Printed in U.S.A Self-perceived Health and 5-Year Mortality Risks among the Elderly in

Studies of the elderly worldwide over the last 3 decades have reported that a self-rating of "poor " compared with "excellent/good " health increases the relative risk of dying. The authors tested the strength of this association by performing age-stratified Cox regression analyses on a 5-year longitudinal study of a repre-sentative sample of noninstitutionalized elderly aged 65 years and older (n = 3,094) in a district of Shanghai, China. More than 20 potential confounders that were only partially controlled in other studies and threats to response validity due to cognitive impairment or diagnosed dementia that were not considered in previous studies were taken into account in this analysis. The results showed that among those aged 65-74 years, "poor " perceived health • increases the adjusted relative risk of death by 1.93 (95 % confidence interval 1.20-3.11) compared with "excellent/good " health. The adjusted relative risk of a "fair " rating of health is 2.16 (95 % confidence interval 1.44-3.25). In the older age group, mortality risks for the ratings of fair as well as poor compared with excellent/good health were not statistically significant. The authors posit that several mech-anisms related to host vulnerability markers and greater-than-expected utilization of health services may explain the association between self-assessed health and mortality risks. Future research should strive to develop more precise measures of these and related variables. Am J Epidemiol 1998;147:880-90. Activities of Daily Living; aged; depression; health status; longevity; mortality; sex; surviva