OBJECTIVE The purpose of this study was to identify risk factors for stroke and to estimate their relative importance in a large, nationally representative sample of very old men and women. METHODS The study was designed as a secondary analysis of the Longitudinal Study on Aging. Baseline (1984) in-person interview data were linked to Medicare hospitalization records for 1984 to 1991. Participants were 6,071 noninstitutionalized adults 70 years old or older at baseline. Hospitalization for ischemic stroke was defined as having one or more episodes with a primary discharge diagnosis containing ICD-9-CM codes of 433.0-434.9, 436, and 437.0-437.1. Multivariable proportional hazards regression was used to estimate the risks associated with previously identified epidemiologic factors. RESULTS Five hundred and three persons (8.3%) had at least one primary discharge diagnosis of ischemic stroke. In descending order of importance-based on the partial r statistics associated with their adjusted hazards ratios (AHRs), the salient risk factors were having a previous history of stroke (AHR = 2.86), age (AHR = 1.04 per year), diabetes (AHR = 1.78), male gender (AHR = 1.42), lower body limitations (AHR = 1.09 per limitation), arthritis (AHR = 0.74), hypertension (AHR = 1.29), and poverty (AHR = 1.33). CONCLUSIONS Patients presenting with the high risk factors identified in this study should be considered for further evaluation and monitoring. Current protocols for the therapeutic management of these higher risk patients should be considered, and compliance should be encouraged.