Determinants of Health: Testing of a Conceptual Model

Systematic exploration of the relative strength of the many determinants of health requires a framework or model for organizing research endeavors. Despite the 150year-old literature on social determinants of health, there have been few efforts at developing such a model. The Lalonde Report (1974), 1 which popularized a prior conceptualization of health as being determined by biological, environmental, life style, and health care organization, is still cited and even repeated in various incarnations as a prototype, despite the fact that much greater specificity is required if policy decisions are to result from scholarly inquiry. In this presentation we provide an example of the evolution of causal diagrams from the early 1970s to the late 1990s. 2–5 This example shows how discrete lines of scholarly endeavors can shape the way these diagrams are formed. For example, in the 1970s and 1980s Aaron Antonovsky’s work 6 on social coherence influenced the specification of pathways; in the 1990s, social cohesion has replace social coherence, even though it is a variable more related to social context than to individual psychological representation of ‘feelings.’ Very few empirical studies address more than two steps in a postulated pathway. 6,7 More recent models (F IG . 1) explicitly represent the ecological context that influences individuals primarily indirectly through more proximate social and environmental phenomena, and even more proximate individual manifestations of these contextual phenomena. Virtually absent from most of the recent literature is consideration of the contextual variables related to health services organization and delivery and their translation into individual experiences of quality of services received. When these more comprehensive models are used, it becomes evident that no single pathway is responsible for manifestation of a population’s health. Rather, the interaction between a variety of types of domains, including the biological, social, behavioral, environmental, and medical become clear. As noted by Greenland et al., 9 conventional multivariate techniques for assessing the strength of relationships become problematic as the interactions among variables overwhelms the ability of these techniques to accurately characterize the interactions. A simple example of these interactions is provided by the work of Shi et al., 10 who used a path analytic strategy informed by F IGURE 1 to examine the relative impact of the contextual variables of income inequality and the supply of primary care versus specialty physicians in the 50 U.S. states. They found that both income inequality and the supply of primary care physicians directly influenced most of the