Development and Validation of a Machine Learning Prediction Model for Textbook Outcome in Liver Surgery: Results From
Jane Wang1, Amir Ashraf Ganjouei1, Taizo Hibi2
1From the Department of Surgery, University of California, San Francisco, San Francisco, CA.
Summary
A machine learning model predicts textbook outcome in liver surgery (TOLS), showing it improves long-term survival after hepatectomy. This tool aids in patient care decisions.
Area of Science:
- Hepatobiliary surgery
- Machine learning in medicine
- Surgical outcomes research
Background:
- Textbook outcome in liver surgery (TOLS) is a composite metric for optimal postoperative recovery.
- A Delphi consensus defined TOLS criteria.
- Validating TOLS is crucial for clinical application.
Purpose of the Study:
- Develop a machine learning (ML) model to predict TOLS using preoperative data.
- Validate TOLS criteria by assessing its association with long-term survival after hepatectomy.
Main Methods:
- Adult hepatectomy patients from a multicenter international cohort (2010-2022) were analyzed.
- Four ML models were trained to predict TOLS; XGBoost was selected.
- Multivariable Cox analysis assessed TOLS association with overall survival (OS).
Main Results:
- 62.8% of 2059 patients achieved TOLS.
- The XGBoost model achieved an AUC of 0.73.
- Predictors of TOLS included minimally invasive approach, fewer/smaller lesions, and lower comorbidity/creatinine.
- TOLS was associated with improved OS (HR 0.82, P=0.015).
Conclusions:
- An ML model accurately predicts TOLS in liver surgery patients.
- TOLS criteria are validated, showing association with improved long-term survival.
- The findings support TOLS use in clinical decision-making.
