Related Experiment Videos
Machine Learning for Predicting Critical Postoperative Interventions: Proof-of-Concept Study Using the INSPIRE
Manan Shukla1, Paul Fodor2, Suresh Yelika3
1Renaissance School of Medicine, Stonybrook University Hospital, 118 Leeward Lane, Port Jefferson, NY, 11777, United States, 1 561-720-8383.
JMIR Perioperative Medicine
|July 28, 2026
Summary
Machine learning models accurately predict critical postoperative interventions, including ventilatory support, ECMO, IABP, and CRRT. These AI tools offer individualized risk assessments to enhance perioperative planning and patient care.
Area of Science:
- Artificial Intelligence in Medicine
- Machine Learning for Healthcare
- Perioperative Medicine
Background:
- Postoperative complications significantly increase patient morbidity and mortality.
- Current surgical risk scores lack accuracy and individualized risk prediction.
- Early prediction of complications can guide timely interventions and improve outcomes.
Purpose of the Study:
- Develop and evaluate machine learning models for predicting critical postoperative interventions.
- Utilize the INSPIRE perioperative dataset for model training and validation.
- Identify patients requiring ventilatory support, ECMO, IABP, or CRRT.
Main Methods:
- Trained and tested artificial neural network and random forest models on the INSPIRE dataset (131,000 cases).
- Used preoperative data, medications, surgery type, and intraoperative vitals as inputs.
- Assessed model performance using accuracy, sensitivity, specificity, PPV, NPV, and AUC.
Main Results:
- Artificial neural network models achieved high accuracy for ECMO (98.9%), ventilatory support (97.7%), IABP (98.0%), and CRRT (94.9%).
- Random forest classifiers demonstrated comparable high performance across all interventions.
- Models reliably identified patients needing life-saving interventions despite class imbalance.
Conclusions:
- Machine learning models provide individualized risk estimates for critical postoperative interventions.
- Accurate predictions support proactive perioperative planning, including ICU resource allocation.
- AI-driven insights can enhance clinical decision-making and patient management.