social determinants of health

Robustly Mapping Social Determinants of Health with XY

Our health is highly influenced by social determinants such as where we live, our social standing, and the vocations we are fortunate to have. While intuitively important, these "social determinants of health" are misunderstood, overlooked, and understudied. Why? Social determinants are highly sensitive, personal, and complex. As a result, individuals and populations who are at need are ignored in healthcare research and potentially do not get the interventions they need.

We all know that the nurture side of the "nature vs. nurture" balance has tremendous influence on our well-being. Our physicians know this too, yet it has been difficult to assemble the massive sources of exposure data and social determinants for patients or healthcare companies.

Healthcare lacks large-scale computational methods to sift signal from noise – what social factors matter, and which ones do not -- in populations and individuals in need.

At, we are striving to build integrative computational tools to map the once elusive social determinants to give decision makers data and tools to map health across the population.

What are social determinants of health?

The seminal Healthy People 2020 initiative defines social determinants of health as

"conditions in the environments in which people are born, live, learn, work, play, worship, and age that affect a wide range of health, functioning, and quality-of-life outcomes and risks."

What makes geography and social determinants of health so complex?

Figure 1. XY's Exposome Data Warehouse mapping social determinants of obesity prevalence in Macon, GA

The relationship between social determinants and health is complex. In other words, a host of specific social determinants – not just one – play a role. This is best illustrated using XY's Exposome Data Warehouse visualization of the “diabetes belt”, a region of the country that is a confluence of individuals with lower incomes, lower rates of completion of high school education, higher exposure to pollution, and lower access to health care. Combine these issues with biological-based disparities as we described in a seminal paper in the New England Journal of Medicine and the challenge is made even more difficult to overcome.

Figure 2. XY's Exposome Data Warehouse mapping social determinants of obesity prevalence in Johns Creek, GA

The difference in social determinants of health across different regions of the country is visible in maps of obesity prevalence using data from the Centers for Disease Control and Prevention (CDC) and the United States Census Bureau. Figure 1 shows a map from XY's Exposome Data Warehouse of social determinants of obesity prevalence in Macon, GA, and to the Figure 2 is the same map for Johns Creek, GA, where the obesity prevalence is much lower.

How best to identify the factors that matter? – enter

Beyond visualization of social determinants, is providing transparent methods based on machine learning to help healthcare companies map all social determinants of health with disease and triangulate findings from diverse public resources, especially non-traditional sources of health and disease risk prediction including large-scale satellite imagery (post coming soon!).

Instead of considering potential social determinants one at a time, we strive to develop both algorithms that consider the totality of all complex and dense social determinants to allow healthcare companies to comprehensively map individual contributions and interactions between social determinants of health. Our models are built to operate in a federated manner, preserving the sensitivity of information. You can learn more about the motivation for this work by reading our publications in JAMA and ARPH. We attempt to deliver models that maximize both interpretability and predictive performance while rigorously vetting performance accuracy in external datasets.

Written by Chirag Patel and Arjun Manrai on October 24, 2019