Abstract

The authors have studied whether area-level socioeconomic status predicts mortality independently of individual-level socioeconomic status in 179,383 persons in the American Cancer Society Nutrition Cohort, followed for mortality from 1992 to 2000 (17,383 deaths). They used an area-level variable based on census blocks that was an average of home value, income, education, and occupation. Education was the individual-level socioeconomic status variable. The authors studied socioeconomic status-mortality trends with each socioeconomic status variable adjusted for the other. For all causes, an individual's education was strongly and inversely associated with mortality; a weak but significant inverse trend was also present for area-level socioeconomic status. A similar pattern was seen for all-vascular disease. For all cancers, there was again a significant inverse trend with education but no trend with area-level socioeconomic status. Adjustment for conventional (non-socioeconomic status) individual-level risk factors diminished the effect of both socioeconomic status variables, although significant trends remained for men between education and all-cause, all-cancer, and all-vascular disease mortality. Study data indicate that the predictive value of area-level socioeconomic status variables varies by cause of death but is less important than individual-level socioeconomic status variables. Multivariate models that consider socioeconomic status as a potential confounder may not need to consider area-level socioeconomic status if data are available on individual-level education.