Background Published evidence on self-rated health’s capacity to predict mortality and its variability across subpopulations lacks consistency. Our objective is to evaluate this predictive association and whether/how it varies by sex, age and educational attainment at the population level in Spain.

Methods Data came from a prospective longitudinal study based on 42 645 individuals aged ≥15 years who participated in the 2011–2012 and 2014 Spanish Health Surveys. Median follow-up time for mortality was 5.4 years. Cox proportional hazards models adjusted for sociodemographic, lifestyle and chronic disease variables were used to estimate the predictive capacity of self-rated health on mortality.

Results Self-rated health was associated with mortality with a dose–response effect (p value for linear trend <0.001). Compared with respondents who rated their health as very good, those rating it as very poor presented an HR of 3.33 (95% CI 2.50 to 4.44). Suboptimal self-rated health was a stronger predictor of mortality among 15–44 year-olds (HR 2.87; 95% CI 1.59 to 5.18), compared with the estimate for 45–64 year-olds (HR 1.86; 95% CI 1.45 to 2.39) (p value for interaction=0.001) and for those 65 and older (HR 1.51; 95% CI 1.36 to 1.68) (p value for interaction <0.001). Regarding educational attainment, the association was stronger for individuals with university studies (HR 2.51; 95% CI 1.67 to 3.76) than for those with only primary or no studies (HR 1.31; 95% CI 1.17 to 1.48) (p value for interaction=0.010). No statistically significant differences were observed between men and women.

Conclusions Self-rated health may be considered a good predictor of all-cause mortality in the population of Spain, although the magnitude of this predictive association varies by age and educational level.