IAGS
39:778-784, 1991
Screening for Frailty: Criteria and Predictors of Outcomes Carol Hutner Winograd, MD,* Meghan B. Gerety, MD,t Maria Chung, MD,$ M a y K. Goldstein, MD,§ Frank Dominguez, Jr., BS, 11 and Robert Vallone, PhDll
Objective: To determine the reliability of rapid screening (P < O.OOOl), nursing home utilization (P < 0.0001), and by clinically derived geriatric criteria in predicting outmortality (P < 0.0001). Multivariate analyses revealed comes of elderly hospitalized patients. that the clinical groups were more predictive of mortality Design: Prospective cohort study of 985 patients and nursing home utilization than were age or Diagnoscreened at the time of hospital admission and followed sis-Related Groups CDRGs). Rehospitalization was unrefor 1 year with respect to the outcomes of mortality, lated to age, clinical group, or DRG, suggesting thaf utilization may not be driven by the clinical factors hospital readmission, and nursing home utilization. measured in this study. Setting: Palo Alto Veterans Affairs Medical Center, a Conclusions: Rapid clinical screening using specific tertiary care teaching hospital. geriatric criteria is effective in identifying frail older subSubjects: Male patients 65 years of age and older admitted to the Medical and Surgical services during the jects at risk for mortality and nursing home utilization. period from October 1, 1985 through September 30, 1986. Our findings suggest that geriatric syndromes are more predictive of adverse outcomes than diagnosis per sc. Results: Patients were grouped by specific screening This well operationalized screening process is inexpencriteria into three groups of increasing frailty: lndependsive as well as effective and could easily be introduced ent, Frail, and Severely Impaired. Each criterion focused into other hospital settings. J Am Geriatr SOC39:778on a geriatric condition and was designed to serve as a 784, 1991. marker for frailty. lncreasing frailty was significantly correlated with increasing length of hospital stay
tudies of multidisciplinary geriatric interventions have had mixed results. Successful trials have been conducted in subacute geriatric evaluation or rehabilitation units,',2 in hospitals using geriatric consultation teams,3or in outpatient settings.*Improved survival, functional status, and mo-
S
rale have been the principal benefits. However, other studes have not found geriatric intervention effect i ~ e . One ~ - ~ reason often cited for the ineffectiveness of these trials has been failure to target the intervention toward those individuals most likely to benefit, the frail elderly.' To date, however, little research has been
* Associate Professor of Medicine, Chief of Clinical Programs in Geriatric Medicine, and Assistant Director, Division of Endocrinology, Gerontology, and Metabolism, Department of Medicine, Stanford University School of Medicine; and Director of Clinical Activities, Geriatric Research, Education and Clinical Center, Veterans Affairs Medical Center, Palo Alto, California. t Assistant Professor of Medicine, Division of Geriatrics and Gerontology and American Federation of Aging ResearchIMerck Foundation Scholar in Geriatric Clinical Pharmacology, University of Texas Health Sciences Center at San Antonio and the Audie L. Murphy Memorial Veterans Hospital, San Antonio, Texas. $ Geriatric Fellow at the Geriatric Research, Education and Clinical Center, Palo Alto Veterans Affairs Medical Center, Palo Alto, CA at the time of this research, now a Clinical Instructor, Division of Geriatric Medicine, Department of Medicine, University of British Columbia, University Hospital, Vancouver, Canada. §Clinical Assistant Professor of Medicine and Director of Graduate Medical Education, Division of Endocrinology, Gerontology, and
Metabolism, Stanford University, and Staff Physician, Geriatric Research, Education and Clinical Center, Palo Alto Veterans Affairs Medical Center, Palo Alto, California. 1) Research Health Scientist at the Geriatric Research, Education and Clinical Center at the time of this research, now a Resource Specialist at the Director's Office, Palo Alto Veterans Affairs h4edical Center, Palo Alto, California. 11 Senior Research Scientist, Software Research Department, Ashton Tate Corporation, Sunnyvale, California. This work was supported in part by National lnstitute on Aging Clinical lnvestigator Award No. KO8 AGO0246 (to CHW), by The Henry J. Kaiser Family Foundation Grant No. 86-3863, and the Veterans Affairs Health Services Research and Development Administration Grant No. I1R 84-062. Address correspondence to Dr. Winograd, GRECC 182-B, Veterans Affairs Medical Center, 3801 Miranda Avenue, Palo Alto, CA 94304.
0002-8614/91/$3.50
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
IAGS-AUGUST 1991-VOL. 39, NO. 8
779
geriatric consultation (short stay), who lived at too great a distance for follow-up (geographic exclusions), who were already enrolled in an interdisciplinary geriatric or rehabilitation program (geriatric services), and who were admitted from a nursing home (nursing home residents). The second step of the screening process used clinical geriatric criteria. The clinical categories defined a prion‘ were: ( 1 ) Independent, (2) Frail and (3) Severely Impaired. To be classified as Independent, a subject had to be independent in all Activities of Daily Living (ADLs) with short-term acute illness, eg pneumococcal pneumonia. Individuals classified as Frail met any one of the clinical criteria described in Table 1. Each criterion focused on a geriatric condition and was designed to serve as a marker for frailty, eg dependence in ADLs, incontinence, impaired mobility, or disabling chronic illness. Severely Impaired subjects were those who METHODS were terminally ill or who had severe dementia and All patients aged 65 years of age and older admitted ADL dependence. These clinical groups correspond to to the Medical and Surgical services of the Palo Alto the previously described’ groups as follows: IndependVeterans Affairs Medical Center (PAVAMC) between ent, Frail, and Severely Impaired correspond respecOctober 1, 1985 and September 30, 1986 were screened tively to Too Well, Appropriate, and Too Sick, respecwithin 96 hours of hospital admission. The PAVAMC tively. The Frail group were the subjects randomized is a tertiary care teaching hospital affiliated with Stan- into the trial of geriatric consultation. Because the ford University. For individuals who were hospitalized consultation process might have produced a change more than once during this period, the first admission within the experimental and control groups, differences was designated the index hospitalization for data with respect to each of the outcomes were analyzed. No difference occurred between experimental and conanalysis. All screening was performed by trained research trol groups with respect to mortality (P = 0.43), hosassistants using a standardized data collection form. pitalization ( P = 0.63), or nursing home utilization Details of the screening process have been published ( P = 0.30). Therefore these groups were included topreviously’ and are illustrated schematically in Figure gether as the Frail group in further analysis. Because all patients admitted to the hospital were to 1. Screening took place in a two-step process. Briefly, the first step of the screening process was administrative and excluded those patients whose expected length TABLE 1. CLINICAL TARGETING CRITERIA of stay was too short to permit an interdisciplinary
done to define targeting criteria to be used to identify the frail elderly in hospital settings. We report the use of a simple clinical screening process with specific geriatric criteria to separate the hospitalized elderly into clinical groups of increasing frailty. We then report the effectiveness of these clinical groups in predicting the negative outcomes of mortality, hospital readmission, and nursing home utilization. The screening criteria for frailty, derived from review of the literature and from clinical experience, were designed to identify subjects for a randomized controlled trial (RCT) of geriatric consultation. The major goals of the RCT were to improve functional status and survival and decrease hospital and nursing home utilization. Our object was to stratify hospitalized patients into groups with increasing risk for these outcomes in order to target the intervention to those at high risk.
-
Admissions 1985 1986
administrative screening
I
clinical screening
t
Geaeraohv n=211
n = 52 n = 24
t
lndeandcnt
:1
n = 251
n = 109 verelv lmoairep n=41
FIGURE 1. Clinical targeting process.
Independent Independent in all ADL’s with short term acute illness Frail (meets any one of the following criteria) Cerebrovascular accident Chronic and disabling illness Confusion Dependence in ADL‘s Depression Falls Impaired mobility Incontinence Malnutrition Polypharmacy Pressure sore Prolonged bedrest Restraints Sensory impairment Socioeconomic/family problems Severely impaired Severe dementia and ADL dependence. Terminal illness. Adapted from Winograd et al, 1988
780
WINOGRAD ET AL
IAGS-AUGUST 1991-VOL. 39, NO, 8
be screened, the screening process had to be rapid and diagnosis responsible for the length of stay. The five inexpensive. Sources of data were arranged in a hier- groups were: surgery, oncology, cardiovascular, pularchical manner to facilitate speed of collection and monary, and any other medical DRG. These five DRG access to information. In order of first use to last use, groups were chosen because prior research has shown the sources of data for the screening process were: these diagnostic groups to have prognostic significance medical record, nursing staff, primary care physician, for response to geriatric To allow sufpatient or the patient’s family. Although performance ficient numbers of subjects in each group, the five DRG on standardized scales such as Activities of Daily Liv- groups were then collapsed into three DRG groups. ing, Mini-Mental State, and Caregiver Burden have The other medical, cardiovascular, and pulmonary been found to be associated with mortality, length of groups were combined into a single “medical” DRG. hospital stay, and nursing home these Before combining these DRG groups, differences with data were not readily available in the medical records. respect to each outcome were examined. The three To administer such scales would have reduced the combined diagnostic categories were not significantly efficiency and clinical utility of the screening process. different from each other with respect to mortality Thus, we did not include such formal measures but ( P = 0.48), hospitalization (P = 0.6), or nursing home rather concentrated on information readily available to utilization (P = 0.68). Subsequent analyses are prethe clinician. All questions about assignment of subjects sented using the three collapsed DRG groups of Surto a category were adjudicated by one of the co- gery, Medicine, and Oncology. investigators (CHW or MBG). Inter-rater reliability was We also wished to determine whether the adminisdetermined with a sample of 53 admissions which were trative step of the screening process, which excluded screened independently by the project coordinator and patients for geographical reasons and because of exthe research assistant during a 1-week period. One- pected short lengths of stay, had not excluded patients hundred percent agreement was achieved for the ad- who were frail and at increased risk for the negative ministrative screening, and 98% agreement was outcomes. When the screening process took place, our achieved in separating patients into the clinical cate- assumption was that the Short Stay and Geography gories. groups, because they were admitted for diagnostic These subjects were followed for 1 year from the procedures and/or were able to come and go from long date of admission of the index hospitalization. Data distances, would have similar mortality and health care were collected for the following outcome measures: utilization to the Independent group. We also predicted length of stay (LOS) of the index hospitalization, num- that the geriatric services group would have similar ber and length of subsequent hospitalizations, number mortality and health care utilization when compared and length of nursing home admissions, and mortality. to the Frail group. All hospital and nursing home utilization data reflect Data Analysis only utilization of VA hospitals, VA Nursing Home Differences among the clinical and administrative Care Units (NHCUs), or community nursing home placement at VA expense. Hospitalization data were groups with respect to length of stay and hospital obtained from the Patient Treatment File (PTF), the mortality were performed using analysis of variance centralized, computerized inpatient data base for the and chi square statistics. Univariate associations of VA. Nursing home utilization data were obtained from Clinical group and DRG groups with survival were the PTF and the records of the Community Nursing examined using Kaplan-Meier survival analysis. lndeHome Program at the Palo Alto VA. Mortality data pendent predictive associations of age, Clinical, and were obtained from the California Automated Mortal- DRG group with mortality were determined using mulity Listings (CAMLIS),l3 which accurately records all tivariate analysis with the Cox proportional hazards deaths of California residents. Of the 985 patients model. To compare the hospital and nursing home admitted during the year, partial data for 14 (1.4%) utilization in the Clinical and DRG groups, utilization were unavailable due to mismatch of social security was converted into a logical variable. Utilization was defined as either no utilization or any utilization, and numbers and/or names with admission dates. To examine the predictive value of (DRGs) and to chi square analysis was used to determine differences. compare the relative efficacy of the DRG groups and Logistic regression was used to determine the relative the clinical screening groups (ie diagnosis versus ger- contribution of age, Clinical, or DRG group to utilizaiatric syndromes) in predicting the outcomes, all pa- tion. tients were classified into one of five major diagnostic RESULTS groups using .the VA DRG designation from the index There were 985 patients over the age of 65 admitted hospitalization. Though the DRG classification system used by the VA is similar to that of Medicare, the final to the PAVAMC during the study period. Of these 584 DRG classification in the VA is based on the principal (59%) were screened out in the administrative screen-
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
/AGS-AUGUST 1991-VOL. 39, NO. 8
ing process, and 401 (41%) remained in the group that was clinically screened (Figure 1). Baseline characteristics of all administratively and clinically screened patients are presented in Table 2. Data regarding outcomes were unavailable for 3 subjects. Therefore, subsequent tables present outcome data for 398 patients. No significant difference in age occurred between the administrative and the clinically screened patients. Significant differences in age existed within each group, however. In the administrative group those receiving geriatric services were older with a mean age of 78.9 compared to 71.9 overall (P < 0.0001). In the clinically screened group, those classified as Frail were also significantly older, with a mean age of 75.2 compared to 71.6 overall (P < 0.0001). Of interest was that 68% of those aged 65-79 fell into the Independent category. Above the age of 79 this proportion fell to 25%. The administrative and clinically screened groups had similar proportions of admissions to Medical and Surgical services (46% vs 44% medicine and 54% vs 56% surgery, respectively). Table 2 also presents the mean length of stay (LOS) for the index admission for both the administrative and the clinically screened groups. Of note is that the average length of stay is more than 10 days in all groups except the short stay. This reflects the longer average length of stay in VA hospitals compared with non-VA h0spita1s.l~Differences in LOS for the index admission among the clinically screened groups are large and significant (P < O.OOOl), with the LOS for the Independent group averaging much less than the other two groups. Table 3 presents nursing home and hospital utilization data for the entire follow-up year. Nursing home utilization was significantly different among the clinically screened groups (P < 0.0001). In the year following the index admission, 13 % of the clinically screened subjects were admitted to a nursing home. Only 3% of the Independent Group were institutionalized, comTABLE 2. BASELINE CHARACTERISTICS OF SCREENED PATIENTS Index LOS* n (%I Age* (days) Administrative Groups Short Stay Geography Geriatric services NH residents Total Clinical Groups Independent Frail Severely impaired Total * P < 0.0001
297 (51) 211 (36) 52 (9) 24 (4) 584 (100)
71.1 (65-93) 71.1 (65-93) 78.9 (65-93) 72.8 (65-93) 71.9 (65-93)
5.1 16.7 11.7 12.9 10.2
251 (63) 109 (27) 41 (10) 401 (100)
70.1 (65-94) 75.2 (65-106) 71.6 (65-88) 71.6 165-106)
12.9 24.8 20.7 16.9
781
TABLE 3. NURSING HOME AND HOSPITAL UTILIZATION IN YEAR FOLLOWING INDEX ADMISSION (FOR SURVIVORS OF INDEX HOSPITALIZATION) Admitted to Readmitted to Survivors Nursing Home* Hospital** n = 357 n ('70) n (%) Independent Frail Severely impaired Total * P < 0.0001. ** P = 0.34.
6 (3) 31 (34) 10 (42) 47 (13)
120 (49) 52 (58) 11 (46) 183 (51)
pared with 34% of the Frail group and 42% of the Severely Impaired group. Multivariate analysis using age, DRG group, and Clinical group as predictors of nursing home utilization in the year following index admission revealed that only the Clinical groups were independently associated with nursing home utilization (P < 0.0001). Calculation of likelihood ratios revealed that among the three clinical groups, the relative likelihood of using a nursing home vaned widely. When compared with the Independent group, the Frail group was 20.75 times and the Severely Impaired group was 28.71 times more likely to enter a nursing home in the year following index admission. Not shown in Table 3, discharge to a nursing home after the index hospitalization was highest among the Severely Impaired group (10.3%)as compared with the Independent (0.4%) and Frail (2.8%) groups (P < 0.0002). In the year following the index admission, 51% of survivors of the index hospitalization were rehospitalized. There was an average of 0.8 hospitalizations (range 0-5, SD 1.03) and 10.8 hospital days (range 0125, SD 20.5). No significant differences occurred in numbers or days of hospitalization among the Clinical Groups or the DRG groups (P = 0.34). Survival curves for the Clinical Groups and the DRG groups are presented in Figure 2. These curves show highly statistically significant differences among the three Clinical groups and the three DRG groups. Survival at one year was inversely associated with increasing frailty, with 87% alive in the Independent group, 55% in the Frail group, and 26% in the Severely Impaired group (P < 0.0001). Survival curves shown for the DRG groups are limited to the 398 patients for whom clinical screening information was available since we wished to compare the relative predictive value of these two classification schemes for mortality. In the three DRG groups survival at 1 year was the highest in the surgical group, with 85% survival, intermediate for the medical group with 67%, and lowest for the oncology group with 41% (P = 0.0001). Within each DRG group, mortality increased with
782
WINOGRAD ET AL
FIGURE
2. Survival
curves
IAGS-AUGUST 1991-VOL. 39, NO. 8
87% Independent (n=218) 85%Surgery (n=148)
for
screened patients by clinical and DRG group. The two sets of survival curves, solid for clinical group and beaded for DRG group, represent the same group of patients. For each of the two sets of curves, the differences between curves are significant with P < 0.0001.
67% Medicine ( ~ 1 % )
55% Frail (n=60) @
40'
41%Oncology (n=16)
20 *
26% Severely Impaired (n=IO)
TABLE 4. TOTAL MORTALITY BY CLINICAL AND DRG GROUP Clinical Groups Independent Frail Severely Impaired n = 398 n (%I n (%I n (%)
Surgery Medicine Oncology Total
13 (10) 16 (17) 3 (21) 32 (13)
11 (36) 32 (47) 6 (60) 49 (45)
2 (100) 13 (59) 14 (93) 29 (74)
Total n (%) 26 (15) 61 (33) 23 (59) 110 (28)
TABLE 5. RESULTS OF MULTIVARIATE ANALYSIS FOR SURVIVAL 95% Confidence Selected Variable Coefficient @ Improvement x2 Interval of j3 Relative Risk
Independent Severely impaired Surgery Medicine
-1.38 0.61 -.99 -.55
76.88 11.90 4.99 4.32
(-1.84, -.42) (.12, 1.1) (-1.61, -.37) (-1.05, -.05)
0.25 1.84 0.37 0.37
For the multivariate analyses, the Frail group was assigned a relative risk of 1.0 for comparison w i t h t h e Independent and Severely Impaired groups. Similarly, the Oncology group was assigned a relative risk of 1.0 for comparision with the Surgery and Medicine groups. These assignments art' for purposes of comparison only. X(t;z) = Xn(t)exp(-l.38 Well + .61 Sick - .99 Surgery - .55 Medicine).
increasing frailty as indicated by the clinical group (Table 4). Moving from left to right in the table, one can see a clear increase in mortality with increasing frailty. Further, multivariate analysis was performed (Table 5 ) using age, Clinical group, and DRG. Age was not a significant predictor of mortality, but the clinical and DRG groups all made independent significant contributions to mortality. The chi square statistic for each term in the model shows that the clinical groups were much more important in the prediction of mortality than were the DRG groups. To answer the question of whether individuals at high risk for mortality, nursing home placement, or hospitalization had been screened out during the administrative phase of screening, two sets of comparisons were made. The Independent, Short Stay, and Geography groups were compared and found to have no significant differences in mortality or rehospitalization. More nursing home admissions occurred in the Short Stay group (11.6%) than in the Geography (5%) or Independent group (3%) (P < 0.001). Similarly, when the Frail and Geriatric Services groups were compared with each other, no significant differences
in mortality or rehospitalization were found. More nursing home admissions occurred in the Geriatric Services group than the Frail group (51% vs 34%, P < 0.05).
DISCUSSION The clinical screening criteria were developed to select individuals who were at high risk for mortality, nursing home placement, and rehospitalization. Each criterion focused on a geriatric condition and was intended to serve as a marker for frailty. This study was undertaken to test our predictions. Our predictions were largely correct. Using 1-year follow-up data, we were able to demonstrate that mortality and nursing home utilization were strongly predicted by the clinical group assigned using the geriatric criteria. Indeed, the clinical group was a better predictor of mortality and nursing home utilization than was either age or DRG group. This finding suggests that geriatric syndromes are more predictive of adverse outcomes than diagnosis per se. Rubenstein's earlier work" also classified hospital inpatients into one of five clinical categories: terminal,
IAGS-AUGUST 1991-VOL. 39, NO. 8
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
acute medical illness, demented, GEU candidate, and independent. He was able to show that this clinical classification separated hospitalized inpatients into groups with differing mortality and utilization of nursing homes and hospitals. His study, however, was designed to select medically stable individuals with geriatric conditions for treatment in a geriatric evaluation unit. Our purpose was to select acutely ill patients who were frail and at high risk for death or nursing home placement. An unexpected finding was that hospital utilization was not associated with Clinical Group, age, or DRG. This was particularly surprising in view of our finding of association with mortality and nursing home placement. Rehospitalization rates were high in all subjects but were not different in subjects who were frailest. Prior studies have not focused on hospital utilization as an outcome, but careful reading reveals that our hospital admission cohort was rehospitalized with a frequency similar to those in previous studies of geriatric 17, Hospitalization does not appear to be influenced by the functional and clinical factors measured in this study. This suggests that other factors must drive hospital utilization. Possibilities include other unmeasured clinical variables, financial, fiscal, family, or environmental factors. Our original assumption was that patients who were hospitalized for short stays for diagnostic procedures (Short Stay) or those who resided a considerable distance from the medical center (Geography) were likely to be the most similar to those in the Independent group. This was the case. The short stay and the geographically excluded groups did have the lowest mortality. Of interest was the fact that more nursing home utilization occurred then expected in the short stay group. These individuals did not undergo the clinical screening meant to detect frailty, and short stay patients may have actually been hospitalized for chemotherapy or diagnostic procedures, thereby representing a group with heavy disease burden or cancer. This may account for the nursing home admissions. Most important, we found that this simple screening process could distinguish the elderly at high risk for mortality and nursing home placement from those at low risk. Our screening process was simple, clearly defined, inexpensive, and used readily available sources of information. The person who screened the patients was an individual with no prior medical training, trained by our research team. Each patient took an average of 10 minutes for screening. All medical, surgical, and neurological admissions were screened by this person on a daily basis using less than 0.25 Full Time Employee Equivalent (FTEE). There are several limitations to our findings, however. It should be pointed out that the health care utilization data is restricted to services provided by the
783
VA. Hospital utilization refers only to Veterans Affairs hospitals, and the nursing home data encompasses VA Nursing Home Care Units or VA Community Based Contract Nursing Homes; placement in nursing homes at private expense is not detailed in this study. Thus, these health care utilization data are limited, but we have no reason to believe they should be different among groups. As part of the Randomized Controlled Trial of geriatric consultation, we attempted to determine the extent of non-VA health care utilization in the Frail cohort. We interviewed 69 elderly subjects in the Frail group at 2 weeks, 3 months, 6 months, and 1 year post-discharge from the index admission. Utilization of a non-VA emergency room, nursing home, or hospital averaged across the time intervals, 7%, 5%, and 9%, respectively. The low utilization of non-VA services in the Frail group suggests these data appear to be relatively complete even though they should be considered lower bound estimates of health care utilization. This well operationalized, simple, brief screening process, using specific clinical geriatric criteria to identify patients most at risk for adverse outcomes, is inexpensive and effective and may be of great value for discharge planners, hospital administrators, and primary care physicians. In addition to providing prognostic information, this strategy may identify individuals in need of specialized geriatric services or discharge planning. Additional research is under way to define the discriminant criteria, prospectively validate this approach, and compare its utility with functional status measures such as Activities of Daily Living. Although these criteria have only been tested in the hospitalized elderly, it is likely that a modified version of our screening process could be developed for the outpatient or nursing home setting. Separation of patients into groups with prognostic significance for mortality and nursing home utilization may allow intervention strategies to be developed to alter these outcomes, and to compare efficacy of interventions for groups at high and low risk. ACKNOWLEDGMENTS The authors wish to thank Lincoln Moses, PhD for his critical review and suggestions, Michael R. Tuley, PhD for his analytical assistance, and Cindy Steams for her assistance in the preparation of this manuscript.
REFERENCES Rubenstein LZ, Josephson KR, Wieland GD et al. Effectiveness of a geriatric evaluation unit. A randomized clinical trial. N Eng J Med 1984;311:1664. Applegate WB, Miller ST, Graney MJ et al. A randomized, controlled trial of a geriatric assessment unit in a community rehabilitation hospital. N Eng J Med 1990;322:1572. Hogan DB, Fox RA, Badley BWD et al. Effect of a geriatric consultation service on management of patients in an acute care hospital. Can Med Assoc J 1987;136:713.
784
WINOGRAD ET AL
IAGS-AUGUST 1991-VOL. 39, NO. 8
4. Williams ME, Williams TF, Zimmer JG et a]. How does the team hospitals and homes: a survey and follow-up. Public Health approach to outpatient geriatric evaluation compare with tradi1984;98:270. tional care: a report of a randomized controlled trail. J Am Geriatr 12. Wachtel TJ, Derby C, Fulton JP. Predicting the outcome of SOC1987;35:1071. hospitalization for elderly persons: Home versus nursing home, 5. Gayton D, Wood-Dauphinee S, de Lorimer M et al. Trial of a South Med J 1984;77:1283. geriatric consultation team in an acute care hospital. J Am Geriatr 13. Arellano MG, Petersen GR, Petitti DB et al. The California SOC1987;35:726. automated mortality linkage system (CAMLIS). Am J Public 6. Campion EW, Jette A, Beckman B. An interdisciplinary geriatrics Health 1984;74:1324. consultation service: a controlled trial. J AM Geriatr SOC1983; 14. Rubenstein LZ, Wieland GD, Josephson KR et al. Improved 31 :792. survival for frail elderly inpatients on a geriatric evaluation unit 7. Becker AM, McVey LJ, Saltz CC et al. Hospital-acquired compli(GEU): who benefits? J Clin Epidemiol 1988;41:441. cations in a randomized controlled clinical trial of a geriatric 15. Randall M, Kilpatrick KE, Pendergast JF et al. Differences in consultation team. JAMA 1987;257:2313. patient characteristics between Veterans Administration and community hospitals. Med Care 1987;125:1099. 8. Rubenstein LZ. Documenting impacts of geriatric consultation. J 16. Rubenstein LZ, Josephson KR, Wieland GD et al. Differential Am Geriatr SOC1987;35:829. 9. Winograd CH, Gerety MB, Brown E et al. Targeting the hospiprognosis and utilization patterns and clinical subgroups of talized elderly for geriatric consultation. J Am Geriatr SOC hospitalized geriatric patients. Health Serv Res 1986;20(11):882. 17. Fethke CC, Smith IM, Johnson N. "Risk" factors affecting read1988;36:1113. 10. Narain P, Rubenstein LZ, Wieland GD et al. Predictors of immission of the elderly into the health care system. Med Care mediate and 6-month outcomes in hospitalized elderly patients. 1986;24:429. The importance of functional status. J Am Geriatr SOC 18. Saltz CC, McVey LJ, Becker PM et al. Impact of a geriatric 1988;36:775. consultation team on discharge placement and repeat hospitali1 1 . Donaldson LJ. Outcome of admissions of elderly people to zation. Gerontol 1989;28:344.
An incorrect version of the article Two-Year Trends in Physical Performance following Supervised Exercise among Community-Dwelling Older Veterans was published in the June 1991 issue of the Journal of the American Geriatric Society. The correct version will appear in the October 1991 issue. The Journal regrets the error.
39:778-784, 1991
Screening for Frailty: Criteria and Predictors of Outcomes Carol Hutner Winograd, MD,* Meghan B. Gerety, MD,t Maria Chung, MD,$ M a y K. Goldstein, MD,§ Frank Dominguez, Jr., BS, 11 and Robert Vallone, PhDll
Objective: To determine the reliability of rapid screening (P < O.OOOl), nursing home utilization (P < 0.0001), and by clinically derived geriatric criteria in predicting outmortality (P < 0.0001). Multivariate analyses revealed comes of elderly hospitalized patients. that the clinical groups were more predictive of mortality Design: Prospective cohort study of 985 patients and nursing home utilization than were age or Diagnoscreened at the time of hospital admission and followed sis-Related Groups CDRGs). Rehospitalization was unrefor 1 year with respect to the outcomes of mortality, lated to age, clinical group, or DRG, suggesting thaf utilization may not be driven by the clinical factors hospital readmission, and nursing home utilization. measured in this study. Setting: Palo Alto Veterans Affairs Medical Center, a Conclusions: Rapid clinical screening using specific tertiary care teaching hospital. geriatric criteria is effective in identifying frail older subSubjects: Male patients 65 years of age and older admitted to the Medical and Surgical services during the jects at risk for mortality and nursing home utilization. period from October 1, 1985 through September 30, 1986. Our findings suggest that geriatric syndromes are more predictive of adverse outcomes than diagnosis per sc. Results: Patients were grouped by specific screening This well operationalized screening process is inexpencriteria into three groups of increasing frailty: lndependsive as well as effective and could easily be introduced ent, Frail, and Severely Impaired. Each criterion focused into other hospital settings. J Am Geriatr SOC39:778on a geriatric condition and was designed to serve as a 784, 1991. marker for frailty. lncreasing frailty was significantly correlated with increasing length of hospital stay
tudies of multidisciplinary geriatric interventions have had mixed results. Successful trials have been conducted in subacute geriatric evaluation or rehabilitation units,',2 in hospitals using geriatric consultation teams,3or in outpatient settings.*Improved survival, functional status, and mo-
S
rale have been the principal benefits. However, other studes have not found geriatric intervention effect i ~ e . One ~ - ~ reason often cited for the ineffectiveness of these trials has been failure to target the intervention toward those individuals most likely to benefit, the frail elderly.' To date, however, little research has been
* Associate Professor of Medicine, Chief of Clinical Programs in Geriatric Medicine, and Assistant Director, Division of Endocrinology, Gerontology, and Metabolism, Department of Medicine, Stanford University School of Medicine; and Director of Clinical Activities, Geriatric Research, Education and Clinical Center, Veterans Affairs Medical Center, Palo Alto, California. t Assistant Professor of Medicine, Division of Geriatrics and Gerontology and American Federation of Aging ResearchIMerck Foundation Scholar in Geriatric Clinical Pharmacology, University of Texas Health Sciences Center at San Antonio and the Audie L. Murphy Memorial Veterans Hospital, San Antonio, Texas. $ Geriatric Fellow at the Geriatric Research, Education and Clinical Center, Palo Alto Veterans Affairs Medical Center, Palo Alto, CA at the time of this research, now a Clinical Instructor, Division of Geriatric Medicine, Department of Medicine, University of British Columbia, University Hospital, Vancouver, Canada. §Clinical Assistant Professor of Medicine and Director of Graduate Medical Education, Division of Endocrinology, Gerontology, and
Metabolism, Stanford University, and Staff Physician, Geriatric Research, Education and Clinical Center, Palo Alto Veterans Affairs Medical Center, Palo Alto, California. 1) Research Health Scientist at the Geriatric Research, Education and Clinical Center at the time of this research, now a Resource Specialist at the Director's Office, Palo Alto Veterans Affairs h4edical Center, Palo Alto, California. 11 Senior Research Scientist, Software Research Department, Ashton Tate Corporation, Sunnyvale, California. This work was supported in part by National lnstitute on Aging Clinical lnvestigator Award No. KO8 AGO0246 (to CHW), by The Henry J. Kaiser Family Foundation Grant No. 86-3863, and the Veterans Affairs Health Services Research and Development Administration Grant No. I1R 84-062. Address correspondence to Dr. Winograd, GRECC 182-B, Veterans Affairs Medical Center, 3801 Miranda Avenue, Palo Alto, CA 94304.
0002-8614/91/$3.50
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
IAGS-AUGUST 1991-VOL. 39, NO. 8
779
geriatric consultation (short stay), who lived at too great a distance for follow-up (geographic exclusions), who were already enrolled in an interdisciplinary geriatric or rehabilitation program (geriatric services), and who were admitted from a nursing home (nursing home residents). The second step of the screening process used clinical geriatric criteria. The clinical categories defined a prion‘ were: ( 1 ) Independent, (2) Frail and (3) Severely Impaired. To be classified as Independent, a subject had to be independent in all Activities of Daily Living (ADLs) with short-term acute illness, eg pneumococcal pneumonia. Individuals classified as Frail met any one of the clinical criteria described in Table 1. Each criterion focused on a geriatric condition and was designed to serve as a marker for frailty, eg dependence in ADLs, incontinence, impaired mobility, or disabling chronic illness. Severely Impaired subjects were those who METHODS were terminally ill or who had severe dementia and All patients aged 65 years of age and older admitted ADL dependence. These clinical groups correspond to to the Medical and Surgical services of the Palo Alto the previously described’ groups as follows: IndependVeterans Affairs Medical Center (PAVAMC) between ent, Frail, and Severely Impaired correspond respecOctober 1, 1985 and September 30, 1986 were screened tively to Too Well, Appropriate, and Too Sick, respecwithin 96 hours of hospital admission. The PAVAMC tively. The Frail group were the subjects randomized is a tertiary care teaching hospital affiliated with Stan- into the trial of geriatric consultation. Because the ford University. For individuals who were hospitalized consultation process might have produced a change more than once during this period, the first admission within the experimental and control groups, differences was designated the index hospitalization for data with respect to each of the outcomes were analyzed. No difference occurred between experimental and conanalysis. All screening was performed by trained research trol groups with respect to mortality (P = 0.43), hosassistants using a standardized data collection form. pitalization ( P = 0.63), or nursing home utilization Details of the screening process have been published ( P = 0.30). Therefore these groups were included topreviously’ and are illustrated schematically in Figure gether as the Frail group in further analysis. Because all patients admitted to the hospital were to 1. Screening took place in a two-step process. Briefly, the first step of the screening process was administrative and excluded those patients whose expected length TABLE 1. CLINICAL TARGETING CRITERIA of stay was too short to permit an interdisciplinary
done to define targeting criteria to be used to identify the frail elderly in hospital settings. We report the use of a simple clinical screening process with specific geriatric criteria to separate the hospitalized elderly into clinical groups of increasing frailty. We then report the effectiveness of these clinical groups in predicting the negative outcomes of mortality, hospital readmission, and nursing home utilization. The screening criteria for frailty, derived from review of the literature and from clinical experience, were designed to identify subjects for a randomized controlled trial (RCT) of geriatric consultation. The major goals of the RCT were to improve functional status and survival and decrease hospital and nursing home utilization. Our object was to stratify hospitalized patients into groups with increasing risk for these outcomes in order to target the intervention to those at high risk.
-
Admissions 1985 1986
administrative screening
I
clinical screening
t
Geaeraohv n=211
n = 52 n = 24
t
lndeandcnt
:1
n = 251
n = 109 verelv lmoairep n=41
FIGURE 1. Clinical targeting process.
Independent Independent in all ADL’s with short term acute illness Frail (meets any one of the following criteria) Cerebrovascular accident Chronic and disabling illness Confusion Dependence in ADL‘s Depression Falls Impaired mobility Incontinence Malnutrition Polypharmacy Pressure sore Prolonged bedrest Restraints Sensory impairment Socioeconomic/family problems Severely impaired Severe dementia and ADL dependence. Terminal illness. Adapted from Winograd et al, 1988
780
WINOGRAD ET AL
IAGS-AUGUST 1991-VOL. 39, NO, 8
be screened, the screening process had to be rapid and diagnosis responsible for the length of stay. The five inexpensive. Sources of data were arranged in a hier- groups were: surgery, oncology, cardiovascular, pularchical manner to facilitate speed of collection and monary, and any other medical DRG. These five DRG access to information. In order of first use to last use, groups were chosen because prior research has shown the sources of data for the screening process were: these diagnostic groups to have prognostic significance medical record, nursing staff, primary care physician, for response to geriatric To allow sufpatient or the patient’s family. Although performance ficient numbers of subjects in each group, the five DRG on standardized scales such as Activities of Daily Liv- groups were then collapsed into three DRG groups. ing, Mini-Mental State, and Caregiver Burden have The other medical, cardiovascular, and pulmonary been found to be associated with mortality, length of groups were combined into a single “medical” DRG. hospital stay, and nursing home these Before combining these DRG groups, differences with data were not readily available in the medical records. respect to each outcome were examined. The three To administer such scales would have reduced the combined diagnostic categories were not significantly efficiency and clinical utility of the screening process. different from each other with respect to mortality Thus, we did not include such formal measures but ( P = 0.48), hospitalization (P = 0.6), or nursing home rather concentrated on information readily available to utilization (P = 0.68). Subsequent analyses are prethe clinician. All questions about assignment of subjects sented using the three collapsed DRG groups of Surto a category were adjudicated by one of the co- gery, Medicine, and Oncology. investigators (CHW or MBG). Inter-rater reliability was We also wished to determine whether the adminisdetermined with a sample of 53 admissions which were trative step of the screening process, which excluded screened independently by the project coordinator and patients for geographical reasons and because of exthe research assistant during a 1-week period. One- pected short lengths of stay, had not excluded patients hundred percent agreement was achieved for the ad- who were frail and at increased risk for the negative ministrative screening, and 98% agreement was outcomes. When the screening process took place, our achieved in separating patients into the clinical cate- assumption was that the Short Stay and Geography gories. groups, because they were admitted for diagnostic These subjects were followed for 1 year from the procedures and/or were able to come and go from long date of admission of the index hospitalization. Data distances, would have similar mortality and health care were collected for the following outcome measures: utilization to the Independent group. We also predicted length of stay (LOS) of the index hospitalization, num- that the geriatric services group would have similar ber and length of subsequent hospitalizations, number mortality and health care utilization when compared and length of nursing home admissions, and mortality. to the Frail group. All hospital and nursing home utilization data reflect Data Analysis only utilization of VA hospitals, VA Nursing Home Differences among the clinical and administrative Care Units (NHCUs), or community nursing home placement at VA expense. Hospitalization data were groups with respect to length of stay and hospital obtained from the Patient Treatment File (PTF), the mortality were performed using analysis of variance centralized, computerized inpatient data base for the and chi square statistics. Univariate associations of VA. Nursing home utilization data were obtained from Clinical group and DRG groups with survival were the PTF and the records of the Community Nursing examined using Kaplan-Meier survival analysis. lndeHome Program at the Palo Alto VA. Mortality data pendent predictive associations of age, Clinical, and were obtained from the California Automated Mortal- DRG group with mortality were determined using mulity Listings (CAMLIS),l3 which accurately records all tivariate analysis with the Cox proportional hazards deaths of California residents. Of the 985 patients model. To compare the hospital and nursing home admitted during the year, partial data for 14 (1.4%) utilization in the Clinical and DRG groups, utilization were unavailable due to mismatch of social security was converted into a logical variable. Utilization was defined as either no utilization or any utilization, and numbers and/or names with admission dates. To examine the predictive value of (DRGs) and to chi square analysis was used to determine differences. compare the relative efficacy of the DRG groups and Logistic regression was used to determine the relative the clinical screening groups (ie diagnosis versus ger- contribution of age, Clinical, or DRG group to utilizaiatric syndromes) in predicting the outcomes, all pa- tion. tients were classified into one of five major diagnostic RESULTS groups using .the VA DRG designation from the index There were 985 patients over the age of 65 admitted hospitalization. Though the DRG classification system used by the VA is similar to that of Medicare, the final to the PAVAMC during the study period. Of these 584 DRG classification in the VA is based on the principal (59%) were screened out in the administrative screen-
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
/AGS-AUGUST 1991-VOL. 39, NO. 8
ing process, and 401 (41%) remained in the group that was clinically screened (Figure 1). Baseline characteristics of all administratively and clinically screened patients are presented in Table 2. Data regarding outcomes were unavailable for 3 subjects. Therefore, subsequent tables present outcome data for 398 patients. No significant difference in age occurred between the administrative and the clinically screened patients. Significant differences in age existed within each group, however. In the administrative group those receiving geriatric services were older with a mean age of 78.9 compared to 71.9 overall (P < 0.0001). In the clinically screened group, those classified as Frail were also significantly older, with a mean age of 75.2 compared to 71.6 overall (P < 0.0001). Of interest was that 68% of those aged 65-79 fell into the Independent category. Above the age of 79 this proportion fell to 25%. The administrative and clinically screened groups had similar proportions of admissions to Medical and Surgical services (46% vs 44% medicine and 54% vs 56% surgery, respectively). Table 2 also presents the mean length of stay (LOS) for the index admission for both the administrative and the clinically screened groups. Of note is that the average length of stay is more than 10 days in all groups except the short stay. This reflects the longer average length of stay in VA hospitals compared with non-VA h0spita1s.l~Differences in LOS for the index admission among the clinically screened groups are large and significant (P < O.OOOl), with the LOS for the Independent group averaging much less than the other two groups. Table 3 presents nursing home and hospital utilization data for the entire follow-up year. Nursing home utilization was significantly different among the clinically screened groups (P < 0.0001). In the year following the index admission, 13 % of the clinically screened subjects were admitted to a nursing home. Only 3% of the Independent Group were institutionalized, comTABLE 2. BASELINE CHARACTERISTICS OF SCREENED PATIENTS Index LOS* n (%I Age* (days) Administrative Groups Short Stay Geography Geriatric services NH residents Total Clinical Groups Independent Frail Severely impaired Total * P < 0.0001
297 (51) 211 (36) 52 (9) 24 (4) 584 (100)
71.1 (65-93) 71.1 (65-93) 78.9 (65-93) 72.8 (65-93) 71.9 (65-93)
5.1 16.7 11.7 12.9 10.2
251 (63) 109 (27) 41 (10) 401 (100)
70.1 (65-94) 75.2 (65-106) 71.6 (65-88) 71.6 165-106)
12.9 24.8 20.7 16.9
781
TABLE 3. NURSING HOME AND HOSPITAL UTILIZATION IN YEAR FOLLOWING INDEX ADMISSION (FOR SURVIVORS OF INDEX HOSPITALIZATION) Admitted to Readmitted to Survivors Nursing Home* Hospital** n = 357 n ('70) n (%) Independent Frail Severely impaired Total * P < 0.0001. ** P = 0.34.
6 (3) 31 (34) 10 (42) 47 (13)
120 (49) 52 (58) 11 (46) 183 (51)
pared with 34% of the Frail group and 42% of the Severely Impaired group. Multivariate analysis using age, DRG group, and Clinical group as predictors of nursing home utilization in the year following index admission revealed that only the Clinical groups were independently associated with nursing home utilization (P < 0.0001). Calculation of likelihood ratios revealed that among the three clinical groups, the relative likelihood of using a nursing home vaned widely. When compared with the Independent group, the Frail group was 20.75 times and the Severely Impaired group was 28.71 times more likely to enter a nursing home in the year following index admission. Not shown in Table 3, discharge to a nursing home after the index hospitalization was highest among the Severely Impaired group (10.3%)as compared with the Independent (0.4%) and Frail (2.8%) groups (P < 0.0002). In the year following the index admission, 51% of survivors of the index hospitalization were rehospitalized. There was an average of 0.8 hospitalizations (range 0-5, SD 1.03) and 10.8 hospital days (range 0125, SD 20.5). No significant differences occurred in numbers or days of hospitalization among the Clinical Groups or the DRG groups (P = 0.34). Survival curves for the Clinical Groups and the DRG groups are presented in Figure 2. These curves show highly statistically significant differences among the three Clinical groups and the three DRG groups. Survival at one year was inversely associated with increasing frailty, with 87% alive in the Independent group, 55% in the Frail group, and 26% in the Severely Impaired group (P < 0.0001). Survival curves shown for the DRG groups are limited to the 398 patients for whom clinical screening information was available since we wished to compare the relative predictive value of these two classification schemes for mortality. In the three DRG groups survival at 1 year was the highest in the surgical group, with 85% survival, intermediate for the medical group with 67%, and lowest for the oncology group with 41% (P = 0.0001). Within each DRG group, mortality increased with
782
WINOGRAD ET AL
FIGURE
2. Survival
curves
IAGS-AUGUST 1991-VOL. 39, NO. 8
87% Independent (n=218) 85%Surgery (n=148)
for
screened patients by clinical and DRG group. The two sets of survival curves, solid for clinical group and beaded for DRG group, represent the same group of patients. For each of the two sets of curves, the differences between curves are significant with P < 0.0001.
67% Medicine ( ~ 1 % )
55% Frail (n=60) @
40'
41%Oncology (n=16)
20 *
26% Severely Impaired (n=IO)
TABLE 4. TOTAL MORTALITY BY CLINICAL AND DRG GROUP Clinical Groups Independent Frail Severely Impaired n = 398 n (%I n (%I n (%)
Surgery Medicine Oncology Total
13 (10) 16 (17) 3 (21) 32 (13)
11 (36) 32 (47) 6 (60) 49 (45)
2 (100) 13 (59) 14 (93) 29 (74)
Total n (%) 26 (15) 61 (33) 23 (59) 110 (28)
TABLE 5. RESULTS OF MULTIVARIATE ANALYSIS FOR SURVIVAL 95% Confidence Selected Variable Coefficient @ Improvement x2 Interval of j3 Relative Risk
Independent Severely impaired Surgery Medicine
-1.38 0.61 -.99 -.55
76.88 11.90 4.99 4.32
(-1.84, -.42) (.12, 1.1) (-1.61, -.37) (-1.05, -.05)
0.25 1.84 0.37 0.37
For the multivariate analyses, the Frail group was assigned a relative risk of 1.0 for comparison w i t h t h e Independent and Severely Impaired groups. Similarly, the Oncology group was assigned a relative risk of 1.0 for comparision with the Surgery and Medicine groups. These assignments art' for purposes of comparison only. X(t;z) = Xn(t)exp(-l.38 Well + .61 Sick - .99 Surgery - .55 Medicine).
increasing frailty as indicated by the clinical group (Table 4). Moving from left to right in the table, one can see a clear increase in mortality with increasing frailty. Further, multivariate analysis was performed (Table 5 ) using age, Clinical group, and DRG. Age was not a significant predictor of mortality, but the clinical and DRG groups all made independent significant contributions to mortality. The chi square statistic for each term in the model shows that the clinical groups were much more important in the prediction of mortality than were the DRG groups. To answer the question of whether individuals at high risk for mortality, nursing home placement, or hospitalization had been screened out during the administrative phase of screening, two sets of comparisons were made. The Independent, Short Stay, and Geography groups were compared and found to have no significant differences in mortality or rehospitalization. More nursing home admissions occurred in the Short Stay group (11.6%) than in the Geography (5%) or Independent group (3%) (P < 0.001). Similarly, when the Frail and Geriatric Services groups were compared with each other, no significant differences
in mortality or rehospitalization were found. More nursing home admissions occurred in the Geriatric Services group than the Frail group (51% vs 34%, P < 0.05).
DISCUSSION The clinical screening criteria were developed to select individuals who were at high risk for mortality, nursing home placement, and rehospitalization. Each criterion focused on a geriatric condition and was intended to serve as a marker for frailty. This study was undertaken to test our predictions. Our predictions were largely correct. Using 1-year follow-up data, we were able to demonstrate that mortality and nursing home utilization were strongly predicted by the clinical group assigned using the geriatric criteria. Indeed, the clinical group was a better predictor of mortality and nursing home utilization than was either age or DRG group. This finding suggests that geriatric syndromes are more predictive of adverse outcomes than diagnosis per se. Rubenstein's earlier work" also classified hospital inpatients into one of five clinical categories: terminal,
IAGS-AUGUST 1991-VOL. 39, NO. 8
SCREENING FOR FRAILTY AND PREDICTORS OF OUTCOMES
acute medical illness, demented, GEU candidate, and independent. He was able to show that this clinical classification separated hospitalized inpatients into groups with differing mortality and utilization of nursing homes and hospitals. His study, however, was designed to select medically stable individuals with geriatric conditions for treatment in a geriatric evaluation unit. Our purpose was to select acutely ill patients who were frail and at high risk for death or nursing home placement. An unexpected finding was that hospital utilization was not associated with Clinical Group, age, or DRG. This was particularly surprising in view of our finding of association with mortality and nursing home placement. Rehospitalization rates were high in all subjects but were not different in subjects who were frailest. Prior studies have not focused on hospital utilization as an outcome, but careful reading reveals that our hospital admission cohort was rehospitalized with a frequency similar to those in previous studies of geriatric 17, Hospitalization does not appear to be influenced by the functional and clinical factors measured in this study. This suggests that other factors must drive hospital utilization. Possibilities include other unmeasured clinical variables, financial, fiscal, family, or environmental factors. Our original assumption was that patients who were hospitalized for short stays for diagnostic procedures (Short Stay) or those who resided a considerable distance from the medical center (Geography) were likely to be the most similar to those in the Independent group. This was the case. The short stay and the geographically excluded groups did have the lowest mortality. Of interest was the fact that more nursing home utilization occurred then expected in the short stay group. These individuals did not undergo the clinical screening meant to detect frailty, and short stay patients may have actually been hospitalized for chemotherapy or diagnostic procedures, thereby representing a group with heavy disease burden or cancer. This may account for the nursing home admissions. Most important, we found that this simple screening process could distinguish the elderly at high risk for mortality and nursing home placement from those at low risk. Our screening process was simple, clearly defined, inexpensive, and used readily available sources of information. The person who screened the patients was an individual with no prior medical training, trained by our research team. Each patient took an average of 10 minutes for screening. All medical, surgical, and neurological admissions were screened by this person on a daily basis using less than 0.25 Full Time Employee Equivalent (FTEE). There are several limitations to our findings, however. It should be pointed out that the health care utilization data is restricted to services provided by the
783
VA. Hospital utilization refers only to Veterans Affairs hospitals, and the nursing home data encompasses VA Nursing Home Care Units or VA Community Based Contract Nursing Homes; placement in nursing homes at private expense is not detailed in this study. Thus, these health care utilization data are limited, but we have no reason to believe they should be different among groups. As part of the Randomized Controlled Trial of geriatric consultation, we attempted to determine the extent of non-VA health care utilization in the Frail cohort. We interviewed 69 elderly subjects in the Frail group at 2 weeks, 3 months, 6 months, and 1 year post-discharge from the index admission. Utilization of a non-VA emergency room, nursing home, or hospital averaged across the time intervals, 7%, 5%, and 9%, respectively. The low utilization of non-VA services in the Frail group suggests these data appear to be relatively complete even though they should be considered lower bound estimates of health care utilization. This well operationalized, simple, brief screening process, using specific clinical geriatric criteria to identify patients most at risk for adverse outcomes, is inexpensive and effective and may be of great value for discharge planners, hospital administrators, and primary care physicians. In addition to providing prognostic information, this strategy may identify individuals in need of specialized geriatric services or discharge planning. Additional research is under way to define the discriminant criteria, prospectively validate this approach, and compare its utility with functional status measures such as Activities of Daily Living. Although these criteria have only been tested in the hospitalized elderly, it is likely that a modified version of our screening process could be developed for the outpatient or nursing home setting. Separation of patients into groups with prognostic significance for mortality and nursing home utilization may allow intervention strategies to be developed to alter these outcomes, and to compare efficacy of interventions for groups at high and low risk. ACKNOWLEDGMENTS The authors wish to thank Lincoln Moses, PhD for his critical review and suggestions, Michael R. Tuley, PhD for his analytical assistance, and Cindy Steams for her assistance in the preparation of this manuscript.
REFERENCES Rubenstein LZ, Josephson KR, Wieland GD et al. Effectiveness of a geriatric evaluation unit. A randomized clinical trial. N Eng J Med 1984;311:1664. Applegate WB, Miller ST, Graney MJ et al. A randomized, controlled trial of a geriatric assessment unit in a community rehabilitation hospital. N Eng J Med 1990;322:1572. Hogan DB, Fox RA, Badley BWD et al. Effect of a geriatric consultation service on management of patients in an acute care hospital. Can Med Assoc J 1987;136:713.
784
WINOGRAD ET AL
IAGS-AUGUST 1991-VOL. 39, NO. 8
4. Williams ME, Williams TF, Zimmer JG et a]. How does the team hospitals and homes: a survey and follow-up. Public Health approach to outpatient geriatric evaluation compare with tradi1984;98:270. tional care: a report of a randomized controlled trail. J Am Geriatr 12. Wachtel TJ, Derby C, Fulton JP. Predicting the outcome of SOC1987;35:1071. hospitalization for elderly persons: Home versus nursing home, 5. Gayton D, Wood-Dauphinee S, de Lorimer M et al. Trial of a South Med J 1984;77:1283. geriatric consultation team in an acute care hospital. J Am Geriatr 13. Arellano MG, Petersen GR, Petitti DB et al. The California SOC1987;35:726. automated mortality linkage system (CAMLIS). Am J Public 6. Campion EW, Jette A, Beckman B. An interdisciplinary geriatrics Health 1984;74:1324. consultation service: a controlled trial. J AM Geriatr SOC1983; 14. Rubenstein LZ, Wieland GD, Josephson KR et al. Improved 31 :792. survival for frail elderly inpatients on a geriatric evaluation unit 7. Becker AM, McVey LJ, Saltz CC et al. Hospital-acquired compli(GEU): who benefits? J Clin Epidemiol 1988;41:441. cations in a randomized controlled clinical trial of a geriatric 15. Randall M, Kilpatrick KE, Pendergast JF et al. Differences in consultation team. JAMA 1987;257:2313. patient characteristics between Veterans Administration and community hospitals. Med Care 1987;125:1099. 8. Rubenstein LZ. Documenting impacts of geriatric consultation. J 16. Rubenstein LZ, Josephson KR, Wieland GD et al. Differential Am Geriatr SOC1987;35:829. 9. Winograd CH, Gerety MB, Brown E et al. Targeting the hospiprognosis and utilization patterns and clinical subgroups of talized elderly for geriatric consultation. J Am Geriatr SOC hospitalized geriatric patients. Health Serv Res 1986;20(11):882. 17. Fethke CC, Smith IM, Johnson N. "Risk" factors affecting read1988;36:1113. 10. Narain P, Rubenstein LZ, Wieland GD et al. Predictors of immission of the elderly into the health care system. Med Care mediate and 6-month outcomes in hospitalized elderly patients. 1986;24:429. The importance of functional status. J Am Geriatr SOC 18. Saltz CC, McVey LJ, Becker PM et al. Impact of a geriatric 1988;36:775. consultation team on discharge placement and repeat hospitali1 1 . Donaldson LJ. Outcome of admissions of elderly people to zation. Gerontol 1989;28:344.
An incorrect version of the article Two-Year Trends in Physical Performance following Supervised Exercise among Community-Dwelling Older Veterans was published in the June 1991 issue of the Journal of the American Geriatric Society. The correct version will appear in the October 1991 issue. The Journal regrets the error.
