Key Findings from Supervised and Unsupervised Analyses
Across our analysis using Random Forest, Logistic Regression models, and K-modes clustering, several key factors consistently emerged as strong predictors of ED readmissions. Critical among them were food security, unmet social needs, and CHW engagement metrics, including time spent with patients and the number of contact attempts. Additionally, health insurance type and the type of referral—particularly high-risk admissions—played a significant role in determining patient outcomes. These findings highlight the importance of addressing social determinants of health (SDoH) alongside medical interventions to effectively reduce readmission rates and improve patient care.
- Type of Referral & High-Risk Patients
- Patients referred as high-risk admissions had consistently higher readmission rates. This emphasizes the need for targeted follow-ups and specialized interventions for these individuals.
- Food Security & Unmet Needs
- Food insecurity was a major predictor of readmissions across all models. CHWs should not only connect patients with food assistance programs but also follow up to ensure proper utilization.
- Patients with unmet social needs were more likely to be readmitted, highlighting the importance of tracking and addressing these concerns through CHW interventions.
- CHW Engagement & Contact Attempts
- While CHW engagement reduced readmissions, patients with more than four contact attempts still faced high readmission risks. This suggests the need to assess the effectiveness of interactions rather than increasing the number of follow-ups.
- Insights from Clustering Analysis
- PCP Visits: Having a primary care provider (PCP) did not necessarily prevent readmissions, indicating a need for better coordination between CHWs and PCPs.
- Employment Status: Socioeconomic stability plays a key role in health outcomes. Patients struggling with employment were at higher risk, suggesting that job assistance programs could be beneficial.
- Utilities & Social Support: Limited access to electricity, heating, or water, as well as a lack of social support networks, contributed to readmissions. CHWs should collaborate with community organizations to provide assistance in these areas.
Recommended Areas for Intervention
- Prioritize Food Security
- CHWs should ensure high-risk patients receive sustained food assistance and monitor their ability to access and use available resources.
- Address Unmet Social Needs
- Strengthen CHW programs to provide support for employment, housing, and utilities, particularly for patients in high-risk clusters.
- Improve CHW Engagement Strategies
- Rather than increasing the number of contact attempts, efforts should focus on enhancing the quality of interactions and tailoring interventions for complex cases.
- Enhance Coordination with Primary Care
- Since PCP visits alone did not significantly reduce readmissions, better integration between CHWs and healthcare providers is needed to ensure continuous patient support and preventive care.
By addressing these social and medical risk factors, CHWs can play a crucial role in reducing ED readmissions and improving overall patient outcomes.
Future Work
Qualitative Analysis of CHW Notes: Future research will involve analyzing CHW notes for qualitative insights, which could further enhance the predictive accuracy and provide richer context for patient needs.