The present paper explores predictors of all-cause mortality based on a longitudinal multidisciplinary study of 422 community residents (197 men, 225 women) aged 69-71 years at baseline examination. The 422 subjects were followed up for 10 years (1976-1986), during which time 102 (59 men, 43 women) died. The Cox proportional hazards model was used to examine the effects of baseline variables on a time-dependent 10-year mortality. Body mass showed a U-shaped relationship to mortality in both sexes. Ex-drinking, current-smoking, history of stroke, history of diabetes mellitus, low ADL and ST and/or T changes in ECG had significant and direct effects on mortality in both sexes or either sex. Whereas, level of education, current-drinking, grip strength, visual retention, and serum albumin revealed significantly inverse relationships to mortality. © 1992.