Can predictive models identify risks arising from social determinants of health (SDH)?

Aug. 25, 2017 / By Herb Fillmore

Strong predictive models will surface the key person, place, and system factors in play for individuals and for populations. These models incorporate environmental as well as personal health status variables to identify risk profiles for various outcomes (e.g., costs, readmissions, ER use, disease progression, etc.) Almost every outcome of interest will be influenced by social determinants of health (SDH).

Good models will delineate how social determinants may be interacting with health status, provider quality and experience, and availability/access of services. Today, SDH variables in large scale models are typically reliant on public data and imputing SDH on a “place basis” to an individual’s home location.

These imputed measures can be thought of as neighborhood stressors and providers with strong community collaborations can directly try to affect those stressors. Providers who want to dig deeper and identify specific SDH for individuals will need more detailed information about a person and his or her environment. However, given current data availability, this can be a big task to accomplish for everyone in the provider’s panel. A good risk model can act as a screen that triggers that deeper investigation into the specific social factors that may be at play for an individual. In other words, a good predictive model will alert the provider to possible SDH factors, but cannot always deliver all the answers for an individual. Getting all the answers will require close engagement with the patient, and the predictive model can start that engagement in the right direction.

We should not only think of this topic in relation to individual patients. The output from predictive models can also be aggregated to examine place-based population health issues. Too often the risk scores are only used for interventions with patients, but just as valuable are aggregate understandings about problems in housing, nutrition, violence, access, etc., that can best be addressed at a system level in collaboration with other agencies in the environment.

Herb Fillmore is senior director, market development for Populations and Payment Solutions at 3M Health Information Systems.@HFillmoreIII