Health-related quality-of-life (HRQOL) measures are becoming increasingly important for evaluating the effectiveness of medical interventions and assessing the health of populations. Preference-based instruments, a subset of HRQOL measures, allow comparisons of overall health status in populations and in clinical settings, and are suitable for economic analyses; but validity studies have used selected samples, mostly examining morbidity. This study compared the performance of a preference-based instrument with self-rated health in predicting subsequent self-rated health, hospitalization, and mortality in a national cohort. A version of the Health Utility Index (HUI), constructed from questions in the 1982 to 1984 National Health and Examination Survey I Epidemiologic Follow-up Study (NHEFS), was used to develop scores for the 1982 to 1984 survey sample. The relationship between both the NHEFS-HUI and self-rated health in 1982 to 1984, and subsequent decline in self-rated health, hospitalizations, and mortality experienced by 1987 were examined using survival analyses. The analyses adjusted for sociodemographic variables (age, sex, race, education, and income), medical conditions, and smoking status reported at the 1982 to 1984 NHEFS interview. Results indicated that NHEFS-HUI and self-rated health scores were worse in older persons, persons with one or more medical conditions, African Americans, and those with less education and lower incomes. The effects of all 19 chronic conditions and smoking were reflected in lower self-rated health scores, whereas the NHEFS-HUI did not capture the effects of two of the conditions or smoking status. Both measures made independent contributions to predicting hospitalizations and mortality by 1987; in addition, the NHEFS-HUI predicted decline in subsequent self-rated health. The NHEFS-HUI also predicted health outcomes in the subgroup of those in initial excellent or very good self-rated health. A preference-based instrument demonstrated predictive validity in three relevant domains of health status outcomes across all sociodemographic groups examined in this cohort. Self-rated health was better able to capture concurrent decrements in health associated with certain chronic illnesses and smoking. It is concluded that preference-based measures capturing both functional status and health perceptions should be incorporated explicitly into national surveys to assess the health of populations.