Perioperative medicine & outcomes
Development and Validation of Models for Preoperative Prediction of Risk and Postoperative Detection of Noninfectious Complications Using Interpretable Machine Learning and Electronic Health Record Data
Authors' conclusion
We developed parsimonious models for estimating the risk of and the detection of postoperative noninfectious complications. Our models showed good to excellent performance
Study context
- Population
- Surgical patients at 5 hospitals of 1 health care system (University of Colorado Health), 2013–2019
- Participants
- 30,639 patients
- Comparison
- Preoperative risk models vs postoperative detection models
- Outcome
- 30-day noninfectious outcomes: mortality, overall morbidity, bleeding, cardiac, pulmonary, renal and VTE morbidity, nonhome discharge, unplanned readmission
- Follow-up
- 30 days
Methodological detailsinterpretable machine learning, multicentre (5 hospitals), electronic health record data, lasso and knockoff filter variable selection
- Interpretable machine learning
- Multicentre (5 hospitals)
- Electronic health record data
- Lasso and knockoff filter variable selection
Performance
As reported in the abstract
Model discrimination (AUC)
| Test or model | AUC range |
|---|---|
| Preoperative risk models | 0.68 to 0.91 |
| Postoperative detection models | 0.92 to 0.97 |
AUC range
Preoperative risk models0.68 to 0.91
Postoperative detection models0.92 to 0.97
- Predictor variables per model
- Between 6 and 22
Further findings
Postoperative complication rates
- Lowest rate (stroke)
- 0.1%
- Highest rate (overall morbidity)
- 10.4%
Original publication
Development and Validation of Models for Preoperative Prediction of Risk and Postoperative Detection of Noninfectious Complications Using Interpretable Machine Learning and Electronic Health Record Data. Ann Surg. Published 26 March 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →