Unmet social needs--including housing, food, utilities, access to care, ability to obtain prescribed medications, and transportation--directly impacts an individual’s health. These unmet needs increase the use of healthcare services and costs, health disparities, and burden on emergency departments (EDs), and contribute to provider burnout. The ED is a prime opportunity to screen these patients and refer them to needed social services. The current model is to screen via a patient interview or questionnaire; however, time constraints, ED workflow issues, and the stigma associated with these questions often limits their use.

This research will compare the effectiveness of predictive modeling versus questionnaires in identifying those with unmet social needs and use the superior method to inform a clinical decision support (CDS) tool, with the overall goal to increase referrals to appropriate social providers. The models will utilize more than 150 data elements from the electronic health record (EHR), health information exchange, State social service organizations, geocoded data sets, and public health data sources. The questionnaire is a 42-item self-administered, vendor-developed, EHR-embedded instrument. The research will take place at an urban safety net hospital, addressing the priority populations of socioeconomically disadvantaged and minority populations.

The specific aims of this research are to:

• Compare the effectiveness of predictive modeling versus questionnaire-based screening in identifying patients in need of social and behavioral services. 
• Identify ED providers’ and patients’ perceptions of screening for unmet social needs using predictive modeling and questionnaire-based screening. 
• Quantify the impact of real-time screening for social needs on subsequent utilization. 

For the first aim, researchers will enhance previously developed and validated predictive models, and conduct a retrospective analysis to identify patients with unmet needs. This model will be compared with the use of prospectively given questionnaires to those seeking care in the ED, with a hypothesis that the former will outperform the latter. For the second aim, a series of focus groups will be conducted with ED patients, physicians, nurses, social workers, and other staff to determine their perceived fit of screening using the two approaches, with the hypothesis that participants will prefer the predictive modeling over the questionnaires.

For the final aim, the researchers will develop a CDS tool informed by the previously identified superior screening method that facilitates referrals to social and behavioral providers by displaying a patient’s need for referral. The researchers hypothesize that this tool will increase referrals, reduce future ED visits, and increase followup with providers. For the purposes of evaluation, effect of usage of the CDS tool will be based on hospitalization for diabetes complications, comparing the CDS intervention site to EDs without the intervention across the State. The findings of this research will inform future efforts to improve social needs screening in EDs and broaden the use of predictive modeling tools in multiple care settings.