Predictive Power: Analyzing the Efficacy of Claims-Based vs. EMR-Based Algorithms for Readmission Risk

Purpose The purpose of this project was to evaluate the effectiveness of two predictive algorithms, the claims-based SHP (Strategic Healthcare Programs) and the electronic medical record-based Epic, in predictive effectiveness in 30-day readmission risk for Medicare and Medicaid patients who were receiving home health services. By determining which algorithm was more accurate, the aim of this project was to identify patients who would most benefit from interventions in a way that could improve resource allocation, minimize hospital readmissions, and improve patient outcomes in home health care. Background Hospital readmissions are a potentially preventable problem for high-risk patients receiving home health care, resulting in significant healthcare costs and penalties under value-based care programs. The Centers for Medicare and Medicaid Services have linked payment structures to readmission rates, which puts financial pressure on healthcare organizations to minimize preventable readmissions. Predictive algorithms can assist by identifying patients at risk of rehospitalization and allowing clinicians to implement targeted interventions. Methods This retrospective quantitative analysis analyzed data from home health patients hospitalized in quarter 4 of 2023, who were insured by Medicare, Medicaid, or Medicare Advantage. Data were acquired from the SHP and Epic algorithms, which gave each patient a percent chance of readmission upon admission. The point biserial correlation was used to assess the relationship between algorithm-predicted scores and actual readmissions, and the Mann-Whitney U test was used to compare score differences between readmitted and non-readmitted groups. Both tests determined which algorithm more accurately predicted readmission risk. Results SHP had a higher association coefficient (0.191) with readmissions than Epic (0.072). The Mann-Whitney U test revealed a significantly greater divergence in SHP risk scores between readmitted and non-readmitted home health patients. SHP's claims-based model, which is based on a larger, cross-system dataset, outperformed the Epic model, which is restricted to single-system data. Conclusion The results indicated that SHP's claims-based algorithm was more effective in predicting readmission risk in home health patients. Integrating SHP into home health processes could increase early detection of high-risk patients, allow for focused interventions, and promote value-based care by lowering hospital readmissions. Future studies could look into combining claims-based and EMR-based algorithms to improve predictive accuracy