Key result
XGBoost best predicts 30-day diabetic readmissions with an AUC of ~0.67.
Why the study?
Hospital readmissions within 30 days remain a critical issue with potential care discontinuities and escalating costs, motivating evaluation of machine learning models to predict readmission risk.
Can machine learning models accurately predict 30-day hospital readmissions in patients with diabetes using EHR data?
Observational (n=101,766)
Yes
Can machine learning models accurately predict 30-day hospital readmissions in patients with diabetes using EHR data?
Absolute Event Rate: 0.667% vs 0.642%
XGBoost models using EHR data show modest predictive ability (AUC 0.667) for 30-day hospital readmissions in patients with diabetes, with previous admissions and medication count as key predictors.
Authors
No takes yet. Share an insight, caveat, or question.
Modest XGBoost performance should not yet alter diabetes readmission risk stratification; leaves open need for validation and refined EHR-based models.
Emi-Johnson et al. (2025) conducted an observational in Diabetes (n=101,766). XGBoost machine learning model vs. Logistic regression, random forest, and deep neural networks was evaluated on Area under the receiver operating characteristic curve (AUC-ROC) for predicting 30-day hospital readmission. The XGBoost machine learning model achieved the highest predictive performance for 30-day hospital readmissions among patients with diabetes, with an AUC-ROC of 0.667.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: