The value of self-rated health in predicting mortality and the incidence of chronic diseases was studied in a cohort of 783 elderly Dutch men in the Zutphen Study. In 1985, 48% of the men felt "healthy," while 12% felt "moderately healthy" or "not healthy." As of 1990, 23% of the 783 men had died. Survival analysis showed that self-rated health was highly predictive of subsequent 5-year mortality from all causes (p < 0.001). When adjusted for the presence of major chronic diseases, age, medication use, smoking, alcohol consumption, physical activity, body mass index, systolic blood pressure, serum cholesterol, education, marital status, and family history of chronic diseases, the relative risk for "moderately healthy" or "not healthy" men compared with "healthy" men was 2.7 (95% confidence interval (CI) 1.8-4.3). Analysis of cause-specific mortality revealed that self-rated health was associated with cardiovascular mortality (crude relative risk (RR) = 2.7), but this finding resulted mainly from confounding by baseline prevalence of cardiovascular diseases (adjusted RR = 1.9, 95% CI 0.9-3.8). However, self-rated health was an independent risk factor for cancer mortality (adjusted RR = 4.2, 95% CI 1.9-9.4) and mortality due to other causes (adjusted RR = 3.0, 95% CI 1.2-7.8). Self-rated health did not independently predict the incidence of chronic diseases. This suggests that self-rated health especially affects fatality from chronic diseases rather than their onset, and this issue should be pursued further.