Key result
Machine learning models show no benefit over regression for predicting 30-day hospital readmissions.
Why the study?
To provide a focused evaluation of predictive modeling using electronic medical record (EMR) data to predict 30-day hospital readmission.
Population
41 studies of predictive models for 28-day or 30-day hospital readmission using EMR data
Comparison
Regression methods vs machine learning methods
Design
Systematic review
Authors
Loading...
Regression models match ML performance for 30-day readmission prediction; confirms simpler methods remain sufficient for EMR-based risk models.
Systematic Review (n=41)
Effect estimate: Difference 0.03 (95% CI -0.0 to 0.07)
Absolute Event Rate: 0.74% vs 0.71%
Predictive models using EMR data for 30-day hospital readmission show modest improvement over administrative data models, with no significant difference in performance between regression and machine learning methods.
Mahmoudi et al. (2020) conducted a systematic review in Hospital readmission (n=41). Machine learning methods vs. Regression models was evaluated on Average C statistics (Difference 0.03, 95% CI -0.0 to 0.07). Machine learning models predicting 30-day hospital readmission using EMR data showed no significant difference in C statistics versus regression models (difference 0.03; 95% CI -0.0 to 0.07).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: