Predictive Modeling to Improve Screening and Referral for Unmet Need

Unmet social needs can significantly affect patient health

Unmet social needs—including housing, food, utilities, access to care, ability to obtain prescribed medications, and transportation—have a negative impact on people’s health. They also increase people’s use and cost of healthcare services, increase burden on EDs, and make health disparities worse. While an ED encounter is a prime opportunity to screen patients and refer them to needed social services, the use of the current interview and questionnaire screening model is limited due to time constraints, ED workflow patterns, and the stigma associated with these questions.

Using predictive models to identify people with unmet social needs

Dr. Joshua Vest and a team of researchers at Indiana University- Purdue University at Indianapolis are studying whether predictive modeling can accurately identify individuals with unmet social needs. The models will use more than 150 data elements from the EHR, HIE, state social service organizations, geocoded data sets, and public health data sources. The team wants to see if the models are more effective than a standard self-administered questionnaire in identifying patients with various social needs. The comparison of results will inform the development of a CDS tool to increase referrals to appropriate social and behavioral services.

“Social determinants of health and unmet need are really what drive health and complicates care. For example, when patients say, “I have to make a choice between paying for food or medication.” These are really difficult problems that the healthcare system is being increasingly asked to address. How do we best help healthcare organizations manage those patient needs? We have to figure out who has those needs, how to address them, and how to best match resources [to meet them].”
- Dr. Joshua Vest

Quantifying the impact of real-time screening for social needs

The team will evaluate the CDS tool in an urban safety net hospital to see if it reduces repeat ED visits and increases referrals to social and behavioral service providers. The research is innovative in that it is applying predictive modeling with personal, social service, and clinical context data, and is shifting social need screening research to the ED setting. In addition, the team is transforming screening results into actionable information for service referral, as organizations are not currently doing this. The use of this real-time CDS tool will provide ED clinicians with a clear course of action and improve current screening practices that often fail to yield actionable information.