Journal of Critical Care (2012) 27, 421.e1–421.e7

Effect of 24-hour mandatory vs on-demand critical carespecialist presence on long-term survival and quality of lifeof critically ill patients in the intensive care unit of ateaching hospital☆,☆☆,★

Martin Reriani MB, ChBa,b, Michelle Biehl MDa,b, Jeff A. Sloan PhDc,Michael Malinchoc MSb,d, Ognjen Gajic MDa,b,⁎

aDivision of Pulmonary and Critical Care Medicine, Rochester, MN, USAbMayo Epidemiology and Translational Research in Intensive Care (METRIC), Rochester, MN, USAcDivision of Mayo Clinic College of Medicine, Rochester, MN, USAdDivision of Biostatistics, Rochester, MN, USA

fimDpa

0d

Keywords:Long-term survival;Quality of life;Critical care staffing

AbstractBackground: Mandatory compared with on-demand intensivist presence improves processes of care anddecreases intensive care unit (ICU) complication rate and hospital length of stay. The effect ofcontinuous mandatory intensivist coverage on long-term patient mortality and quality of life (QOL) isnot known.Methods: We compared the long-term survival before (year 2005) and after (year 2006) the interventionwhen the staffing model changed from on-demand presence to mandatory 24-hour staff-critical carespecialist presence in the medical ICU. Baseline and 6-month QOL surveys (SF-36 [short form 36health survey]) were compared in subgroups of patients admitted before and after the staffing change.Cox proportional hazard and paired statistical analyses were used for survival and QOL comparisons.Results: The baseline characteristics did not differ significantly between the 2 groups except for raceand Acute Physiology and Chronic Health Evaluation III score (median, 30 vs 37; P b .001 before andafter the staffing model change). Long-term survival was not significantly different before and after thestaffing change—adjusted hazard ratio, 1.05; 95% confidence interval, 0.95 to 1.16; P = .3. In a subsetof ICU survivors, SF-36 physical component score improved significantly at 6 months compared with

☆ Author contributions: Dr Reriani contributed to developing the underlying study hypothesis, interpretation of the results, manuscript preparation, andnal approval of the submitted manuscript. Dr Biehl contributed to the interpretation of the results, manuscript preparation, and final approval of the submittedanuscript. Michael Malinchoc contributed to the performance of data extraction, statistical analysis results, and final approval of the submitted manuscript.r Gajic contributed to developing the underlying study hypothesis, identifying the appropriate study population, interpretation of the results, manuscriptreparation, and final approval of the submitted manuscript. He is the guarantor of the manuscript and takes responsibility for the integrity of the work aswhole.

☆☆ Financial/nonfinancial disclosures: None.★ Funding/support: None.⁎ Corresponding author. Mayo Clinic, Rochester, MN 55905. Tel.: +1 507 255 6276; fax: +1 507 255 4267.E-mail address: gajic.ognjen@mayo.edu (O. Gajic).

883-9441/$ – see front matter © 2012 Elsevier Inc. All rights reserved.oi:10.1016/j.jcrc.2011.10.001

421.e2 M. Reriani et al.

baseline after the staffing model change—Δ mean (SD) 8 (14) vs 2 (11), P = .03. However, there wasno difference in the Δ mean mental component score of the SF-36 between the 2 groups (P = .77).Conclusions: Introduction of an additional night shift to provide mandatory as opposed to on-demand24-hour staff critical care specialist coverage did not affect long-term survival of medical ICU patients.© 2012 Elsevier Inc. All rights reserved.

1. Introduction

Several studies have demonstrated that staffing the inten-sive care units (ICUs) by critical care physicians improvespatient outcomes [1-3]. A systematic review on physicianstaffing in the ICU showed that high-intensity staffing(mandatory intensivist consultation or closed ICU) improvesICU and hospital mortalities and reduces ICU and hospitallength of stay (LOS) when compared with low-intensitystaffing (no intensivist or elective intensivist consultation)[4]. We recently showed that having a 24-hour mandatorystaffing presence in the ICU further improves processes ofcare and staff satisfaction and decreases ICU complicationrate and hospital LOS when compared with on-demandcritical care specialist presence [5].

Traditionally, the assessment of critical care outcomes hasfocused largely on in-hospital mortality. However, as morepatients survive critical illness, assessing other outcomessuch as long-term survival and quality of life (QOL) hasbecome a priority [6]. A recent systematic review assessingQOL in adult survivors of critical illness showed that QOL inthis population is lower before ICU admission, and afterhospital discharge, it improves but remains lower than thegeneral population level [7]. Assessment of these outcomemeasures could help families and health care providers insetting goals of ICU care and potentially help in ICUdecision making.

We undertook to improve patient care in our institution byimplementing an intervention in the critical care specialiststaffing model from on-demand to mandatory nighttimecoverage [5]. The new model introduced an additional nightshift critical care specialist staffing in week-long blocks inaddition to the residents and critical care fellows who werepresent before the intervention. The current follow-up studywas designed to extend our previous observations of theimpact of continuous 24-hour on-site presence of a criticalcare specialist on long-term survival and QOL of critically illpatients in an academic institution.

2. Methods

In the follow-up of a previously published prospectivestudy [5], we compared long-term survival and QOL at 6months from baseline (as measured by SF-36 [Short Form 36Health Survey]) before and after the intervention when the

staffing model changed from on-demand presence tomandatory 24-hour staff-critical care specialist presence.This study was approved by the Mayo FoundationInstitutional Review Board, Mayo Institutional ReviewBoard number 2150-05.

Critically ill patients consecutively admitted to themedical ICU and their care providers were included in thestudy. Patients who denied research authorization and thosewho were admitted for low-risk monitoring were excluded[8]. The study design and general characteristics of themedical ICU have been previously described [5].

2.1. Brief description of the ICU setting andstaffing intervention

The study was conducted in a 24-bed medical ICU withan average daily admission rate of 7 and an average midnightcensus of 16 patients. Before the change in staffing model,2 ICU teams, each led by a staff critical care specialist,provided care during the daytime with alternate admissiondays and nighttime coverage. Although 1 ICU fellow and2 internal medicine residents provided continuous in-housenighttime coverage, staff critical care specialists on callwere available by pager and were expected to come to theICU on demand within 15 to 30 minutes after being called.The on-call staff critical care specialist communicated withthe on-call fellow via pager and telephone, and decisionsabout the need to see a specific patient before the nextmorning were based on the presentation given by the criticalcare fellow and the severity of illness. Other factors, such astotal ICU acuity and activity, could influence the nocturnalin-house presence of the staff critical care specialist. The newstaffing model was introduced on January 3, 2006, andconsisted of an additional night-shift staff critical carespecialist who was to attend to all patient care needs on sitebetween 7 PM and 7 AM. This attendance served the samefunction as the daytime staff critical care specialist includingindependent physical examination, review of the medicaldatabase, review of the plan of care, supervision of allinvasive procedures, and trainee education. The schedule ofthe night-shift staff critical care specialist was arranged as ablock of seven 12-hour night shifts followed by 5 daysliberated from on-site work duty. Individual critical carespecialists were allowed to shorten the 1-week blockschedule as necessary for personal or professional reasons.Multidisciplinary ICU rounds typically occurred every

421.e3Mandatory vs. on-demand critical care specialist presence

morning during both periods. Standardized order sets wereavailable for short- and long-term sedation, ventilatormanagement, electrolyte replacement, and sepsis manage-ment. No other major practice model interventions weremade during the study period.

2.2. Baseline characteristics

Baseline characteristics and ICU and hospital outcomeswere assessed from the prospective data collection by thebedside ICU nurses into the Acute Physiology and ChronicHealth Evaluation (APACHE) III database [8] for 12 monthsbefore and after the intervention. The severity of illness andthe predicted lengths of ICU and hospital stay werecalculated using the APACHE III prognostic model [8,9].

2.3. Long-term survival

Patients were followed up from the time of discharge fromthe hospital to June 2010. Death was ascertained for allpatients who were followed up at our institution. We usedseveral different medical databases to determine the vitalstatus of these patients including Registration File Database,Institutional Planning Database, Hospital Utilization ReviewDatabase, decision support, Survey Research Center (SRC),follow-up, and electronic medical records from St Mary'sHospital and Rochester Methodist Hospital. For thosewho were not followed up at our institution or got lost tofollow-up, we enlisted the help of the survey research center(SRC) in our institution to determine the vital status of thepatients as well as the last date that the patients were knownto be alive or their date of death.

2.4. Quality of life

In a subset of patients admitted in the last 2 months ofthe control period (November and December 2005) and2 months of the intervention period (March and April 2006,after a 2-month run-in period elapsed), QOL was assessedusing the SF-36 Quality of Life Survey administered atbaseline and 6 months after ICU discharge. Baseline(premorbid) functional status and QOL were determined byin-hospital survey of the patients or their surrogates. Afterobtaining informed consent, trained study coordinatorsestablished if the patient was competent to complete theentire questionnaire by administering the minimental test. Ifthe patient was deemed incompetent or too ill to complete thesurvey, a surrogate was identified to help fill the question-naires. Follow-up contact information was obtained, and thepatients or their surrogate who successfully completed thebaseline survey were contacted by telephone 6 months afterindex hospitalization. The prespecified primary end pointwas the change at 6 months from baseline of QOL asmeasured by the overall SF-36 QOL survey.

2.5. Statistical methods

2.5.1. Survival analysisThe starting point for all survival analyses was date of

admission to the ICU. Death from any cause was the endpoint, and the patient's survival time was the interval fromICU date to death date. For patients who had not died,survival time was the interval from the ICU date to the lastdate they were known to be alive. This interval was censored.The Kaplan-Meier method was used to estimate the patient'ssurvival functions both before and after the staffing change,with the zero time point being the ICU admission. Thesefunctions were estimated for the total sample and for low-,moderate-, and high-risk patients. These risk classes weredefined by the APACHE score (measured within 1 hour ofadmission) using the 25th and 75th quantiles of theAPACHE score. The log-rank test was used to comparethe survival functions before and after the staffing change.The Cox proportional hazards model was used to estimatethe hazard ratio (HR) and confidence intervals and to adjustthe HRs for imbalance in baseline risk factors. The factorsin the Cox model were the staffing change, APACHEclasses, and the interactions between staffing and APACHEclasses. These analyses were performed using the SAS (SASInstiture Inc. Carry NC USA) and S-Plus (Productdevelopment Company, Needham MA) statistical software.

2.5.2. Quality of lifeThe responses to the SF-36 items for the 8 SF-36 scales were

scored using the publisher's normative methods that, in turn,generated the summary physical and mental components'scores. All of these scores are standardized such that, in anormal age- and sex-matched reference population, the meanscore is equal to 50 and SD of the scores is equal to 10. Patientswho have SF-36 scores greater than 50 report better health-related QOL than the normal population, and scores less than 50indicate poorer QOL. For patients who completed the SF-36 atthe ICU and at 6-month follow-up, the within-patient changescore was calculated as the difference between the score at ICUand their score at 6-month follow-up. Means and SD of eachscales and change score were calculated. The significance ofthe within-patient change (from ICU to 6 months thereafter)was assessed using paired t tests. The significance of thestaffing change was assessed using independent t tests on thechange scores. For all analyses, 2-tailed P values are reportedwith the conventional cutoff of .05 used to determinenonchance findings.

3. Results

Study outline is depicted in Fig. 1.The total number of ICU admissions increased from 1582

before to 1697 after the staffing change. The baselinecharacteristics did not differ significantly between the 2

Fig. 1 Flow chart of patients and reason for exclusion from study.

421.e4 M. Reriani et al.

groups except for baseline severity of illness as measured byAPACHE III and race (Table 1). Compared with before,patients admitted after the staffing model had a higherAPACHE III median (37 vs 30, P b .001). The ICUand overall hospital mortalities were similar before and afterthe staffing change—number of deaths (percentage), 131

Table 1 Comparison of baseline characteristics before andafter the staffing model intervention

Before,n = 1584

After,n = 1697

P

Age (y), median (IQR) 66.5 (50.5-78) 66.0 (49-78) .599Male sex, n (%) 872 (55.1) 885 (52.2) .096White race, n (%) 1424 (89.9) 1463 (86.2) b.001Admission source, n (%) .790Direct admission 164 (10.4) 182 (10.7)ED 807 (50.9) 845 (49.8)Floor 381 (24.1) 412 (24.3)ICU 14 (0.9) 21 (1.2)OR 8 (0.5) 4 (0.2)Other hospital 205 (12.9) 224 (13.2)DNR order on ICUadmission (%)

20.5 19.9 .669

APACHE III score,median (IQR)

30 (17-45) 37 (24-50) b.001

ED indicates emergency department; OR, operating room.

(8.3%) vs 140 (8.2%), P = .97; and 228 (14.4%) vs 255(15%), P = .62.

3.1. Survival analysis

Long-term survival analysis continued for a total of 3279patients for median (interquartile range [IQR]) period of

0.0

0.2

0.4

0.6

0.8

1.0

Pro

port

ion

0 1 2 3 4Years

time 0 is ICU admit

after.staffing.model=0after.staffing.model=1

P = .3

Fri Aug 20 16:03:37 CDT 2010

Fig. 2 Kaplan-Meier survival before and after mandatory staffingchange baseline characteristics.

421.e5Mandatory vs. on-demand critical care specialist presence

347 (16-1003) days. Long-term survival was similar afterthe staffing change as compared with before, unadjusted—(Fig. 2); HR, 0.92; 95% confidence interval, 0.84 to 1.02. Inunivariate analysis do-not-resuscitate (DNR) status, age,nonwhite race, baseline Charlson score, APACHE score atthe time of ICU admission, and floor transfer (vs directadmission from ICU) were significantly associated withlong-term mortality. In an adjusted model for the abovecovariates, long-term survival did not differ significantlyafter the staffing model compared with before.

3.2. Quality of life

A total of 80 SF-36 QOL surveys were obtained atbaseline from patients before and 85 from patients after thestaffing model change (165/377 patients, response rate of44%). Thirty-four surveys were obtained at 6 months before(response rate of 43% of eligible patients) and 47 at 6 monthsafter (response rate of 55% of eligible patients) the staffingchange model, respectively.

Physical composite score improved significantly at 6 monthscompared with baseline in patients after the staffing model ascompared with before—Δmean (SD), 8 (14) vs 2 (11),P = .03.The individual SF-36 scales comprising the physical compo-nent score that improved significantly after the change instaffing model as compared with before were the following:role limitations due to physical problems (11 [14] vs 3 [12], P =.01) and vitality (10 [12] vs 3 [11], P = .01). There was nodifference in the mental composite score of patients after thestaffing model change as compared with before (Table 2).

4. Discussion

The present study did not show any difference in thelong-term survival of critically ill patients after a change inthe staffing model from on-demand to continuous 24-houron-site presence by a staff academic critical care specialist.

Table 2 Quality of life outcomes before and after mandatory staffing

SF-36 scale On demand

BaselineICU,n = 85

6-monthfollow-up,n = 34

Within-patiechange,n = 34

Physical functioning 37 ± 12 44 ± 12 4 ± 10Role limitations, physical 38 ± 9 44 ± 12 3 ± 12Bodily pain 41 ± 11 48 ± 10 6 ± 12General health perception 37 ± 11 42 ± 12 2 ± 11Vitality 38 ± 11 45 ± 9 3 ± 11Social functioning 36 ± 11 48 ± 11 10 ± 12Role limitations, emotional 43 ± 13 52 ± 9 10 ± 14Mental health 43 ± 13 53 ± 9 7 ± 13Mental component 43 ± 12 54 ± 8 10 ± 13Physical component 36 ± 10 40 ± 12 2 ± 11

We have previously demonstrated that such a change wasassociated with improved processes of care and staffsatisfaction and decreased ICU complication rate, hospitalLOS, and hospital costs but no change in ICU or hospitalmortality [5,10]. After the staffing change, a subset of6-month survivors had greater improvement in physicalcomponents of SF-36, but these findings need to beinterpreted with caution because of the small sample size.

Several studies have demonstrated that mandatory intensi-vist consultation or closed ICU where all care is directed by theintensivists is associated with improved outcomes, both lowerICU and hospital mortalities [1-4]. To our knowledge, there areno studies assessing the impact of intensivist staffing on long-term survival and QOL of critically ill patients of 24-hourmandatory physician staffing model in the ICU. Unlike theprevious studies comparing staffing of ICU with intensivistsand nonspecialist coverage, our study did not demonstrate anyadded benefit of having continuous 24-hour staff intensivistcoverage on either hospital mortality or long-term survivalcompared with on-demand staff intensivist coverage with in-house critical care fellows and residents. This finding has animportant implication for the planning of critical care servicesgiven the projected shortage of ICU work force [11,12]. Thewidespread adoption of this model may not be justified fromthe perspective of patient survival. Preliminary data suggestthat 24-hour intensivist staffing may be cost-effective in aquaternary ICU institution such as the Mayo Clinic [10];however, the findings from this secondary analysis of ourcohort need to be independently confirmed.

There are several possible explanations for the lack oflong-term outcome benefit with 24-hour intensivist staf-fing in our institution. First of all, daytime interventionsincluding protocols to minimize sedation and facilitateearly mobilization may be more appropriate targets forpreventing complications and long-term disability of patientswho survive to hospital discharge. We reasoned thatcontinuous presence of staff intensivists would minimizediagnostic error and facilitate timely interventions (golden

change

Mandatory 24 h

nt BaselineICU,n = 80

6-monthfollow-up,n = 47

Within-patientchange,n = 47

Comparechange scoreP value

35 ± 16 43 ± 12 7 ± 18 .3436 ± 11 47 ± 10 11 ± 14 .0141 ± 12 51 ± 9 10 ± 13 .1240 ± 10 45 ± 12 3 ± 12 .5438 ± 10 50 ± 10 10 ± 12 .0137 ± 13 50 ± 11 11 ± 17 .6344 ± 13 54 ± 8 7 ± 14 .5045 ± 12 54 ± 10 8 ± 13 .8245 ± 13 55 ± 10 9 ± 14 .7734 ± 12 42 ± 11 8 ± 14 .03

421.e6 M. Reriani et al.

hours), therefore limiting the development of multipleorgan failure and its consequences including long-termsurvival and QOL. However, we were not able to provethis hypothesis.

The intervention itself presented only a minor incremen-tal change over already-advanced staffing model withcontinuous physical presence of critical care fellows. Evenin the on-demand model, the resident and fellow physicianhad access to and facilitated staff attending to patientinteraction, albeit to a more limited scale. The incrementalvalue of such a staffing change from on-demand tocontinuous staffing on long-term survival is thus likely tobe minimal at best.

Patients admitted during the period with continuous ICUstaffing coverage may have had an earlier and morecomprehensive discussion of the prognosis of their illnessand determination of the goals of care. This potentially couldhave lead to higher mortality of patients who were sicker atbaseline, as attending staff compared with residents andfellows were more likely to address the DNR status earlierand, in consultation with the family, make the decisions tolimit medical interventions. Any survival benefits of thestaffing model in patients with lower APACHE scores atbaseline could thus have been offset by the increasedmortality in patients with higher APACHE scores. In fact, ina post hoc analysis, severity of illness as measured byAPACHE was shown to interact significantly with thestaffing model change. In a model adjusting for severity ofillness and the interaction term, the staffing model changewas associated with significantly improved survival. Insubgroup analysis of patients in quartiles of APACHE,staffing model change was associated with decreasedmortality in all quartiles except the fourth, in which therewas a higher mortality. This is further supported by the factthat the effect of the staffing model change on survival wasmore pronounced in the subgroup of patients admitted to thehospital at night (between 7 PM and 7 AM) when the nightattending physician took over care of ICU patients. Theresults of these post hoc analyses, although intriguing, areonly hypothesis raising.

Finally, it is possible that intensivist presence per se doesnot necessarily lead to improved patient outcomes andstandardized processes of care within a multidisciplinaryenvironment may be more important [13].

The finding of improved physical QOL by the changesmade in the staffing model to continuous mandatory staffingneeds to be interpreted with caution. There is the potential ofselection bias because only a small subset of patients wasincluded in evaluation and analysis of the QOL. We,however, did show that the baseline characteristics of the2 groups did not differ significantly at baseline. Becausethis is a cohort study, there is still the potential of otherfactors confounding these results. It, however, still remainsinteresting and hypothesis generating for future, moredefinite studies on QOL and mandatory continuous intensi-vist coverage of ICU.

4.1. Study limitations

There are several limitations to be mentioned. First, thestudy design is an observational cohort with all limitations ofbefore-and-after study designs. This clearly precludesstronger inference about the cause and effect of any findings.Regression to the mean and the effect of unmeasuredconfounders may have contributed to the observed QOLresults after the intervention. Although no other qualityimprovement project was implemented in our ICU during thesame period, variation in ICU practice between the 2 yearsmay have influenced the results.

Second, there were many patients who were lost tofollow-up. We tried to minimize this number by enlisting thehelp of an independent commercial company through theSRC in our institution to determine the vital status of patientslost to follow-up. Despite these measures, we were not ableto ascertain the vital status of 249 patients, which represented6.7% of the whole cohort.

Third, the QOL analysis was done in a subset of patientswho were admitted before and after the staffing change. Thepreplanned 2-month run-in period may have been too short,and the seasonal variations could have influenced theanalysis of the QOL data. Both baseline and follow-upQOL data were available, allowing for statistical analysis ofthe change in QOL (Δ), to some extent, mitigating theconcerns of different case mix in the 2 periods. Noadjustment was made for multiple comparisons, furtherlimiting the interpretation of QOL results.

Fourth, the SF-36 questionnaires addressing QOL werefilled, in part, by relatives upon ICU admission when patientswere deemed too sick to do so. Although previous studiesshowed that family members, especially proxies, can reliablyanswer QOL queries on behalf of the patient [14-16], thismay not always be the case.

Finally, although the continuous presence of attendingphysicians may have affected end-of-life care in this patientpopulation, we did not measure the quality of death anddying in this population.

5. Conclusion

In our academic center's ICU, the introduction of anadditional night shift to provide mandatory as opposed toon-demand 24-hour staff critical care specialist coveragewas not associated with improvement in long-term survivalof critically ill medical patients.

References

[1] Brown JJ, Sullivan G. Effect on ICU mortality of a full-time criticalcare specialist. Chest 1989;96(1):127-9.

421.e7Mandatory vs. on-demand critical care specialist presence

[2] Reynolds HN, Haupt MT, Thill-Baharozian MC, Carlson RW. Impactof critical care physician staffing on patients with septic shock ina university hospital medical intensive care unit. JAMA 1988;260(23):3446-50.

[3] Li TC, Phillips MC, Shaw L, Cook EF, Natanson C, Goldman L. On-site physician staffing in a community hospital intensive care unit.Impact on test and procedure use and on patient outcome. JAMA1984;252(15):2023-7.

[4] Pronovost PJ, Angus DC, Dorman T, Robinson KA, Dremsizov TT,Young TL. Physician staffing patterns and clinical outcomes in criti-cally ill patients: a systematic review. JAMA 2002;288(17):2151-62.

[5] Gajic O, Afessa B, Hanson AC, Krpata T, Yilmaz M, Mohamed SF,et al. Effect of 24-hour mandatory versus on-demand critical carespecialist presence on quality of care and family and providersatisfaction in the intensive care unit of a teaching hospital. CritCare Med 2008;36(1):36-44.

[6] Angus DC, Carlet J. Surviving intensive care: a report from the 2002Brussels Roundtable. Intensive Care Med 2003;29(3):368-77.

[7] Dowdy DW, Eid MP, Sedrakyan A, Mendez-Tellez PA, Pronovost PJ,Herridge MS, et al. Quality of life in adult survivors of critical illness:a systematic review of the literature. Intensive Care Med 2005;31(5):611-20.

[8] Knaus WA, Wagner DP, Draper EA, Zimmerman JE, Bergner M,Bastos PG, Sirio CA, et al. The APACHE III prognostic system. Riskprediction of hospital mortality for critically ill hospitalized adults.Chest 1991;100(6):1619-36.

[9] Knaus WA, Wagner DP, Zimmerman JE, Draper EA. Variations inmortality and lengthof stay in intensive care units.Ann InternMed1993;118(10):753-61.

[10] Banerjee R, Naessens JM, Seferian EG, Gajic O, Moriarty JP, JohnsonMG, et al. Economic implications of nighttime attending intensivistcoverage in a medical intensive care unit. Crit Care Med 2011;39(6):1257-62.

[11] Fink MP, Suter PM. The future of our specialty: critical care medi-cine a decade from now. Crit Care Med 2006;34(6):1811-6.

[12] Kelley MA, Angus D, Chalfin DB, Crandall ED, Ingbar D, JohansonW, et al. The critical care crisis in the United States: a report fromthe profession. Chest 2004;125(4):1514-7.

[13] Levy MM, Rapoport J, Lemeshow S, Chalfin DB, Phillips G, Danis M.Association between critical care physician management and patientmortality in the intensive care unit. Ann Intern Med 2008;148(11):801-9.

[14] Hofhuis J, Hautvast JL, Schrijvers AJ, Bakker J. Quality of life onadmission to the intensive care: can we query the relatives? IntensiveCare Med 2003;29(6):974-9.

[15] Capuzzo M, Grasselli C, Carrer S, Gritti G, Alvisi R. Qualityof life before intensive care admission: agreement betweenpatient and relative assessment. Intensive Care Med 2000;26(9):1288-95.

[16] Rogers J, Ridley S, Chrispin P, Scotton H, Lloyd D. Reliability of thenext of kins' estimates of critically ill patients' quality of life.Anaesthesia 1997;52(12):1137-43.