0
We're unable to sign you in at this time. Please try again in a few minutes.
Retry
We were able to sign you in, but your subscription(s) could not be found. Please try again in a few minutes.
Retry
There may be a problem with your account. Please contact the AMA Service Center to resolve this issue.
Contact the AMA Service Center:
Telephone: 1 (800) 262-2350 or 1 (312) 670-7827  *   Email: subscriptions@jamanetwork.com
Error Message ......
Original Contribution |

Importance of Functional Measures in Predicting Mortality Among Older Hospitalized Patients FREE

Sharon K. Inouye, MD, MPH; Peter N. Peduzzi, PhD; Julie T. Robison, PhD; John S. Hughes, MD; Ralph I. Horwitz, MD; John Concato, MD, MS, MPH
[+] Author Affiliations

From the Departments of Internal Medicine (Drs Inouye, Peduzzi, Hughes, Horwitz, and Concato) and Epidemiology and Public Health (Drs Peduzzi and Robison), Yale University School of Medicine, New Haven, Conn, and the Department of Veterans Affairs Cooperative Studies Program Coordinating Center (Dr Peduzzi) and Medical Service (Drs Hughes and Concato), Veterans Affairs Connecticut Healthcare System, West Haven.


JAMA. 1998;279(15):1187-1193. doi:10.1001/jama.279.15.1187.
Text Size: A A A
Published online

Context.— Measures of physical and cognitive function are strong prognostic predictors of hospital outcomes for older persons, but current risk adjustment and burden of illness assessment indices do not include these measures.

Objective.— To evaluate and validate the contribution of functional measures to the ability of 5 standard burden of illness indices (Charlson, Acute Physiology and Chronic Health Evaluation [APACHE] II, Disease Staging, All Patient Refined Diagnosis Related Groups, and a clinician's subjective rating) in predicting 90-day and 2-year mortality among older hospitalized patients.

Design.— Two prospective cohort studies.

Setting.— General medicine service, university teaching hospital.

Patients.— For the development cohort, 207 consecutive patients aged 70 years or older, and for the validation cohort, 318 comparable patients.

Main Outcome Measure.— Death within 90 days and 2 years from the index admission.

Results.— In the development cohort, 29 patients (14%) and 81 patients (39%) died within 90 days and 2 years, respectively. A functional axis was developed using 3 independent risk factors: impairment in instrumental activities of daily living, Mini-Mental State Examination score of less than 20, and shortened Geriatric Depression Scale score of 7 or higher, creating low-, intermediate-, and high-risk groups with associated mortality rates of 20%, 32%, and 60%, respectively (P<.001); the C statistic for the final model was 0.69. The corresponding mortality rates in the validation cohort, in which 59 (19%) and 138 (43%) died within 90 days and 2 years, respectively, were 24%, 45%, and 60% (P<.001); the C statistic for the final model was 0.66. For each burden of illness index, the functional axis contributed significantly to the predictive ability of the model for both 90 days and 2 years. When the functional axis and each burden of illness measure were analyzed in cross-stratified format, mortality rates increased progressively from low-risk to high-risk functional groups within strata of burden of illness indices (double-gradient phenomenon). The contributions of functional and burden of illness measures were substantive and interrelated.

Conclusions.— Functional measures are strong predictors of 90-day and 2-year mortality after hospitalization. Furthermore, these measures contribute substantially to the prognostic ability of 5 burden of illness indices. Optimal risk adjustment for older hospitalized patients should incorporate functional status variables.

Figures in this Article

RISK ADJUSTMENT recently has acquired increasing importance for medical effectiveness and health policy evaluations. For evaluating medical effectiveness, meaningful comparison of outcomes between treatment groups requires controlling for patients' baseline risk.1 For health policy evaluation, risk-adjusted outcome measures are being scrutinized by organizations, including government agencies, managed care companies, and consumer groups.2 Given the current importance of these risk adjustment indices, examination of their content and validity is crucial. Burden of illness indices, defined as measures incorporating acute and chronic health and comorbidity, have been the primary means of health risk adjustment. In the evaluation of older persons, lack of inclusion of functional status has been a major limitation of these indices. Functional status—defined as everyday behaviors necessary to maintain daily life (eg, activities of daily living [ADLs]) and encompassing areas of physical, cognitive, and social functioning1—is of key importance to health outcomes of the elderly.

Previous studies in elderly patients have demonstrated that functional status measures are important predictors of hospital outcomes.37 Moreover, functional measures are stronger predictors of hospital outcomes—including functional decline, length of stay, institutionalization, and death—than admitting diagnoses, diagnosis related groups, and standard indices of illness burden.3,810 For example, Davis et al11 demonstrated that a measure of physical functioning was the best single predictor of hospital mortality, surpassing acute physiologic measures. Despite their importance, most burden of illness indices do not incorporate functional measures.

Our goal was to evaluate how measures of physical and cognitive function contribute to the predictive ability of traditional burden of illness indices in hospitalized older patients. Our hypothesis was that functional measures are important independent prognostic factors and also contribute substantially to the prognostic ability of existing indices. Specific aims were (1) to identify functional measures that predict 90-day and 2-year mortality in older hospitalized patients; (2) to develop a predictive model for mortality based on these functional measures—the functional axis—and to validate this model in an independent sample; and (3) to examine the effect of adding the functional axis to existing burden of illness indices.

Development Study

Subjects. Potential participants were 342 patients, aged 70 years or older, with no clinical evidence of delirium at enrollment, admitted consecutively on weekdays to the 6 general medicine (nonintensive care) floors at Yale-New Haven Hospital (YNHH), New Haven, Conn, between November 6, 1989, and June 22, 1990. The YNHH is an 800-bed urban teaching hospital with 200 medical beds, serving a large community and referral population. Patients were excluded if they could not be interviewed for reasons including intubation, coma, severe aphasia, or terminal condition (n=61); if they were discharged in less than 48 hours (n=44); if they or their physicians declined participation (n=29); or if they had been enrolled in the study on a previous admission (n=1). The final sample included 207 participants.

Clinical Evaluation. Trained clinician-researchers carried out structured interviews with the patients, primary nurses, and family members within 48 hours of admission. The patient interview included demographic information, current living situation, self-reported basic ADLs12 and instrumental activities of daily living (IADLs)13 2 weeks before admission, mobility assessment,14 Mini-Mental State Examination (MMSE),15 the shortened Geriatric Depression Scale (GDS),16 standard near-vision (Jaeger-type) and hearing (whisper)17 tests, and the Confusion Assessment Method (CAM) rating.18 The nurse interview included a subjective rating of burden of illness.19 The family interview included the modified Blessed Dementia Rating scale (mBDRS).20

A separate group of trained medical record abstractors conducted detailed review of hospital records for medical diagnoses, laboratory results, vital signs, and pertinent clinical data needed to rate the medical record–based burden of illness indices. The diagnosis-based indices were created using administrative databases from the hospital information system (see below).

All data were obtained using standardized instruments. The clinician-researchers and medical record abstractors were blind to the research question and hypotheses. Informed consent for participation in the study was obtained from the patient or, for those with substantial cognitive impairment, from the closest relative. This study was approved by the Institutional Review Board of Yale University School of Medicine.

Definition of Functional Variables. Definitions and cutpoints for variables were selected a priori based on clinically meaningful thresholds or previously published studies. The ADL impairment was defined as needing physical assistance with at least 1 of 7 basic care skills (ie, feeding, bathing, grooming, dressing, toileting, transferring, or walking) by patient self-report at baseline; this cutpoint has been used previously.3,4 The IADL impairment was defined as needing assistance with at least 1 of 7 instrumental activities (ie, using the telephone, grocery shopping, using transportation, cooking, housekeeping, taking medications, and handling finances); this cutpoint has been used previously.21 To approximate the patient's preillness status, ADLs and IADLs were referent to 2 weeks prior to admission. Predictive validity for these measures (ie, ability to predict future mortality, rehospitalization, and institutionalization) has been documented.22,23 Mobility impairment was defined as impairment in 1 or both mobility items (ie, ability to walk half a mile or up and down a flight of stairs). Baseline delirium was defined by CAM diagnostic criteria,18 which included acute onset and fluctuating course, inattention, and disorganized thinking or altered level of consciousness. The delirium cases identified were detected with cognitive testing after enrollment. The cutpoint of a score of less than 20 on the MMSE was selected to account for the low mean score in this elderly hospitalized population and to increase specificity for dementia; this cutpoint has been used previously.3,24 The cutpoint of a score higher than 4 on the mBDRS has been used to indicate dementia previously.3,4,20 The composite dementia variable, incorporating both the MMSE and mBDRS, is a highly specific measure for the presence of dementia and has been used as a reference standard.25 It is defined as (1) mBDRS score higher than 4, or (2) mBDRS score higher than 2 and MMSE score less than 20 and duration of cognitive symptoms longer than 6 months. Vision impairment was defined as corrected binocular near vision worse than 20/70.4 Hearing impairment was defined as hearing correctly less than 6 of 12 numbers with both ears on the whisper test.3 The cutpoint of 7 or higher on the shortened 15-item GDS was selected based on recommendations for a cutpoint of 14 or higher on the original 30-item GDS for specific identification of depression.16

Selection of Burden of Illness Indices and Cutpoints. We considered 5 global burden of illness indices, chosen as representative examples of the types of indices currently in widespread use. We wanted to include examples of medical record–based and diagnosis-based indices, as well as a subjective index. The Charlson index26 is a medical record–based system, designed to predict death in longitudinal studies, with an integer score (usually coded as 0, 1, or ≥2) representing increasing levels of illness burden. The Acute Physiology and Chronic Health Evaluation (APACHE) II27 is a medical record–based system (with an integer score from 0-71), which combines acute physiologic measurements and chronic illness factors designed to predict in-hospital and intensive care unit death. To indicate high burden of illness, the cutpoint of more than 16 is used on the APACHE II index, as supported by previous studies.4,7 The Disease Staging scale28,29 is a diagnosis-based system generated from computerized discharge abstract data, designed to predict in-hospital mortality, scored as stages 1.0, 2.0, and 3.0 with substages (eg, 1.1, 1.2) according to diagnosis. Stage 3 is defined as multiple-site involvement or generalized systemic involvement, with associated poor prognosis29; thus, the cutpoint score of 3 or higher is selected to indicate severe disease and high burden of illness. All Patient Refined Diagnosis Related Groups (APR-DRGs)30 is a diagnosis-based system generated from computerized discharge abstract data, designed to predict hospital utilization, with an ordinal score from 1 to 4 representing increasing levels of illness burden. This scale is trichotomized as 1, 2, or 3 or higher to create risk gradients with sufficient frequency distribution in each group. The nurse's severity rating is a subjective rating scale adapted from Charlson et al19 given by the primary nurse on admission yielding an overall rating of burden of illness, scored as low, moderate, or high. This scale is dichotomized as high or low-moderate, as in previous studies where this scale predicted hospital outcomes.3,4 The comorbidity index of Greenfield et al31 was not used because its "physical impairment" subscale includes elements that we considered outside the domain of functional status (eg, circulatory, respiratory, neurologic) and excludes areas we considered to be central aspects of functional status (eg, ADLs and IADLs).

Outcome. The outcome for this study was death within 2 years from the index hospital admission date. As a secondary analysis, 90-day mortality was also examined. Vital status was determined initially by telephone interviews at 90 days and 2 years after the index hospital admission. All reported deaths and dates of death were confirmed by review of medical records, state death certificates, or both. For those subjects unavailable for follow-up or with incomplete data, vital status was determined using the National Death Index (NDI, National Center for Health Statistics, Hyattsville, Md). Deaths were considered confirmed if an NDI record matched the subject on name (first and last), sex, date of birth (month, day within 1 day, and year), and Social Security number (if available).

Validation Study

Subjects. Patients potentially eligible for the validation study included 801 patients aged 70 years or older admitted consecutively during the week to the medicine service at YNHH between July 9, 1990, and July 31, 1991. The inclusion and exclusion criteria were identical to those of the development study. Patients were excluded if they could not be interviewed (n=147); if they were discharged in less than 48 hours (n=118); if they or their physicians declined participation (n=56); if they had been enrolled in the study on a previous admission (n=100); or for other reasons (n=62). The final sample included 318 participants.

Procedure. The clinical evaluation, definition of variables, and outcomes were identical to those in the development study. The same clinician-researchers carried out the study using identical data collection instruments, blinded to the research questions and hypotheses.

Statistical Analyses

For development of the functional axis, the goal was to select functional measures that individually and in combination would be predictive of 2-year mortality. To select among potential candidate functional variables (Table 1), which were grouped into 3 general conceptual categories (physical, cognitive, and other), we adopted the following a priori criteria: (1) clinical relevance; (2) risk factor more prevalent than 15%; (3) relative risk (RR) at least 1.5; (4) P value less than .10 in bivariate Cox proportional hazards analyses; and (5) absence of time-dependent effects (ie, variable does not demonstrate differential impact over time). Independent variables from each category were then combined in a functional axis. The axis approach is an accepted means of variable reduction, providing a strategy to narrow down and select important variables.3,4,32,33

Table Graphic Jump LocationTable 1.—Variables Considered for Functional Axis (N=207)*

To evaluate the factors considered for inclusion in the functional axis, the relationship of each variable with 2-year mortality was assessed in a bivariate Cox proportional hazards analysis. The proportionality assumption was examined for each variable by including a time-dependent effect in the model, ie, a time by variable interaction term. Relative risks (hazard ratios) and 95% confidence intervals (CIs) were calculated from the proportional hazards model regression coefficients and SEs. Four variables were selected for multivariable Cox analysis based on the a priori established criteria, and backward elimination (P<.05 to remove) was used to select the final set of variables.

The predictive accuracy of the final model in both the development and validation cohorts was examined using 2 methods34: (1) by plotting observed vs predicted survival in groups of patients defined by the functional axis (calibration); and (2) by calculating the C statistic (discrimination). Cumulative survival rates were calculated by the Kaplan-Meier method. Differences in cumulative survival rates among the functional risk groups were evaluated by the log-rank statistic. A χ2 test of linear trend was used to test for differences in crude mortality rates among the functional risk groups.

The quantitative effect of the final variables included in the model was assessed by determining mortality rates for groups defined as low, intermediate, or high risk. The size of the observed mortality gradient determines the combined impact of these variables. The statistical effect of adding the functional axis to each burden of illness index was evaluated by calculating an improvement in model χ2 for both 2-year and 90-day mortality. First, a Cox proportional hazards model was fit for each burden of illness index alone and model χ2 was determined. Subsequently, the functional axis was added to the model, and the improvement in model χ2 was calculated with 1 df to test the added effect of the functional axis. Using a separate Cox model for each measure, we also examined the interaction of the functional axis with each burden of illness measure.

Baseline characteristics of the development study population (n=207) were mean ± SD age of 79±6.0 years; 59% female; 91% white; 43% married; 7% admitted from a nursing home; and mean ± SD educational level of 11.3±3.4 years. The median length of hospital stay was 8 days (range, 3-67 days). A total of 81 subjects (39%) died during the 2 years of follow-up, with 17 in-hospital deaths, 29 deaths within 90 days, and 52 deaths within 1 year. Demographic factors, including age, sex, race, education, marital status, living arrangements, and admission source, had no statistically significant impact on mortality and thus were omitted from the models.

Development of the Functional Axis

The 10 candidate variables considered for the functional axis—divided into 3 categories—are shown in Table 1. From the physical functioning category, any IADL impairment was selected because of its superior RR and importance as a predictor of hospital outcomes in previous studies.21 From the cognitive functioning category, an MMSE score of less than 20 was selected because of the relatively low prevalence (<20%) of the other variables, as well as the time-dependent nature of baseline delirium. From the other category, both the shortened GDS score of 7 or higher and vision impairment were selected. The shortened GDS was selected because it achieved the highest RR; vision impairment fulfilled the selection criteria and was chosen over hearing impairment because of its importance in previous studies.4,35

In multivariable analysis, 3 variables—any IADL impairment, an MMSE score less than 20, and a shortened GDS score of 7 or higher—retained quantitatively and statistically significant impact on mortality; the vision impairment variable was eliminated (Table 1). Thus, the final model included 1 variable from each category: physical functioning, cognitive functioning, and other.

Performance of the Functional Axis

Development Cohort. We developed a risk stratification system by adding 1 point for each of the final functional risk factors present, as shown in Table 2. The low-risk group included patients with no risk factors present, the intermediate-risk group had 1 risk factor present, and the high-risk group had 2 or 3 risk factors present. The 2-year mortality rates for the low-, intermediate-, and high-risk groups were 20%, 32%, and 60%, respectively (χ2 test of linear trend, 22.9; P<.001), representing a 3.1-fold increase in mortality between the low-risk and high-risk groups. Figure 1 shows the observed and predicted cumulative survival rates by functional risk group and indicates excellent calibration of the model over the entire 2-year period. The value of the C statistic for the final model was 0.69.

Table Graphic Jump LocationTable 2.—Impact of Functional Axis on Mortality
Graphic Jump Location
Figure 1.—Cumulative survival rates for functional risk groups, created by our risk stratification system in the development cohort. The dashed lines indicate the predicted survival rates, and the solid lines, the observed survival rates. The numbers on the curves represent the actual (observed) number remaining at risk per the original number in the risk group at 12 and 24 months for low-, intermediate-, and high-risk groups (log rank, P<.001).

Validation Cohort. Of the 318 patients in the validation cohort, 138 (43%) died during the 2 years of follow-up, with 17 in-hospital deaths, 59 deaths within 90 days, and 103 deaths within 1 year. Cumulative 2-year survival did not differ significantly between the 2 cohorts. Baseline characteristics of the validation study population were mean ± SD age of 79±6.0 years; 54% female; 91% white; 49% married; 7% admitted from a nursing home; and mean ± SD educational level of 11.3 ± 3.5 years; none of these characteristics differed significantly from the development cohort. The median length of hospital stay was 9 days (range, 2-79 days), which was significantly longer than in the development cohort (Wilcoxon P=.02).

Applying the risk stratification system to the validation cohort (Table 2), the 2-year mortality rates for the low-, intermediate-, and high-risk groups were 24%, 45%, and 60%, respectively (χ2 test of linear trend, 26.4; P<.001), representing a 2.5-fold increase in mortality between the low-risk and high-risk groups. Figure 2 shows the observed and predicted cumulative survival rates by functional risk group in the validation cohort and indicates good calibration of the model. The value of the C statistic for the final model was 0.66.

Graphic Jump Location
Figure 2.—Cumulative survival rates for functional risk groups, created by our risk stratification system in the validation cohort. The dashed lines indicate the predicted survival rates, and the solid lines, the observed survival rates. The numbers on the curves represent the actual (observed) number remaining at risk per the original number in the risk group at 12 and 24 months for low-, intermediate-, and high-risk groups (log rank, P<.001).
Effect of Adding Functional Axis to Burden of Illness Indices

Development Cohort. The improvement in model χ2 for 2-year survival when adding the functional axis to each burden of illness index is shown in Table 3. The corresponding values for improvement in model χ2 for 90-day survival are shown in Table 4. For every burden of illness index, at both 90 days and 2 years, the functional axis adds substantially and significantly to the baseline model χ2, with the improvement in χ2 ranging from 6.4 to 10.2 at 90 days and from 15.0 to 26.0 at 2 years. Thus, in each case, the functional axis contributes significantly to the predictive ability of burden of illness indices.

Table Graphic Jump LocationTable 3.—Improvement in Model χ2 for 2-Year Survival When Adding Functional Axis to Burden of Illness Indices*
Table Graphic Jump LocationTable 4.—Improvement in Model χ2 for 90-Day Survival When Adding Functional Axis to Burden of Illness Indices*

Table 5 shows the mortality gradients produced when adding the functional axis to each burden of illness index in the development cohort. The data show that adding the functional axis creates a mortality gradient within each level of the burden of illness index (eg, the 48% mortality rate for the Charlson index ≥2 now ranges from 20% to 68%). Moreover, the total gradient created by the burden of illness index is further enhanced with the addition of the functional status measures (eg, the Charlson index alone creates a gradient from 6% to 48%, or 42% overall, whereas the gradient increases from 0% to 68%, or 68% overall, with the functional axis added). This effect is demonstrated for each burden of illness index studied. In addition, Table 4 demonstrates a pronounced "double-gradient" phenomenon,36 with mortality rates increasing progressively from low-risk to high-risk groups in most directions—across rows, down columns, or diagonally. This phenomenon indicates that each measure exerts a distinct prognostic effect in the presence of the other, and taken together they contribute to greater impact and risk stratification than either measure alone. Finally, no interaction terms for the functional axis and burden of illness measures were statistically significant at P≤.05.

Table Graphic Jump LocationTable 5.—Mortality Gradient When Adding Functional Axis to Burden of Illness Indices in Development Study*

Validation Cohort. The findings in the development cohort are replicated in the validation cohort. For every burden of illness index, the functional axis adds substantially and significantly to the baseline model χ2 in the validation cohort (Table 3 and Table 4). The addition of the functional axis again creates the double-gradient phenomenon and enhances risk stratification by each burden of illness index (Table 6).

Table Graphic Jump LocationTable 6.—Mortality Gradient When Adding Functional Axis to Burden of Illness Indices in Validation Study

In these prospective cohort studies, we have identified 3 independent functional variables that predict 90-day and 2-year mortality in older hospitalized medical patients: any impairment in IADLs, cognitive impairment (MMSE score <20), and depressive symptoms (shortened GDS score ≥7). These 3 measures represent conceptually and clinically distinct domains of functional status—physical (IADLs), cognitive (MMSE), and psychological (GDS). A simple predictive model based on these 3 factors can be used to predict which elderly medical patients are at high, intermediate, and low risk of mortality during 2-year follow-up. The overarching goals in selection of the final variables were simplicity and clinical relevance; thus, we opted to use as few variables as possible—while still capturing the critical domains of interest—and to use a priori cutpoints that were clinically meaningful and used in previous studies. Most importantly, the addition of these functional measures to any of the 5 widely used burden of illness indices tested substantially and significantly improves their ability to predict 2-year mortality in this population. Moreover, the greater spread in mortality rates created with the addition of functional measures enhances the use of the burden of illness indices as clinical prediction tools, allowing clinicians and clinical investigators to identify groups at lower and higher risk for mortality.

The double-gradient phenomenon demonstrated in Table 5 and Table 6 lends strong support for the importance of both burden of illness and functional measures in predicting mortality in older hospitalized patients. Both types of measures contribute in substantive and interrelated ways. Since burden of illness and functional measures each contribute independent information to forecasting mortality among hospitalized older patients, mortality prediction models that do not include both types of measures are incomplete. Since these models are intended to assist with risk adjustment for populations or patient groups, eg, for purposes of evaluation of medical effectiveness or comparisons of clinical performance across health systems, these models should not be used for mortality prediction in individuals.

Strengths of the current study include the clinically rich prospective data collection on a large group of hospitalized older patients, the complete 2-year follow-up for mortality, the wide range of burden of illness indices examined, and the validation in an independent patient population. Direct comparison of the different burden of illness indices was not a goal of this study; such comparison would involve overinterpretation of our findings. Although each index performed differently in mortality prediction, as anticipated from previous studies,37 replication of the contribution of the functional axis across all indices is an important strength of our study. Finally, validation of all results in an independent data set provides confirmation of the stability of the findings.

The issue of generalizability of the study deserves comment. First, validation in an independent sample lends strong support for the external validity of the study results. Second, the use of all medical patients admitted to YNHH provides a representative group of acutely ill older patients. Despite YNHH's serving as a referral center, the vast majority of patients are admitted to YNHH from the greater New Haven community. Previous studies have verified that patients hospitalized at YNHH reflect the surrounding community according to racial and ethnic composition, socioeconomic status, and educational level.3,4

The present study focuses on the prediction of long-term (2-year) mortality. Prediction of short-term mortality has been the primary goal of most current risk adjustment systems, which focus primarily on the acute hospitalization and short-term follow-up. The reality, however, is that systems are increasingly assuming total health care risks for older populations. Thus, issues of long-term prognosis and mortality prediction will assume increasing importance. Moreover, from the clinician's point of view, longer-term prediction of outcomes has always been the paramount goal to guide clinical care and to choose between different potential treatment strategies. Moreover, it is important to reiterate that the functional axis works well in mortality prediction for both short-term and long-term time frames, as demonstrated in Figure 1 and Figure 2.

A potential limitation in the application of the study findings is the difficulty in obtaining information on functional variables. A first priority for future work will be to find feasible and practical ways for measuring functional status across health care settings. Valid and reliable collection of functional status information by trained lay data collectors has been demonstrated in a variety of studies in hospital and community settings.38,39 Although many hospitals are developing computerized nursing databases that incorporate functional measures, such as the Nursing Acuity Assessment,11 these assessments are not consistently available. Previous investigators have used chart review to obtain information on ambulatory status, functional impairment, and urinary continence.6,7,31 We hope that the present study will provide impetus to hospitals, clinicians, and managed care organizations to improve the systematic collection of both physical and cognitive functional status information on all older hospitalized patients.

Future work is needed to extrapolate the current findings to other settings—including hospital, institutional, outpatient, and community—and to test alternate functional status measures that may be more readily available. In addition, future studies can address the mechanisms by which functional status influences mortality, the impact on mortality of changes in physical and cognitive function over time, and whether modifying these factors decreases mortality. Development of an optimal risk adjustment system for older hospitalized patients should incorporate components of acute and chronic health, comorbidity, and functional status and will benefit from further extension of the present work.

Iezzoni LI. Risk Adjustment for Measuring Health Outcomes.  Ann Arbor, Mich: Health Administration Press; 1994.
Iezzoni LI. Examining the validity of severity measures in today's health policy context.  J Gen Intern Med.1995;10:406-408.
Inouye SK, Wagner DR, Acampora D.  et al.  A predictive index for functional decline in hospitalized elderly medical patients.  J Gen Intern Med.1993;8:645-652.
Inouye SK, Viscoli CM, Horwitz RI, Hurst LD, Tinetti ME. A predictive model for delirium in hospitalized elderly medical patients based on admission characteristics.  Ann Intern Med.1993;119:474-481.
Fields SD, MacKenzie CR, Charlson ME, Sax FL. Cognitive impairment: can it predict the course of hospitalized patients?  J Am Geriatr Soc.1986;34:579-585.
Keeler EB, Kahn KL, Draper D.  et al.  Changes in sickness at admission following the introduction of the prospective payment system.  JAMA.1990;264:1962-1968.
Daley J, Jencks S, Draper D, Lenhart G, Thomas N, Walker J. Predicting hospital-associated mortality for Medicare patients.  JAMA.1988;260:3617-3624.
Winograd CH, Gerety MB, Chung M, Goldstein MK, Dominguez F, Vallone R. Screening for frailty: criteria and predictors of outcomes.  J Am Geriatr Soc.1991;39:778-784.
Narain P, Rubenstein LZ, Wieland GD.  et al.  Predictors of immediate and 6-month outcomes in hospitalized elderly patients: the importance of functional status.  J Am Geriatr Soc.1988;36:775-783.
Gillick MR, Serrell NA, Gillick LS. Adverse consequences of hospitalization in the elderly.  Soc Sci Med.1982;16:1033-1038.
Davis RB, Iezzoni LI, Phillips RS, Reiley P, Coffman GA, Safran C. Predicting in-hospital mortality: the importance of functional status information.  Med Care.1995;33:906-921.
Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged: the index of ADL, a standardized measure of biological and psychosocial function.  JAMA.1963;185:914-919.
Lawton MP, Brady EM. Assessment of older people: self-maintaining instrumental activities of daily living.  Gerontologist.1969;9:179-186.
Rosow I, Breslau N. A Guttman health scale for the aged.  J Gerontol.1966;21:556-559.
Folstein MF, Folstein SE, McHugh PR. The Folstein Mini-Mental State Examination: a practical method for grading the cognitive state of patients for the clinician.  J Psychiatr Res.1975;12:189-198.
Sheikh JI, Yesavage JA. Geriatric Depression Scale (GDS): recent evidence and development of a shorter version. In: Brink TL, ed. Clinical Gerontology: A Guide to Assessment and Intervention. New York, NY: Haworth Press; 1986:165-173.
MacPhee GJA, Crowther JA, McAlpine CH. A simple screening test for hearing impairment in elderly patients.  Age Ageing.1988;17:347-351.
Inouye SK, vanDyck CH, Alessi CA, Balkin S, Siegal AP, Horwitz RI. Clarifying confusion: the Confusion Assessment Method: a new method for detection of delirium.  Ann Intern Med.1990;113:941-948.
Charlson ME, Frederic LS, MacKenzie CR, Fields SD, Braham RL, Douglas RG. Assessing illness severity: does clinical judgment work?  J Chronic Dis.1986;39:439-452.
Uhlmann RF, Larson EB, Buchner DM. Correlations of Mini-Mental State and modified Dementia Rating Scale to measures of transitional health status in dementia.  J Gerontol.1987;42:33-36.
Sager MA, Rudberg MA, Jalaluddin M.  et al.  Hospital Admission Risk Profile (HARP): identifying older patients at risk for functional decline following acute medical illness and hospitalization.  J Am Geriatr Soc.1996;44:251-257.
Sager MA, Franke T, Inouye SK.  et al.  Functional outcomes of acute medical illness and hospitalization in older persons.  Arch Intern Med.1996;156:645-652.
Landefeld CS, Palmer RM, Kresevic DM, Fortinsky RH, Kowal J. A randomized trial of care in a hospital medical unit especially designed to improve the functional outcomes of acutely ill older patients.  N Engl J Med.1995;332:1338-1344.
Tinetti ME, Inouye SK, Gill TM, Doucette JT. Shared risk factors for falls, incontinence, and functional dependence: unifying the approach to geriatric syndromes.  JAMA.1995;273:1348-1353.
Auerswald KB, Charpentier PA, Inouye SK. The informed consent process in older patients who developed delirium: a clinical epidemiologic study.  Am J Med.1997;103:410-418.
Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis.1987;40:373-383.
Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system.  Crit Care Med.1985;13:818-829.
Gonnella JS, Hornbrook MC, Louis DZ. Staging of disease: a case-mix measurement.  JAMA.1984;251:637-644.
Markson LE, Nash DB, Louis DZ, Gonnella JS. Clinical outcomes management and disease staging.  Eval Health Prof.1991;14:201-227.
Edwards N, Honemann D, Burley D, Navarro M. Refinement of Medicare diagnosis-related groups to incorporate a measure of severity.  Health Care Financ Rev.1994;16:45-64.
Greenfield S, Blanco DM, Elashoff RM, Ganz PA. Patterns of care related to age of breast cancer patients.  JAMA.1987;257:2766-2770.
Inouye SK, Charpentier PA. Precipitating factors for delirium in hospitalized elderly persons: predictive model and interrelationship with baseline vulnerability.  JAMA.1996;275:852-857.
Viscoli CM, Horwitz RI, Singer BH. Beta-blockers after myocardial infarction: influence of first-year clinical course on long-term effectiveness.  Ann Intern Med.1993;118:99-105.
Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.  Stat Med.1996;15:361-388.
Rudberg MA, Furner SE, Dunn JE, Cassel CK. The relationship of visual and hearing impairments to disability: an analysis using the Longitudinal Study of Aging.  J Gerontol.1993;48:M261-M265.
Feinstein AR, Wells CK. A clinical-severity staging system for patients with lung cancer.  Medicine (Baltimore).1990;69:1-33.
Iezzoni LI, Ash AS, Shwartz M, Daley J, Hughes JS, Mackiernan YD. Predicting who dies depends on how severity is measured: implications for evaluating patient outcomes.  Ann Intern Med.1995;123:763-770.
Hirsch CH, Sommers L, Olsen A, Mullen L, Winograd CH. The natural history of functional morbidity in hospitalized older patients.  J Am Geriatr Soc.1990;38:1296-1303.
Cornoni-Huntley J, Brock DB, Ostfeld AM, Taylor JO, Wallace RB. Established Populations for Epidemiologic Studies of the Elderly: Resource Data Book.  Bethesda, Md: National Institutes of Health, National Institute on Aging; 1986. NIH publication 86-2443.

Figures

Graphic Jump Location
Figure 1.—Cumulative survival rates for functional risk groups, created by our risk stratification system in the development cohort. The dashed lines indicate the predicted survival rates, and the solid lines, the observed survival rates. The numbers on the curves represent the actual (observed) number remaining at risk per the original number in the risk group at 12 and 24 months for low-, intermediate-, and high-risk groups (log rank, P<.001).
Graphic Jump Location
Figure 2.—Cumulative survival rates for functional risk groups, created by our risk stratification system in the validation cohort. The dashed lines indicate the predicted survival rates, and the solid lines, the observed survival rates. The numbers on the curves represent the actual (observed) number remaining at risk per the original number in the risk group at 12 and 24 months for low-, intermediate-, and high-risk groups (log rank, P<.001).

Tables

Table Graphic Jump LocationTable 1.—Variables Considered for Functional Axis (N=207)*
Table Graphic Jump LocationTable 2.—Impact of Functional Axis on Mortality
Table Graphic Jump LocationTable 3.—Improvement in Model χ2 for 2-Year Survival When Adding Functional Axis to Burden of Illness Indices*
Table Graphic Jump LocationTable 4.—Improvement in Model χ2 for 90-Day Survival When Adding Functional Axis to Burden of Illness Indices*
Table Graphic Jump LocationTable 5.—Mortality Gradient When Adding Functional Axis to Burden of Illness Indices in Development Study*
Table Graphic Jump LocationTable 6.—Mortality Gradient When Adding Functional Axis to Burden of Illness Indices in Validation Study

References

Iezzoni LI. Risk Adjustment for Measuring Health Outcomes.  Ann Arbor, Mich: Health Administration Press; 1994.
Iezzoni LI. Examining the validity of severity measures in today's health policy context.  J Gen Intern Med.1995;10:406-408.
Inouye SK, Wagner DR, Acampora D.  et al.  A predictive index for functional decline in hospitalized elderly medical patients.  J Gen Intern Med.1993;8:645-652.
Inouye SK, Viscoli CM, Horwitz RI, Hurst LD, Tinetti ME. A predictive model for delirium in hospitalized elderly medical patients based on admission characteristics.  Ann Intern Med.1993;119:474-481.
Fields SD, MacKenzie CR, Charlson ME, Sax FL. Cognitive impairment: can it predict the course of hospitalized patients?  J Am Geriatr Soc.1986;34:579-585.
Keeler EB, Kahn KL, Draper D.  et al.  Changes in sickness at admission following the introduction of the prospective payment system.  JAMA.1990;264:1962-1968.
Daley J, Jencks S, Draper D, Lenhart G, Thomas N, Walker J. Predicting hospital-associated mortality for Medicare patients.  JAMA.1988;260:3617-3624.
Winograd CH, Gerety MB, Chung M, Goldstein MK, Dominguez F, Vallone R. Screening for frailty: criteria and predictors of outcomes.  J Am Geriatr Soc.1991;39:778-784.
Narain P, Rubenstein LZ, Wieland GD.  et al.  Predictors of immediate and 6-month outcomes in hospitalized elderly patients: the importance of functional status.  J Am Geriatr Soc.1988;36:775-783.
Gillick MR, Serrell NA, Gillick LS. Adverse consequences of hospitalization in the elderly.  Soc Sci Med.1982;16:1033-1038.
Davis RB, Iezzoni LI, Phillips RS, Reiley P, Coffman GA, Safran C. Predicting in-hospital mortality: the importance of functional status information.  Med Care.1995;33:906-921.
Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaffe MW. Studies of illness in the aged: the index of ADL, a standardized measure of biological and psychosocial function.  JAMA.1963;185:914-919.
Lawton MP, Brady EM. Assessment of older people: self-maintaining instrumental activities of daily living.  Gerontologist.1969;9:179-186.
Rosow I, Breslau N. A Guttman health scale for the aged.  J Gerontol.1966;21:556-559.
Folstein MF, Folstein SE, McHugh PR. The Folstein Mini-Mental State Examination: a practical method for grading the cognitive state of patients for the clinician.  J Psychiatr Res.1975;12:189-198.
Sheikh JI, Yesavage JA. Geriatric Depression Scale (GDS): recent evidence and development of a shorter version. In: Brink TL, ed. Clinical Gerontology: A Guide to Assessment and Intervention. New York, NY: Haworth Press; 1986:165-173.
MacPhee GJA, Crowther JA, McAlpine CH. A simple screening test for hearing impairment in elderly patients.  Age Ageing.1988;17:347-351.
Inouye SK, vanDyck CH, Alessi CA, Balkin S, Siegal AP, Horwitz RI. Clarifying confusion: the Confusion Assessment Method: a new method for detection of delirium.  Ann Intern Med.1990;113:941-948.
Charlson ME, Frederic LS, MacKenzie CR, Fields SD, Braham RL, Douglas RG. Assessing illness severity: does clinical judgment work?  J Chronic Dis.1986;39:439-452.
Uhlmann RF, Larson EB, Buchner DM. Correlations of Mini-Mental State and modified Dementia Rating Scale to measures of transitional health status in dementia.  J Gerontol.1987;42:33-36.
Sager MA, Rudberg MA, Jalaluddin M.  et al.  Hospital Admission Risk Profile (HARP): identifying older patients at risk for functional decline following acute medical illness and hospitalization.  J Am Geriatr Soc.1996;44:251-257.
Sager MA, Franke T, Inouye SK.  et al.  Functional outcomes of acute medical illness and hospitalization in older persons.  Arch Intern Med.1996;156:645-652.
Landefeld CS, Palmer RM, Kresevic DM, Fortinsky RH, Kowal J. A randomized trial of care in a hospital medical unit especially designed to improve the functional outcomes of acutely ill older patients.  N Engl J Med.1995;332:1338-1344.
Tinetti ME, Inouye SK, Gill TM, Doucette JT. Shared risk factors for falls, incontinence, and functional dependence: unifying the approach to geriatric syndromes.  JAMA.1995;273:1348-1353.
Auerswald KB, Charpentier PA, Inouye SK. The informed consent process in older patients who developed delirium: a clinical epidemiologic study.  Am J Med.1997;103:410-418.
Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis.1987;40:373-383.
Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system.  Crit Care Med.1985;13:818-829.
Gonnella JS, Hornbrook MC, Louis DZ. Staging of disease: a case-mix measurement.  JAMA.1984;251:637-644.
Markson LE, Nash DB, Louis DZ, Gonnella JS. Clinical outcomes management and disease staging.  Eval Health Prof.1991;14:201-227.
Edwards N, Honemann D, Burley D, Navarro M. Refinement of Medicare diagnosis-related groups to incorporate a measure of severity.  Health Care Financ Rev.1994;16:45-64.
Greenfield S, Blanco DM, Elashoff RM, Ganz PA. Patterns of care related to age of breast cancer patients.  JAMA.1987;257:2766-2770.
Inouye SK, Charpentier PA. Precipitating factors for delirium in hospitalized elderly persons: predictive model and interrelationship with baseline vulnerability.  JAMA.1996;275:852-857.
Viscoli CM, Horwitz RI, Singer BH. Beta-blockers after myocardial infarction: influence of first-year clinical course on long-term effectiveness.  Ann Intern Med.1993;118:99-105.
Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.  Stat Med.1996;15:361-388.
Rudberg MA, Furner SE, Dunn JE, Cassel CK. The relationship of visual and hearing impairments to disability: an analysis using the Longitudinal Study of Aging.  J Gerontol.1993;48:M261-M265.
Feinstein AR, Wells CK. A clinical-severity staging system for patients with lung cancer.  Medicine (Baltimore).1990;69:1-33.
Iezzoni LI, Ash AS, Shwartz M, Daley J, Hughes JS, Mackiernan YD. Predicting who dies depends on how severity is measured: implications for evaluating patient outcomes.  Ann Intern Med.1995;123:763-770.
Hirsch CH, Sommers L, Olsen A, Mullen L, Winograd CH. The natural history of functional morbidity in hospitalized older patients.  J Am Geriatr Soc.1990;38:1296-1303.
Cornoni-Huntley J, Brock DB, Ostfeld AM, Taylor JO, Wallace RB. Established Populations for Epidemiologic Studies of the Elderly: Resource Data Book.  Bethesda, Md: National Institutes of Health, National Institute on Aging; 1986. NIH publication 86-2443.

Letters

CME
Meets CME requirements for:
Browse CME for all U.S. States
Accreditation Information
The American Medical Association is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The AMA designates this journal-based CME activity for a maximum of 1 AMA PRA Category 1 CreditTM per course. Physicians should claim only the credit commensurate with the extent of their participation in the activity. Physicians who complete the CME course and score at least 80% correct on the quiz are eligible for AMA PRA Category 1 CreditTM.
Note: You must get at least of the answers correct to pass this quiz.
You have not filled in all the answers to complete this quiz
The following questions were not answered:
Sorry, you have unsuccessfully completed this CME quiz with a score of
The following questions were not answered correctly:
Commitment to Change (optional):
Indicate what change(s) you will implement in your practice, if any, based on this CME course.
Your quiz results:
The filled radio buttons indicate your responses. The preferred responses are highlighted
For CME Course: A Proposed Model for Initial Assessment and Management of Acute Heart Failure Syndromes
Indicate what changes(s) you will implement in your practice, if any, based on this CME course.

Multimedia

Some tools below are only available to our subscribers or users with an online account.

Web of Science® Times Cited: 342

Related Content

Customize your page view by dragging & repositioning the boxes below.

Articles Related By Topic
Related Collections
PubMed Articles