Perioperative Risk Assessment in Pancreatic Surgery Using Machine Learning.

Area of Science:

• Surgical Oncology
• Medical Informatics
• Critical Care Medicine

Background:

• Pancreatic surgery carries significant risks of postoperative complications and mortality.
• Complications often arise after patients leave intensive care unit (ICU) monitoring.
• Early identification of at-risk patients is crucial for improving outcomes.

Purpose of the Study:

• To develop and evaluate machine learning models for risk stratification in pancreatic surgery patients.
• To identify patients at high risk for postoperative death or major complications.
• To enhance patient monitoring and survival rates through predictive analytics.

Main Methods:

• Trained multiple machine learning models using pre-, intra-, and short-term postoperative data.
• Utilized data from patients undergoing pancreatic resection at Charité - Universitätsmedizin Berlin.
• Evaluated model performance using the area under the precision-recall curve (AUPRC).

Main Results:

• The best-performing model, a logistic regression, achieved an AUPRC of 0.51 for predicting patient death.
• The models achieved an AUPRC of 0.53 for predicting major complications.
• Model performance improved as more perioperative patient data became available.

Conclusions:

• Risk stratification models can aid in identifying high-risk patients after pancreatic surgery.
• Logistic regression demonstrated effectiveness for predicting adverse events in this cohort.
• Repeated computation of risk scores throughout the perioperative period is recommended for optimal patient management.