Five approaches to modeling estimates—and their associated weaknesses

The County Health Rankings provides a unique opportunity to empirically examine the association between health outcomes (i.e., mortality and morbidity) and health factors (i.e., health behaviors, clinical care, social and economic factors, and the physical environment) at the county level. The aim of this study is to provide potential approaches to examine these associations and outline how to further utilize the Rankings in health services research. Data from the 2010 and 2013 County Health Rankings for 3,141 U.S. counties were used. Five methodological approaches were applied to the Rankings to assess the relative contributions of health factors on mortality including OLS Regression modeling, fixed effect (FE) modeling, hierarchical linear modeling (HLM), confirmatory factor analyses (CFA), and structural equation modeling (SEM). Large variations of the resulting estimates are due to the different assumptions of each model's specification. The ability to connect the empirical results to interpretation and policy interventions may depend heavily on which approach is used. Cautions and limitations will be discussed for deciding on the appropriate model specification, given specific research questions.