Using data from the 1987 National Medical Expenditure Survey, a representative sample of US civilians, and their 5-year mortality, we examined the adjusted relationships among baseline self-reported health, derived from SF-20 subscales (health perceptions, physical function, role function and mental health) and sociodemographics (age, sex, race/ethnicity, income and education) and subsequent mortality. Included were 21,363 persons aged 21 and over, with complete follow-up on 19,812. Physical function showed the greatest decline with age, whereas mental health increased slightly. Women reported lower health for all scales except role function. Greater income was associated with better health, least marked for mental health. Greater education was associated with better health, most marked for health perceptions. Compared with whites, blacks reported lower health, whereas Latinos reported higher health. Lower self-reported health predicted increased adjusted mortality. After adjustment for baseline self-rated health, the relationships between income and education and mortality were greatly attenuated, whereas the relationships between age, gender, race/ethnicity and mortality were not. Self-rated health exhibited more profound relationships with mortality in younger persons, those with more education, and whites. In conclusion, lower socioeconomic status (SES), and being black are associated with lower reported health status and higher mortality; women report lower health status but exhibit lower mortality; and Latinos report higher health status and exhibit lower mortality. The effects of SES on mortality are largely explained by their associations with self-rated health, whereas, the effects of gender and race/ethnicity on mortality appear to act through independent pathways. Because of these differential sociodemographic relationships caution is urged when using self-rated health measures in research, clinical, and policy settings.