Machine learning-based prediction of postoperative mortality risk after abdominal surgery

Area of Science:

• Medical Informatics
• Surgical Oncology
• Machine Learning in Healthcare

Background:

• Preoperative risk assessment is crucial for identifying patients at high risk of postoperative mortality.
• Traditional scoring systems for risk assessment can be time-consuming.
• Machine learning (ML) models offer a potential solution for rapid and accurate risk prediction.

Purpose of the Study:

• To evaluate the efficacy of various machine learning algorithms in predicting mortality risk following abdominal surgery.
• To compare the performance of ML models against traditional methods for postoperative mortality prediction.

Main Methods:

• A retrospective study involving 230 patients who underwent abdominal surgery.
• Development of nomogram, decision-tree, random-forest, gradient-boosting, support vector machine (SVM), and naïve Bayesian models.
• Model performance was assessed using receiver operating characteristic (ROC) curves and the DeLong test for comparison.

Main Results:

• The study included 230 patients, with 52 deaths and 178 survivors.
• The Support Vector Machine (SVM) model achieved the highest Area Under the Curve (AUC) of 0.983 (95% CI: 0.959-1.000).
• Other models, including random-forest (0.928) and nomogram (0.908), also demonstrated strong predictive capabilities.

Conclusions:

• Nomogram, random-forest, gradient-boosting tree, and support vector machine models show significant potential for predicting postoperative mortality.
• These ML models provide a rapid and accurate alternative to traditional risk assessment tools.
• The choice of model can be tailored to specific clinical scenarios for optimal patient management.