Predicting Self-Reported Social Risk in Medically Complex Adults Using Electronic Health Data.

Area of Science:

• Health Informatics
• Social Determinants of Health
• Predictive Modeling

Background:

• Social barriers (food insecurity, financial distress, housing instability) significantly impede chronic illness management.
• Systematic strategies are crucial for identifying individuals needing proactive social resource outreach.

Purpose of the Study:

• To develop and validate a predictive model identifying adults with multiple chronic conditions at higher risk for social needs.
• The model aims to predict likelihood of food insecurity, financial distress, and/or housing instability.

Main Methods:

• Developed and validated a predictive model using electronic health record data and census tract data for adults with ≥2 chronic conditions.
• Model performance was assessed using area under the receiver operating characteristic curves (AUCs) and evaluated for algorithmic bias.
• External validation was performed on a separate cohort of non-Medicaid health plan members.

Main Results:

• The final model, with 30 predictors, achieved an AUC of 0.68 in identifying patients with social needs.
• The model demonstrated robustness across different race/ethnic groups.
• Key predictors included utilization, diagnosis, behavior, insurance, neighborhood, and pharmacy variables.

Conclusions:

• A predictive model utilizing medical record and public census data can effectively identify patients needing social needs assessment.
• The model enables prioritization of outreach efforts to patients with the greatest social needs.
• Adjustable risk thresholds can optimize screening and referral for social resources based on population prevalence.