Using admission characteristics to predict return to the community from a post-acute geriatric evaluation and management unit

OBJECTIVE: To compare the Cumulative Illness Rating Scale (CIRS) and the Nu rsing Severity Index (NSI) as independent predictors of discharge outcome f rom a post-acute GEM unit and to define a multivariate model for predicting the same outcome. DESIGN: Retrospective chart review for the entire sample. The sample was sp lit into two cohorts, a derivation cohort (n = 298) and a validation cohort (n = 154). SETTING: A 20-bed, post-acute GEM unit in a nonproprietary skil led nursing facility. PARTICIPANTS: All 452 patients admitted to the GEM from the unit's inceptio n in December 1994 until January 1998. MEASUREMENT: Demographics, CIRS, NSI, functional status, and social support variables were measured using data available on admission to the GEM unit. The discharge outcome was dichotomized as return to the community or not. RESULTS: A total of 99.7% of the individuals in the derivation cohort were living in the community before the index hospitalization; 75.8% of patients in the derivation cohort returned to the community. The NSI, individual "s evere" items from the CIRS, age, and social support were in the final logis tic regression model fitted to the derivation cohort. A total of 87.7% of t he observed discharge outcomes were predicted when the model was applied to the validation cohort and the calculated probability of return to the comm unity. CONCLUSIONS: Variables for severity of illness, function, social support, a nd age combined into a logistic regression equation that predicted more tha n 80% of the dichotomized discharge outcome in the derivation cohort. The p roportion of discharge outcomes that were predicted with the validation coh ort remained high at 87.7%. The NSI and CIRS were each important to a model that is anticipated to refine the selection of geriatric patients for post -acute services.