Self-rated health trajectories and mortality among older adults.

OBJECTIVES: For this article, we evaluated whether measures of prior self-rated health (SRH) trajectories had associations with subsequent mortality that were independent of current SRH assessment and other covariates.

METHODS: We used multivariable logistic regression that incorporated four waves of interview data (1993, 1995, 1998, and 2000) from the Asset and Health Dynamics Among the Oldest Old Survey in order to predict mortality during 2000-2002. We defined prior SRH trajectories for each individual based on the slope estimated from a simple linear regression of their own SRH between 1993 and 1998 and the variance around that slope. In addition to SRH reported in 2000, other covariates included in the mortality models reflected health status, health-related behaviors, and individual resources.

RESULTS: Among the 3,129 respondents in the analytic sample, SRH in 2000 was significantly (p <.0001) associated with mortality, but the measures of prior SRH trajectories were not. Prior SRH trajectory was, however, a significant determinant of current SRH. We observed significant independent associations with mortality for age, sex, education, lung disease, and having ever smoked.

DISCUSSION: Although measures of prior SRH trajectories did not have significant direct associations with mortality, they did have important indirect effects via their influence on current SRH.