Modeling biological age and its link with the aging process.

Differences in health status at older ages are a result of genetic predispositions and physiological responses to exposure accumulation over the lifespan. These vary across individuals and lead to health status heterogeneity as people age. Chronological age (CA) is a standard indicator that reflects overall risks of morbidity and mortality. However, CA is only a crude proxy for individuals' latent physiological deterioration. An alternative to CA is biological age (BA), an indicator of accumulated age-related biological change reflected in markers of major physiological systems. We propose and validate two BA estimators that improve upon existing ones. These estimators (i) are based on a structural equation model (SEM) that represents the relation between BA and CA, (ii) circumvent the need to impose arbitrary assumptions about the relation between CA and BA, and (iii) provide tools to empirically test the validity of assumptions the researcher may wish to invoke. We use the US National Health and Nutrition Examination Survey 1988-1994 and compare results with three commonly used methods to compute BA (principal components-PCA, multiple regression-MLR, and Klemera-Doubal's method-KD). We show that SEM-based estimates of BA differ significantly from those generated by PCA and MLR and are comparable to, but have better predictive power than KD's. The proposed estimators are flexible, allow testing of assumptions about functional forms relating BA and CA, and admit a rich interpretation as indicators of accelerated aging.