Key result
Machine learning models using gradient-boosted trees accurately predicted 30-day postoperative emergency department readmissions (AUC 0.866), with reliable predictions at 36 hours after surgery.
Why the study?
Prediction of hospital readmissions had received little attention in surgical populations, and existing predictors required data only available at discharge.
Can machine learning models accurately predict 30-day postoperative emergency department readmissions in surgical patients?
Population
34,532 surgical hospital admissions at a tertiary care academic medical center
Comparison
Different machine learning models and feature permutations for predicting readmission
Design
Cohort study
Follow-up
30 days
Authors
Loading...
ML models enable early postoperative readmission risk prediction; leaves open whether interventions based on these models improve outcomes.
Cohort (n=34,532)
No
Can machine learning models accurately predict 30-day postoperative emergency department readmissions in surgical patients?
Effect estimate: AUC 0.866
Machine learning models can accurately predict 30-day postoperative emergency department readmissions as early as 36 hours after surgery, without requiring discharge-level data.
Mišić et al. (2020) conducted a cohort in Postoperative surgical patients (n=34,532). Machine learning prediction models was evaluated on future hospital readmission originating from the emergency department within 30 days of surgery (AUC 0.866). Machine learning models using gradient-boosted trees accurately predicted 30-day postoperative emergency department readmissions (AUC 0.866), with reliable predictions at 36 hours after surgery.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: