The purpose of this study was to determine to what extent a single measure, self-rated health (SRH), independently predicts long-term hospitalizations due to all causes and to cardiovascular diseases by using both the standard Cox proportional hazards model and a more robust events model. The study cohort consisted of 2,812 elderly subjects residing in New Haven, Connecticut, who were followed from 1982 to 1996 as part of the Established Populations for Epidemiologic Study of the Elderly. After adjustment for baseline risk factors, using the Cox model, a favorable SRH was associated with a significantly lowered risk for a first hospitalization for all causes (risk ratio (RR) = 0.850, 95% confidence interval (CI): 0.774, 0.934) and congestive heart failure (RR = 0.599, 95% CI: 0.426, 0.841) but not for myocardial infarction (RR = 0.882, 95% CI: 0.565, 1.379). With the adjusted robust events model, a positive SRH was associated with a decreased risk in both a first (RR = 0.813, 95% CI: 0.744, 0.889) and a second (RR = 0.870, 95% CI: 0.782, 0.968) hospitalization for any cause. These results indicate that a single measurement of SRH predicts long-term patterns of hospitalization, especially for heart failure, among older adults.