Aligning method with theory: a comparison of two approaches to modeling the social determinants of health

Abstract

There is increasing interest in the study of the social determinants of maternal and child health. While there has been growth in the theory and empirical evidence about social determinants, less attention has been paid to the kind of modeling that should be used to understand the impact of social exposures on well-being. We analyzed data from the nationwide 2006 Canadian Maternity Experiences Survey to compare the pervasive disease-specific model to a model that captures the generalized health impact (GHI) of social exposures, namely low socioeconomic position. The GHI model uses a composite of adverse conditions that stem from low socioeconomic position: adverse birth outcomes, postpartum depression, severe abuse, stressful life events, and hospitalization during pregnancy. Adjusted prevalence ratios and 95% confidence intervals from disease-specific models for low income (<20,000/year) compared to high income (≥ 80,000/year) ranged from a low of 1.43 (1.09-1.85) for adverse birth outcomes to a high of 5.69 (3.59-8.84) for stressful life events. Estimates from the GHI model for experiencing three to five conditions yielded a prevalence ratio of 18.72 (9.29-35.77) and a total population attributable fraction of 78%. While disease-specific models are important for uncovering etiological factors for specific conditions, models that capture GHIs might be an attractive alternative when the focus of interest is on measuring and understanding the myriad consequences of adverse social determinants of health.