“What data do you use?” We get asked this question by curious customers, partners and industry observers virtually every day. We certainly understand where is comes from it, but it is tough to answer without first addressing the most important consideration: the domain model.
Socially Determined is a “model-first” company. For us, this means we start by creating a set of analytic building blocks that describe fundamental relationships within the social determinants of health—or plainly, how social risk impacts one’s health and behavior. Only then do we seek and analyze the data that provides the best evidence to support the models.
Having a strong foundation in models that reflect the way the real world operates allows us to focus on sources of evidence that are going to have a high likelihood of revealing answers—without being tied to one specific data source or another.
Here’s an exercise to illustrate this point: let’s look at the impact of food insecurity on one’s health. Food insecurity is a complex issue and there is no single dataset available to establish a person’s risk for food insecurity. The evidence cannot be found solely in an individual’s health record or insurance claims file.
But our customers are trying to find the answers to important questions, like:
What is the link between food insecurity and diabetes progression?
How does food insecurity impact maternal health and birth outcomes?
What positive effects do fruit and vegetable prescription programs have on spending and health outcomes among program participants?
By following our model first and then evaluating the right datasets, we can answer these questions. Our models consider:
The use of commercial business data and open data about the person’s neighborhood as evidence of access to healthy or unhealthy food.
Harvesting unstructured data from clinical notes as evidence of what providers assess about a patient’s eating habits, susceptibility to hunger, or food literacy.
Screening the individual for direct evidence of hunger or their inability to afford food.
Using commercial pattern of life data on the individual as evidence of their ability to reliably afford food for their family.
In this example and in all of our domain areas (see our previous post on our social risk metrics), we find that our model-first approach focuses our data exploration and harvesting to find the answers our customers are looking for. We build confidence and trust with our customers that our data strategy is correct when we can establish that (a) our models consider the right influencers and (b) we are using the best available evidence to support those models.
While we curate many sources (federal, state, and local open data; commercial data from multiple vendors; screening, clinical, and claims data from our customers) in our SocialScape® platform, the simple fact is there is no “killer” data set for measuring the impact of the social determinants of health. That’s why we believe in a model-first context for the discussion of data—and our customers understand that the real question around SDOH isn’t “what data sources do you use?” It’s “what is your model?”