OBJECTIVE: To study the relationship between self-rated health (SRH) and prevalence of chronic diseases, and all-cause mortality in the elderly population. METHODS: In 1992, a cohort of 3257 people > or = 55 years old was selected from Beijing, the information of SRH and other related variables were collected from 3 157 subjects at the baseline survey. Three follow-up surveys were conducted in 1994, 1997 and 2000, respectively. RESULTS: The SRH was influenced by age, gender, marriage status and satisfaction on their own economic condition. Comparing the subjects with excellent SRH, the prevalence rates of chronic diseases, stroke, heart diseases and respiratory system diseases were almost doubled among those with average and poor SRH. By 2000, 993 death occurred. All-cause mortality was negatively associated with SRH, i.e. the risk of death was 12% which was 53% higher for the subjects with average SRH (HR = 1.12, 95% CI: 0.93 - 1.35) and poor SRH (HR = 1.53, 95% CI: 1.25 - 1.88) than those with excellent SRH, respectively. The risks of death from stroke and heart disease were 2.25 (HR = 2.25, 95% CI: 1.67 - 3.04) and 2.22 (HR = 2.22, 95% CI: 1.61 - 3.07) times higher among the subjects with poor SRH than those with excellent SRH respectively. After adjustment for age, gender, resident place, marriage status, education, satisfaction on their own economic condition, seeing doctors or hospitalized within the last 1 year, history of chronic disease, cognition function, body mass index, activities of daily living and depression, as well as deleted the subjects died within first or third year of the baseline survey respectively, poor SRH remained a significantly independent predictor to all-cause death as well as to the death of stroke and heart diseases. CONCLUSIONS: The frequency of poor SRH was influenced by age, gender, marriage status and satisfaction on their own economic condition. Poor SRH was associated with the prevalence of chronic conditions and mortality among the elderly. The findings suggested that SRH might have served as an important indicator in the evaluation on health status among the elderly.