Journal of Critical Care (2012) 27, 694–701

Predictors of mortality of mechanically ventilated patientsin internal medicine wards☆

Moshe Hersch MD, MSc a,⁎,1, Gabriel Izbicki MDb,1, David Dahan MDb,1,Gabriel S. Breuer MDb, Gideon Nesher MDb, Sharon Einav MDa

aIntensive Care Unit, Shaare Zedek Medical Center (affiliated with the Hebrew University Medical School-HadassahJerusalem), Jerusalem 91031, IsraelbDepartment of Internal Medicine, Shaare Zedek Medical Center (affiliated with the Hebrew University MedicalSchool-Hadassah Jerusalem), Jerusalem 91031, Israel

6

0h

Keywords:Mechanical ventilation;Internal medicine ward;Outcome;Intensive care unit;Survival

AbstractPurpose: Budget restrictions have led to shortage of intensive care unit (ICU) beds in several countries.Consequently, ventilated patients are often kept on the wards. This study examined survival likelihoodamong patients ventilated on thewards and the predictive value of commonly used severity-of-illness scores.Methods: This study is a prospective observation and characterization of consecutive, mechanicallyventilated patients in 3 internal medicine wards of a single hospital who were denied ICU admission.Outcome measures are as follows: 28-day mortality, survival to hospital discharge, and 3 monthspostdischarge.Results: Eighty-six patients were examined. The patients were 78.9 ± 8.9 years old; 53% were independentpreadmission. Respiratory insufficiency due to infection was the main reason for mechanical ventilation(58%). Charlson and acute physiology scores (APS) averaged 4 ± 2.2 and 91.8 ± 26.7, respectively. Twenty-eight-day mortality was 71%, whereas in-hospital mortality was 74% and 3 months postdischarge mortalitywas 79%. Survivors were significantly younger than nonsurvivors (74.4 ± 8.5 years vs 80.4 ± 8.6 years,P b .01), were more likely to be ventilated for cardiac causes (41% vs 11%, P = .04), and hadsignificantly higher initial mean blood pressure (79.4 mm Hg vs 58.2 mm Hg, P = .02) and bloodalbumin levels (29.8 g/L vs 25.7 g/L, P = .05). Death rate was 10 times more likely, with an APS greaterthan 90 on the day of intubation as compared with an APS less than 90.Conclusion: Mortality in patients ventilated on the ward was high, especially in the subgroup of patientswith an APS score greater than 90. The early calculation of APS may assist in focusing therapeuticefforts on patients with better survival chances.© 2012 Elsevier Inc. All rights reserved.

☆ Conflicts of interest: The authors declare that they do not have any conflict of interest in relation to the subject of this manuscript.⁎ Corresponding author. Shaare Zedek Medical Center, Intensive Care Unit, POB 3235, Jerusalem 91031, Israel. Tel.: +972 2 6555564; fax: +972 2

555144.E-mail address: hersch@szmc.org.il (M. Hersch).1 The 3 authors have contributed equally to this manuscript and share the first authorship.

883-9441/$ – see front matter © 2012 Elsevier Inc. All rights reserved.ttp://dx.doi.org/10.1016/j.jcrc.2012.08.020

695Mechanically ventilated patients in internal medicine wards

1. Introduction

Changes in health care financing designed to curb medicalexpenditure are becoming increasingly prevalent. This andother factors have led to a relative lack of intensive care unit(ICU) beds in several European countries [1] and an evolvingcrisis in the US critical care services [2-4]. Like in manyother countries, Israel has been suffering from a shortage ofICU beds for the last decade. Ward physicians in de-partments of internal medicine and geriatric medicine in mostof the hospitals ventilate patients requiring intubation andventilation as medically/acutely indicated, independent ofwhether the patient is admitted to an ICU or not. This leads toa situation where ventilated patients are often kept on theward because the number of ventilated patients exceeds byfar the number of available ICU beds [5]. In Israel, it isestimated that most ventilated patients are kept on the wardin departments of internal medicine and geriatric medicine[6,7]. This situation, in varying rates, is prevalent in most ofthe hospitals in Israel, thus transferring ventilated patients toanother hospital's ICU is not feasible in most cases.

Mostly as of lack of place and/or being deemed to lesslikely benefit from intensive care (usually because of severityof disease and very high predicted mortality) than others, it isoften the patients who are perceived to be the most severelyill that are not admitted to ICU. This situation is particularlyrelevant to nonsurgical elderly patients with multiple chronicillnesses [5]. Decisions whether to refer an intubated wardpatient to ICU are not anchored upon clear criteria [7-10]; nomodels for prediction of survival have been developed forcritically ill ventilated patients who remain on the ward, andprediction models that exist for ICU patients requirevalidation in this population before clinical implementation.

The current study describes the course and outcome ofmechanically ventilated patients in medical wards, which is,in a main part, predicated on scarcity of resources. It was notour intention to compare the course and outcome ofmechanically ventilated patients in medical wards with thecourse and outcome in ICU because this was previously done[6,7]. Rather, we aimed to explore some parameters that willprospectively identify those patients ventilated on the wardwho have a better chance than others to survive. The value ofthe severity of acute illness (using the acute physiology score[APS], part of the Acute Physiology and Chronic HealthEvaluation [APACHE] III) [11], the burden of chronic illness(using the Charlson Index) [12,13], and the functional status(using the Lawton scale for activities of daily living [ADLs])[14,15] for predicting survival were used for this purpose. Bydoing so, we aimed to provide to the internal medicinephysicians a practical key that will assist them to focus theirefforts on patients with a better prognosis. The aim of anyseverity of illness scoring (beyond research purposes) is tohelp the treating physician to construct a more realisticprognostication. Because it was obvious that our studypopulation will consist mainly of elderly patients with poor

prognosis, in the face of our limited resources and difficultyto provide them on the wards (monitoring, blood sampling,vasopressors, imaging, intrahospital transport, surgical in-terventions, dialysis, cardiopulmonary resuscitation, etc),any piece of information about prognosis will help intherapeutic decisions and in resource allocation.

We preferred this score over the traditional completeAPACHE II score because we were looking for a “new,” notICU-related score. We wanted this score to reflect morewidely our study population's acute physiological changesand also the multiple comorbidities and functional limita-tions that this aged population most probably will have. TheAPS component of the APACHE III contains a largernumber of acute physiological parameters than the APACHEII (20 vs 12) and is not affected by the patient beingpostoperative elective/emergency (which is not relevant toour study population, but affects APACHE II), and theCharlson Index contains more chronic illnesses than theAPACHE II (19 detailed conditions vs 5 general ones). Theindex of ADL is a well-recognized reflector of biological andpsychosocial function translated into survival rate [14,15].

2. Methods

2.1. Clinical setting

The Shaare ZedekMedical Center is a 550-bed university-affiliated acute care hospital. It has 3 critical care units (22beds: 6 general [mixed medical and surgical], 6 cardiac, and10 post–cardiothoracic surgery), 2 departments of internalmedicine (65 beds), and 1 department of acute care geriatricmedicine (60 beds). In-hospital care in these wards is similarto that throughout the country; nurse-to-patient ratios are 1:8(morning shift) to 1:15 (night shift), whereas this ratio in theICU is 1:2 at all shifts. Residents and specialists in internalmedicine or geriatric medicine who are fully employed by thehospital are responsible for daily in-hospital patient care.Each team (1 senior physician [a specialist in internalmedicine and/or geriatric medicine] and 2-3 residentphysicians) is responsible for the care of 15 to 20 patients.On average, each team takes care of 1 to 3 mechanicallyventilated patients in each ward. Their expertise to managemechanically ventilated patients is at-the-best limited to a3-months rotation in the ICU. This is (sadly) a verycommon practice. Neither ethical committee of any hospitalnor the Ministry of Health was at any time involved in aformal decision to approve mechanical ventilation in thewards. Intensive care unit ICU beds are scarce in almostevery hospital, and for this reason, there is no alternativebut to intubate, ventilate, and treat patients outside the ICU.This is just a reality of the Israeli daily in-hospital routine.

Intensivists and pulmonologists are involved in thesepatients' therapy at the discretion of the treating team but noton a regular basis. In the medical wards, all the required

696 M. Hersch et al.

facilities (as UPS and O2 reserves) are present. Daily follow-up by respiratory therapists (who are not licensed to changeventilator settings in this country) includes documentation ofventilation parameters and proper functioning of theventilators, but nothing concerning the endotracheal tubessuch as proper depth, tube cuff pressure, and so on. Theyprovide consultation to ward physicians about the ventilatorsettings and alarms. Suction is performed by ward nurses inaccordance with the patient's needs. Physiotherapy isprovided on demand. Monitoring for ventilated patientsincludes vital signs and pulse oximetry once to twice per8-hour shift. Closer monitoring is applied if necessary, suchas automatic, frequent noninvasive blood pressure measure-ments, and continuous pulse oximetry/heart rate display,done by a mobile bedside monitoring device. Blood samplingis performed at least once daily and includes mostly arterialblood gas analysis. More sampling is performed whennecessary. Mobile radiology services are available at thedemand of the treating team. Most medical therapiesincluding vasopressors are provided when necessary, butsome other such as nitric oxide inhalation, insertion of arterialline, and prone position are never provided on the internalmedicine ward. Hemodialysis is available when required butrequires mobilization of the patient to the dialysis unit.Tracheostomy is performed by ear-nose-and-throat physi-cians usually after at least 2 weeks of intubation.

2.2. Referral to intensive care

The decision to refer an acutely ventilated patient to theICU is not affected by patient/family preference, unless a do-not-attempt-resuscitation (DNAR) decision is made accord-ing to patients'/families' will, in which case this decisiondefers ICU admission. The possible admission process toICU is 2 staged: selection of the patient by the ward medicalteam for referral to the ICU consultant and the choicemade bythis consultant regarding admission to the ICU (ICU triage).Selection for referral to the ICU consultant is performed bythe senior physician of the medical ward team. This physiciangathered the information about patients'/families' will(including DNAR status), cognitive function, and cancerstatus before making the decision where to locate the patient.However, usually, this information was not collected beforeintubation because of lack of opportunity/time to do so andalso because, practically, it had no affect on the decision tointubate the patient because of religious-cultural factors.

2.3. Inclusion/exclusion criteria

The study was prospective and observational. Afterinstitutional review board approval and during a period of6 months (January 1, 2008–June 30, 2008), all patients olderthan 18 years who required acute invasive (via endotrachealtube) mechanical ventilation for life support purposesadmitted in the 2 departments of internal medicine or the

department of acute care geriatric medicine at the ShaareZedek Medical Center were screened for inclusion. Includedwere patients who underwent tracheal intubation en route tothe medical center by the ambulance team, in the emergencydepartment (ED) by an ED senior physician, or duringadmission on the ward by a senior ward physician and whowere not accepted to the ICU. Usually, the ICU specialistwas not involved in the decision to intubate a patient neitherin the ED nor and on the wards. These patients were notadmitted to the ICU mostly because the senior physicianinvolved in the process of intubation did not refer them toICU triage at all. This happened because the ICU was fullmost of the time (the ICU occupancy information was readilyavailable online) and/or the senior physician considered thepatient to have a grave prognosis with probable futility ofICU therapy. Patients who occupied an ICU bed at any timeduring the same admission, patients who were chronicallyventilated before hospital admission or study initiation, orpatients with a tracheotomy before admission and those witha documented DNAR order were excluded from the presentstudy. Patients were recruited either in the first 24 hours fromintubation or from the time of their arrival to the hospital (forpatients intubated en route), whereas it was clear that they arenot admitted to the ICU.

2.4. Data collection

The following data were collected upon inclusion:demographic details (age, sex, and place and facility ofresidence), cause of deterioration, Lawton scale for ADLsbefore admission, and detailed Charlson score. The worstvalues of data included in the acute physiology scorecomponent of the APACHE III (including the neurologicstatus of the patient) [9,16] were recorded at inclusion (first 24hours) and 72 hours later.

The primary outcome measure was survival to hospitaldischarge, although follow-up included the recording of28-day mortality. Secondary outcome measures weresurvival 3 months after discharge from hospital, whenrelevant postdischarge data were gathered by telephonecommunication, and the association between survival andthe APS, Charlson, and ADL scores.

No changes were made in patients' treatment, for thepurposes of the study.

2.5. Statistical analysis

In the first step, descriptive statistics were used to examineall independent variables and patient outcome (the dependentvariable). Categorical variables (eg, mortality, sex) arepresented in percentages. Numerical variables (eg, APS,Charlson, and ADL scores) were examined for their distribu-tion and are expressed as means, SDs, medians, and ranges. Inthe second step (bivariate analysis), the possible differences inthe independent variables between patients who survived to

Table 1 The demographic characteristics of the studypopulation

Characteristic Patients(n = 86)

n %

Male sex 45 52Residence Home a 60 70

Nursing home 26 30ICD-9 diagnosisleading to intubation b

Sepsis—pulmonary 30 35Sepsis—nonpulmonary 20 23COPD c 9 10Cardiac d 20 23Neurologic e 6 7

ADL score—completelyindependent before admission

46 53

ICD-9 indicates International Classification of Diseases, Ninth Edition;

697Mechanically ventilated patients in internal medicine wards

hospital discharge and those who did not were examined. Theresults and their significance (P value) are presented. The χ2

test or Fisher exact test was used for categorical variables, andStudent t test was used for numerical variables.

Each score (APS, Charlson, ADL) comprises severalvariables; therefore, the final scores achieved by individualpatients are all numerical. To establish a cutoff value for eachscore that is meaningful for survival, the distribution of thesevalues was also examined in survivors and nonsurvivors.Univariate logistic regression was then used to evaluate therelationship of the cutoff value to the dependent variable(survival to hospital discharge). Results of the logisticregression are presented as odds ratios (ORs) with associated95% confidence intervals (CI) and P values. Statisticalanalyses were performed using SPSS version 12 software(SPSS Inc, Chicago, Ill).

COPD, chronic obstructive pulmonary disease.a Including patients with home care.b Data missing for 1 patient.c Chronic obstructive pulmonary disease and asthma.d Congestive heart failure, myocardial infarction, and arrhythmia.e Subarachnoid hemorrhage, ischemic stroke, and hemorrhagic

stroke.

3. Results

3.1. Study population

During the study period, 90 patients were ventilated in the3 wards and screened for inclusion, whereas 155 patients(medical and surgical, ventilated or not) were admitted to theICU. Four patients were excluded due to prior chronicventilation status, leaving 86 patients eligible for inclusion.These patients were distributed among the 3 medical wardsas follows: internal medicine A, 35 (of 588 total admissions);internal medicine B, 29 (of 650 total admissions); andgeriatric departments, 22 (of 770 total admissions). Mostpatients were admitted from home, and the rest wereresidents of chronic care facilities (Table 1). In 62 patients,ventilation was initiated on the wards, whereas in 24, it wasin the ED or by a paramedic out-of-hospital ventilation. Themean age was 78.9 ± 8.9 years (range, 55-95 years; median,79.5 years) (Table 2). Most patients were older than 75 years(72%; n = 62). Forty-six patients (53%) were functionallyindependent before hospital admission (Table 1). Theaverage Charlson score of the patient's population was 4 ±2.2 (range, 0-11; median, 4), and the average APS was 91.8 ±26.7 (range, 30-169; median, 90) (Table 3).

3.2. Reasons for mechanical ventilation

The primary International Classification of Diseases,Ninth Edition diagnoses that triggered acute deterioration tomechanical ventilation were infectious diseases (A00-B99,58% [n = 50]) and diseases of the circulatory system (I00-I99, 23% [n = 20]) (Table 1).

3.3. Patient outcomes

Mortality. The 28-day mortality was 71%, whereas the in-hospital mortality rate was 74% (n = 64), with 22 patients

surviving to hospital discharge. Four additional patients diedwithin 3 months of hospital discharge, yielding an overallmortality rate of 79% (68/86) for the follow-up period. Morethan half of the deaths occurred within 3 days of intubation(36/64; 56%). The in-hospital mortality rates for patients 70years or younger, 71 to 80 years old, and older than 80 yearswere 50%, 68%, and 90%, respectively, and were signifi-cantly higher for those older than 70 years as compared withthose 70 years or younger (50% vs 80%, P = .02). None ofthe 12 nonagenarians who were included in the studysurvived to hospital discharge.

3.4. Mechanical ventilation

The patients were ventilated for an average of 9 ± 18 days(range, 1-90 days; median, 3 days). Forty-four patients(51%) continued to require mechanical ventilation more than3 days postintubation, and 26 patients (30%) requiredventilation for 6 days or more. Extubation was feasible in11 (50%) of the survivors within 6 days.

3.5. Functional status at hospital discharge

Among the 22 patients who survived to hospitaldischarge, only 41% (n = 9) returned home. The rest,59% (n = 13), were discharged to chronic care facilities.However, only 3 of these 13 patients had initially arrivedfrom home. Fifteen of the 18 patients who survived to 3months after hospital discharge returned to their preadmis-sion functional state.

Table2

Com

ponentsof

theAPSthat

werefoun

dto

besign

ificantly

differentbetweensurvivorsandno

nsurvivo

rsto

hospitaldischargein

bivariateanalysis

Variable

Total

popu

latio

nDied

Survived

P95

%CI

nMean(SD)

Range

Median

nMean(SD)

Median

nMean(SD)

Median

Age

(y)

8678

.9(8.9)

55-95

79.5

6480

.4(8.6)

81.5

2274

.4(8.5)

76b.01

1.78

to10

.22

24hafterinclusion

Meanarterial

bloo

dpressure

(mm

Hg)

(reference

rang

e,80

-123

mm

Hg)

8663

.6(37.4)

0-15

664

6458

.2(37.9)

6322

79.4

(31.6)

70.5

.02

−39

to−3.2

Album

in(g/L)(reference

rang

e,35

-55g/L)

7026

.7(7.3)

4-44

26.5

5425

.7(7.3)

26.5

1629

.8(6.6)

27.5

.05

−0.8to

0.00

48Bilirubin(μmol/L)(reference

rang

e,5-17

μmol/L)

7914

.2(17.1)

1.7-12

6.5

8.5

6015

.9(19.9)

10.3

199.2(5.3)

8.5

.02

−0.72

to0.06

72hafterinclusion

Meanarterial

bloo

dpressure

(mm

Hg)

4278

.1(18.8)

40-109

75.5

2873

.2(18.5)

70.5

1488

(15.9)

88.5

.01

−3to

−26

.5Creatinine(μmol/L)(reference

rang

eb13

3μm

ol/L)

4217

2(133

)41

-774

142

2820

3.3(141

)17

514

106(80)

80.02

0.15

-2

Urine

output

(mL/24h

)42

1389

(115

5)0-42

0011

0028

1066

(105

2)65

014

2035

(111

2)19

50b.01

−16

78to

−26

0

698 M. Hersch et al.

3.6. Comparison between survivors andnonsurvivors to hospital discharge

Survivors did not differ from nonsurvivors in either male-to-female ratio or type of residence before admission.However, survivors were significantly younger (74.4 ± 8.5years) than nonsurvivors (80.4 ± 8.6 years; P b .01; Table 2),were less likely to have undergone intubation due to sepsis(41% vs 64%, respectively; P = .045), and more likely tohave undergone intubation for cardiac causes (41% vs 11%,respectively; P = .04). Survivors had significantly higherinitial mean blood pressures and blood albumin levels andlower bilirubin levels (Table 2). Shock (mean arterialpressure of ≤64 mm Hg) was evident in only 32% (7/22)of the survivors, whereas it was evident in 58% (37/64) ofthe nonsurvivors.

No differencewas found between the groupswithin the first24 hours postinclusion in terms of pulse rate (P = .54),temperature (P = .81), respiratory rate (P = .25), blood pH (P =.26), hematocrit (P = .40), blood white cell count (P = .47),serum creatinine (P = .22), sodium (P = .86), and glucose (P =.71) or in urine output (P = .11).

Seventy-two hours after inclusion, nonsurvivors hadsignificantly lower blood pressures and urine outputs andhigher creatinine levels than survivors (Table 2), but nosignificant differences were found between the groups inother components of the APS.

3.7. Relationships between patient survival and scores

Nonsurvivors had significantly higher APSs both 24 and72 hours after intubation (Table 3). Univariate logisticregression analysis demonstrated that when the initial (withinthe first 24 hours) APS was greater than 80, the probability ofdeath was 5 times higher, and when it was greater than 90,the probability of death was 10 times higher (Table 4). Otherscores were less predictive of mortality.

4. Discussion

The current study demonstrates that only about a quarterof patients who undergo intubation but remain in internalmedicine or geriatric medicine wards survive to hospitaldischarge. Nevertheless, despite this very high mortality rate,a relatively high percentage of those who survive, namely,68% (15/22), returned to preadmission functional state 3months after initiation of ventilation. We showed thatsurvivors to hospital discharge were younger than non-survivors, although they were still very old (74.4 ± 8.5 years)compared with the mean age of all patients admitted to theICU during the study period (60.2 ± 21.8 years). Moreover,none of the patients older than 90 years survived. Higherblood pressures in the 24 hours postintubation and higher

Table 3 Bivariate analysis of the differences in the various scores between survivors and nonsurvivors to hospital discharge (24 and 72hours after admission)

Score Total population Died Survived P 95% CI

n Mean (SD) Range Median n Mean (SD) Median n Mean (SD) Median

24 h Charlson 86 4 (2.2) 0-11 4 64 4.1 (2.3) 4 22 3.7 (2) 3 .46 −0.7 to 1.5ADL 86 3.7 (2.7) 0-6 6 64 3.7 (2.7) 6 22 3.6 (2.9) 6 .92 −1.4 to 1.3APS 86 91.8 (26.7) 30-169 90 64 98.6 (25.2) 98 22 71.8 (20.9) 70 b.001 14.9-38.7

72 h APS 42 57.1 (20.9) 16-104 55.5 28 65.8 (17.6) 61.5 14 39.7 (15.6) 39 b.001 14.9-37.3

699Mechanically ventilated patients in internal medicine wards

blood albumin were associated with survival to hospitaldischarge. Shock in the 24 hours postintubation (defined asmean arterial pressure of ≤64 mm Hg) consisted an ominousprognostic sign because it was much more prevalent innonsurvivors than in survivors (58% vs 32%). Theprobability of death increased with higher APS scores.

Reports of mortality in acutely ill patients requiringmechanical ventilation range from 28% to 69% [17-19]. Thefindings of this study support those of our prior report [6] thatfound an overall 80% mortality rate among ventilatedpatients in internal medicine wards, whereas ICU mortalityof nonsurgical ventilated patients was 62%. Sprung et al [20]demonstrated lower in-hospital mortality in ICU-admittedpatients (14%) compared with patients who were refusedICU admission (46%). Simchen et al [5] associated ICUadmission with a protective effect when compared withtreatment in the ward in the initial 3 days after deterioration.Joynet et al [21] described significantly higher standardizedmortality ratios for patients who were refused admission toan ICU when compared with those who were admitted (1.24vs 0.93, respectively). This finding has recently beensupported in a review showing increased hospital mortalityrates in patients who were refused admission to an ICUcompared with those admitted (OR, 3.04; 95% CI, 1.49-6.17) [22].

The current study examined a unique group of patients—those who, despite receiving mechanical ventilation,were left on the ward and were not admitted to the ICU.It is also the first to examine the prognostic value of theAPS, Charlson score, and ADL in such patients. With thebackground of shortage of ICU beds, the burden ofchronic illness and the baseline functional status of these

Table 4 Univariate logistic regression analysis of therelationship between mortality and the cutoff values for thevarious scores at 24 hours after inclusion

Score Range of values Likelihood of death

OR P 95% CI

Charlson ≤5 vs N5 2.3 .22 0.9-8.7ADL Completely dependent

vs others1.1 .85 0.4-3.1

APS ≤90 to N90 10.5 b.001 2.8-39.5≤80 to N80 5.5 .001 1.9-15.6

patients are both prominent factors in the decision not toadmit to an ICU [5]. This is the first attempt to use these2 variables, along with the physiological score in a non-ICU population (a value that is known to correlate withsurvival of critically ill ICU-admitted patients), usingscores that have been associated with patient survival inlarge cohorts.

The variables that compile the APS are well known to beassociated with mortality in ICU patients [11]; a directrelationship exists between higher APS and increasedmorbidity and probability of death [23,24]. In this study,APS above certain values predicted substantially increasedmortality rates in this non-ICU group of ventilated patients.Conversely, a high Charlson index or a poor ADL score wasnot associated with increased mortality, in contrast withprevious studies showing that both the burden of comorbid-ity [25-27] and the degree of functional independence [14]predict survival in critically ill, although mainly ICUventilated patients. This inconsistency may stem from thesample size. Alternatively, the presence of multiple comor-bidities may have been the cause for nonreferral to ICU apriori; selection bias would diminish the likelihood offinding meaningful differences in these scores betweensurvivors and nonsurvivors. Finally, the burden of acuteillness in this patient cohort may be such that the impact ofany other factor on survival is significantly less important.

In Israel (and perhaps in other countries), because ofcultural/religious/legal concerns, the choice not to intubate apatient with poor prognosis is not given to the medical team.Moreover, the “Terminally Ill law” [28], which waslegislated in the Israeli parliament in 2005, makeswithdrawal of a ventilator illegal. Unfortunately, this resultin the awkward situation where many patients are ventilatedin the internal/geriatric medicine wards, although as shownin this study, it is a nonbeneficial intervention because mostpatients are dying within 3 days of ventilation. However,because of the reasons stated earlier, the physicians areobliged to give them the most appropriate therapy in theavailable circumstances. The only practical way that we cansuggest, not to prevent but to decrease the harm of thisnonbeneficial practice, is to identify those patients ventilatedon the ward who have the best prognosis vs those with theworst and to try to focus the allocation of therapeuticresources in the first group. Based on the data in this study,it is possible to estimate which of the onward mechanically

700 M. Hersch et al.

ventilated patients will have the best prognosis, thus helpingthe medical team to focus their limited therapeutic resourcesin the ward on these patients. Moreover, the results of thelargest study to date on admitted and refused patients tothe ICU [29], which showed a greater mortality benefit forthe elderly than for the younger admitted to the ICU, giveour results an additional value, enabling us to identifythose elderly patients ventilated on the ward who will havethe best chances to benefit from our therapy including ademand for ICU admission.

It seems likely, as our study has showed, that slightlyyounger patients who have undergone intubation for cardiaccauses (and not for sepsis) and have normal albumin levels,normal blood pressure, and low APS after intubation willhave favorable survival rates in internal medicine or geriatricmedicine wards. These patients should be treated asaggressively as we can even on the wards and demand anICU bed for them as soon as possible. Conversely, olderpatients with very high APS (APS N90) have the worstprognosis, and we suggest to carefully consider lessaggressive therapy to this group. Nevertheless, it should beemphasized the undoubted limitation of any severity-of-illness scoring in individual prognostication. What wesuggest is to carefully consider the initiation of complexand expensive therapies in this group, especially when theymight deprive other patients with better chances to survivefrom these therapies because of resource and capabilitylimitations on the wards. We and others [6,7] showed informer studies that survival is generally higher in ICUintubated patients as compared with internal medicine wards.The present study was not controlled for comparison withICU patients but rather compared survivors with nonsurvi-vors among ward patients. Controlling for ICU admissionswould have remained limited by patient selection; random-ization would not be ethically acceptable.

There are several limitations to this study. First, it washospital single-center study, albeit in 3 wards. Validation ofits findings with larger sample sizes elsewhere is requiredto substantiate the prognostic factors associated with highersurvival rate found in this study. The described practice oftreating mechanically ventilated patients in medical wardsmay still be incomprehensible to physicians in othernations, yet some will likely experience their owninsufficiency of resources in the not-too-distant future.Second, the cutoff points associated with poor survivalwere defined post hoc, a method which is associated withhigh likelihoods of type I errors. Prospective validation inlarger cohorts is therefore necessary. The third limitingfactor of this study is that potential confounders includedvariability in actual ICU bed availability (eg, seasonal andweekly variation) and the effect of clustering on triagedecision making. In the absence of formal adjudicated ICUreferral/triage rules, there may be variability in practice.The analysis would be strengthened by a hierarchical modelthat includes physician effect, but such modeling is beyondthe scope of this study.

5. Conclusions

In conclusion, paucity of ICU beds is increasinglyprevalent [5,6,30], a situation that will probably not beamended in the foreseeable future and might result in anincreasing number of patients ventilated outside the ICU.These patients have remarkably high mortality rates. Ourscoring system, which appropriately weighs the likelihood ofsurvival of mechanically ventilated patients in internalmedicine wards, may be an important tool in the guidanceof the management for these patients.

References

[1] Vincent JL. Forgoing life support in western European intensive careunits: the results of an ethical questionnaire. Crit Care Med 1999;27:1686-7.

[2] Angus DC, Kelley MA, Schmitz RJ, White A, Popovich Jr J. Caringfor the critically ill patient. Current and projected workforcerequirements for care of the critically ill and patients with pulmonarydisease: can we meet the requirements of an aging population?Committee on Manpower for Pulmonary and Critical Care Societies(COMPACCS). JAMA 2001;285:1016-7.

[3] Kelley MA, Angus D, Chalfin DB, Crandall ED, lngbar D, JohansonW, et al. The critical care crisis in the United States: a report from theprofession. Chest 2004;125:1514-7.

[4] Council on Graduate Medical Education. Physician workforce policyguidelines for the United States, 2000–2020. Washington, DC: HRSA,DHHS; 2003 [Report No. 16].

[5] Simchen E, Sprung CL, Galai N, Zitser-Gurevich Y, Bar-Lavi Y,Gurman G, et al. Survival of critically ill patients hospitalized in andout of intensive care units under paucity of intensive care unit beds.Crit Care Med 2004;32:1654-61.

[6] Hersch M, Sonnenblick M, Karlic A, Einav S, Sprung CL, Izbicki G.Mechanical ventilation of patients hospitalized in medical wardsversus the intensive care unit—an observational comparative study. JCrit Care 2007;22:13-7.

[7] Liebermann D, Nachson L, Miloslavsky O, Dvorkin V, Shimoni A,Zelinger J, et al. Elderly patients undergoing mechanical ventilation inand out of intensive care units: a comparative, prospective study of 579ventilations. Crit Care 2010;14:176-84.

[8] Strauss MJ, LoGerfo JP, Yeltatzie JA, Temkin N, Hudson LD.Rationing of intensive care unit services. An everyday occurrence.JAMA 1986;255:1143-6.

[9] Singer DE, Carr PL, Mulley AG, Thibault GE. Rationing intensivecare-physician responses to a resource shortage. N Engl J Med1983;309:1155-60.

[10] Franklin C, Rackow EC, Mamdani B, Burke G, Weil MH. Triageconsiderations in medical intensive care. Arch Intern Med 1990;150:1455-9.

[11] Knaus WA, Wagner DP, Draper EA, Zimmerman JE, Bergner M,Bastos PG, et al. The APACHE III prognostic system. Risk predictionof hospital mortality for critically ill hospitalized adults. Chest1991;100:1619-36.

[12] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method ofclassifying prognostic comorbidity in longitudinal studies: develop-ment and validation. J Chronic Dis 1987;40:373-83.

[13] Poses RM, McClish DK, Smith WR, Bekes C, Scott WE. Prediction ofsurvival of critically ill patients by admission comorbidity. J ClinEpidemiol 1996;49:743-7.

[14] Stauffer JL, Fayter NA, Graves B, Cromb M, Lynch JC, Goebel P.Survival following mechanical ventilation for acute respiratory failurein adult men. Chest 1993;104:1222-9.

701Mechanically ventilated patients in internal medicine wards

[15] Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies ofillness in the aged. The index of ADL: a standardized measure ofbiological and psychosocial function. JAMA 1963;185:914-9.

[16] Zimmerman JE, Wagner DP, Draper EA, Wright L, Alzola C, KnausWA. Evaluation of acute physiology and chronic health evaluation IIIpredictions of hospital mortality in an independent database. Crit CareMed 1998;26:1317-26.

[17] Knaus WA. Prognosis with mechanical ventilation: the influence ofdisease, severity of disease, age, and chronic health status on survivalfrom an acute illness. Am Rev Respir Dis 1989;140:S8-S13.

[18] Castella X, Artigas A, Bion J, Kari A. A comparison of severity ofillness scoring systems for intensivecare unit patients: results of amulticenter, multinational study. The European/North AmericanSeverity Study Group. Crit Care Med 1995;23:1327-35.

[19] Esteban A, Anzueto A, Frutos F, Alia I, Brochard L, Stewart TE, et al.Mechanical Ventilation International Study Group. Characteristics andoutcomes in adult patients receiving mechanical ventilation: a 28-dayinternational study. JAMA 2002;287:345-55.

[20] Sprung CL, Geber D, Eidelman LA, Baras M, Pizov R, Nimrod A,et al. Evaluation of triage decisions for intensive care admission. CritCare Med 1999;27:1073-9.

[21] Joynet GM, Gomersall CD, Tan P, Lee A, Cheng CA, Wong EL.Prospective evaluation of patients refused admission to an intensivecare unit: triage, futility. Intensive Care Med 2001;27:1459-65.

[22] Sinuff T, Kahnamoui K, Cook DJ, Luce JM, Levy MM. Values Ethicsand Rationing in Critical Care Task Force. Rationing critical care beds:a systematic review. Crit Care Med 2004;32(7):1588-97.

[23] Somogyi-Zalud E, Zhong Z, Hamel MB, Lynn J. The use of life-sustaining treatments in hospitalized persons aged 80 and older. J AmGeriatr Soc 2002;50:930-4.

[24] Seneff MG, Zimmerman JE, Knaus WA, Wagner DP. Predicting theduration of mechanical ventilation: the importance of disease andpatient characteristics. Chest 1996;110:469-79.

[25] Quality of Life After Mechanized Ventilation in the Elderly StudyInvestigators. 2-Month mortality and functional status of critically illadult patients receiving prolonged mechanical ventilation. Chest2002;121:549-58.

[26] Needham DM, Bronskill SE, Sibbald WJ, Pronovost PJ, Laupacis A.Mechanical ventilation in Ontario, 1992-2000: incidence, survival, andhospital bed utilization of noncardiac surgery adult patients. Crit CareMed 2004;32:1504-9.

[27] Inouye SK, Peduzzi PN, Robison JT, Hughes JS, Horwitz RI, ConcatoJ. Importance of functional measures in predicting mortality amongolder hospitalized patients. JAMA 1998;15(279):1187-93.

[28] Steinberg A, Sprung CL. The dying patient: new Israeli legislation.Intensive Care Med 2006;32:1234-47.

[29] Sprung CL, Artigas A, Kesecioglu J, Pezzi A, Wiis J, PirracchioR, et al. The Eldicus prospective, observational study of triagedecision making in European intensive care units. Part II:intensive care benefit for the elderly. Crit Care Med 2012;40(1):132-8.

[30] Lyons RA, Wareham K, Hutchings HA, Major E, Ferguson B.Population requirement for adult critical-care beds: a prospectivequantitative and qualitative study. Lancet 2000;355:595-8.