the purpose of this study was to develop (phase I) and validate (phase II) a mortality prognostic index, based on the annual clinical data base, for the men of this Veterans Administration extended care facility. The study population during phase I consisted of 123 men who were residing in three of the seven wards of the facility in August 1984. Sixty-six of these individuals were institutionalized because of a chronic neurologic (50) or medical (15) disorder ("nonpsychiatric group"). In 57 men, the reason for institutionalization was a chronic psychosis (schizophrenia, 53; manic depressive illness, 4) ("psychiatric group"). During August to October 1984, a comprehensive clinical data base comprising 70 attributes (including diagnoses and drugs) was collected. Deaths were recorded during the next 14 months. Death rate during the 14 months of observation was 33.3% in the nonpsychiatric group, and only 1.7% in the psychiatric group. In the nonpsychiatric men, univariate analysis yielded six attributes significantly correlated with death rate: serum cholesterol level, hematocrit, hemoglobin, midarm muscle circumference, triceps skinfold, and number of morbidity episodes. After serum cholesterol and hematocrit had been entered into a multivariate analysis model, none of the other four attributes contributed significant information about death rate. The multivariate analysis led to a mortality risk index (MRI) for nonpsychiatric patients, MRI = [hematocrit in %] + 10% [serum cholesterol in mg/dl]. As MRI varied from less than 50 to greater than 65, death rate in the nonpsychiatric group varied in parallel from 86-11%.(ABSTRACT TRUNCATED AT 250 WORDS)