Sources of Variation in U.S. Mortality: A Latent Variable Analysis

Andrew Stokes, University of Pennsylvania
Chris Tencza, University of Pennsylvania

Modifiable risk factors exert a strong influence on health and mortality in the United States. Estimation of the effects of modifiable risk factors is challenging as US mortality data do not provide information on the risk factor(s) that gave rise to a particular disease and surveys are rarely large enough to permit detailed analyses. We explore a new indirect method for identifying and estimating mortality effects of leading risk factors, which relies on vital statistics data. Our basic assumption is that most spatial variation in cause-specific mortality rates is a manifestation of a small number of latent variables, variation in which give rise to the observed mortality patterns. We apply factor analysis to extract the major components of mortality variation. We find that the majority of US mortality variation is explained by a small number of factors and that these factors can be meaningfully interpreted in terms of known modifiable risk factors.