RESUMEN
OBJECTIVES: To identify a set of predictors of mortality among residents in the community, before any physical, biochemical, or image examination is performed, that could be collected on a routine standardized basis, to help the clinician define a patient follow-up strategy and the health planner make decisions regarding the care of older people. DESIGN: A household follow-up study, with an evaluation at baseline and 2 years later. SETTING: Residential area, with a low rate of in- and outmigration, in Sao Paulo, a large industrialized urban center in southeastern Brazil. PARTICIPANTS: One thousand six hundred sixty-seven older urban residents in the community (65+), from different socioeconomic backgrounds, enrolled after a study area census. INTERVENTION: Structured home interview with Brazilian Older Americans Resources and Services Multidimensional Functional Assessment Questionnaire, previously validated in Portuguese. MEASUREMENTS: A logistic regression model for the risk of dying in the period was developed, having as independent variables, sociodemographic characteristics plus six other dimensions: subjective self-evaluation of health, past medical history, use of health services, dependence in activities of daily living (ADLs), mental health, and cognitive status. RESULTS: There were 146 deaths (9%) in the cohort during the follow-up interval. The variables that appeared as independent predictors of death in the final logistic regression model were: gender (relative risk (RR) = 2.8 (males)), age (RR = 2.0 (80+ vs. 65-69)), hospitalization in the previous 6 months (RR = 2.4 (at least one)), dependence in ADLs (RR = 3.0 (assistance required for 7 vs. 0 ADLs)); and cognitive impairment (RR = 1.9 (Mini-Mental State Examination 18 vs. 30)). CONCLUSION: These findings suggest that in developing countries such as Brazil, an assessment of dependence in daily living and cognitive status should be an essential part of any health evaluation of an older person, not only because these variables represent potentially high independent mortality risks, but also because they can be easily and reliably assessed, using well-validated instruments, and may be susceptible to intervention.