Comparison of simple efficient clinical and self-reported predictors of mortality in a National United States Cohort

Abstract

Purpose: To compare the relative value of simple efficient methods for predicting mortality.

Methods: We compared three clinical (blood pressure, blood glucose, and low density lipoprotein cholesterol [LDL] cholesterol) and three self-reported (smoking, educational attainment and self-reported health) prospectively measured predictors of mortality in a cohort of 30,239 white and black adults who were 45 years of age or older at enrollment between 2003-2007. Survival was modeled using proportional hazards analysis and the c-statistic was used to evaluate information provided by each measure.

Results: Information on all variables and follow up was available for 27,482 (91%), and among these, there were 4,409 (16%) deaths over an average of 7.6 years. The clinical measures of blood pressure, blood glucose, and cholesterol were modestly good predictors of short-term survival (for each, p < .01). However, simple one-item self-reports provided better prediction of mortality than the clinical indicators. The Age-Sex Race (ASR) adjusted Hazard Ratios (HR) for self-reported current smoking in contrast to not smoking (2.43, CI: 2.25-2.63) self-rating of health as poor in contrast to excellent (HR = 6.26, CI: 5.42-7.23), and less than high school education versus collage graduation (HR = 2.21, CI: 2.01-2.42) were all highly significant.

Conclusion: Simple one-item self-reports may be undervalued as meaningful predictors of longevity.