Introduction Self-reported health is a perception based on an individual interpretation of physical aspects, mental status and expectations and is considered a good predictor of mortality among old people in developed countries. The aim of this study was to evaluate the role of self-reported health as a predictor of mortality in elderly men in a developing country.

Methods The study population consisted of 2875 elderly men of a medium size city in Southeast Brazil, who were followed for 4 years or until the date of their death, whichever occurred first. Individuals alive at the end of follow-up were censored. Multivariate analysis was performed through Cox regression models. Variables presenting statistically significant associations with mortality in bivariate analysis where entered into the models.

Results During the follow-up, 298 deaths occurred. Elders with self-reported poor health presented a greater risk of death, compared to those with self-reported good/excellent health in almost all stratus of the analysed variables. In the final model, poor self-reported health (HR 1.54 95% CI 1.21 to 1.96), age (HR 1.07 95% CI 1.06 to 1.09), marital status (HR 1.32 95% CI 1.04 to 1.69), current use of cigarette smoking (HR 1.94 95% CI 1.24 to 2.62), cardiovascular disease (HR 1.62 95% CI 1.06 to 2.47), diabetes (HR 1.53 95% CI 1.14 to 2.04) and recent hospitalisation (HR 1.50 95% CI 1.15 to 1.95) were independently associated to mortality.

Conclusion Self-reported health was a good predictor of mortality in this population of elderly men, even when adjusted for other independent variables. It is important that healthcare services incorporate this indicator in the health evaluation of old people.