Predicting Outcomes in Patients Undergoing Pancreatectomy Using Wearable Technology and Machine Learning: Prospective...

Area of Science:

• Surgical oncology
• Digital health
• Machine learning in medicine

Background:

• Pancreatic cancer is a leading cause of cancer death.
• Pancreatectomy is the only cure but has high morbidity.
• Predicting outcomes after pancreatectomy is crucial for patient management.

Purpose of the Study:

• To evaluate wearable telemonitoring for predicting pancreatectomy outcomes.
• To assess the utility of patient activity metrics and machine learning.
• To compare machine learning model performance against the ACS-NSQIP SRC.

Main Methods:

• Prospective study of 48 patients undergoing pancreatectomy.
• Wearable telemonitoring devices collected pre-operative activity data.
• Machine learning models integrated clinical data and activity metrics.
• Performance compared with the American College of Surgeons National Surgical Quality Improvement Program surgical risk calculator (ACS-NSQIP SRC) using AUROC curves.

Main Results:

• Machine learning models incorporating activity data outperformed ACS-NSQIP SRC.
• The best model achieved an AUROC of 0.7875 for predicting textbook outcomes.
• ACS-NSQIP SRC had an AUROC of 0.6333.
• Patients not achieving textbook outcomes often experienced severe complications or readmission.

Conclusions:

• Machine learning models integrating patient activity metrics enhance prediction of pancreatectomy outcomes.
• Wearable telemonitoring offers a promising tool for pre-operative risk stratification.
• This approach can potentially improve patient selection and post-operative care planning.