A Comparison of Biomarkers across Two Older Populations: Assessment of Consistency and Generalizability

Christopher L Seplaki, Johns Hopkins University
Noreen Goldman, Princeton University
Maxine Weinstein, Georgetown University
Tara Gruenewald, University of California, Los Angeles
Arun Karlamangla, University of California, Los Angeles
Teresa E. Seeman, University of California, Los Angeles

Researchers continue to expand the array of biomarkers of physiological system function collected alongside traditional self-reported measures in population-based social surveys. Still, little is known about the variability of many biomarkers across populations or the consistency of associations between biomarkers and self-reported health measures. We take steps to address this gap using a wide array of traditional clinical and non-traditional biomarkers of physiological system function, together with two self-reported health outcomes (one physical, one cognitive), from two large, comparable, population-based studies of older adults from the U.S. and Taiwan. We provide a comprehensive descriptive assessment of the variability of biomarker distributions across populations, and we test two hypotheses regarding the consistency of 1) associations with the self-reported outcomes and 2) predictive performance both within and out-of-sample. We find mixed support regarding consistency of associations, and evidence suggesting variability in predictive performance.