What determines health? That question drives all our work, and while we’ve observed a lot of progress in untangling the many answers to that question, we have very few tools for measuring different health effects.

In a recent JAMA Viewpoint, authors Jose Figueroa, Austin Frakt, and Ashish Jha have proposed an approach to address that gap: the polysocial risk score. Akin to the polygenic risk score used in genetics to predict predisposition to certain diseases based on the interaction of many different genes, the “individualized ‘polysocial risk score’ could help predict the risk that varying combinations of social conditions are related to specific health outcomes.”

The authors continue,

If these scores are developed and validated in rigorous studies, such an approach could obviate the need to specifically measure the individual influence of each social factor, an effort that has proven to be far more difficult than anticipated and continues to yield inadequate actionable results.

Polysocial risk scores could give us more a meaningful understanding of how factors like nutrition and substandard housing interact to produce health outcomes rather. Because many factors relate to the same outcome, focusing on a single factor (or a set of single factors, each individually) won’t capture the complexity of the underlying mechanisms.

However, it’s a massive challenge to develop these proposed risk scores. Data would need to be aggregated on a massive scale and would have to include individual factors (e.g., income and education), identity factors (e.g., religion, gender), and social/community/environment factors (e.g., availability of good housing, access to nutritious food, social support, racism). These data would then have to be linked to health outcomes to develop models to predict how these factors affect health status and outcomes.

There are many important considerations when developing a tool like this. Who is represented in the data? How is privacy protected? How would we track change over time, given these factors are dynamic? How would the scores be deployed and how would we ensure they are useful? These are important questions that need careful thought.

Read the full piece here.