daily living; ICU " intensive care unit; IQR " interquartilerange.* Out of 126 patients, unless noted otherwise.† Out of 108 patients, because 18 (14%) patients did not answer.‡ Categories include respiratory (pneumonia, aspiration, and pulmonary embolus),neurologic (ischemic stroke, subarachnoid hemorrhage, the Guillain–Barré syn-drome, and status epilepticus), trauma, nontrauma surgical (immediate postoper-ative general and cardiothoracic), and cardiac (myocardial infarction and out-of-hospital cardiac arrest).§ From APACHE (Acute Physiology and Chronic Health Evaluation) II classifi-cation (12).! Out of 125 patients, because 1 patient did not have tracheostomy.

ArticleTrajectories of Care in Prolonged Mechanical Ventilation

www.annals.org 3 August 2010 Annals of Internal Medicine Volume 153 • Number 3 169

RESULTSBaseline Characteristics and Hospital Course

Patients were middle-aged, insured, and well-educatedand had few premorbid functional limitations or medicalcomorbid conditions on average (Table 1). Admission di-agnoses were nearly equivalent in proportion amongtrauma, nontrauma surgical, neurologic, and medicalcauses. Patients had a median of 27 ventilator days (IQR,18 to 27 days) and a median hospital stay of 39 days. Atotal of 86 (68%) patients were ultimately weaned fromventilation (Appendix Table 1, available at www.annals.org). The 23 patients (18%) who died in the hospital hada greater length of stay than did survivors (53 days [IQR,33 to 82 days] vs. 28 days [IQR, 27 to 52 days]) andreceived more ventilator days (46 days [IQR, 32 to 81 days]vs. 25 days [IQR, 20 to 34 days]; P " 0.001). All patientswho died during the initial hospitalization were receiving ven-tilation via tracheostomy at the time of death. Most (74%)survivors were discharged to a postacute care facility.

One-Year Outcomes and Trajectories of CareAt 1 year, 70 patients (56%) were alive, although only

11 (9%) were independently functioning and only 19(27%) had a good quality of life (Table 2). Of these sur-vivors, 68 were ultimately decannulated, all within 1month of ventilator weaning. Those who died duringfollow-up lived a median of only 79 days (IQR, 46 to 125days). Patients had 457 transitions in care location (median, 4transitions [IQR, 3 to 5 transitions]) during follow-up (Fig-ure 1), receiving a total of 14 552 days of inpatient hospi-tal and facility care. There were 150 readmissions in 68(67%) of the 103 hospital survivors. Most readmissions(96 [65%] patients) occurred within 3 months; nearly halfwere related to sepsis (Appendix Table 2, available at www.annals.org). The average patient spent 74% (95% CI,68% to 80%) of all days alive in a hospital or postacutecare facility or was receiving home health care; 61% (CI,54% to 68%) of study days were facility-based. Only 3patients (2%) were both initially discharged to their homeand remained there, whereas only 3 of 54 previously em-ployed patients (6%) ever returned to work. A total of 19surrogates (27%) reported a good quality of life for 1-yearsurvivors.

At 1 year, 11 patients (9%) had a good health out-come, 33 (26%) had a fair outcome, and 82 (65%) had apoor outcome (Table 2). Only 23 patients (18%) eitherimproved in outcome category or maintained a good out-come between 3 and 12 months (Figure 2). All 1-yearsurvivors in the fair- and good-outcome groups were resid-ing at home, except for 1 fair-outcome survivor, who re-sided in a nursing facility. Compared with patients withpoor outcomes, those with fair outcomes had more transi-tions (median, 5 transitions [IQR, 4 to 7 transitions] vs. 2transitions [IQR, 0 to 5 transitions]) and were more likelyboth to be readmitted (53% vs. 27%; P " 0.02) and to re-ceive facility-based postacute care (81% vs. 61%; P " 0.03).

Table 2. Patient and Process-of-Care Outcomes at 1 YearAfter Discharge*

Outcome Value

Survival, n (%) 70 (56)

Quality of life in survivors at 1 y, n (%)†Good 19 (27)Fair 17 (24)Poor 34 (49)

Median total quality-adjusted life-days over 1 y (IQR)‡ 66 (16–223)

Disposition at 1 y, n (%)Home without paid home care 11 (9)Home with paid home care 47 (37)Skilled nursing facility 10 (8)Inpatient rehabilitation facility 1 (1)Residing in a hospital, never discharged 1 (1)Dead 56 (44)

Location of death over 1 y, n (%)§Withdrawal of mechanical ventilation in hospital 22 (39)Receiving full support on ventilator in hospital or facility 24 (43)Hospice 7 (13)Home 3 (5)

Mechanical ventilation outcomes, n (%)Weaned from ventilator 86 (68)

During initial hospitalization 69At long-term acute care facility 13At skilled nursing facility 4At another hospital 1

Not weaned from ventilator 40 (32)Alive, still ventilated at 1 y 3Died, ventilator-dependent 37

Median duration of ventilator support (IQR), d "If weaned from ventilator 21 (16–37)If never weaned from ventilator 42 (33–74)

Transitions among care locations over 1 y, n (%)¶To a lower or an equal level of care 305 (65)To a higher level of care 152 (35)Readmissions** 150Inpatient rehabilitation to skilled nursing facility 2

Hospital readmissions, by diagnostic category, n (%)††Sepsis 56 (45)Respiratory failure 26 (21)Surgical complications 22 (17)Neurologic complications 7 (6)Other medical 39 (31)

Median postacute care utilization (IQR), d‡‡Long-term acute care facility 29 (21–70)Skilled nursing facility 42 (27–228)Rehabilitation facility 28 (21–45)Home health services 84 (34–250)

IQR " interquartile range.* Out of 126 patients, unless noted otherwise.† Out of 70 patients. Quality of life was determined by surrogate assessment usingEuroQol-5D index score. A score "0.44 is poor, 0.45 to 0.79 is fair, and !0.80 is good.‡ Calculated by adjusting all days alive during 1 y with correspondingEuroQol-5D index scores obtained at 3- and 12-mo follow-up.§ Out of 56 patients.! 86 patients were eventually weaned from ventilator support, and 40 patients werenever weaned from ventilator support.¶ Out of 457 transitions. A total of 88 patients moved to a lower or an equal levelof care, and 69 patients moved to a higher level of care. The total patients whomoved to either lower or higher levels of care includes 58 patients who hadtransitions to both higher and lower levels of care.** In 68 of the 102 patients discharged from the index hospitalization to a desti-nation other than hospice.†† Out of 150 readmissions. Readmission categories include sepsis (urinary tractinfection, pneumonia, catheter infections, and other), respiratory failure (conges-tive heart failure, pneumonia, pulmonary embolism), surgical complications orrepeated surgery related to primary admission, neurologic complications (seizure,intracerebral hemorrhage, subdural hematoma), or other medical conditions (de-hydration, mental status change, gastrointestinal hemorrhage, admission for che-motherapy, fall, renal failure, pancreatitis, gastrostomy complication).‡‡ 43 admissions to long-term acute care facilities in 38 patients, 63 admissions toskilled nursing facilities in 36 patients, 54 admissions to rehabilitation facilities in28 patients, and 44 episodes of home health service in 36 patients.



Patients with a poor outcome were older, had morecomorbid conditions, and were more frequently dischargedwhile still receiving mechanical ventilation than those withbetter outcomes (Table 3). Those with fair outcomes weremost frequently uninsured and were less severely ill on theday of tracheostomy placement. Patients with fair and pooroutcomes were discharged in equal proportions to long-term acute care and skilled nursing facilities (42 [51%] vs.16 [48%]). Patients with good outcomes tended to be ad-mitted with trauma and were more frequently dischargedhome than other patients. The proportion of patients withmoderate or greater severity of illness (acute physiologyscores $15) on the day of tracheostomy was the same inthose with poor outcomes (64%) and good outcomes(63%). The proportion of surrogates with high expecta-tions for either survival or functional independence did notdiffer on the basis of subsequent health outcome categori-zation (P $ 0.05 for all). Physicians were somewhat lessoptimistic about patients’ survival, although they reportedhigh expectations for functional independence for 36(44%) patients who later had poor outcomes, 17 (52%)

who later had fair outcomes, and 11 (100%) who later hadgood outcomes (P " 0.002).

Resource UtilizationThe mean total 1-year cost of health care for cohort

members was $306 135 (SD, $285 467), whereas cost forthe entire cohort exceeded $38.5 million (Table 4). Mostof the cost ($28.1 million, or 73%) was incurred by theinitial hospitalization. The highest mean cost for postacutecare was for persons receiving long-term acute care($91 277), followed by those receiving care in a skillednursing facility ($31 892), care in an inpatient rehabilita-tion facility ($21 244), and home health service care($6669). Outpatient costs averaged $551, although trans-portation costs exceeded $10 000 per patient. The 1-yearcosts did not differ by health outcome (P " 0.40) (Appen-dix Table 3, available at www.annals.org).

DISCUSSION

Our study offers new insights into a growing popula-tion of relatively understudied critically ill patients. Over 1

Figure 1. Trajectories of care for patients in the prolonged mechanical ventilation cohort over the first year after discharge.

Initial hospitalization (n = 126)Residing at 1 year: 1Transition to inpatient hospice: 3*Died: 23

Other hospital (n = 3)Total patients: 3Residing at 1 y: 0Died: 0

99 patients

1transition

1transition

2transitions

13transitions

22transitions

3 patientsAlive at 1 y: 3





Long-term acute-care facility (n = 43)

Total patients: 38Residing at 1 y: 0Died: 6*

Skilled nursingfacility (n = 63)†

Total patients: 36Residing at 1 y: 10Died: 6*

Inpatient rehabilitationfacility (n = 54)

Total patients: 28Residing at 1 y: 1Died: 0

Home (n = 136)Total patients: 71Residing at 1 y: 58

Independent: 11Dependent: 47

Died: 4*

75 transitions74 transitions

26 transitions

29 transitions


6 transitions

12 transitions

2 transitions


11 transitions

Hospital readmission (n = 150)Total patients: 68Died: 13*

Arrows between care locations indicate both the direction of patient transitions and the total number of patients transferred between locations over 1 year.Solid lines represent initial transitions between the hospital and other locations. Dashed lines represent subsequent hospital readmissions and dischargesinvolving postdischarge care locations. Dotted lines represent transitions among postdischarge care locations, including home. Each box summarizes thetotal numbers of both readmissions and patients admitted, as well as how many patients remained or died in each location of care at 1 year.* 7 transitions to inpatient hospice and death not shown (3 from the acute hospitalization and 1 each from home, long-term acute care facility, skillednursing facility, and hospital readmission).† 1 transition from skilled nursing facility to skilled nursing facility not shown.



year, 126 patients who received prolonged mechanical ven-tilation had a median of 4 transitions of care location eachand spent nearly 75% of all days in hospitals, in postacutecare facilities, or at home receiving paid home care. Survi-vors at 1 year were left with a serious burden of pervasive,persistent disability despite aggressive care that cost a totalof $38 million, or approximately $3.5 million for each1-year survivor without serious functional dependencies.

These findings are important for patients, families, cli-nicians, and policymakers. First, the effect on the U.S.health care system of patients receiving prolonged mechan-ical ventilation has probably been substantially under-estimated (2, 3, 21). Past estimates (22, 23) of these pa-tients’ resource utilization have not focused on cumulativeacute and postdischarge care. We found that whereas theinitial hospitalization accounted for most costs, postacutecare facilities and readmissions contributed substantially to

resource utilization. The pattern of patient death may haveaccentuated this distribution of costs, because patients whodied during hospitalization had an average length of staymore than 2 weeks longer than that of those who survived.Because the risk for death remained high throughoutfollow-up, the opportunity to utilize postdischarge re-sources was attenuated. Still, the high cost of acute criticalcare is clearly a major factor, because cohort members’hospital costs alone were 15 times greater than those of anaverage Medicare patient with critical illness (24). Also, thereadmission rate we observed was nearly 50% higher thanthat reported in Medicare beneficiaries who survive a hos-pitalization that includes mechanical ventilation (25). It istherefore troubling that the number of patients receivingprolonged mechanical ventilation is expected to increasesubstantially over the coming decade (2).

These data also are relevant to efforts to reform post-acute care payment, initiated with the Deficit ReductionAct of 2005 (26). Payments for prolonged mechanical ven-tilation are widely disparate, with acute care hospitals oftenreceiving far less than postacute care facilities (27). Somehave proposed basing payment for the treatment of pa-tients with chronic, critical illness on the quality of longi-tudinal care, to reward lower-cost providers who canreduce costly transitions and readmissions (28, 29).However, the singularly high readmission rate from post-acute care may be associated with patient characteristicsthat are impossible to modify, such as age and comorbidconditions, and should be investigated further before abenchmark rate is considered as a quality modifier of pay-ment (30). The complexity of patients’ trajectories of carehighlights the need to define quality indicators for thispopulation that are transportable across institution type,with the goals of improving patients’ overall care and theefficiency with which care is delivered (26).

Several studies (7, 31–33) have shown the extensiveeffect of critical illness on the physical, mental, and finan-cial well-being of patients and their families. However,both the magnitude of disability and the infrequency ofpostdischarge recovery are noteworthy. Our findings thatpatients with poor outcomes were more likely to be elderly,have comorbid conditions, and be receiving ventilation atdischarge are generally similar to those of other studies (34,35), as is our observation that illness severity scores at thetime of tracheostomy do not accurately discriminate be-tween patients with good or poor outcomes. Similarly, weobserved that most patients with a good functional recov-ery were admitted because of trauma (36). However, pa-tients with intermediate outcomes—those who are alivebut with moderate functional dependency—may be themost challenging to manage, because of the perceived un-certainty associated with their prognosis. These previouslyhigh-functioning patients were less severely ill than otherpatients. Despite their decision makers’ initial optimism,however, they rarely improved over time, instead cyclingfrequently between postacute care facilities and hospitals.

Figure 2. Patient trajectories at 1 year, by health outcome.

3 Months

9 (7%)6 (4.5%)*2 (1.5%)1 (1%)

5 (4%)*

18 (14%)

4 (3%)

7 (6%)

12 (9.5%)*

23 (18%)

12 (9.5%)

34 (27%)

47 (37%)

12 Months

Good Outcome

Alive with nofunctional

dependencies

Fair Outcome

Alive with somefunctional

dependencies

Alive withcompletefunctional

dependency

Poor Outcome

Dead

Each bar shows patients at 3- and 12-mo intervals, grouped by survivaland number of functional limitations in basic activities of daily living.The arrows indicate group members’ subsequent longitudinal transitionsto other health outcomes. For example, between 3 and 12 months, 34patients with a fair 3-month outcome improved to a good outcome (n "5), remained at fair outcome (n " 18), or worsened to a poor outcome(n " 4) or died (n " 7). Percentages in both 3- and 12-mo outcomescategories are calculated by including 36 (29%) patients (not shown)who were dead at 3 mo.* These 23 (18%) patients improved or remained in the good-outcomegrouping between 3 and 12 mo.



Our study confirms that prolonged mechanical venti-lation is a highly resource-intensive condition with a gen-erally poor outcome. However, the circumstances underwhich decision making about prolonged mechanical venti-lation occur are not ideal, probably favoring the pursuitof aggressive care (8). First, the content of physician–surrogate communication is inadequate for fully shared de-cision making (37). Nelson and colleagues (8) reportedthat 80% and 93% of surrogates of patients with pro-longed mechanical ventilation received no informationabout possible functional dependency or expected 1-yearsurvival, respectively. Second, both clinicians and surro-gates substantially overestimate prospects for recovery anddo not anticipate the amount and intensity of caregivingthat will be required (5). A new prognostic model (34) hasshown promise for this population but requires further val-idation. Third, previous research (38) has shown that mostinternists are uncomfortable discussing uncertain prog-noses, as may be the case for a patient who survives anacute critical illness but still requires life support. However,

Table 4. Resource Utilization at 1 Year*

Resource Mean IndividualCost (SD), $

Cohort TotalCost, $

Total costs at 1 y 306 135 (285 467) 38 577 935

Initial hospitalization 223 406 (278 165) 28 149 128

Postacute care 57 730 (77 735) 5 504 902Long-term acute care facility 91 277 (103 017) 3 468 519Skilled nursing facility 31 892 (34 727) 1 148 122Rehabilitation facility 21 244 (15 718) 594 839Home health services 6669 (4754) 293 422

Hospital readmissions 54 818 (87 204) 3 727 631

Transportation costs† 10 906 (6951) 1 155 998

Outpatient costs‡ 551 (585) 40 276

* Out of 126 patients.† Includes air and ground transport between acute and postacute care facilities.‡ Includes emergency department visits, clinic visits, and outpatient surgeries.

Table 3. Baseline and Hospital Characteristics Associated With 1-Year Health Outcomes*

Characteristic Health Outcome at 1 Year† P Value‡

Poor (n ! 82) Fair (n ! 33) Good (n ! 11)

Mean age (SD), y 59 (14) 47 (17) 51 (22) 0.001

>1 dependency in activities of daily living, n (%) 13 (16) 6 (18) 1 (9) 0.19

Median comorbid conditions (IQR), n 2 (1–4) 1 (0–3) 0 (0–2) 0.001

Insurance status, n (%) 0.04Private 43 (53) 19 (58) 10 (91)Government (Medicare or Medicaid) 29 (35) 4 (12) 1 (9)Self-pay 10 (12) 10 (30) 0 (0)

Primary ICU admission diagnostic category, n (%)§ 0.02Trauma 9 (11) 13 (39) 5 (45)Nontrauma surgical 28 (34) 5 (15) 3 (27)Medicine 45 (55) 15 (45) 3 (27)

Acute Physiology Score >15 on day of tracheostomy, n (%) 52 (63) 12 (36) 7 (64) 0.03

Physicians with high expectations for, n (%)Survival 36 (44) 24 (73) 11 (100) !0.001Functional independence 36 (44) 17 (52) 11 (100) 0.002

Outcome for which surrogates had high expectations, n (%)Survival 74 (90) 32 (97) 11 (100) 0.30Functional independence 54 (66) 25 (76) 11 (100) 0.05

Mechanical ventilation at hospital discharge, n (%) 42 (52) 5 (15) 1 (9) !0.001

Discharge disposition, n (%) !0.001Home 8 (10) 6 (18) 6 (55)Long-term acute care, skilled nursing facility, other hospital 42 (51) 16 (48) 2 (18)Inpatient rehabilitation facility 9 (11) 11 (33) 3 (27)Dead 23 (28) 0 (0) 0 (0)

Median number of transitions of care location (IQR) 2 (0–5) 5 (4–7) 2 (2–5) 0.001

ICU " intensive care unit; IQR " interquartile range.* Out of 126 patients.† Health outcome categories are defined as good (alive with no dependencies in activities of daily living), fair (alive but with 1 to 5 dependencies in activities of daily living),and poor (either alive and completely dependent or dead).‡ Based on Pearson chi-square test, Fisher exact test, 1-way analysis of variance, or Kruskal–Wallis test.§ Categories include trauma, nontrauma surgical (immediate postoperative general and cardiothoracic), and medicine (pulmonary, infectious disease, neurologic, and cardiacdiagnoses).



surrogates acknowledge the inevitable uncertainty in criti-cal illness outcomes and still desire prognostic estimates inthe setting of end-of-life decisions (39). Finally, the com-plexities of critical illness can be hard for providers to ex-plain in terms that surrogates understand and value. Thesimple health outcome groupings we have reported mayhelp in this regard and may also lend themselves to incor-poration in future decision-support tools for this popula-tion. El-Jawahri and colleagues (40) have shown that deci-sion tools that use simple categorizations of choices andoutcomes are more effective than verbal descriptions alonein end-of-life considerations.

We enrolled critically ill patients near the time of tra-cheostomy. This is when the physician determines thattimely ventilator liberation is unlikely and the surrogatedecision maker acknowledges that the patient would desireprolonged life support. Although tracheostomy is beingperformed increasingly earlier in the course of ventilation,little persuasive evidence suggests that either early (!1week) or late ($2 to 3 weeks) timing confers importantclinical benefit (41–43). This uncertainty has probablycontributed to the substantial variation in practice seenacross physicians, hospitals, and regions (44).

Our study has several limitations. We used participantself-reports to quantify the duration of postdischarge care.Although this strategy may result in inaccuracies, data sug-gest that costs would be underestimated rather than in-flated (45, 46). Similarly, we could not quantify the nota-ble financial strain of critical illness on patients and theircaregivers, also reducing its true economic effect. In addi-tion, although we enrolled participants consecutively andfew participants (20%) declined, our findings may not rep-resent patients receiving prolonged mechanical ventilationat other institutions or those who have different sociocul-tural or linguistic backgrounds. The Durham, North Caro-lina, area has a relatively high long-term acute care facilitypenetration compared with other regions of the UnitedStates, which may lead to more care transitions. Furtherstudy in larger data sets may allow a more robust charac-terization of potentially modifiable risk factors for resourceutilization.

The incidence of prolonged mechanical ventilationwill probably increase in the coming years, consuming sub-stantial health care resources in the process. Given thedisproportionately high costs and associated disability ofprolonged mechanical ventilation, clinicians need toreconsider their approach to its provision. Currently, thedecision-making process for prolonged mechanical ventila-tion is marked by unrealistic expectations and poor com-munication. It seems prudent that, in the context of pro-longed mechanical ventilation, physicians not only discusslong-term outcomes with surrogates in terms that they caneasily understand but also explicitly convey the probabledemands of treatment and the future functional depen-dence patients will probably have.

From Duke University, Durham, and University of North Carolina,Chapel Hill, North Carolina, and Leonard Davis Institute of HealthEconomics, University of Pennsylvania, Philadelphia, Pennsylvania.

Grant Support: By the National Institutes of Health (grants K23HL081048, K23 HL082650, and K23 HL067068).

Potential Conflicts of Interest: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum"M09-2795.

Reproducible Research Statement: Study protocol and data set: Avail-able from Dr. Cox (e-mail, [email protected]). Statisticalcode: Not available.

Requests for Single Reprints: Christopher E. Cox, MD, MPH, DukeUniversity Medical Center, Division of Pulmonary and Critical CareMedicine, Box 102043, Durham, NC 27710; e-mail, [email protected].

Current author addresses and author contributions are available at www.annals.org.

References1. Carson SS, Bach PB. The epidemiology and costs of chronic critical illness.Crit Care Clin. 2002;18:461-76. [PMID: 12140908]2. Zilberberg MD, Shorr AF. Prolonged acute mechanical ventilation and hos-pital bed utilization in 2020 in the United States: implications for budgets, plantand personnel planning. BMC Health Serv Res. 2008;8:242. [PMID: 19032766]3. Cox CE, Carson SS, Govert JA, Chelluri L, Sanders GD. An economicevaluation of prolonged mechanical ventilation. Crit Care Med. 2007;35:1918-27. [PMID: 17581479]4. Nelson JE, Meier DE, Litke A, Natale DA, Siegel RE, Morrison RS. Thesymptom burden of chronic critical illness. Crit Care Med. 2004;32:1527-34.[PMID: 15241097]5. Cox CE, Martinu T, Sathy SJ, Clay AS, Chia J, Gray AL, et al. Expectationsand outcomes of prolonged mechanical ventilation. Crit Care Med. 2009;37:2888-94; quiz 2904. [PMID: 19770733]6. Nelson JE, Tandon N, Mercado AF, Camhi SL, Ely EW, Morrison RS.Brain dysfunction: another burden for the chronically critically ill. Arch InternMed. 2006;166:1993-9. [PMID: 17030833]7. Van Pelt DC, Milbrandt EB, Qin L, Weissfeld LA, Rotondi AJ, Schulz R,et al. Informal caregiver burden among survivors of prolonged mechanical ven-tilation. Am J Respir Crit Care Med. 2007;175:167-73. [PMID: 17068327]8. Nelson JE, Mercado AF, Camhi SL, Tandon N, Wallenstein S, August GI,et al. Communication about chronic critical illness. Arch Intern Med. 2007;167:2509-15. [PMID: 18071175]9. Deficit Reduction Act of 2005. Pub. L. No. 109-171, S. 1932 (2006). Ac-cessed at www.govtrack.us/congress/billtext.xpd?bill"s109-1932 on 11 June2010.10. MacIntyre NR, Epstein SK, Carson S, Scheinhorn D, Christopher K,Muldoon S; National Association for Medical Direction of Respiratory Care.Management of patients requiring prolonged mechanical ventilation: reportof a NAMDRC consensus conference. Chest. 2005;128:3937-54. [PMID:16354866]11. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index foruse with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613-9.[PMID: 1607900]12. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: aseverity of disease classification system. Crit Care Med. 1985;13:818-29. [PMID:3928249]13. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practicalmethod for grading the cognitive state of patients for the clinician. J PsychiatrRes. 1975;12:189-98. [PMID: 1202204]14. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illnessin the aged. the index of ADL: a standardized measure of biological and psycho-social function. JAMA. 1963;185:914-9. [PMID: 14044222]



15. The EuroQol Group. EuroQol—a new facility for the measurement ofhealth-related quality of life. Health Policy. 1990;16:199-208. [PMID:10109801]16. Angus DC, Carlet J; 2002 Brussels Roundtable Participants. Survivingintensive care: a report from the 2002 Brussels Roundtable. Intensive Care Med.2003;29:368-77. [PMID: 12536269]17. Kind P, Hardman G, Macran S. UK Population Norms for the EQ-5D.York, United Kingdom: University of York; 1999. Accessed at www.york.ac.uk/inst/che/pdf/DP172.pdf on 1 June 2010.18. Luo N, Johnson JA, Shaw JW, Feeny D, Coons SJ. Self-reported healthstatus of the general adult U.S. population as assessed by the EQ-5D and HealthUtilities Index. Med Care. 2005;43:1078-86. [PMID: 16224300]19. Taira DA, Seto TB, Siegrist R, Cosgrove R, Berezin R, Cohen DJ. Com-parison of analytic approaches for the economic evaluation of new technologiesalongside multicenter clinical trials. Am Heart J. 2003;145:452-8. [PMID:12660668]20. Ellis SG, Miller DP, Brown KJ, Omoigui N, Howell GL, Kutner M, et al.In-hospital cost of percutaneous coronary revascularization. Critical determinantsand implications. Circulation. 1995;92:741-7. [PMID: 7641352]21. Heyland DK, Konopad E, Noseworthy TW, Johnston R, Gafni A. Is it‘worthwhile’ to continue treating patients with a prolonged stay ($ 14 days) inthe ICU? An economic evaluation. Chest. 1998;114:192-8. [PMID: 9674469]22. Cox CE, Carson SS, Lindquist JH, Olsen MK, Govert JA, Chelluri L;Quality of Life After Mechanical Ventilation in the Aged (QOL-MV) Investi-gators. Differences in one-year health outcomes and resource utilization by defi-nition of prolonged mechanical ventilation: a prospective cohort study. Crit Care.2007;11:R9. [PMID: 17244364]23. Dewar DM, Kurek CJ, Lambrinos J, Cohen IL, Zhong Y. Patterns in costsand outcomes for patients with prolonged mechanical ventilation undergoingtracheostomy: an analysis of discharges under diagnosis-related group 483 in NewYork State from 1992 to 1996. Crit Care Med. 1999;27:2640-7. [PMID:10628603]24. Cooper LM, Linde-Zwirble WT. Medicare intensive care unit use: analysisof incidence, cost, and payment. Crit Care Med. 2004;32:2247-53. [PMID:15640637]25. Wunsch H, Guerra C, Barnato AE, Angus DC, Li G, Linde-Zwirble WT.Three-year outcomes for Medicare beneficiaries who survive intensive care.JAMA. 2010;303:849-56. [PMID: 20197531]26. RTI International. Post-Acute Care Payment Reform Demonstration. Ac-cessed at www.pacdemo.rti.org on 7 June 2010.27. Centers for Medicare & Medicaid Services. Overview of the prospectivepayment system. Accessed at www.cms.hhs.gov/ProspMedicareFeeSvcPmtGenon 7 June 2010.28. Davis K. Paying for care episodes and care coordination [Editorial]. N EnglJ Med. 2007;356:1166-8. [PMID: 17360996]29. Bodenheimer T. Coordinating care—a perilous journey through the healthcare system. N Engl J Med. 2008;358:1064-71. [PMID: 18322289]30. Nasraway SA, Button GJ, Rand WM, Hudson-Jinks T, Gustafson M.Survivors of catastrophic illness: outcome after direct transfer from intensive careto extended care facilities. Crit Care Med. 2000;28:19-25. [PMID: 10667494]31. Angus DC, Musthafa AA, Clermont G, Griffin MF, Linde-Zwirble WT,Dremsizov TT, et al. Quality-adjusted survival in the first year after the acute

respiratory distress syndrome. Am J Respir Crit Care Med. 2001;163:1389-94.[PMID: 11371406]32. Covinsky KE, Goldman L, Cook EF, Oye R, Desbiens N, Reding D, et al.The impact of serious illness on patients’ families. SUPPORT Investigators.Study to Understand Prognoses and Preferences for Outcomes and Risks ofTreatment. JAMA. 1994;272:1839-44. [PMID: 7990218]33. Herridge MS, Cheung AM, Tansey CM, Matte-Martyn A, Diaz-GranadosN, Al-Saidi F, et al; Canadian Critical Care Trials Group. One-year outcomesin survivors of the acute respiratory distress syndrome. N Engl J Med. 2003;348:683-93. [PMID: 12594312]34. Carson SS, Garrett J, Hanson LC, Lanier J, Govert J, Brake MC, et al. Aprognostic model for one-year mortality in patients requiring prolonged mechan-ical ventilation. Crit Care Med. 2008;36:2061-9. [PMID: 18552692]35. Carson SS, Bach PB. Predicting mortality in patients suffering from pro-longed critical illness: an assessment of four severity-of-illness measures. Chest.2001;120:928-33. [PMID: 11555531]36. Engoren M, Arslanian-Engoren C, Fenn-Buderer N. Hospital and long-term outcome after tracheostomy for respiratory failure. Chest. 2004;125:220-7.[PMID: 14718444]37. White DB, Braddock CH 3rd, Bereknyei S, Curtis JR. Toward shareddecision making at the end of life in intensive care units: opportunities for im-provement. Arch Intern Med. 2007;167:461-7. [PMID: 17353493]38. Christakis NA, Iwashyna TJ. Attitude and self-reported practice regardingprognostication in a national sample of internists. Arch Intern Med. 1998;158:2389-95. [PMID: 9827791]39. Evans LR, Boyd EA, Malvar G, Apatira L, Luce JM, Lo B, et al. Surrogatedecision-makers’ perspectives on discussing prognosis in the face of uncertainty.Am J Respir Crit Care Med. 2009;179:48-53. [PMID: 18931332]40. El-Jawahri A, Podgurski LM, Eichler AF, Plotkin SR, Temel JS, MitchellSL, et al. Use of video to facilitate end-of-life discussions with patients withcancer: a randomized controlled trial. J Clin Oncol. 2010;28:305-10. [PMID:19949010]41. Cox CE, Carson SS, Holmes GM, Howard A, Carey TS. Increase in tra-cheostomy for prolonged mechanical ventilation in North Carolina, 1993-2002.Crit Care Med. 2004;32:2219-26. [PMID: 15640633]42. Terragni PP, Antonelli M, Fumagalli R, Faggiano C, Berardino M, Pal-lavicini FB, et al. Early vs late tracheotomy for prevention of pneumonia inmechanically ventilated adult ICU patients: a randomized controlled trial. JAMA.2010;303:1483-9. [PMID: 20407057]43. Griffiths J, Barber VS, Morgan L, Young JD. Systematic review and meta-analysis of studies of the timing of tracheostomy in adult patients undergoingartificial ventilation. BMJ. 2005;330:1243. [PMID: 15901643]44. Nathens AB, Rivara FP, Mack CD, Rubenfeld GD, Wang J, Jurkovich GJ,et al. Variations in rates of tracheostomy in the critically ill trauma patient. CritCare Med. 2006;34:2919-24. [PMID: 16971852]45. Kahn JM, Rubenfeld GD, Rohrbach J, Fuchs BD. Cost savings attributableto reductions in intensive care unit length of stay for mechanically ventilatedpatients. Med Care. 2008;46:1226-33. [PMID: 19300312]46. Rapoport J, Teres D, Zhao Y, Lemeshow S. Length of stay data as a guideto hospital economic performance for ICU patients. Med Care. 2003;41:386-97.[PMID: 12618642]



Current Author Addresses: Drs. Unroe, Govert, Martinu, Chia, Gray,and Cox: Division of Pulmonary and Critical Care Medicine, Box 2629,Duke University, Durham, NC 27710.Dr. Kahn: University of Pennsylvania Medical Center, Blockley Hall,Room 723, 423 Guardian Drive, Philadelphia, PA 19104-6160.Dr. Carson: Division of Pulmonary and Critical Care Medicine, Univer-sity of North Carolina, 4134 Bioinformatics Building, Chapel Hill, NC27599.Dr. Sathy: Division of Pulmonary and Critical Care Medicine, Univer-sity of Washington, Harborview Medical Center, 325 Ninth Avenue,Seattle, WA 98104.Dr. Clay: Department of Surgery, Duke University, Box 2945 Durham,NC 27710.Dr. Tulsky: Center for Palliative Care, Hock Plaza, Suite 1105, 2424Erwin Road, Durham, NC 27705.

Author Contributions: Conception and design: S.S. Carson, J.A.Govert, T. Martinu, A.S. Clay, J. Chia, J.A. Tulsky, C.E. Cox.Analysis and interpretation of the data: M. Unroe, J.M. Kahn, S.S. Car-son, J.A. Govert, J.A. Tulsky, C.E. Cox.Drafting of the article: M. Unroe, S.S. Carson, J.A. Govert, A.S. Clay,J.A. Tulsky, C.E. Cox.Critical revision of the article for important intellectual content: M.Unroe, J.M. Kahn, S.S. Carson, J.A. Govert, S.J. Sathy, J. Chia, A. Gray,J.A. Tulsky, C.E. Cox.Final approval of the article: M. Unroe, J.M. Kahn, S.S. Carson, J.A.Govert, T. Martinu, A.S. Clay, J. Chia, A. Gray, J.A. Tulsky, C.E. Cox.Provision of study materials or patients: T. Martinu.Statistical expertise: C.E. Cox.Obtaining of funding: J.A. Tulsky, C.E. Cox.Administrative, technical, or logistic support: A.S. Clay, C.E. Cox.Collection and assembly of data: J.M. Kahn, T. Martinu, S.J. Sathy, A.S.Clay, J. Chia, A. Gray, C.E. Cox.

47. U.S. Bureau of Labor Statistics. Consumer price index. Accessed at www.bls.gov/cpi/home.htm on 7 June 2010.48. U.S. Department of Health and Human Services. Medicare Program;Home Health Prospective Payment System Refinement and Rate Update forCalendar Year 2008; Final Rule. Washington, DC; U.S. Department of Healthand Human Services: 2007;49762-945. Accessed at http://edocket.access.gpo.gov/2007/pdf/07-4184.pdf on 7 June 2010.49. Lewarski JS, Gay PC. Current issues in home mechanical ventilation. Chest.2007;132:671-6. [PMID: 17699139]50. American Medical Association. Current Procedual Terminology Code On-line Search. Accessed at http://catalog.ama-assn.org/Catalog/cpt/cpt_search.jspon 7 June 2010.

APPENDIX: COST CALCULATIONSAll costs are adjusted for inflation by the medical compo-

nent of the consumer price index for the U.S. Southeastern re-gion, urban, to 2007 U.S. dollars (47).

Initial HospitalizationCosts for the primary hospitalization (including physician

fees) during which patients were enrolled in the study were de-termined by using itemized charges from each patient’s adminis-trative billing record and were converted to costs by usingdepartment-specific cost-to-charge ratios obtained from the Cen-ters for Medicare & Medicaid Services Healthcare Cost Informa-tion System (19).

Hospital ReadmissionsCosts for subsequent hospital admissions were obtained by

combining reported ICU and hospital lengths of stay with esti-mated average daily ICU and hospital costs by using the Centersfor Medicare & Medicaid Services Medicare Provider AnalysisReview. First, we used all 2006 Medicare Provider Analysis Re-view to identify all discharged patients in the Durham, NorthCarolina, hospital referral region (as defined by the DartmouthAtlas) involving intensive care. Next, we estimated total costs foreach hospitalization by multiplying departmental charges withdepartmental cost-to-charge ratios from the Medicare cost re-ports. We then used observed ICU and hospital lengths of stay toestimate average daily costs, weighting initial, second, and subse-quent ICU days by using a previously validated approach (45,46). After adjustment by the medical component of the con-sumer price index, daily hospital ward costs were $1303. Forpatients requiring ICU care, costs were estimated at $8545 forday 1, $4126 for day 2, and $2231 for subsequent days of care.Professional fees were estimated by adding 17% of hospital costs(20).

Long-Term Acute Care FacilitiesUsing the Medicare Provider Analysis Review database, we

evaluated all patients transferred to a long-term acute care facilityfrom an acute care hospital in the Durham, North Carolina,hospital referral region in 2006. Total daily costs were estimatedat $1657 from the sum of department-specific charges multipliedby department-specific cost-to-charge ratios obtained from 2006Medicare costs reports (45, 46).

Skilled Nursing FacilitiesUsing the Medicare Provider Analysis Review database, we

first identified all 2006 admissions to skilled nursing facilities(out of 8122 admissions to 39 facilities) among persons living inthe 3-digit ZIP code 277xx, which encompasses Durham, NorthCarolina, and the surrounding area. Next, we obtained averagedaily costs ($260) by multiplying total charges for the 8112Medicare admissions to these facilities by the skilled nursingfacility–specific cost-to-charge ratio taken from the MedicareHealthcare Cost Report Information System database, dividingby the length of stay.

Rehabilitation FacilitiesWe used a procedure similar to that described for skilled

nursing facilities to determine daily rehabilitation facility costs($458).

Home Health Care CostsWe calculated home health care costs per Medicare guide-

lines described in the 2007 Federal Register (48). In general, abase payment ($2337 in 2007) is made for the first 60 days ofcare that is itself adjusted for clinical status, functional status, andrecent service utilization, as well as geographical differences inwage. This overall case-mix assessment, done using the Outcomeand Assessment Information Set instrument, is a composite mea-sure of clinical status, functional status, and recent service utili-zation. In this model, we assumed patients fit the case-mix cate-gory of C1F2S2, representing low to moderate disability.

Annals of Internal Medicine

www.annals.org 3 August 2010 Annals of Internal Medicine Volume 153 • Number 3 W-57

Therefore, we multiplied the 2007 base payment ($2337) by thecase-mix adjustment factor corresponding to C1F2S2 (0.9393) toobtain the standard 60-day rate ($2195). Next, we multiplied thecase-mix–adjusted rate by the labor factor (0.77082) and ad-justed this to the wage index of Durham County, North Carolina(0.9816), to calculate the adjusted labor component of the totalcost ($1661). The nonlabor component was calculated by mul-tiplying the nonlabor proportion (0.22918) by the case-mix–adjusted rate ($2195). Finally, the total home health care costswere calculated to be $2164 by adding the labor and nonlaborcomponents. For persons receiving 4 or more days of homehealth care, a per diem rate of $48 was applied, assuming the careof a home health care aide (rather than nurse). Home ventilationwas assigned a cost of $950 per month (49).

Interfacility TransportationHospital data were reviewed to account for all episodes of

air (helicopter and fixed-wing airplane) travel to and from thestudy hospital during the initial hospitalization. Ground am-bulance transport was assumed to occur during all transitions

between hospitals and postacute care facilities. Operationalcosts (labor, supplies, and vehicle) were obtained from insti-tutional billing sources to estimate transportation episodecosts per episode ($2983 for ground transportation and $9270for air transportation).

Outpatient Clinic and Ambulatory Surgery CostsPatient charts were abstracted to quantify episodes of care in

the emergency department, outpatient clinics, and ambulatorysurgeries during follow-up. Costs were derived from averageNorth Carolina service payments based on Current ProceduralTerminology codes listed in the American Medical Associationdatabase (50). For emergency department and clinic visits, weassigned costs based on codes 99284 and 99213 (level 3 estab-lished patient return visit), respectively. Costs for ambulatorysurgical procedures were estimated on the basis of proceduresdocumented in the medical record by using relevant CurrentProcedural Terminology codes. These data were incomplete for 8(6%) patients and were missing for 4 (3%) patients.

W-58 3 August 2010 Annals of Internal Medicine Volume 153 • Number 3 www.annals.org

Appendix Table 1. Mechanical Ventilation and 1-Year Outcomes, by Ventilator Status*

Characteristic All Patients (n ! 126) Weaned FromVentilator (n ! 86)

Not Weaned FromVentilator (n ! 40)

Median days of mechanical ventilation (IQR) 27 (18–24) 21 (16–37) 43 (33–74)

Duration of mechanical ventilation, n (%)!7 d 3 (2) 3 (3) 07–13 d 13 (10) 13 (15) 014–20 d 24 (19) 22 (26) 2 (5)!21 d 86 (68) 48 (56) 38 (95)

Duration of ventilation before tracheostomy, d†Median (IQR) 11 (8–17) 10 (7–16) 14 (9–17)Range 3–47 3–47 3–30

Tracheostomy decannulated, n (%) 82 (65) 82 (95) 0

Median length of hospital stay (IQR), d 39 (28–57) 38 (28–52) 43 (31–77)

Hospital discharge disposition, n (%)Home without paid home health care 6 (5) 6 (7) 0Home with paid home health care 14 (11) 12 (14) 2 (5)Long-term acute care facility 36 (29) 24 (28) 12 (30)Skilled nursing facility 17 (13) 15 (17) 2 (5)Rehabilitation facility 23 (18) 23 (27) 0Other hospital 3 (2) 3 (3.5) 0Still in acute care hospital at 1 y 1 (1) 0 1 (3)Inpatient hospice facility 3 (2) 3 (3.5) 0Dead 23 (18) 0 23 (57)

Median transitions in care location (IQR), n‡ 4 (2–6) 5 (3–7) 2 (1–3)

Percentage of all days alive receiving facility-basedcare or home health care (95% CI)§

74 (68–80) 62 (54–70) 100 (100–100)

1-y survival, n (%) 70 (56) 67 (78) 3 (8)

1-y health outcome, n (%)!Good 9 (7) 9 (11) 0Fair 34 (27) 33 (38) 1 (3)Poor 83 (66) 44 (51) 39 (97)

IQR " interquartile range.* Out of 126 patients.† Does not include 1 patient who never had tracheostomy.‡ Transitions of care location over 1 y for 103 hospital survivors overall. There were 86 hospital survivors among those weaned from ventilators and 17 hospital survivorsamong those who were never weaned from ventilators.§ Includes acute hospitalization, postacute care facilities, and home health care received during the entire study period (1 y)." Health outcome categories measured at 1 y and defined as good (alive with no dependencies in activities of daily living), fair (alive but with 1 to 5 dependencies in activitiesof daily living), and poor (either alive and completely dependent or dead).


Appendix Table 2. Patient and Process-of-Care Outcomes at3 Months*

Outcome Value

Survival, n (%) 90 (71)Disposition, n (%)

Home without paid home care 13 (10)Home with paid home care 19 (15)Long-term acute care facility 10 (8)Skilled nursing facility 14 (11)Inpatient rehabilitation facility 12 (10)Other hospital (transfer or readmission) 13 (10)Residing in acute care hospital, never discharged 6 (5)Dead 36 (29)

Liberated from ventilator and alive, n (%) 78 (87)Patients readmitted, n (%)† 62 (60)Quality of life in 3-mo survivors, n (%)‡

Good 12 (13)Fair 24 (27)Poor 54 (60)

Median total quality-adjusted life-days (IQR)§ 21 (14–35)Health outcome, n (%)"

Good 9 (7)Fair 47 (37)Poor 70 (56)

Mean total costs (SD), $¶ 278 733 (279 855)

IQR " interquartile range.* Out of 126 patients, unless noted otherwise.† Out of 103 hospital survivors. These 62 patients were readmitted for 96 separateepisodes.‡ Out of 90 patients. We determined quality of life by assessment of 3-mo survi-vors’ surrogates by using the EuroQol-5D index score. Scores "0.44 were consid-ered poor, 0.45 to 0.79 were considered fair, and !0.80 were considered good.§ Calculated by adjusting all days alive during 3 mo with correspondingEuroQol-5D index scores obtained at 3-mo follow-up." Health outcome categories measured at 3 mo and defined as good (alive with nodependencies in activities of daily living), fair (alive but with 1 to 5 dependenciesin activities of daily living), and poor (either alive and completely dependent ordead).¶ Includes acute and postacute care facilities.

W-60 3 August 2010 Annals of Internal Medicine Volume 153 • Number 3 www.annals.org

Appendix Table 3. Trajectories of Care and Resource Use at 1 Year, by Ventilator and Health Outcomes*

Characteristic MedianTransitionsof Care(IQR), n

P Value† MedianReadmissions(IQR), n

P Value† MedianPostacuteCare Days(IQR), n

P Value† Mean Total Costs(SD), $

P Value†

Mechanical ventilation characteristics andoutcomes

Mechanical ventilation !21 d (n " 40) 4 (2–6) 0.10 1 (0–2) 0.35 86 (25–285) 0.08 226 114 (167 101) 0.03Mechanical ventilation !21 d (n " 86) 3 (0–6) 1 (0–2) 51 (0–213) 343 819 (320 161)Mechanical ventilation !28 d (n " 65) 4 (2–6) 0.02 1 (0–2) 0.10 80 (23–304) 0.007 235 229 (141 057) 0.004Mechanical ventilation !28 d (n " 61) 2 (0–6) 0 (0–2) 43 (0–116) 381 690 (370 540)!14 ventilator days before tracheostomy

(n " 72)3 (1–5) 0.99 1 (0–2) 0.59 51 (17–250) 0.50 305 741 (356 426) 0.99

!14 ventilator days before tracheostomy(n " 54)

3 (1–6) 1 (0–2) 62 (0–237) 306 660 (148 362)

Weaned from ventilation (n " 86)‡ 5 (3–7) 0.001 2 (0–2) 0.001 82 (32–290) 0.001 274 823 (157 475) 0.08Not weaned from ventilation (n " 40) 0 (0–2) 0 (0) 0 (0–41) 373 456 (447 585)

Health outcome§Good (n " 9) 2 (2–5) 0.001 0 (0–2) 0.005 22 (14–45) 0.02 316 476 (326 988) 0.40Fair (n " 34) 5 (4–7) 2 (1–3) 80 (49–258) 321 757 (185 483)Poor (n " 83) 2 (0–5) 0 (0–2) 44 (0–264) 151 751 (80 534)

Quality of life!Good (n " 19) 4 (2–5) 0.05 1 (0–2) 0.04 55 (21–112) 0.04 230 835 (119 741) 0.50Fair (n " 17) 6 (4–7) 2 (2–3) 288 (38–327) 320 884 (179 454)Poor (n " 34) 5 (2–7) 2 (1–3) 145 (56–310) 346 374 (469 166)

IQR " interquartile range.* Out of 126 patients.† P values based on 1-way analysis of variance or Kruskal–Wallis tests.‡ During the 1-y follow-up.§ Health outcome categories measured at 1-y follow-up and defined as good (alive with no dependencies in activities of daily living), fair (alive but with 1 to 5 dependenciesin activities of daily living), and poor (either alive and completely dependent or dead)." Quality of life determined by assessment of 1-y survivors’ surrogates using the EuroQol-5D index score. Scores "0.44 were considered poor, 0.45 to 0.79 were consideredfair, and !0.80 were considered good.


ORIGINAL INVESTIGATION

Brain DysfunctionAnother Burden for the Chronically Critically Ill

Judith E. Nelson, MD, JD; Nidhi Tandon, MD; Alice F. Mercado, RN, MBA;Sharon L. Camhi, MD; E. Wesley Ely, MD, MPH; R. Sean Morrison, MD

Background: Chronic critical illness is a devastating syn-drome of prolonged respiratory failure and other derange-ments.Toourknowledge,nopreviousresearchhasaddressedbrain dysfunction in the chronically critically ill, althoughthis topic is important for medical decision making.

Methods: We studied a prospective cohort of 203 con-secutive, chronically critically ill adults transferred to ourhospital’s respiratory care unit (RCU) after tracheotomy forfailure to wean. We measured prevalence and duration ofcoma and delirium during RCU treatment using the Con-fusion Assessment Method for the Intensive Care Unit withthe Richmond Agitation-Sedation Scale. To assess survi-vors (at 3 and 6 months after RCU discharge), we used avalidated telephone Confusion Assessment Method.

Results: Before hospitalization, most (153 [75.4%]) ofthe 203 patients in the study were at home, completelyindependent (115 [56.7%]), and cognitively intact (116[82.0%]). In the RCU, 61 (30.0%) were comatose through-

out the stay. Approximately half of patients (66 of 142)who were not in coma were delirious. Patients spent anaverage of 17.9 days (range, 1-153 days) in coma or de-lirium (average RCU stay, 25.6 days). Half of survivors(79 of 160) had one of these disturbances at RCU dis-charge. At 6 months, three fourths (151) of the study pa-tients were dead or institutionalized; of 85 survivors, 58(68.2%) were too profoundly impaired to respond to tele-phone cognitive assessment, and 53 (62.4%) were de-pendent in all activities of daily living.

Conclusions: Severe, prolonged, and permanent braindysfunction is a prominent feature of chronic critical ill-ness. These data, together with previous reports of symp-tom distress and rates of mortality and institutionaliza-tion, describe burdens for chronically critically ill patientsreceiving continued life-prolonging treatment and for theirfamilies.

Arch Intern Med. 2006;166:1993-1999

I NCREASING USE OF INTENSIVE CAREunit (ICU) resources by an agingpopulation has given rise to a newand devastating medical condi-tion: chronic critical illness.1 Al-

though prolonged dependence on mechani-cal ventilation is itshallmark, chroniccriticalillness is not simply an extended period ofacute critical illness but a discrete syn-drome encompassing distinctive derange-ments of metabolism, organ physiologictraits, andendocrineand immunologic func-tion.1 This syndrome has been defined foradministrative, research, and clinical pur-poses by the placement of tracheotomy forfailure to wean in the ICU.2

Chronic critical illness is not a rare con-dition but rather a serious national healthproblem affecting a large and rapidly grow-ing population3,4 who are cared for by gen-eral and specialty clinicians across mul-tiple disciplines. The US health systemalready treats more than 100 000 such pa-tients each year, with annual expendi-tures of approximately $24 billion.2-4

Outcomes reported from a variety of in-stitutions and health care settings (includ-

ing in-patient respiratory care units, long-term acute care facilities, and nursinghomes) remain poor: most chronicallycritically ill patients are dead within 6months, and survivors typically requiretotal care in skilled nursing facilities.5-9

Before their critical illness, most of thesepatients lived independently in the com-munity, despite advanced age and comor-bid medical conditions.5 Most had no medi-cal history of cognitive impairment.5 Ascritical illness became chronic, however,they developed severe weakness, nutri-tional deficits, and significant symptom dis-tress, along with persistent dysfunction ofmultiple organs.5-9 They spend weeks tomonths in hospital, including prolongedICU stays, receiving numerous medica-tions. For all of these reasons, the risk forbrain dysfunction is high.10-12

Prognosis for long-term cognitive func-tion is one of the most important factorsin medical decision making by patients andsurrogates.13,14 Many patients consider cog-nitive impairment to be less acceptablethan death as an outcome.13 It is knownthat in mechanically ventilated patients

Author Affiliations: Division ofPulmonary and Critical CareMedicine, Department ofMedicine (Drs Nelson, Tandon,and Camhi and Ms Mercado),and the Hertzberg PalliativeCare Institute of the BrookdaleDepartment of Geriatrics andAdult Development(Drs Nelson and Morrison),Mount Sinai School ofMedicine, New York, NY;Geriatric Research, Educationand Clinical Center, BronxVeterans AdministrationMedical Center, New York(Dr Morrison); and the Divisionof Allergy/Pulmonary/CriticalCare Medicine, Department ofMedicine, and Center forHealth Services Research,Vanderbilt University School ofMedicine, and the VeteransAdministration TennesseeValley Geriatric Research,Education and Clinical Center,Nashville (Dr Ely).

(REPRINTED) ARCH INTERN MED/ VOL 166, OCT 9, 2006 WWW.ARCHINTERNMED.COM1993

©2006 American Medical Association. All rights reserved.

with acute critical illness, delirium is common,15 has amean duration of 2.4 ICU days,15 and is associated withhigher hospital mortality and morbidity.16 Yet, to ourknowledge, no prior research has addressed cognitive dis-turbance in the context of chronic critical illness. To ex-pand knowledge and enhance decision making, we un-dertook this study describing the prevalence and durationof brain dysfunction (coma and delirium) among hos-pitalized, chronically critically ill patients requiring pro-longed mechanical ventilation and other treatments.

METHODS

RESPIRATORY CARE UNIT

We conducted a prospective study in the respiratory care unit(RCU) of our 1100-bed, university-affiliated, urban tertiary re-ferral hospital. The RCU is a 14-bed, in-patient unit for chroni-cally critically ill patients from the hospital’s 5 adult ICUs (medi-cal, surgical, cardiothoracic surgical, cardiac, and neurosurgical);it is analogous to the increasing numbers of long-term acute carefacilities across the country. Patients are accepted if they no longerneed acute ICU care, are mechanically ventilated via trache-otomy, and, in the clinical opinion of the referring ICU physi-cian, are likely to be liberated from the ventilator. Other chroni-cally critically ill patients may not be transferred to the RCU eitherbecause of death in the ICU after tracheotomy or because the tra-cheotomy was placed in preparation for permanent ventilator sup-port, with no expectation that the patient would ever wean. (From2001 to 2005, when 535 patients were treated in the RCU, 511patients underwent tracheotomy and were not admitted to theRCU; of the latter group, 230 [45%] died in the hospital.)

As supervised by pulmonary and critical care physicians withnurse practitioners, RCU treatment is standardized by a de-tailed health care map and weaning protocol,17 but regular careduring the study did not include a protocol for sedation or cog-nitive assessment. Surviving patients are discharged from theRCU on successful liberation from the ventilator or a clinicaldetermination that further efforts to achieve ventilator inde-pendence will not succeed.

PATIENT ENROLLMENT

From September 2003 to January 2005, we conducted dailyscreening to identify eligible patients among all new RCU ad-missions. We defined eligibility by elective tracheotomy for ICUweaning failure.2,18 Tracheotomy under these circumstances re-flects the clinical impression that the patient is expected nei-ther to wean nor to die in the immediate future and thereforeit serves as a practical point of demarcation between acute andchronic critical illness.19 It is also the defining criterion of na-tional diagnosis related groups 541 and 542 (formerly diagno-sis related group 483),20 in which a large national dataset ofsimilar patients across many institutions and settings are clas-sified and can be compared. We excluded patients with previ-ous ventilator dependence or RCU admission, patients fromother institutions, and those without English proficiency. Theinstitutional review board of Mount Sinai School of Medicine,New York, approved this study, and we obtained informed con-sent from all subjects or appropriate surrogates.

COGNITIVE ASSESSMENT

We conducted cognitive assessments at multiple, longitudinaltime points: at study entry (after tracheotomy and transfer fromICU to RCU), biweekly in the RCU, at RCU discharge, and, for

surviving patients, at 3 and 6 months after RCU discharge. Theseintervals included assessments while patients were mechani-cally ventilated and after ventilator liberation.

For RCU cognitive assessments, we used the Confusion As-sessment Method for the ICU (CAM-ICU),15,21 which includesa preliminary evaluation of consciousness. A trained researchnurse (A.F.M.) performed these assessments.22 To evaluate con-sciousness, we used the Richmond Agitation-Sedation Scale(RASS),23,24 which encompasses 10 levels (0 to !4 indicates in-creasing agitation and 0 to !5 indicates decreasing conscious-ness). We deferred delirium assessment for patients initially ratedas deeply sedated (RASS level !4) or unarousable (RASS level!5), whom we classified together as comatose; we approachedthese patients later the same day and performed delirium as-sessment as possible.22 Next, we evaluated the patients for 4basic features of delirium: (1) acute onset or fluctuating course,(2) inattention (evaluated by the Vigilance A Random LetterTest21), and (3) disorganized thinking or (4) altered level ofconsciousness. We defined delirium by the presence of bothfeatures 1 and 2 and either feature 3 or feature 4.15

We conducted postdischarge cognitive assessments by tele-phone. Because patients able to respond could communicateorally, we used the validated telephone CAM25 for this pur-pose.26 For patients who could not respond to telephone in-terviews, we asked responding surrogates about the nature ofthe patient’s inability to participate (cognitive or physical limi-tation); surrogates reported the patient’s level of conscious-ness and other aspects of cognitive status.

OTHER DATA COLLECTION

We recorded administration of analgesic, sedative, antidepres-sant, antipsychotic, and psychostimulant medications in the RCUto provide a context for cognitive assessments during RCU treat-ment. We collected additional information including diag-noses, comorbid conditions, and other data for Acute Physi-ology and Chronic Health Evaluation II27 and CharlsonComorbidity Index28 scores; lengths of stay; and site and ven-tilator status at discharge and follow-up. To measure func-tional status at multiple time points, we used the motor scaleof the Functional Independence Measure (FIM Motor),29 whichincludes 13 items (eg, eating, bathing, and walking) that areeach scored from 1 (needs total assistance) to 7 (complete in-dependence); total FIM Motor scores range from 13 (com-pletely dependent) to 91 (completely independent). Re-sponses to telephone interviews (used for postdischargefunctional assessments) and to in-person interviews (used forin-hospital assessments) are strongly correlated for the FIM Mo-tor,30 which is well validated. In addition, we calculated the KatzIndex of Activities of Daily Living.31

As a baseline for comparison with our prospective assess-ments of cognition and function in the hospital and after dis-charge, we used interviews with families or other appropriateand knowledgeable surrogates at study entry about the pa-tients’ status during the 2 weeks before hospitalization. Cog-nitive status interviews focused on medical history of demen-tia, decreased level of consciousness, confusion, and memoryloss; we also reviewed patients’ medical records for evidenceof these or other cognitive disturbances. For assessment of base-line functional status, we interviewed families with respect toFIM Motor items, on which surrogate reports are well corre-lated with information from patients themselves.32

STATISTICAL ANALYSIS

We calculated mean and median daily doses of opioids (in mor-phine equivalents) and of sedatives (in lorazepam equivalents) as



the mean and median, respectively, of the individual mean dailydosesgiventoeachpatientduring theRCUstay.Wecomparedpa-tients with delirium or coma and patients without these forms ofbraindysfunctionwithrespect to thepercentageofpatients receiv-ingmedications intheseclassesandtheaveragedailydosesof thesemedications ("2 test and t test, respectively). We calculated dura-tion of delirium or coma in the RCU as the number of consecu-tive days from the first day on which CAM-ICU assessment docu-mented either condition. For each patient, we also calculated thepercentage of total RCU days for each cognitive disturbance un-der study. We used multivariate logistic (discrete outcomes) andmultivariate linear (continuousoutcomes) regression toexaminethe association of number of days spent in delirium or coma withRCU length of stay, ventilator liberation, hospital mortality, siteofdischargeforsurvivors(homevspost–acutecarefacility),3-monthand 6-month residence (home vs nursing home), functional sta-tus,andmortality, controlling forage, sex,ethnicity,CharlsonCo-morbidity Index, Acute Physiology and Chronic Health Evalua-tion II score, and number of acute comorbidities.

RESULTS

Among 330 patients who had undergone tracheotomy andwere consecutively admitted to the RCU, 230 were eli-gible (37, 30, 20, and 13 patients were excluded for priorventilator dependence or RCU treatment, transfer from an-other hospital, language barrier, or other reasons, respec-tively), and 203 (88.2%) were enrolled in the study. Con-sent was refused by 24 patients and was not sought for 3because death supervened. Eighty-nine surviving patientswere discharged from the hospital, and all but 4 (who werelost to follow-up) were followed up for 6 months.

PATIENT CHARACTERISTICS

Table1 details patient characteristics. Study patients weremainly older adults (median age, 72 years) with multiplemedical illnesses and organ failures. Most, however, hadbeen living at home and were functioning independentlybefore their hospitalization. More than 80% (166 of 203)were reported by their families and confirmed by medicalrecords to have no cognitive impairment. The mean (SD)Charlson Comorbidity Index score, Acute Physiology andChronic Health Evaluation II score, and number of acutecomorbid conditions at study entry were 4.5 (2.7), 20.5(5.1), and 2.3 (2.1), respectively (ie, there was a high se-verity illness and comorbidity burden). By then, most pa-tients had already been hospitalized for a long period (me-dian duration of illness [interquartile range], 25.0 days[19.0-34.0 days]), including a prolonged ICU stay (16.0days [11.0-22.0 days]). These patients were treated in theRCU for even longer periods (23.0 days [13.0-35.0 days]).Total duration of exposure to an intensive care environ-ment (ICU plus RCU stay) averaged 45.0 days.

PATIENT OUTCOMES

Patient outcomes are presented in Table 2. Of the 203subjects, 58 (28.6%) died in hospital and less than half(89 [43.8%]) were alive at 6 months. Of 145 hospital sur-vivors, 15 (10.3%) were discharged to home and 107(73.8%) to a nursing home or another acute care hospi-tal; 48 (33.1%) remained mechanically ventilated. Me-

dian (interquartile range) survival of those who died af-ter hospital discharge (n=56) was 55 days (8-92 days).At 6 months, 32 (15.8%) of the 203 original subjects were

Table 1. Characteristics of 203 Chronically Critically IllStudy Patients

Characteristic Value*

Age, median (range), y 72 (21-99)Male 117 (58)Race/ethnicity

White, non-Hispanic 106 (52)Black, non-Hispanic 53 (26)Hispanic 36 (18)Other 8 (4)

Residence before hospital admission†Private home 153 (75)Rehabilitation facility 8 (4)Nursing home 24 (12)Other 16 (8)

FIM Motor score at hospital admission,median (IQR)‡

85 (41-91)

Independent in all activities of daily living32

at hospital admission115 (57)

Cognitively impaired at hospital admission§ 37 (18)Cognitively impaired at study entry! 143 (75)Primary diagnosis at ICU admission

Cardiovascular 45 (22)Pulmonary 75 (37)Neurologic 33 (16)Surgical 31 (15)Other 19 (9)

ICU transferring to RCUMedical 128 (63)Surgical 23 (11)Cardiothoracic 2 (1)Neurosurgical 40 (20)Cardiac care 10 (5)

ICU length of stay, median (IQR), d 16 (11-22)Causes of prolonged respiratory failure¶

Acute lung disease 94 (46)Chronic lung disease 35 (17)Cardiac disease 34 (16)Neurologic disease 57 (28)Surgical or postoperative condition 35 (17)Sepsis/multiorgan dysfunction 55 (27)Other 26 (13)

APACHE II Score (at study entry), mean (SD)† 20.5 (5.1)Charlson Comorbidity Index Score, mean (SD) 4.5 (2.7)

Abbreviations: APACHE II, Acute Physiology and Chronic Health EvaluationII; FIM, Functional Independence Measure; ICU, intensive care unit;IQR, interquartile range; OMCT, Orientation-Memory-Concentration Test;RCU, respiratory care unit.

*Data are given as number (percentage) except where noted.†Two hundred one patients (data were unavailable for 2 patients).‡One hundred ninety-nine patients (prehospital functional status was

unavailable for 4 patients). As described in the “Methods” section, scores(1 [total assistance] to 7 [completely independent]) on 13 individual motoritems in the FIM29 are summed to form the FIM Motor score, which rangesfrom 13 (completely dependent) to 91 (completely independent).

§Defined as reported by family or evidence in medical records of history ofdementia, decreased level of consciousness, confusion, memory loss, orother cognitive impairment.

!Defined as a score greater than 10 on the 6-item OMCT33 (n = 19) orlacking sufficient cognitive capacity to respond to screen (n = 124); 13others refused administration of the OMCT screen; 47 patients had an OMCTscore of 10 or less at study entry.

¶Percentages total more than 100% because prolonged respiratory failurewas typically multifactorial.



living at home (vs 75.4% before hospitalization), 2 ofwhom were still mechanically ventilated. Approxi-mately two thirds (53 of 85) of 6-month survivors (in-cluding 14 [43.8%] of 32 patients at home) were depen-dent in all activities of daily living.

RESULTS OF COGNITIVE ASSESSMENTS

Prevalence of coma and delirium is detailed inTable 3. At RCU admission, 50.2% of patients werecomatose, and 14.8% were delirious. Approximately 1in 3 (61 of 203) study patients were comatose through-out the entire RCU stay. Thirteen (21.3%) of these 61patients were admitted to the ICU for stroke whereasmost of the patients in a persistent coma did not havean underlying, structural injury to the central nervoussystem. Among patients not in coma, delirium wasdiagnosed in almost half (66 of 142). Patients spent amean time of 17.9 days (range, 1-153 days) in coma ordelirium (mean RCU length of stay, 26 days), andamong RCU survivors, half were comatose or deliriousat discharge. Thus, after prolonged, aggressive treat-ment for chronic critical illness, 58.1% of our originalstudy group either were dead (19.2%) or severely cogni-tively impaired at RCU discharge (38.9%); 45 (59.2%) of

the remaining 76 were functionally dependent in allactivities of daily living and required nursing homeplacement.

The Figure plots prevalence of coma and deliriumthrough week 4 in the RCU. The overall burden of braindysfunction did not diminish significantly during this pe-riod. Delirium tended to be persistent or recurrent forpatients who were delirious at RCU admission: 17 (56.7%)of 30 patients who were initially delirious never re-gained normal consciousness prior to discharge or death.For those whose delirium resolved, the mean (SD) timeto resolution was 11.6 (10.4) days. Of 62 patients withnormal consciousness on admission, one quarter subse-quently developed delirium (11 patients) or coma (5 pa-tients), and of these 16, 7 (43.7%) failed to regain a nor-mal state of consciousness.

POSTDISCHARGE

Follow-up data on brain dysfunction are presented inTable 3. At 6 months, 60 (70.6%) of 85 survivors,including 15 (46.9%) of 32 then living at home,remained profoundly cognitively impaired to the pointthat they could not participate in telephone mental sta-tus testing. Of 25 patients who could communicate bytelephone, none was cognitively impaired as measuredby the CAM.

MEDICATIONS

Administration of medications with the number (percent-age) of 203 study patients is given in the following tabu-lation:

Opioid 160 (78.8)Sedative 128 (63.1)Antidepressant 43 (21.2)Antipsychotic 125 (61.6)Psychostimulant 19 (9.4)

The median daily doses of opioid in parenteral mor-phine equivalents, and of benzodiazepine (the only seda-tive used in the RCU) in lorazepam equivalents,36 were5.1 mg/d and 1.0 mg/d, respectively. Among 62 patientswho presented to the study with intact cognition—thatis, neither comatose according to the RASS nor deliri-ous according to the CAM-ICU—35 (92.1%) of 38 whoremained intact received a sedative or opioid comparedwith 16 (100%) of 16 patients who developed brain dys-function during the RCU stay (P=.34) (8 patients diedin the RCU). On average, patients who were comatoseor delirious received 2.4 mg/d of morphine equivalentsand 1.2 mg/d of lorazepam equivalents in the 2 days priorto brain dysfunction, compared with daily doses of 6.4mg of morphine equivalents and 1.3 mg of lorazepamequivalents in the cognitively intact group (P=.30 andP=.76, respectively). Of the 16 patients who were ini-tially intact but developed coma or delirium in the RCU,11 (68.7%) received sedatives, opioids, or antipsychoticdrugs during the 24 hours before the first episode of braindysfunction whereas 9 (56.2%) of 16 received these medi-cations after the episode.

Table 2. Outcomes of 203 Chronically Critically IllStudy Patients

Outcome Value*

Length of stay, median (IQR), dRCU 23 (13-35)Hospital 54 (13-182)

Liberated from mechanical ventilationat RCU discharge†

97 (48)

Hospital discharge siteHome 15 (7)Nursing home or acute care hospital 104 (51)Rehabilitation facility 25 (12)Died in hospital 58 (29)

Cumulative mortalityRCU 39 (19)Hospital 58 (29)3 mo after discharge 103 (51)6 mo after discharge 114 (56)

FIM Motor score, median (IQR)‡At RCU discharge§ 13 (13-24)3 mo after discharge! 29 (13-66)6 mo after discharge¶ 56 (13-87)

Dependent in all activities of daily living31

at 6 months after discharge53 (62)

Abbreviations: FIM, Functional Independence Measure; IQR, interquartilerange; RCU, respiratory care unit.

*Data are given as number (percentage) except where noted.†Defined as 168 hours of unassisted breathing, regardless of any subsequent

mechanical ventilation.5,34 Excludes patients who died in the RCU.‡FIM Motor score29 ranges from 13 (completely dependent) to 91

(completely independent), as explained in the “Methods” section.§One hundred sixty-four patients (39 patients died in the RCU).!Ninety-eight patients (103 patients died, and 1 patient withdrew from study

participation before 3 months; 1 could not be contacted).¶Eighty-five patients (114 patients died, and 2 patients had withdrawn from

the study by the 6-month follow-up; time after discharge was less than 6months for 1 patient as of close of database for these analyses; 1 patient couldnot be contacted).



MULTIVARIATE ANALYSES

The number of days spent in delirium or coma was sig-nificantly associated with an increased likelihood of beingdischarged to a post–acute care facility as opposed to home(odds ratio, 1.09; 95% confidence interval [CI], 1.00-1.20; P=.047), longer length of hospital stay (parameterestimate, 0.03; 95% CI, 0.02-0.03; P#.001 for associa-tion with longer length of stay), and lower FIM Motorscores indicating poorer functional status at 3- and6-month follow-ups for survivors (for 3-month follow-up: parameter estimate, !0.47; 95% CI, !0.78 to !0.15;P=.004; for 6-month follow-up: parameter estimate, !0.92;95% CI, –1.4 to !0.43; P#.001). The association of thenumber of days spent in delirium or coma with 3-monthsurvival for those who survived hospitalization was of bor-derline significance (odds ratio, 0.98. 95% CI, 0.95-1.00; P=.056). Days spent in delirium or coma was notsignificantly associated with in-hospital mortality, like-lihood of liberation from the ventilator, 6-month mor-tality, or 3- or 6-month site of residence.

COMMENT

There are now estimated to be over 100 000 chronically criti-cally ill patients in the United States each year,2,3,37 and num-bers are steadily increasing.4 These patients, who have sur-vived but not recovered from acute critical illness, have adiscrete syndrome: ongoing ventilator dependence, se-vere debility, and characteristic abnormalities of meta-bolic and immunologic function, with resulting failure ofotherorgansandrecurrentnosocomial infections.1 Thepres-

ent study suggests that brain dysfunction, manifest as de-lirium (assessed by CAM-ICU) and coma (levels !4 and!5 on the RASS scale), is another prominent feature ofchronic critical illness. In addition, brain dysfunction inthese patients is prolonged, and its duration is associatedwith longer lengths of stay and poorer functional status.Six months after aggressive treatment in a specialized hos-pital unit, three fourths of our study patients were eitherdead or institutionalized, and two thirds of survivors re-

Table 3. Prevalence of Brain Dysfunction (Coma or Delirium) During Chronic Critical Illness

Cognitive Assessment*

Patients, No. (%)

RCU Treatment PeriodPostdischargeFollow-up, Mo

RCUAdmission

BiweeklyAssessment

RCUDischarge 3 6

Comatose or otherwise unresponsiveCAM-ICU (RASS !4 or !5) 102 (50) 77 (38)† 58 (36)CAM (telephone) 75 (77)‡ 60 (71)‡

Delirious 30 (15) 56 (28)§ 21 (13) 0 0Always 7 (4)$50% and #100% of assessments 22 (11)#50% of assessments 27 (13)

Not delirious or comatose 62 (30) 63 (31) 76 (48) 23 (23) 25 (29)Refused to respond 8 (4) 6 (3) 5 (3) 0 0Total patients ! 202 202 160 98 85

Abbreviations: CAM, Confusion Assessment Method; ICU, intensive care unit; RASS, Richmond Agitation-Sedation Scale; RCU, respiratory care unit.*We assessed brain dysfunction using the CAM-ICU,15 incorporating the RASS.23,24 Patients rated as level !5 (unarousable) or !4 (deeply sedated) on the RASS

(classified together as comatose) lacked sufficient consciousness for further evaluation of delirium at that assessment.†We categorized as comatose those patients who were rated as RASS level !4 or !5 at 75% or more of our biweekly approaches.‡We used the validated telephone CAM25 for postdischarge evaluations.§For patients able to respond to CAM-ICU (ie, not comatose on the initial RASS evaluation) at more than 25% of the biweekly approaches, we categorized as

delirious those patients found to have at least 1 episode of delirium, and we categorized as not delirious those patients who were never found to be delirious.Some patients categorized as delirious or not delirious for the biweekly assessments may have had 1 or more episodes of coma (but #75% of the CAM-ICUapproaches) during this period.

!One patient was not approached for cognitive assessment at RCU admission or during biweekly approaches, and 4 among 164 RCU survivors were dischargedbefore completion of the assessment. The numbers of patients for 3-month and 6-month follow-ups were 98 and 85, respectively, as explained in the previous2 footnotes.

60

40

50

20

30

10

00 1

Not Delirious Delirious Coma

2 3 4Week of RCU Stay, No.

Ass

essm

ents

, %

Figure. We used the Richmond Agitation-Sedation Scale23,24 and theConfusion Assessment Method for the Intensive Care Unit15,35 for biweeklycognitive assessments, as detailed in the “Methods” section. The bar graphshows results of these assessments for 202 study patients through week 4in the respiratory care unit (RCU) (average length of RCU stay, 3 to 4 weeks).



mained profoundly cognitively impaired whereas most ofthese patients had been at home, independent, and with-out cognitive impairment prior to hospitalization. Less than10% were living at home without brain dysfunction as mea-sured by CAM at 6 months. Findings of this study have im-portant implications for decision making.

Delirium occurs commonly in ICU patients with acutecritical illness15,38 and is associated with longer hospitaland ICU lengths of stay35,39,40 and higher mortality at 6months.39 Our data suggest that patients who survive theICU but remain ventilator dependent with chronic criti-cal illness have similarly high rates of delirium as wellas coma. However, unlike ICU delirium (which has a me-dian duration of 48 hours),15 these conditions persist inthe chronically critically ill for a prolonged period. Thisoccurs despite the fact that doses of sedating drugs aregenerally lower after tracheotomy and during ventilatorweaning.41 In our study, 65.3% of patients were admit-ted to the RCU with coma or delirium, more than 25%of those who were initially without coma or delirium de-veloped these conditions in the RCU, and less than halfof both groups combined regained normal cognition.

Cognitive impairment is often considered to be theheaviest burden of illness and is an outcome that maybe even less acceptable than death.13 Patients, families,and clinicians consider information about expected cog-nitive outcome to be important specifically for decisionmaking about treatment when critical illness becomeschronic.14,42 Our documentation herein of the preva-lence and duration of brain dysfunction may help to in-form discussions and decisions at that pivotal juncture.

There are several limitations to this study. We per-formed our research in a single hospital’s in-patient RCU.Our findings may not apply to patients who have under-gone tracheotomy and whose illness is so severe that theycannot survive outside of an ICU (like those in our hos-pital who underwent tracheotomy but died in the ICU be-fore transfer to the RCU) or those who undergo trache-otomy with the understanding that they are not expectedto be liberated but rather to remain mechanically venti-lated indefinitely (like those in our hospital who are notreferred to the RCU because they are clinically deter-mined to be unweanable). Our eligibility criteria, how-ever, were designed to encompass a broad and generaliz-able group of patients requiring prolonged mechanicalventilation via tracheotomy after failure to wean in the ICU.Although treated in various care settings (long-term acutecare facilities and skilled nursing facilities as well as in-patient hospital units), chronically critically ill patients havecomparable characteristics and outcomes, as reported frommultiple institutions and reflected in the national diagno-sis related group 541-542/483 database.6-9,20

We have described our experience with a large seriesof patients initially treated in a variety of ICUs, of whomwe made over 1700 delirium assessments using instru-ments (CAM-ICU and RASS) that were rigorously testedin prior work. Because the CAM-ICU was already vali-dated in 2 patient groups,15,21 we did not retest it against areference standard but instead focused on use with a dis-crete group of patients with prolonged illness and me-chanical ventilation. In follow-up after RCU discharge, weused the original CAM (validated telephone version),25

which is more sensitive than CAM-ICU for diagnosing de-lirium in nonintubated patients.26 The number of pa-tients who could respond to cognitive assessment after hos-pital discharge was limited by the high rates of mortalityand institutionalization; patients able to respond to tele-phone interview were less likely to have delirium and otherforms of cognitive impairment than nonresponders, whichwould have underestimated the true prevalence of post-discharge delirium in the entire study population.

CONCLUSIONS

As the general population ages and intensive care treat-ments are offered to older and sicker patients, cliniciansin all fields and disciplines will encounter chronically criti-cally ill patients with increasing frequency. Most of thesepatients die within 6 months, and most survivors are in-stitutionalized with severe functional impairments. Ourprevious research5 indicates that most patients withchronic critical illness spend weeks to months in the hos-pital experiencing a multitude of distressing physical andpsychological symptoms. The present study suggests thatfew patients with chronic critical illness avoid deliriumor coma and that most of them spend significant time dur-ing treatment and thereafter with these severe forms ofbrain dysfunction. Although regular, comprehensive as-sessment and treatment of symptoms may reduce suf-fering, it is less clear that effective treatments exist to ad-dress the high prevalence of brain dysfunction observedhere. Nor can we be certain that such treatments wouldimprove mortality rates or functional outcomes.

Thechoice tocontinue life-prolongingtreatmentswhencritical illness enters achronicphasemaybemadebysomepatients and families despite poor prognosis for survivaland for cognitive and functional recovery. However, ac-cumulating evidence of symptom distress and long-termcognitive and functional impairment calls attention towhether they are truly making informed decisions at thetime of tracheotomy and whether ongoing goals of careare routinelydiscussedduring thecourseof critical illness.Futureresearchshouldfocusonstrategiestoalleviatesymp-tom burden and to address delirium and other brain dys-function. In addition, it is essential to examine how in-formation is communicated and how decisions are madeabout the benefits and burdens of life-prolonging thera-pies (eg, mechanical ventilation) for the chronically criti-cally ill. In the meantime, options presented to decisionmakers for these patients should include a time-limitedtrial of continued respiratory support, with a plan for dis-continuation of the ventilator if cognitive and functionalrecovery do not occur within a reasonable period of time.The levels of disease complexity, physical and psychologi-cal suffering, and cognitive impairment also suggest thatquality care for these patients should include not only ex-pert pulmonary and critical care but also involvement ofa palliative care consultation team.

Accepted for Publication: June 23, 2006.Correspondence: Judith E. Nelson, MD, JD, Box 1232,Mount Sinai School of Medicine, One Gustave L. Levy Place,New York, NY 10029 ([email protected]).



Author Contributions: Dr Nelson had full access to allof the data in the study and takes responsibility for theintegrity of the data and the accuracy of the data analy-sis. Study concept and design: Nelson and Morrison. Ac-quisition of data: Nelson, Mercado, Camhi, and Morrison.Analysis and interpretation of data: Nelson, Tandon, Camhi,Ely, and Morrison. Drafting of the manuscript: Nelson,Camhi, and Morrison. Critical revision of the manuscriptfor important intellectual content: Nelson, Tandon,Mercado, Ely, and Morrison. Statistical analysis: Tandonand Morrison. Obtained funding: Nelson. Administrative,technical, and material support: Tandon, Mercado, Ely, andMorrison. Study supervision: Nelson and Morrison.Financial Disclosure: Dr Ely has received investigator-initiated, unrestricted grants from Pfizer Inc and fromHospira Inc to evaluate pharmacologic strategies for pre-vention and treatment of delirium; those projects and theirsponsors have no connection to the study reported herein.Dr Ely is also the recipient of similar grants from Eli Lillyand has received honoraria from all 3 companies.Funding/Support: This study was supported by grant R01AG21172 from the National Institute on Aging (NIA). DrNelson is the recipient of an Independent Scientist Award(K02 AG024476) from the NIA. Dr Morrison is the recipi-ent of a Midcareer Investigator Award in Patient-OrientedResearch from the NIA (K24 AG022345). Dr Ely is the re-cipient of a Mentored Patient-Oriented Research Career De-velopment Award from the NIA (K23 AG01023).Role of the Sponsors: The sponsors for this study hadno role in the design or conduct of the study; collection,management, analysis, or interpretation of the data; orpreparation, review, or approval of the manuscript.Acknowledgment: We are grateful to the chronically criti-cally ill patients in our RCU, to their families, and to thededicated nurse practitioners and RCU staff for their co-operation in this research.

REFERENCES

1. Nierman DM, Nelson JE, eds. Chronic critical illness. Crit Care Clin. 2002;18(themeissue):461-715.

2. Carson SS, Bach PB. The epidemiology and costs of chronic critical illness. CritCare Clin. 2002;18:461-476.

3. US Department of Health and Human Services. CMS programs & information.Centers for Medicare & Medicaid Web site. http://www.cms.hhs.gov. AccessedJune 15, 2006.

4. Dewar DM, Kurek CJ, Lambrinos J, Cohen IL, Zhong Y. Patterns in costs andoutcomes for patients with prolonged mechanical ventilation undergoing tra-cheostomy: an analysis of discharges under diagnosis-related group 483 in NewYork State from 1992 to 1996. Crit Care Med. 1999;27:2640-2647.

5. Nelson JE, Meier DE, Litke A, Natale DA, Siegel RE, Morrison RS. The symptomburden of chronic critical illness. Crit Care Med. 2004;32:1527-1534.

6. Carson SS, Bach PB, Brzozowski L, Leff A. Outcomes after long-term acute care:an analysis of 133 mechanically ventilated patients. Am J Respir Crit Care Med.1999;159:1568-1573.

7. Nasraway SA, Button GJ, Rand WM, Hudson-Jinks T, Gustafson M. Survivorsof catastrophic illness: outcome after direct transfer from intensive care to ex-tended care facilities. Crit Care Med. 2000;28:19-25.

8. Dasgupta A, Rice R, Mascha E, Litaker D, Stoller JK. Four-year experience witha unit for long-term ventilation (respiratory special care unit) at the ClevelandClinic Foundation. Chest. 1999;116:447-455.

9. Scheinhorn DJ, Chao DC, Stearn-Hassenpflug M, LaBree L, Heltsley DJ. Post-ICU mechanical ventilation. Chest. 1997;111:1654-1659.

10. Inouye SK, Bogardous ST, Charpentier P, et al. A multicomponent interventionto prevent delirium in hospitalized older patients. N Engl J Med. 1999;340:669-676.

11. Marcantonio ER, Juarez G, Goldman L, et al. The relationship of postoperativedelirium with psychoactive medications. JAMA. 1994;272:1518-1522.

12. Inouye SK, Charpentier PA. Precipitating factors for delirium in hospitalized el-derly persons: predictive model and interrelationship with baseline vulnerability.JAMA. 1996;275:852-857.

13. Fried TR, Bradley EH, Towle VR, Allore H. Understanding the treatment prefer-ences of seriously ill patients. N Engl J Med. 2002;346:1061-1066.

14. Nelson JE, Kinjo K, Meier DE, Ahmad K, Morrison RS. When critical illness be-comes chronic: informational needs of patients and families. J Crit Care. 2005;20:79-89.

15. Ely EW, Inouye SK, Bernard GR, et al. Delirium in mechanically ventilated pa-tients: validity and reliability of the Confusion Assessment Method for the Inten-sive Care Unit (CAM-ICU). JAMA. 2001;286:2703-2710.

16. Ely EW, Shintani A, Truman B, et al. Delirium as a predictor of mortality in mechani-cally ventilated patients in the intensive care unit. JAMA. 2004;291:1753-1762.

17. Carasa M, Nespoli G. Nursing the chronically critically ill patient. Crit Care Clin.2002;18:493-508.

18. Nierman DM, Mechanick JI. Bone hyperresorption is prevalent in chronically criti-cally ill patients. Chest. 1998;114:1122-1128.

19. Nierman DM. A structure of care for the chronically critically ill. Crit Care Clin.2002;18:477-491.

20. US Department of Health and Human Services. Acute inpatient prospective pay-ment system overview. Centers for Medicare and Medicaid Services Web site.http://www.cms.hhs.gov/AcuteInpatientPPS. Accessed July 28, 2006.

21. Ely EW, Margolin R, Francis J, et al. Evaluation of delirium in critically ill pa-tients: validation of the Confusion Assessment Method for the Intensive Care Unit(CAM-ICU). Crit Care Med. 2001;29:1370-1379.

22. Brain dysfunction in critically ill patients. ICU Delirium and Cognitive Impair-ment Study Group Web site. http://www.icudelirium.org. Accessed June 15, 2006.

23. Sessler CN, Gosnell MS, Grap MJ, et al. The Richmond Agitation-Sedation Scale:validity and reliability in adult intensive care unit patients. Am J Respir Crit CareMed. 2002;166:1338-1344.

24. Ely EW, Truman B, Shintani A, et al. Monitoring sedation status over time in ICUpatients: reliability and validity of the Richmond Agitation-Sedation Scale (RASS).JAMA. 2003;289:2893-2991.

25. Marcantonio ER, Michaels M, Resnick NM. Diagnosing delirium by telephone.J Gen Intern Med. 1998;13:621-623.

26. McNicoll L, Pisani MA, Ely EW, Gifford D, Inouye SK. Detection of delirium in theintensive care unit: comparison of confusion assessment method for the inten-sive care unit with confusion assessment method ratings. J Am Geriatr Soc. 2005;53:495-500.

27. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of dis-ease classification system. Crit Care Med. 1985;311:818-829.

28. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifyingprognostic comorbidity in longitudinal studies: development and validation.J Chronic Dis. 1987;40:373-383.

29. Stineman MG, Hamilton BB, Goin JE, Granger CV, Fiedler RC. Functional gainand length of stay for major rehabilitation impairment categories. Am J Phys MedRehabil. 1996;75:68-78.

30. Smith PM, Illig SB, Fielder RC, Hamilton B, Ottenbacher K. Intermodal agree-ment of follow-up telephone functional assessment using the Functional Inde-pendence Measure in patients with stroke. Arch Phys Med Rehabil. 1996;77:431-435.

31. Katz S, Downs TD, Cash HR, Grotz RC. Progress in the development of the indexof ADL. Gerontologist. 1970;10:20-30.

32. Magaziner J, Bassett SS, Hebel JR, Gruber-Baldini A. Use of proxies to measurehealth and functional status in epidemiologic studies of community-dwelling womenaged 65 years and older. Am J Epidemiol. 1996;143:283-292.

33. Katzman R, Brown T, Fuld P, Peck A, Schechter R, Schimmel H. Validation of ashort Orientation-Memory-Concentration Test of cognitive impairment. Am JPsychiatry. 1983;140:734-739.

34. Scheinhorn DJ, Hassenpflug M, Artinian BM, LaBree L, Catlin JL. Predictors ofweaning after 6 weeks of mechanical ventilation. Chest. 1995;107:500-505.

35. Ely EW, Gautam S, Margolin R, et al. The impact of delirium in the intensive careunit on hospital length of stay. Intensive Care Med. 2001;27:1892-1900.

36. Chan JD, Treece PD, Engelberg RA, et al. Narcotics and benzodiazepine use af-ter withdrawal of life support. Chest. 2004;126:286-293.

37. Healthcare Cost and Utilization Project Web site. http://hcup.ahrq.gov/HCUPnet.asp. Accessed June 15, 2006.

38. McNicoll L, Pisani MA, Zhang Y, Ely EW, Siegel MD, Inouye SK. Delirium in theintensive care unit: occurrence and clinical course in older persons. J Am Geri-atr Soc. 2003;51:1-9.

39. Ely EW, Stephens RK, Jackson JC, et al. Current opinions regarding the impor-tance, diagnosis, and management of delirium in the intensive care unit: a sur-vey of 912 healthcare professionals. Crit Care Med. 2004;32:106-112.

40. Thomason JW, Shintani A, Peterson JF, Pun BT, Jackson JC, Ely EW. Intensivecare unit delirium is an independent predictor of longer hospital stay: a prospec-tive analysis of 261 non-ventilated patients. Crit Care. 2005;9:R375-R381.

41. Nieszkowska A, Combes A, Luyt CE, et al. Impact of tracheotomy on sedativeadministration, sedation level, and comfort of mechanically ventilated intensivecare unit patients. Crit Care Med. 2005;33:2527-2533.

42. Camhi SL, Morrison RS, Mercado A, Tandon N, Nelson JE. Communication aboutchronic critical illness. Proc Am Thor Soc. 2005;2:A594.



A prognostic model for one-year mortality in patients requiringprolonged mechanical ventilation*

Shannon S. Carson, MD; Joanne Garrett, PhD; Laura C. Hanson, MD, MPH; Joyce Lanier, RRT;Joe Govert, MD; Mary C. Brake, MD; Dante L. Landucci, MD; Christopher E. Cox, MD, MPH;Timothy S. Carey, MD, MPH

As patient management strategiesin the intensive care unit con-tinue to advance, more patientsare surviving the early acute

phases of critical illness. However, when mul-tiorgan failure fails to resolve or leads to sub-sequent complications such as critical illnesspolyneuropathy, prolonged mechanical venti-lation (PMV) can result (1, 2). The number ofpatients requiring PMV has been increasingover the last decade and promises to increasedramatically when members of the baby

boomer generation reach advanced age andbecome particularly susceptible to this com-plication (3).

Patients requiring PMV consume adisproportionately high amount ofhealthcare resources both in the inten-sive care unit and after hospital discharge(4, 5). Their short-term and long-termmortality is high (6), and they experiencea very heavy symptom burden for pro-longed periods (7, 8). Hospital survivorshave a significant degree of functionaland cognitive limitations and a high re-admission rate (9). Some remain at highrisk for death after hospital discharge,but not all. Prolonged hospitalization forpatients on PMV who are at high risk ofdeath does not meet current standards ofcost-effectiveness (10). Considering thehigh symptom burden of this populationand often poor outcomes, a mortality pre-diction model that identifies patients onPMV with the highest and lowest risk fordeath would be useful to inform discus-sions of prognoses among clinicians andpatients or their surrogate decision-makers. Such a model could also stan-dardize illness severity in cohort studies

examining outcomes and interventions inthis resource-intensive group of patients.

A consensus conference defined PMVas patients requiring invasive mechanicalventilation for at least 21 days after acuteillness (1). We conducted a prospectivecohort study to develop and validate amortality prediction model for adult pa-tients meeting this definition. Our inten-tion was to develop a model that would bepractical for use in the clinical settingand have very high specificity in patientsat highest risk of death.

MATERIALS AND METHODS

Patients

A total of 300 adult patients were prospec-tively enrolled from University of North Caro-lina Hospitals, a 640-bed university-based ter-tiary care medical center with 65 adultintensive care unit beds that can accommo-date mechanically ventilated patients. Twohundred patients were consecutively enrolledfrom November 2001 to January 2004 for thedevelopment set of the prognostic model. Onehundred patients were consecutively enrolledfrom February 2004 to June 2005 to form the

*See also p. 2200.From the Department of Medicine (SSC, JG, LCH,

JL, TSC), University of North Carolina School of Med-icine, Chapel Hill, NC; the Department of Medicine (JG,CEC), Duke University School of Medicine; and theDepartment of Medicine (MCB, DLL), East CarolinaSchool of Medicine.

Supported by National Institutes of Health grantsK23 HL067068 (SSC) and K23 HL081048 (CEC).

The authors have not disclosed any potential con-flicts of interest.

For information regarding this article, E-mail:[email protected]

Copyright © 2008 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e31817b8925

Objective: A measure that identifies patients who are at highrisk of mortality after prolonged ventilation will help physicianscommunicate prognoses to patients or surrogate decision mak-ers. Our objective was to develop and validate a prognostic modelfor 1-yr mortality in patients ventilated for 21 days or more.

Design: The authors conducted a prospective cohort study.Setting: The study took place at a university-based tertiary

care hospital.Patients: Three hundred consecutive medical, surgical, and

trauma patients requiring mechanical ventilation for at least 21days were prospectively enrolled.

Measurements and Main Results: Predictive variables weremeasured on day 21 of ventilation for the first 200 patients andentered into logistic regression models with 1-yr and 3-mo mor-tality as outcomes. Final models were validated using data from100 subsequent patients. One-year mortality was 51% in thedevelopment set and 58% in the validation set. Independent

predictors of mortality included requirement for vasopressors,hemodialysis, platelet count <150 ! 109/L, and age >50 yrs.Areas under the receiver operating characteristic curve for thedevelopment model and validation model were .82 (SE .03) and .82(SE .05), respectively. The model had sensitivity of .42 (SE .12) andspecificity of .99 (SE .01) for identifying patients who had >90%risk of death at 1 yr. Observed mortality was highly consistentwith both 3- and 12-mo predicted mortality. These four predictivevariables can be used in a simple prognostic score that clearlyidentifies low-risk patients (no risk factors, 15% mortality) andhigh-risk patients (three or four risk factors, 97% mortality).

Conclusions: Simple clinical variables measured on day 21 ofmechanical ventilation can identify patients at highest and lowestrisk of death from prolonged ventilation. (Crit Care Med 2008; 36:2061–2069)

KEY WORDS: mechanical ventilation; illness severity scores; out-comes; statistical model; critical illness; prognosis

2061Crit Care Med 2008 Vol. 36, No. 7

model’s validation set. Enrollment criteria in-cluded requirement of mechanical ventilationafter acute illness for at least 21 days afterinitial intubation. If patients were extubatedwithin that initial 21-day period but neededreintubation, they were enrolled only if theperiod of spontaneous breathing was !72 hrs.Exclusion criteria included age !18 yrs, se-vere burns, chronic neuromuscular diseases,chronic mechanical ventilation before admis-sion, receipt of "7 days of mechanical venti-lation before transfer from a referral center,prisoners, and refusal of consent.

Patients in adult medical and surgical in-tensive care units were screened on a dailybasis. All eligible patients were enrolled forreview of existing medical records. We re-quested permission through primary physi-cians to approach patients or, as was usuallynecessary, their surrogates to request consentfor interviews and telephone follow-up. If pa-tients or surrogates refused consent to partic-ipate, they were excluded from the study, in-cluding review of existing data. If a surrogatewas not available, follow-up was achieved byreview of medical records and the NationalDeath Index. The research protocol was ap-proved by the University of North CarolinaInstitutional Review Board.

Data Collection

On day 21 of mechanical ventilation, med-ical records were abstracted for demographicdata, diagnoses, comorbidities, and premorbidfunctional status. Physiological variables wererecorded from the first day of intensive careunit admission and from day 21 of mechanicalventilation. Acute Physiology and ChronicHealth Evaluation II scores were calculatedusing data from the first 24 hrs of intensivecare unit admission (11). Sequential OrganFailure Assessment scores were calculated us-ing data from the first 24 hrs of intensive careunit admission and from data collected on day21 of mechanical ventilation (12). The Charl-son Index score, a measure of medical comor-bidities, was calculated from medical recorddata based on conditions present at day 21(13). Premorbid functional status was assessedby the surrogate’s perception of whether thepatient needed assistance with any activity ofdaily living (ADL) during the 2 wks beforeacute illness.

Patients were followed during the rest oftheir hospitalization for duration of mechani-cal ventilation, mortality, intensive care unitand hospital disposition, and length of stay.Patients or surrogates who consented to tele-phone follow up were contacted at 3 mos, 6mos, and 12 mos from the time of enrollment(day 21 of mechanical ventilation). They wereinterviewed regarding the patient’s vital sta-tus, place of residence, number of hospitalreadmissions, requirement for mechanical

ventilation, tracheostomy, feeding tubes, andthe patient’s functional status. Performance ofsix basic ADLs was assessed by questionnairesasking how much assistance patients neededwith feeding, getting out of bed, walking,dressing, toileting, and bathing (14). A writtennotification that a phone call was going to bemade was mailed 2 wks before the scheduledcontact, and multiple telephone calls were at-tempted until the patient or surrogate wasreached. Patients and surrogates were re-minded that they had the option of not an-swering any or all of the questions. For hos-pital survivors who were enrolled for reviewof existing data only, and for patients whowere lost to telephone follow-up, 1-yr mor-tality was assessed by review of the NationalDeath Index.

All data collection instruments were pre-tested using records from ten patients whowere not part of the study sample. Revisionswere made after clarifications by all investiga-tors. Data on the first ten patients enrolled inthe study were collected by both the primarydata collector and the principal investigator toensure concordance. Similar quality checks

were conducted on a random sample of 10% ofthe first 100 patients enrolled. Subjective vari-ables such as primary and secondary diagnosesand comorbidities were made by both theprimary data collector and the principal in-vestigator on all patients, and discrepancieswere settled together. Interview instru-ments were pretested on a sample of tenpatients and surrogates, and revisions weremade accordingly. Telephone interviewersreceived full instruction from the principalinvestigator, and mock interviews were con-ducted until performance was consistentand reproducible. Analysis on the develop-ment set model was not begun until followup on the validation set was completed andthe database was closed.

Statistical Analysis

Summary analyses were performed on demo-graphic and physiological variables and ex-pressed as mean # SD for normally distributeddata and median, interquartile range for non-normal data. Power analyses indicated that alogistic regression model with 200 patients

Figure 1. Enrollment and follow-up data. MV, mechanical ventilation; NDI, National Death Index.

2062 Crit Care Med 2008 Vol. 36, No. 7

and 130 expected deaths would have sufficientpower to include 13 variables. These predictorvariables were chosen a priori based on clini-cal judgment and previous studies in differentsettings (15–18). All were measured on day 21of mechanical ventilation. The variables in-cluded age, premorbid independence in ADL,PaO2/FIO2, inability to lift the upper extremityfrom the bed, requirement for any dose ofpressor (dopamine, norepinephrine, phenyl-ephrine), platelet count, requirement for he-modialysis (any patient receiving hemodialysisbetween days 19 and 22 of mechanical venti-lation or any patient with renal failure forwhom hemodialysis had been indicated butwithheld), and specific comorbidities (severechronic pulmonary disease, peripheral vascu-lar disease, diabetes mellitus with chroniccomplication, congestive heart failure). Biva-riate analysis of associations between the pri-mary outcome, death at 1 yr, and the prese-lected predictor variables were performed fordescriptive purposes. Potential collinearitywas assessed by examining pairwise correla-tions and measuring variance inflation factors.Collinearity was not found to be an issue inour data and therefore did not affect our mod-eling strategy.

All variables identified a priori as potentialpredictors were included in a logistic regres-sion model with death at 1 yr as the primaryoutcome. The maximal model was reduced bythe investigators by eliminating variables se-quentially and comparing each new model bylikelihood ratio tests and by comparing thearea under the receiver operating characteris-tic curve for each new model. Calibration ofthe model was assessed using Pearson’s chi-square goodness-of-fit test (GoF). Odds ratiosand 95% confidence intervals associated witheach variable in the final model are reported. Asimilar model was constructed using 3-monthmortality (90 days after day 21 of mechanicalventilation) as the primary outcome.

For the validation phase of the study, val-ues for predictive variables measured for the100 patients in the validation set were enteredinto logistic regression models using the betavalues from the reduced logistic regressionmodel from the 200-patient developmentphase. Validation of the prediction model wasestablished by comparing the area under thereceiver operating characteristic curve, sensi-tivity, and specificity of the development andvalidation study models. Calibration of the val-idation study model was assessed using Pear-son’s GoF test. Similar analyses were per-formed using 3-month mortality as theprimary outcome.

A clinical prediction rule was adapted fromthe final prediction model by assigning pointsto each predictive variable based on regressioncoefficients from the development model. Per-formance of this clinical prediction rule was

assessed by comparing observed to predictedoutcomes for 1-yr and 3-month mortality and bycomparison of area under the receiver operatingcharacteristic curve to that of the final model.

Data are presented as mean # SD or me-dian (interquartile range). Area under receiveroperating characteristic curves, sensitivity,and specificity are presented as value (SE). All

analyses were performed using Stata 8.0 soft-ware (Stata, College Station, TX).

RESULTS

Of 336 consecutive patients who wereeligible for the study, 36 were excluded(Fig. 1). Vital status 1 yr after enrollment

Table 1. Patient characteristics

Variable Development Validation p Value

n $ 200 n $ 100Age, mean # SD 55.7 # 16.7 55.5 # 16.6 .92Age, median (IQR) 58 (42–69) 57 (44–66) .82Male, n (%) 120 (60) 49 (49) .07Race, n (%) .31

White 124 (62) 61 (61)African American 61 (31) 36 (36)Hispanic 9 (5) 1 (1)Asian 4 (2) 1 (1)Native American 1 (0.5) 1 (1)

Premorbid statusResidence, n (%) n $ 181 n $ 81 .27

Home 171 (94) 75 (93)Assisted living facility 3 (2) 4 (5)Skilled nursing facility 7 (4) 2 (2)

Independent in ADLs, n (%) n $ 175 n $ 79 .29143 (82) 60 (76)

APACHE II ICU Admit, mean # SD 20.6 # 7.3 25.3 # 6.8 .0001SOFA day 1 MV, mean # SD 9.9 # 3.4 9.8 # 2.9 .78

Day 21 measurementsAdvance directives, n (%) .94

Do-not-resuscitate order 12 (6) 6 (6)Advanced power of attorney 9 (5) 6 (6)Living will 5 (3) 2 (2)None 168 (87) 81 (85)

Service, n (%) .20Medicine 79 (40) 46 (46)General surgery/trauma 56 (28) 26 (26)Cardiac surgery 20 (10) 5 (5)Thoracic surgery 18 (9) 10 (10)Neurosurgery 17 (9) 4 (4)Transplant surgery 7 (4) 5 (5)

SOFA day 21 MV, mean # SD 7.6 # 3.9 7.7 # 3.4 .94PaO2/FIO2, mean # SD 219 # 95.7 216 # 110 .82WBC, mean # SD 13.3 # 8.6 11.7 # 7.1 .10Platelet count (%109/L), mean # SD 307 # 216 243 # 159 .009Pressors, n (%) 33 (16) 28 (28) .02Hemodialysis, n (%) 49 (25) 32 (32) .15Albumin, median (IQR) 2.0 (1.7–2.4) 2.0 (1.8–2.5) .58BMI, mean # SD 30.2 # 8.2 33.5 # 11.9 .009Charlson index score, mean # SD 2.7 # 2.2 3.1 # 2.3 .16Specific comorbidities, n (%)

Severe chronic pulmonary disease 22 (11) 14 (14) .45Chronic vascular disease 16 (8) 7 (7) .78Diabetes with chronic complications 23 (12) 10 (10) .70Congestive heart failure 30 (15) 15 (15) 1.0

Upper extremity strength, n (%) .10Against gravity 129 (66) 59 (60)Withdraw to pain 42 (21) 17 (17)No movement 25 (13) 22 (22)

Lower extremity strength, n (%) .10Against gravity 94 (48) 42 (43)Withdraw to pain 63 (32) 26 (27)No movement 38 (19) 30 (31)

Tracheostomy, n (%) 167 (84) 77 (77) .38Days to tracheostomy, n (%) 17 (12–22) 19 (12–26) .11

ADLs, activities of daily living; APACHE II, Acute Physiology and Chronic Health EvaluationSystem; SOFA, Sequential Organ Failure Assessment; WBC, white blood cell count; BMI, body massindex; IQR, interquartile range; MV, mechanical ventilation; ICU, intensive care unit.


was confirmed by telephone follow up ormedical record review in 263 patients andby National Death Index review in 25patients. One-year mortality was un-known for 12 patients for an overall fol-low-up rate of 96%.

Patient demographics and outcomesfor the development and validation setsare presented in Tables 1 and 2. Intensivecare unit admission diagnoses are shownin Appendix 1. The groups were mostlysimilar, but patients in the validation sethad higher admission Acute Physiologyand Chronic Health Evaluation II scoresand higher hospital and 3-month mortal-ity. Thirty-seven of the patients who sur-vived hospitalization were not confirmedto have died by review of medical recordsor telephone follow-up. National DeathIndex records were available for 25 ofthem. All were noted to have survived theyear. The remaining 12 patients forwhom National Death Index records werenot yet available were counted as survi-vors based on survival of the 25 patientslost to telephone follow up who did haveNational Death Index data available.

Results of bivariate analyses are pre-sented for descriptive purposes in Table

Table 2. Outcomes

VariableDevelopment

N $ 200ValidationN $ 100 p Value

Hospital disposition, n (%) .56Died 82 (41) 50 (51)Long-term acute care 21 (11) 13 (13)Rehabilitation 59 (30) 24 (24)Skilled nursing facility 15 (8) 5 (5)Home with assistance 19 (10) 5 (5)Home independent 2 (1) 1 (1)

If diedReceived CPR at time of death, n (%) 10 (12) 3 (6) .02Days DNR to death, median (IQR) 1 (0–3) 2.5 (1–7) .27

Liberated from MV in hosp, n (%) 106 (53) 43 (43) .23Liberated, hospital survivors n $ 118 n $ 50 .88

95 (81) 39 (78)Reintubated, n (%) 14 (8) 14 (14) .10Liberated from MV in one year, n (%) 114 (58) 47 (49) .14Ventilator days, median (IQR) 35 (26–51) 35 (27–54) .71

Ventilator days, survivors 39 (29–58) 38 (29–52) .72Ventilator days, nonsurvivors 32 (25–45) 35 (25–59) .29

ICU length of stay, median (IQR) 37 (28–52) 36 (30–54) .46Hospital length of stay, median (IQR) 51 (36–72) 50 (37–74) .91Mortality, n (%)

Three months 83 (42) 52 (52) .08One Year: known follow-up n $ 175 n $ 84 .11

103 (59) 58 (69)One Year: includes NDI dataa n $ 200 n $ 100 .18

103 (52) 58 (58)

aAll hospital survivors in the Development set who did not have a record of subsequent death atUniversity of North Carolina survived the year based upon National Death Index (NDI) records. Thisassumption was made for 12 similar patients in the Validation set; CPR, cardiopulmonary resuscitation;DNR, do-not-resuscitate; MV, mechanical ventilation; IQR, interquartile range; ICU, intensive care unit.

Table 3. Bivariate analysis of associations between predetermined predictive variables and one-year mortality in development set

Variable n Survived Died RR (95% CI) p Value

Age"50 years 128 51 (40) 77 (60) 1.66 (1.19, 2.34) .001!50 years 72 46 (64) 26 (36)

ADLsNeeds assistance with 1 ADL 32 11 (34) 21 (66) 1.47 (1.07, 2.0) .03No assistance needed 143 79 (55) 64 (45)

PaO2/FIO2, mean # SD 181 229 # 96 208 # 103 — .21Upper extremity strength

Cannot lift against gravity 61 15 (25) 46 (75) 1.84 (1.44, 2.36) .0001Can lift against gravity 137 81 (59) 56 (41)

VasopressorsRequired 33 2 (6) 31 (94) 2.18 (1.79, 2.66) .0001Not required 165 94 (57) 71 (43)

Platelets!150 % 109/L 48 4 (8) 44 (92) 2.41 (1.93, 3.01) .0001"150 % 109/L 150 93 (62) 77 (38)

HemodialysisRequired 49 10 (20) 39 (80) 1.88 (1.49, 2.37) .0001Not required 151 87 (58) 64 (42)

Chronic pulmonary diseasePresent 22 10 (45) 12 (55) 1.07 (0.71, 1.60) .76Absent 178 87 (49) 91 (51)

Peripheral vascular diseasePresent 16 8 (50) 8 (50) 0.97 (0.58, 1.61) .90Absent 184 89 (48) 95 (52)

Diabetes with chronic complicationPresent 23 7 (30) 16 (70) 1.41 (1.03, 1.93) .06Absent 177 90 (51) 87 (49)

Congestive heart failurePresent 30 12 (40) 18 (60) 1.20 (0.86, 1.67) .31Absent 170 85 (50) 85 (50)

ADL, activities of daily living; RR, relative risk; CI, confidence interval.


3. All predetermined predictor variableswere included in the initial maximal lo-gistic regression model. Requirement ofvasopressors, platelets !150 % 109/L, age"50 yrs, requirement of hemodialysis,and upper extremity weakness were inde-pendent predictors of death at 1 yr in areduced model. Clinically, upper extrem-ity weakness was considered difficult toreproduce because the use of sedatives,which strongly affected this measure-ment, varied significantly between pa-tients. This issue has affected the reliabil-ity of other illness severity models (19).Therefore, models with and without thisvariable were compared. The area underthe receiver operating characteristiccurve for the final reduced model shownin Table 4 (vasopressors, platelets

!150 % 109/L, age "50 yrs, requirementof hemodialysis) was .82 (SE .03). Thiscompares with .84 (SE .02) for the modelwith the final four variables plus upperextremity weakness and .85 (SE .03) forthe maximal model (p $ .46 for compar-ison of all three). For the final reducedmodel, sensitivity for identifying patientsat "50% risk of death was .58 (SE .16),and specificity was .91 (SE .16). Sensitivityfor identifying patients at "90% risk ofdeath was .42 (SE .12), and specificity was.99 (SE .01). The model had good fit basedon its nonsignificant GoF test (&2

10df $6.72, p $ .75).

Using values measured in patientsfrom the validation set, the same modelhad an area under the receiver operatingcharacteristic curve of .82 (SE .05) (p $

.93 compared with development set) andagain demonstrated good fit (GoF &2

16df $18.31, p $ .31). Comparisons of observedto predicted values for development andvalidation sets are shown in Figure 2.Reliability of the model was very consis-tent in the validation set. These four vari-ables were also independent predictors of3-mo mortality in a separate model (GoF&2

10df $ 11.39, p $ .33) and showed con-sistent performance in the validation set(GoF &2

16df $ 23.99, p $ .09) (Table 4).As a sensitivity analysis, patients in the

validation set who were lost to follow upwere assumed to have all died (rathersurvive, as was the case in the develop-ment set, confirmed by the NationalDeath Index). The area under the receiveroperating characteristic curve for that

Figure 2. Comparison of observed and predicted 1-yr mortality for patients divided into five equal-sized groups from the Development set (A) and Validationset (B). CI, confidence interval.

Table 4. Model performance

Variable

Three-Month MortalityDevelopmentOR (95% CI)

One-Year MortalityDevelopmentOR (95% CI)

Vasopressor 4.2 (1.2, 14.2) 8.8 (1.6, 48.4)Platelets !150 % 109/L 7.1 (2.7, 18.6) 14.5 (4.1, 50.8)Age "50 years old 3.5 (1.6, 7.8) 5.6 (2.4, 12.9)Requiring hemodialysis 3.1 (1.3, 7.5) 2.9 (1.1, 7.7)

Three-Month Mortality One-Year Mortality

Model Development Validation Development Validation

Area under ROC (SE) 0.81 (0.03) 0.79 (0.05)a 0.82 (0.03) 0.82 (0.05)b

Sensitivityc (SE) 0.31 (0.10) 0.32 (0.13) 0.42 (0.12) 0.44 (0.20)Specificityc (SE) 0.97 (0.01) 0.95 (0.02) 0.99 (0.01) 0.95 (0.02)

ap $ .75 for comparison with Development set; bp $ .93 for comparison with Development set; csensitivity and Specificity determined for 90% risk ofdeath. Presented as value (standard error).

CI, confidence interval; OR, odds ratio; ROC, receiver operating characteristic.


model was .76 (SE .05). Sensitivity was .37(SE .18) and specificity remained high at.93 (SE .02) for "90% likelihood of death.

To create a prognostic scoring systemthat could ultimately be used by clini-cians in daily practice, we assigned pointsto each of the four predictive variables inproportion to the regression coefficientsfrom the development model. The regres-sion coefficients were of similar magni-tude, so we assigned 1 point for each riskfactor resulting in a range of scores from0 to 4. Performance of the 4-point prog-

nostic scoring system (Prognosis for Pro-longed Ventilation [ProVent] score) isshown in Table 5. Predicted and observedmortality for patients in the developmentset and observed mortality for patients inthe validation set are included. In thedevelopment set, patients with the Pro-Vent score of 0, representing no risk fac-tors (n $ 41 [21%]) had a 1-yr mortalityof only 15%. Patients with the score of 1,representing one risk factor (n $ 98[50%]), had a 1-yr mortality rate of 42%.Patients with the score of 2, representing

two risk factors (n $ 26 [13%]), hadmortality rate of 77% at 3 mos and 88%at 1 yr. Patients with the three or fourrisk factors had 3-mo mortality of 90%and 1-yr mortality of 97%. This highestrisk group (scores 3 or 4) represents 16%of the development set and 24% of thevalidation set. The area under the re-ceiver operating characteristic curve forProVent score for the combined cohort is.82 (SE .03; 95% confidence interval, .75–.88). For patients with "50% risk ofdeath, sensitivity is .58 (SE .12) and spec-ificity is .95 (SE .07). For patients with"90% risk of death, sensitivity is .32 (SE

.20) and specificity is .99 (SE .01). Sur-vival according to ProVent score riskgroup is shown in Figure 3.

Data on functional status were avail-able for 57% of 1-yr survivors. There wereno differences between patients with andwithout available data for age (p $.4),Sequential Organ Failure Assessmentscore at day 21 (p $ .30), Charlson score(p $ .88), or premorbid independence inADLs (p $ .68). Only 24% of survivorswere independent in all ADLs after 1 yr.Thirty-nine percent of survivors withProVent scores of 0 and 18% of survivorswith ProVent scores of 1 were indepen-dent in all ADLs. None of the patientswith ProVent scores of 2 or greater wereboth alive and independent in all ADLsafter 1 yr.

DISCUSSION

This prospective cohort study con-firms that four easily measured variablesrecorded at day 21 of ventilation canidentify patients who are both at high risk

Figure 3. Kaplan-Meier curves by risk group for combined cohort: low $ ProVent score 0 (no riskfactors), n $ 55 (18% of cohort); intermediate $ ProVent score 1 (one risk factor), n $ 137 (47% ofcohort); high $ Provent score 2 (two risk factors), n $ 47 (16% of cohort); highest $ ProVent score3 or 4 (three or four risk factors), n $ 52 (17% of cohort). Day 0 is the time of intubation.

Table 5. Prognosis for Prolonged Ventilation (ProVent) score variables measured on Day 21 of mechanical ventilation: Age "50 $ 1 point Vasopressor $1 point platelets !150 % 109/L $ 1 point requires hemodialysis $ 1 point

ProVent Score

Development Set Validation Set

n (%) Predicted 1-Year Mortality (95% CI) Observed 1-Year Mortality n (%) Observed 1-Year Mortality

0 41 (21) 0.12 (0.06, 0.21) 0.15 14 (14) 0.141 98 (50) 0.44 (0.36, 0.53) 0.42 42 (42) 0.432 26 (13) 0.83 (0.71, 0.90) 0.88 21 (21) 0.863 22 (11) 0.97 (0.90, 0.99) 0.95 13 (13) 1.04 9 (5) 0.99 (0.97, 1.0) 1.0 8 (8) 1.0

Development Set Validation Set

Predicted 3-Month Mortality 95% (CI) Observed 3-Month Mortality Observed 3-Month Mortality

0 41 (21) 0.10 (0.05, 0.17) 0.12 14 (14) 0.071 98 (50) 0.32 (0.25, 0.40) 0.29 42 (42) 0.382 26 (13) 0.67 (0.55, 0.77) 0.77 21 (22) 0.803 22 (11) 0.90 (0.78, 0.95) 0.91 13 (14) 0.854 9 (5) 0.97 (0.91, 0.99) 0.89 8 (8) 1.0

CI, confidence interval.


and low risk of mortality during pro-longed mechanical ventilation. Thisprognostic model has very high specific-ity, limiting the possibility of inappropri-ately poor prognoses. The model per-formed well during validation in a cohortthat was enrolled during a different timeperiod than the development set and thathad higher illness severity. Three of thefour variables that are independent pre-dictors of mortality—requirement ofpressors, requirement of hemodialysis,and platelet count !150 % 109/L, reflectongoing systemic inflammation and mul-tiorgan failure. The other prognosticvariable, age 50 or older, likely reflectslower physiological reserve independentof acute organ failure and specific comor-bidities. It may also reflect less willing-ness on the part of older patients or sur-rogates to endure weeks and months ofinvasive care when progress does notseem apparent (20).

Much has been written about how pa-tients on PMV require a unique approachto care as a result of differences in phys-iology (1, 21–23). However, few studies ofinterventions in this patient populationhave been published. The ability to stan-dardize illness severity would facilitatethe design of cohort studies evaluatinginterventions to improve process of careand survival. For example, as a result ofissues of high costs and limited re-sources, hospitals are compelled to dis-charge patients on PMV to variouspostintensive care unit settings, includ-ing respiratory care units, long-term carehospitals (LTCH), or even skilled nursingfacilities, for continued weaning andmanagement (24 –26). These facilitieshave been proliferating at a high pace tomeet increasing demand (27). Althoughit is possible that these facilities decreasehospital costs, it is not clear whether out-comes are affected. This prognosticmodel was developed and validated in apopulation with relatively limited accessto postacute care weaning facilities.Therefore, this model provides an acutecare baseline against which outcomesfrom care in different settings can becompared. Variables for the model aremeasured before most LTCH transfers oc-cur (25) so illness severity can be stan-dardized before transfer to alternativecare settings.

Two prognostic models have beenpublished for patients on PMV managedin LTCH (15, 16), but neither have beenvalidated and only one included long-term follow up. In one study, existing

illness severity scores demonstrated poordiscrimination and calibration for hospi-tal mortality in patients on PMV at anLTCH. When measured on the day of ad-mission to the LTCH, the area under thereceiver operating characteristic curvewas !.70 for Acute Physiology andChronic Health Evaluation II, SimplifiedAcute Physiology Score II, MortalityProbability Model II, and Logistic OrganDystunction System (28).

The majority of patients with ad-vanced illnesses do not want to be keptalive on life support when there is littlehope for a meaningful recovery (29). Fo-cus group studies involving patients onPMV and their families have revealed thatthey would benefit from more directcommunication with healthcare provid-ers, especially with regard to prognosis(30). Another study of prognosticationduring physician–family discussionsabout limiting life support revealed thatprognoses for long-term survival weregiven in only 12% of conferences (31). Inthe Study to Understand Prognoses andPreferences for Outcomes and Risks ofTreatment (SUPPORT) study, of the 1,494patients who spent more than 14 days inthe intensive care unit, fewer than 40%reported that their physicians had talkedwith them about their prognoses or pref-erences for life-sustaining treatment(32). The simple prognostic model devel-oped in this study could enhance com-munication of prognosis to these patientsand their surrogates by providing objec-tive estimates of short-term and long-term outcome.

A major strength of this study is theheterogeneous patient population, in-cluding medical and surgical patients aswell as patients with major trauma. Theprognostic model does not require as-signment of a specific diagnosis, which isusually difficult in critically ill patientswith multiple active processes. Nor doesit require assessments of neurologicfunction, which can be unreliable in pa-tients receiving sedation (19). Selectionbias was limited by consecutive and pro-spective enrollment and a high follow-uprate for mortality.

This study has several important lim-itations. Differences in management atother centers or communities could re-sult in worse performance of this modelin those settings. External validation us-ing multiple tertiary care centers in di-verse regions is warranted before clinicalor research application of this model isconsidered (33). Although the study was

large enough to have sufficient power toassess the preselected variables in thestudy protocol, other potential predictorsmay not have been examined. However,the variables that were studied produceda model that is simple, reproducible, andhighly specific.

It is possible that objective prognosticinformation will not change physicianpractice. In the SUPPORT trial, an inter-vention using a sophisticated prognosticmodel designed to facilitate discussion ofprognosis and wishes for aggressive carein acutely ill patients had no significantimpact on these outcomes (34). There areseveral reasons why the prognostic modelin this study could have a more signifi-cant impact than that of the SUPPORTtrial. The prognostic score is simple tounderstand and can be assessed by theclinician at the bedside within secondsrather than relying on an intermediarywith a complicated formula. The prog-nostic information comes later in the pa-tient’s clinical course when extensive ef-forts have been made on the patient’sbehalf, yet progress has stalled and re-serve is limited. Both clinicians and sur-rogates may be more likely to accept achange in the course of care when pooroutcomes are expected despite weeks ofmaximal treatment.

The majority of physicians find prog-nostication to be stressful and difficult,and they feel that they have inadequatetraining in this area (35). They are par-ticularly concerned about being wrong,especially when withholding or with-drawing life support is a possible out-come of decision making. Variables wereselected a priori for this model with anaim to identify the patients at highest riskof death. Consequently, the model hasvery high specificity (.99) for patientswith a 90% mortality risk. There is min-imal chance of misclassifying a patient asvery high risk (false-positive). Measuringspecificity at a high mortality risk comesat the expense of lower sensitivity. Asmany as 58% of patients who ultimatelydied were not classified in the highestrisk group (false-negatives). When prog-nosticating, however, most clinicians areworried more about giving negative prog-noses for patients who would otherwisesurvive (35), favoring a mortality modelwith high specificity.

Of course, objective prognostic in-formation will not change physicianpractice in isolation. Other importantfactors are necessary to improve pa-tient/family communication about end-


of-life issues (36, 37). Finally, such ascoring system should not be used toreplace clinician judgment regardinglikely outcomes, but rather to informthose judgments (38).

CONCLUSIONS

Patients receiving PMV who are at thehighest risk of death can be identifiedbased on the requirement of either vaso-pressors or hemodialysis or the presenceof platelet counts !150 % 109/L or ageover 50 yrs. After external validation, aprognostic scoring system using theserisk factors could facilitate earlier andmore definitive discussions between cli-nicians and patients or surrogates regard-ing appropriate goals of care.

REFERENCES

1. MacIntyre NR, Epstein SK, Carson S, et al:Management of patients requiring prolongedmechanical ventilation: Report of a NAMDRCconsensus conference. Chest 2005; 128:3937–3954

2. Carson SS, Bach PB: The epidemiology andcosts of chronic critical illness. Crit CareClin 2002; 18:461–476

3. Cox CE, Carson SS, Holmes GM, et al: In-crease in tracheostomy for prolonged me-chanical ventilation in North Carolina,1993–2002. Crit Care Med 2004; 32:2219–2226

4. Wagner DP: Economics of prolonged me-chanical ventilation. Am Rev Respir Dis1989; 140:S14–18

5. Douglas SL, Daly BJ, Gordon N, et al: Sur-vival and quality of life: Short-term versuslong-term ventilator patients. Crit Care Med2002; 30:2655–2662

6. Cox CE, Carson SS, Lindquist JAH, et al:Differences in one-year health outcomes andresource utilization by definition of pro-longed mechanical ventilation: A prospectivecohort study. Crit Care 2007; 11:R9

7. Nelson JE, Meier DE, Litke A, et al: Thesymptom burden of chronic critical illness.Crit Care Med 2004; 32:1527–1534

8. Nelson JE, Tandon N, Mercado AF, et al:Brain dysfunction. Another burden for thechronically critically ill. Arch Intern Med2006; 166:1993–1999

9. Carson SS: Outcomes of prolonged mechan-ical ventilation. Curr Opin Crit Care 2006;12:405–411

10. Cox CE, Carson SS, Govert JA, et al: Aneconomic evaluation of prolonged mechani-cal ventilation. Crit Care Med 2007; 35:1918–1927

11. Knaus WA, Draper EA, Wagner DP, et al:APACHE II. A severity of disease classifica-tion system. Crit Care Med 1986; 311:818–829

12. Vincent JL, de Mendonca A, Cantraine F, etal: Use of the SOFA score to assess the inci-dence of organ dysfunction/failure in inten-sive care units: Results of a multicenter, pro-spective study. Working group on ‘sepsis-related problems’ of the European Society ofIntensive Care Medicine. Crit Care Med 1998;26:1793–1800

13. Charlson ME, Pompei P, Ales KL, et al: A newmethod of classifying prognostic comorbidityin longitudinal studies: Development andvalidation. J Chron Dis 1987; 40:373–383

14. Scott WK, Macera CA, Cornman CB, et al:Functional health status as a predictor ofmortality in men and women over 65. J ClinEpidemiol 1997; 50:291–296

15. Carson SS, Bach PB, Brzozowski L, et al:Outcomes after long-term acute care. Ananalysis of 133 mechanically ventilated pa-tients. Am J Respir Crit Care Med 1999; 159:1568–1573

16. Dematte D’Amico JE, Donnelly HK, MutluGM, et al: Risk assessment for inpatient sur-vival in the long-term acute care setting afterprolonged critical illness. Chest 2003; 124:1039–1045

17. Chao DC, Scheinhorn DJ, Stearn-Hassenpflug M:Impact of renal dysfunction on weaning fromprolonged mechanical ventilation. Crit Care1997; 1:101–104

18. Cook DJ, Rocker D, Marshall MD, et al: With-drawal of mechanical ventilation in anticipa-tion of death in the intensive care unit.N Engl J Med 2003; 349:1123–1132

19. Polderman KH, Jorna EM, Girbes AR: Inter-observer variability in APACHE II scoring:Effect of strict guidelines and training. In-tensive Care Med 2001; 27:1365–1369

20. Suchyta MR, Clemmer TP, Elliott CG, et al:Increased mortality of older patients withacute respiratory distress syndrome. Chest1997; 111:1334–1339

21. Scheinhorn DJ, Stearn-Hassenpflug M: Pro-vision of long-term mechanical ventilation.Crit Care Clin 1998; 14:819–832

22. Scheinhorn DJ, Chao DC, Hassenpflug MS,et al: Post-ICU weaning from mechanicalventilation: The role of long-term facilities.Chest 2001; 120(Suppl):482S–484S

23. Nierman DM, Nelson JE: A structure of carefor the chronically critically ill. Crit CareClin 2002; 18:477–492

24. Rudy EB, Daly BJ, Douglas S, et al: Patientoutcomes for the chronically critically ill:Special care unit versus intensive care unit.Nurs Res 1995; 44:324–331

25. Scheinhorn DJ, Hassenpflug MS, Votto JJ, etal; Ventilator Outcomes Study Group: Venti-

lator-dependent survivors of catastrophic ill-ness transferred to 23 long-term care hospi-tals for weaning from prolonged mechanicalventilation. Chest 2007; 131:76–84

26. Corrado A, Roussos C, Ambrosino N, et al:Respiratory intermediate care units: A Euro-pean survey. Eur Respir J 2002; 20:1343–1350

27. Medicare Payment Advisory Commission: Defin-ing long-term care hospitals (Chapter 5, June2004 report). Available at: www.medpac.gov/publications. Accessed November 2004

28. Carson SS, Bach PB: Predicting mortality inpatients suffering from prolonged critical ill-ness. Chest 2001; 120:928–933

29. Heyland DK, Dodek P, Rocker G, et al; for theCanadian Researchers, End-of-Life Network(CARENET): What matters most in end-of-life care: Perceptions of seriously ill patientsand their family members. CMAJ 2006; 174:1–9 [online]

30. Nelson JE, Kinjo K, Meier DE, et al: Whencritical illness becomes chronic: Informa-tional needs of patients and families. J CritCare 2005; 20:79–89

31. White DB, Engelberg RA, Wenrich MD, et al:Prognostication during physician–family dis-cussions about limiting life support in inten-sive care units. Crit Care Med 2007; 35:442–448

32. Teno JM, Fisher E, Hamel MB, et al: Deci-sion-making and outcomes of prolonged ICUstays in seriously ill patients. J Am GeriatrSoc 2000; 48(Suppl):S70–74

33. Justice AC, Covinsky KE, Berlin JA: Assessingthe generalizability of prognostic informa-tion. Ann Intern Med 1999; 130:515–524

34. A controlled trial to improve care for seri-ously ill hospitalized patients. The Study toUnderstand Prognoses for Outcomes andRisks of Treatments (SUPPORT). TheSUPPORT Principal Investigators. JAMA1995; 274:1541–1548

35. Christakis NA, Iwashyna TJ: Attitude andself-reported practice regarding prognostica-tion in a national sample of internists. ArchIntern Med 1998; 158:2389–2395

36. Lautrette A, Darmon M, Megarbane B, et al: Acommunication strategy and brochure forrelatives of patients dying in the ICU. N EnglJ Med 2007; 356:469–478

37. Curtis JR, Patrick DL, Shannon SE, et al: Thefamily conference as a focus to improve com-munication about end-of-life care in the in-tensive care unit: Opportunities for improve-ment. Crit Care Med 2001; 29(Suppl 2):N26–N33

38. Lynn J, Teno JM, Harrell FE: Accurate prog-nostications of death. Opportunities andchallenges for clinicians. West J Med 1995;163:250–257


APPENDIX

Intensive Care UnitAdmitting Diagnosis

n $ 200Development Set

n $ 100 Validation Set

Pulmonary fibrosis 3 (2) 0Chronic obstructive pulmonary disease 4 (2) 2 (2)Acute respiratory distress syndrome 10 (5) 5 (5)Respiratory arrest 3 (2) 5 (5)Cystic fibrosis 5 (3) 1 (1)Pneumonia 13 (7) 14 (14)Sepsis 14 (7) 8 (8)Congestive heart failure 2 (1) 4 (4)Myocardial infarction 3 (2) 0Cardiac arrest 5 (3) 2 (2)Intracranial hemorrhage, nonoperative 3 (2) 0Overdose 1 (!1) 2 (2)Neuromuscular weakness 8 (4) 0Hepatic failure 1 (!1) 2 (2)Gastrointestinal hemorrhage 3 (2) 2 (2)Pancreatitis 6 (3) 1 (1)Other GI condition 1 (!1) 2 (2)Hematologic malignancy 2 (1) 0Other malignancy 1 (!1) 1 (1)Other medical 4 (2) 1 (1)Multiple trauma 32 (16) 10 (10)Head trauma 4 (2) 0C-spine injury 3 (2) 1 (1)Coronary artery bypass graft 5 (3) 1 (1)Heart valve surgery 8 (4) 2 (2)Thoracic surgery 13 (7) 6 (6)GI perforation/obstruction 7 (4) 6 (6)Other GI surgery 6 (3) 7 (7)Vascular surgery 7 (4) 2 (2)Surgery for intracranial hemorrhage 8 (4) 2 (2)Heart transplant 3 (2) 2 (2)Lung transplant 1 (!1) 1 (1)Liver transplant 8 (4) 6 (6)Other surgery 3 (2) 2 (2)

Data presented as n (%). Percentages do not add to 100 due to rounding. Other Medical includesasthma, diabetic ketoacidosis, pulmonary embolus, meningitis, acute renal failure, 1 case each.

GI, gastrointestinal.


table of contents-five key articles for sit cliniciansfigure 1. the syndrome of chronic critical...

Documents