JAGS 51:529–533, 2003© 2003 by the American Geriatrics Society 0002-8614/03/$15.00

Predictive Factors of In-Hospital Mortality in Older Patients Admitted to a Medical Intensive Care Unit

Mario Bo, MD, Massimiliano Massaia, MD, Silvio Raspo, MD, Francesca Bosco, MD,Paola Cena, MD, Mario Molaschi, MD, AP, and Fabrizio Fabris, MD, FP

OBJECTIVES:

To identify prognostic factors that are in-dependently predictive of in-hospital mortality in older pa-tients hospitalized in a medical intensive care unit (MICU).

DESIGN:

Prospective cohort study.

SETTING:

A MICU in an Italian university hospital.

PARTICIPANTS:

Patients aged 65 and older consecu-tively admitted to the MICU directly from the first-aidunit.

MEASUREMENTS:

Upon admission, the following vari-ables were examined: demographics, clinical history (dia-betes mellitus, active neoplasm, cognitive impairment, im-mobilization, pressure ulcers, use of nutritional support,home oxygen therapy), physiopathology (Acute Physiol-ogy and Chronic Health Evaluation (APACHE) II), andcognition/function (activity of daily living (ADL), instru-mental activity of daily living (IADL), Short Portable Men-tal Status Questionnaire (SPMSQ)). The vital status of thepatient at the end of hospitalization was recorded.

RESULTS:

Over a period of 10 months, 659 patientswere recruited (mean age

�

standard deviation

�

76.6

�

7.5; 352 men and 307 women). There were 97 deaths(14.71%). The following factors proved to be significantlyassociated with in-hospital mortality: old age, low bodymass index (BMI) values, low values of albumin, high scoreson APACHE II, functional impairment (ADL, IADL), cogni-tive impairment (SPMSQ), history of cognitive deteriora-tion, history of confinement to bed, and presence of pres-sure ulcers. Using multivariate analysis, the followingvariables were independently predictive of in-hospitalmortality: lack of independence in ADLs (

P

�

.001),moderate-to-severe cognitive impairment on SPMSQ (

P

�

.001), score on APACHE II (

P

�

.002), and low BMIvalues (

P

�

.031).

CONCLUSION:

The prognosis of older patients hospital-ized in medical intensive care units depends not only on

the acute physiological impairments, but also on a series ofpreexisting conditions, such as loss of functional indepen-dence, severe and moderate cognitive impairment, and lowBMI.

J Am Geriatr Soc 51:529–533, 2003.

Key words: in-hospital mortality; risk factors; older people

n Italy, as in most Western countries, persons aged 65and older represent 18% of the population but account

for 45.5% of hospital admissions.

1

With advanced years, ahigher percentage increase in unfavorable outcomes ofhospitalization accompanies the greater frequency of hos-pitalization. Therefore, it appears particularly importantto identify specific prognostic indicators for reliable riskstratification of older patients.

Several studies have evaluated short- and long-termprognostic indicators in older adults admitted to generalmedical wards.

2–8

Even though an increased risk of mortal-ity accompanies old age, most studies suggest that age alonedoes not represent a strong predictor for mortality

5,9–12

andthat there exist other, more-reliable predictive parameters,such as level of functional impairment

3,4,11,13,14

and cogni-tive deterioration.

2,7

Although this approach is now gener-ally accepted in these settings, the risk stratification of pa-tients in medical intensive care units (MICUs) is stillmainly based on the use of several acute physiological pa-rameters. Older patients admitted to MICUs usually havemore-critical health conditions, more-severe physiologicalimpairments, and higher mortality rates than patients ingeneral medical wards.

5,15

Some studies, which did notconsider prognostic indicators specific to older people,have shown that the severity of acute physiological impair-ment is an important predictor of in-hospital death inthese patients,

12,15

but functional, cognitive, and nutri-tional measures have not been included in models of pre-dictors of in-hospital mortality in older patients admittedto MICUs.

The aim of this study was to identify the factors pre-dictive of in-hospital mortality in older patients hospital-ized in a MICU. Inclusion of functional measures could

From the Department of Medical and Surgical Science, Section of Geriatrics, University of Turin, Turin, Italy.

Address correspondence to Mario Bo, MD, University of Turin, Department of Medical and Surgical Science, Section of Geriatrics, C.so Bramante, 88. 10100 Turin, Italy. E-mail: fabris@ddmc.unito.it

I

530

BO ET AL.

APRIL 2003–VOL. 51, NO. 4 JAGS

improve the prediction of in-hospital mortality beyondthat achieved with currently available mathematicalmodels such as the Acute Physiology and Chronic HealthEvaluation (APACHE) II. Knowledge of these predictorscould prove useful to clinicians who decide whether toadmit older patients to a MICU and to researchers whostudy the effectiveness of clinical interventions on therisk of mortality in this setting. For this purpose, an anal-ysis was performed on clinical and demographic vari-ables and a number of parameters of physiological, func-tional, and cognitive impairments investigated usingstandardized scales.

METHODS

This prospective study consecutively recruited 659 pa-tients aged 65 and older over a period of 10 months(from August 1, 2000 to May 1, 2001) who were admit-ted to the MICU of the “Molinette” Hospital of Turin.They were admitted directly from the hospital first-aidunit. Patients coming from other hospitals, from other in-tensive care units, or from other departments of this hos-pital and patients who died or were discharged within 24hours of admission were excluded from the study. Pa-tients admitted to MICU usually have critical health con-ditions requiring noninvasive or invasive monitoring andvital support. The primary diagnoses on admission in-cluded acute coronary syndromes (17.6%), pulmonaryedema (10.6%), gastrointestinal bleeding (7.5%), acuterespiratory failure (6.8%), and cardiac arrhythmia(6.2%).

Two doctors from the staff of the Geriatrics Divisionof the same hospital gathered the data using predefinedforms within the first 24 hours of patient admission. Thefollowing demographic variables were recorded: name andsurname, sex, date of birth, age, education, marital status,and living conditions (e.g., whether the patient was livingalone, with a spouse, with his or her own children or rela-tives, with paid personnel, in long-term facilities). Theabove information was obtained in part directly from thehospital admission form and in part from the patient andwas always validated through an interview with relatives,caregivers, or persons responsible for providing assistance.

The clinical-history data were obtained from theclinical documentation produced at the time of admissionand from the medical history gathered from the patient,whenever possible. Data obtained from surrogates wereused in those patients (15.4%) who were unable to be in-terviewed.

5

The clinical- and medical-history data wereinvestigated with the aim of identifying the presence (be-fore hospitalization) of diabetes mellitus, active neo-plasm, or history of cognitive-behavioral deterioration(for

�

1 year). Also assessed were the need for assistancefrom or supervision by a caregiver, immobilization orconfinement to bed, urinary or urinary and fecal inconti-nence (or the use of an indwelling urethral catheter), useof nutritional aids by oral route or nutrition byparenteral or enteral route, and the need to use oxygen athome.

The following data were gathered from the physicalexamination, which was performed upon patient admis-sion and registered in the patient’s clinical record: weightand height (from which the body mass index (BMI)

16

was

calculated), body temperature in degrees centigrade, sys-tolic and diastolic arterial pressure in mmHg in the recum-bent position and in the sitting position, and heart rate andrespiratory rate over 1 minute. Also noted were the presenceand severity of possible pressure ulcers (scoring: 0

�

absent,1

�

skin redness, 2

�

involvement as far as the subcutis, 3

�

muscle involvement, 4

�

bone involvement).In all patients, main blood chemistry examinations

(serum creatinine, serum sodium, serum potassium, hema-tocrit, white blood cell count) were performed upon ar-rival at the first-aid unit. Fasting serum albumin and totalcholesterol values were collected the first morning afterthe patient’s admission into the hospital ward. Blood-gasanalysis of arterial blood was performed on all the pa-tients in basal conditions at arrival in the hospital ward.Gathered data were used to fill in the APACHE II form.

The APACHE II score is the sum of three scores: agepoints, chronic health points, and acute physiology scorepoints. The correction coefficient according to illness,which was drawn from the list of coefficients published inthe original Knaus et al. paper,

17

was applied to the maindiagnosis upon hospitalization.

The data about functional and cognitive evaluationwere obtained by applying the forms currently in use andpreviously validated for acceptance of patients into thecare of the university’s Geriatric Division.

18

Assessment offunctional status in the 2 weeks before admission was per-formed using activity of daily living (ADLs)

19

and the in-strumental activity of daily living (IADLs) scales.

20

Dataon ADLs and IADLs were obtained from the patient,whenever possible, and integrated with the informationgathered from a close relative, caregiver, or person respon-sible for providing assistance.

5

The cognitive status of thepatient upon admission was assessed using the Short Por-table Mental Status Questionnaire (SPMSQ).

21

Surrogateresponses were used in cognitively impaired patients(SPMSQ

�

5).

3,13

The outcome of the hospitalization wasobtained from the hospital discharge summary. The fol-lowing variables were considered: vital status of the pa-tient upon discharge (whether alive or dead), type of dis-charge (e.g., ordinary, voluntary, transfer to protectedhome/hospitalization, nursing home, long-term care facility,transfer to another center), duration of hospitalization,main diagnosis, and concomitant illnesses or complicationsaccording to the

International Classification of Diseases,Ninth Revision, Clinical Modification.

The normal distribution of the quantitative variableswas evaluated using a graphical method and the Kolmog-orov-Smirnov test. The distribution was considered gauss-ian when the probability associated to the Kolmogorov-Smirnov test was high (

P

�

.01).Using univariate analysis, the variables associated

with in-hospital death were identified; the nonparametriccontinuous variables were analyzed using the Mann-Whitney

U

test, whereas the dichotomous variables were analyzedusing contingency tables with chi-square test and oddsratio (OR) with a 95% confidence interval (CI). The levelof significance used was

P

�

.05. Significant variableswere then introduced into the multivariate analysis bymeans of logistic regression using the stepwise-forwardmethod, to identify the variable independently associatedwith in-hospital mortality.

JAGS APRIL 2003–VOL. 51, NO. 4

PREDICTORS OF IN-HOSPITAL MORTALITY

531

RESULTS

During the period under examination, 1,112 patients aged65 and older were admitted to the MICU; 346 subjectswere not included in the study because they came fromother intensive care units or from intermediate care (stepdown) units. One hundred seven (13.9%) of the 766 eligi-ble patients were not enrolled because the outcome of thehospitalization was not yet available when data collectionfor this study was completed. However, their demographiccharacteristics were similar to those of subjects enrolled.

Table 1 illustrates the main characteristics of the 659patients enrolled. The mean age

�

standard deviation was76.6

�

7.5 (range 65–99); there were 352 men (mean age75.0

�

7) and 307 women (mean age 78.4

�

7.7). Of thesubjects examined, 24% had diabetes mellitus and 14.1%had an active neoplasm. A history of cognitive deteriora-tion was present in 15.3% of the patients. One-quarter ofthe subjects reported a history of incontinence (urinary in-continence, urinary-fecal incontinence, or the presence ofan indwelling urethral catheter). Patients without signs ofcognitive impairment (SPMSQ 0–2 errors) upon admissionaccounted for 78.1% of the population. Complete inde-pendence in ADLs (0 functions lost) and IADLs (10–14functions preserved) was present, respectively, in 68.5%and in 63.7% of the hospitalized patients, and the meanAPACHE II score in the sample was 13.2

�

5.3. The meanstay in the MICU was 7.1

�

6 days, and the mean dura-tion of stay in the hospital was 19.4

�

15.3 days. Therewere 97 deaths (14.7%), 51 (52.6%) in the MICU and theremaining 46 (47.4%) in other departments.

Table 2 shows the rates of in-hospital mortality ac-cording to the various predictors. Of the continuous clinical-demographic variables, age, low body mass index (BMI),low albumin values, and high APACHE II scores were sig-nificantly associated with in-hospital mortality. Of the

clinical variables, a medical history of cognitive deteriora-tion (OR

�

2.28, 95% CI

�

1.32–3.71), confinement tobed (OR

�

4.35, 95% CI

�

1.98–9.17), and presence ofpressure ulcers (OR

�

3.64, 95% CI

�

1.29–10.25) weresignificantly associated with risk of in-hospital mortality.

The indices of functional impairment (ADL, IADL)and cognitive impairment (SPMSQ) were evaluated ascontinuous variables and in terms of dichotomous param-eters. For all the variables, the level of impairment observedwas significantly associated with in-hospital mortality (

P

�

.001). For the same variables, analyzed dichotomously, theassociation was also statistically significant. The presenceof moderate to severe cognitive impairment (4–10 errors)was associated with a significantly higher risk of deaththan slight to absent impairment (0–3 errors) (OR

�

7.02,95% CI

�

4.38–11). Dependent patients (measured byADLs,

�

1 functions lost) and patients autonomous in at leastfive functions (measured by IADLs) were five times morelikely to die than independent patients and patients autono-mous in four or fewer functions (OR

�

5.08, 95% CI

�

3.22–8 and OR

�

4.91, 95% CI

�

2.83–8.57, respectively).For cognitive-functional parameters, lack of indepen-

dence in ADLs, the presence of moderate to severe cogni-tive impairment on the SPMSQ, the APACHE II score, andlow BMI upon admission were independently predictive ofin-hospital mortality (Table 3). Similar results were ob-served when APACHE II corrected according to illnesswas entered in the model instead of the APACHE II score.

DISCUSSION

The results of the present study indicate that the in-hospi-tal prognoses of older patients admitted to a MICU are de-pendent not only on the severity of the physiological im-pairments, but also on the presence of other specificpredictors, such as the lack of independence, serious cog-

Table 1. Demographic, Anamnestic, and Clinical Characteristics of the Population Examined (N

�

659)

Characteristic Value

Male, n (%) 352 (53.4)Age, mean

�

SD 76.6

�

7.5Duration of hospitalization, days, mean

�

SD 19.4

�

15.3Duration of stay in medical intensive care unit, days, mean

�

SD 7.1

�

6Body mass index, kg/m

2

, mean

�

SD 24.4

�

4.6Serum albumin, g/dL, mean

�

SD (normal range 3.5–5.2) 3.1

�

0.5Serum cholesterol, mg/dL, mean

�

SD (normal range 1–240) 171.6

�

49.4Mechanically ventilated or intubated, n (%) 82 (12.4)Diabetes mellitus, n (%) 158 (24.0)Metastasized active neoplasm, n (%) 93 (14.1)Pressure ulcers, n (%) 16 (2.4)Incontinence, n (%) 166 (25.2)Confinement to bed, n (%) 30 (4.6)Home oxygen therapy, n (%) 42 (6.4)Cognitive deterioration, n (%) 101 (15.3)Activities of daily living, functions lost, mean

�

SD 0.8

�

1.6Instrumental activities of daily living, functions preserved, mean

�

SD 10.4

�

3.9Short Portable Mental Status Questionnaire, errors, mean

�

SD 2.8

�

2.9Acute Physiology and Chronic Health Evaluation II score, mean

�

SD 13.2

�

5.3

SD

�

standard deviation.

532

BO ET AL.

APRIL 2003–VOL. 51, NO. 4 JAGS

nitive impairment, and low BMI values. Although these re-sults reinforce current evidence on predictors of mortalityin older hospitalized patients,

2–5,8,9,12,13,15,22–25

this is the firststudy that has shown the predictive value of these vari-ables with regard to hospital mortality in a MICU setting.MICUs are costly and often-crowded hospital units whereolder patients often experience prolonged immobilizationfor close monitoring and intensive life support. Older pa-tients admitted to MICUs usually have greater physiologi-cal alterations

10,15

and higher mortality rates

4,6

than pa-tients in general medical wards. The risk stratification ofthe patients admitted to MICUs is currently based mainlyupon the use of several acute physiological scales. Thisstudy found that cognitive, functional, and nutritionalmeasures have prognostic relevance for hospital mortality.The inclusion of these specific prognostic indicators into amathematical model might provide a reliable predictor of

individual risk for in-hospital mortality, which might per-mit more-appropriate selection of patients for costlyMICU care. Fewer predictive factors have been identifiedthan in other studies.

2,5,8

The inclusion of these predictivevariables (ADL, SPMSQ, and BMI) appear to improvemortality risk stratification, thereby justifying the addi-tional effort required to obtain this information. Determi-nation of ADL capabilities and SPMSQ takes approxi-mately 10 minutes or less by specifically trained staff.

This study differs from other studies of mortality inMICU patients in that it enrolled older subjects

17,26–30

andexamined the effect of functional impairments before ad-mission on in-hospital mortality.

Similar to other studies,

2,5,10,15

it was found that age isnot predictive of in-hospital mortality. Studies that haveevaluated mortality in patients aged 50 and older con-cluded that, in the absence of functional limitations, olderage does not involve a greater risk of death.

11

Although thefunctional impairment detected by the ADL and IADLscales was predictive of unfavorable prognosis by univari-ate analysis, only lack of independence in ADLs proved tobe independently predictive of in-hospital mortality. Thisfinding, consistent with other studies,

2,4,9,24

likely reflectsthe different domains explored by the two scales. Indeed,loss of independence in ADLs usually indicates greaterfrailty of the patient than loss of independence in IADLs,which may not necessarily be differentiated by reviewingthe underlying critical health conditions.

The mean APACHE II score observed in this study(13.2

�

5.3) is lower than that reported in other studiesconducted on patients admitted to emergency depart-ments.

15,28

The different case mix of other intensive careunits because of separation of patients into MICUs, coro-nary care units, and surgical intensive care units may ex-plain this difference.

These findings are consistent with previous studiesthat have found that cognitive impairment is a predictor ofin-hospital and long-term mortality

2,6,14,31

and is associatedwith longer length of stay in the hospital and worse func-tional outcome.

2,14,22,25,32

Similar to other studies,

16,23 it wasfound that low BMI values are also predictive of in-hospitalmortality.

Some limitations of this study must be addressed. Thestudy population of this research consisted of older pa-tients admitted to a MICU of a university teaching hospi-tal and may not be generalized to patients hospitalized at

Table 2. Rates of Mortality in Older MICU Subjects UsingUnivariate Analysis

CharacteristicMortality

% P-value

Age65–70 6.871–76 12.1 �.000177–82 15.983–99 25.5

Body mass index, kg/m2*�21.7 18.821.7–24.2 18 �.0524.3–26.7 11.7�26.7 10.4

Serum albumin, g/dL*�2.8 21.82.9–3.1 19.4 �.053.2–3.4 2.3�3.5 —

Acute Physiology and Chronic HealthEvaluation II score*

�9 2.810–12 15 �.000113–17 18.5�17 25

Short Portable Mental StatusQuestionnaire errors

0–4 8.25–7 22.4 �.00018–10 55.9

Activities of daily living functions lost0 (independent) 7.8 �.00011–6 (dependent) 30

Instrumental activities of daily living score14–10 (independent) 6.99–5 (partially dependent) 24.1 �.00014–0 (dependent) 40

History of cognitive deterioration 24.8 �.01History of confinement to bed 40 �.0001Presence of pressure ulcers 37.5 �.01

* Sample divided into four equally populated subgroups.

Table 3. Variables Independently Predictive of In-HospitalMortality by Logistic Regression

VariableOddsRatio

95%Confidence

Interval

Body mass index, kg/m2 0.93 0.88–0.99Activities of daily living (dependence) 2.84 1.71–4.74Short Portable Mental Status Questionnaire

(moderate to severe impairment) 3.98 2.41–6.58Acute Physiology and Chronic Health

Evaluation II score 1.07 1.03–1.12

JAGS APRIL 2003–VOL. 51, NO. 4 PREDICTORS OF IN-HOSPITAL MORTALITY 533

other institutions. Although it is a single-site study with asmall sample, the number of older subjects enrolled isgreater than in most other studies32–35 on older patients.Medical care that occurs after the patient is transferred toanother hospital unit could affect mortality risk.

In conclusion, the in-hospital prognosis of older pa-tients admitted to a MICU depends not only on acutephysiological impairments, but also on a series of preexist-ing conditions such as loss of functional independence, se-vere cognitive impairment, and low BMI. If these findingsare validated in future studies, evaluations for these preex-isting conditions should be incorporated into models ofmortality when older patients are treated in a MICU or arebeing considered for MICU admission.

REFERENCES1. Annuario Statistico Italiano. Rome, Italy: ISTAT. November 2000.2. Narain P, Rubenstein LZ, Wieland GD et al. Predictors of immediate and 6-

month outcomes in hospitalized elderly patients. J Am Geriatr Soc 1988;36:775–783.

3. Satish S, Winograd CH, Chavez C et al. Geriatric targeting criteria as predic-tors of survival and health care utilization. J Am Geriatr Soc 1996;44:914–921.

4. Incalzi RA, Capparella O, Gemma A et al. The interaction between age andcomorbidity contributes to predicting the mortality of geriatric patients inthe acute-care hospital. J Intern Med 1997;242:291–298.

5. Teno JM, Harrell FE, Knaus W et al. Prediction of survival for older hospi-talized patients: The HELP survival model. J Am Geriatr Soc 2000;48:S16–S24.

6. Inouye S, Peduzzi P, Robison J et al. Importance of functional measures inpredicting mortality among older hospitalized patients. JAMA 1998;279:1187–1193.

7. Callahan C, Hendrie H, Tierney W. Documentation and evaluation of cogni-tive impairment in elderly primary care patients. Ann Intern Med 1995;122:422–429.

8. Walter L, Brand R, Counsell S et al. Development and validation of a prog-nostic index for 1-year mortality in older adults after hospitalization. JAMA2001;285:2987–2994.

9. Chelluri L, Pinsky MR, Grenvik AN. Outcome of intensive care of the‘oldest-old’ critically ill patients. Crit Care Med 1992;20:757–761.

10. Chelluri L, Pinsky MR, Donahoe MP et al. Long-term outcomes of criticallyill elderly patients requiring intensive care. JAMA 1993;269:3119–3123.

11. Mayer-Oakes S, Oye R, Leake B. Predictors of mortality in older patients fol-lowing medical intensive care. The importance of functional status. J AmGeriatr Soc 1991;39:862–868.

12. McClish DK, Powell SH, Montenegro H et al. The impact of age on utiliza-tion of intensive care resources. J Am Geriatr Soc 1987;35:983–988.

13. Covinsky KE, Palmer RM, Counsell SR et al. Functional status before hospi-talization in acutely ill older adults. Validity and clinical importance of retro-spective reports. J Am Geriatr Soc 2000;48:164–169.

14. Alarcon T, Barcena A, Gonzales-Montalvo JI et al. Factors predictive of out-come on admission to an acute geriatric ward. Age Ageing 1999;28:429–432.

15. Van Den Noortgate N, Vogelaers D, Afschrift M et al. Intensive care for very

elderly patients. Outcome and risk factors for in-hospital mortality. Age Age-ing 1999;28:253–256.

16. Landi F, Onder G, Gambassi G et al. Body mass index and mortality amonghospitalized patients. Arch Intern Med 2000;160:2641–2644.

17. Knaus WA, Draper EA, Wagner D et al. APACHE II. A severity of diseaseclassification system. Crit Care Med 1985;13:818–829.

18. Fabris F, Macchione C, Molaschi M et al. La cartella clinica geriatrica: Unaproposta di valutazione funzionale multidimensionale. Minerva Med 1989;80:1–56.

19. Katz S, Ford AB, Moskowitz RW et al. Studies of illness in the aged. The in-dex of ADL, a standardized measure of biological and psychosocial function.JAMA 1963;185:914–919.

20. Lawton MP, Brody EM. Assessment of older people: Self-maintaining and in-strumental activities of daily living. Gerontologist 1969;9:179–186.

21. Pfeifer E. A short portable mental status questionnaire for the assessment oforganic brain deficit in elderly patients. J Am Geriatr Soc 1975;23:433–441.

22. Wu AW, Yasui Y, Alzola C et al. Predicting functional status outcomes inhospitalized patients aged 80 years and older. J Am Geriatr Soc 2000;48:S6–S15.

23. Galanos AN, Pieper CF, Kussin PS et al. Relationship of body mass to subse-quent mortality among seriously ill hospitalized patients. SUPPORT investi-gators. The study to understand prognosis and preferences for outcome andrisks treatments. Crit Care Med 1997;25:1962–1968.

24. Antonelli Incalzi R, Capparella O, Gemma A et al. A sample method of rec-ognizing geriatric patients at risk for death and disability. J Am Geriatr Soc1992;40:34–38.

25. Sager MA, Rudberg MA, Jalaluddin M et al. Hospital Admission Risk Pro-file (HARP). Identifying older patients at risk for functional decline followingacute medical illness and hospitalization. J Am Geriatr Soc 1996;44:251–257.

26. Zimmerman JE, Wagner DP, Draper EA et al. Evaluation of acute physiol-ogy and chronic health evaluation III predictions of hospital mortality in anindependent database. Crit Care Med 1998;26:1317–1326.

27. Markgraf R, Deutschinoff G, Pientka L et al. Comparison of acute physiol-ogy and chronic health evaluations II and simplified acute physiology scoreII. A prospective cohort study evaluating these methods to predict outcomein a German interdisciplinary intensive care unit. Crit Care Med 2000;28:26–33.

28. Escarce J, Kelley M. Admission source to the medical intensive care unit pre-dicts hospital death independent of APACHE II score. JAMA 1990;264:2389–2394.

29. Moreno R, Morais P. Outcome prediction in intensive care. Results of a pro-spective, multicentre, Portuguese study. Intensive Care Med 1997;23:177–186.

30. Goldhill D, Sumner A. Outcome of intensive care patients in a group of Brit-ish intensive care units. Crit Care Med 1998;26:1337–1345.

31. Hanson LC, Danis M, Lazorick S. Emergency triage to intensive care. Can weuse prognosis and patient preferences? J Am Geriatr Soc 1994;42:1277–1281.

32. Lamont C, Sampson S, Matthias R et al. The outcome of hospitalization foracute illness in the elderly. J Am Geriatr Soc 1983;31:282–288.

33. Inui TS, Stevenson KM, Plorde D et al. Identifying hospital patients whoneed early discharge planning for special dispositions. A comparison of alter-native techniques. Med Care 1981;19:922.

34. Wachtel TJ, Derby C, Fulton JP. Predicting the outcome of hospitalizationfor elderly persons: Home versus nursing home. South Med J 1984;77:1283–1285.

35. Maguire PA, Taylor C, Stout RW. Elderly patients in acute medical wards.Factors predicting length of stay in hospital. BMJ 1986;292:1251–1253.