Machine Learning Prediction Algorithm for In-Hospital Mortality following Body Contouring.

Area of Science:

• Medical Informatics
• Surgical Outcomes
• Machine Learning in Healthcare

Background:

• Body contouring procedures carry risks, including serious complications and mortality.
• Identifying predictors of mortality is crucial for patient safety.

Purpose of the Study:

• To identify key predictors of mortality after body contouring.
• To develop and compare machine learning models for predicting in-hospital death risk.

Main Methods:

• Utilized the National Inpatient Sample database (2015-2017) for patients undergoing body contouring.
• Included demographics, comorbidities, complications, and operative features as predictors.
• Evaluated eight machine learning models using AUC, accuracy, sensitivity, specificity, and decision curve analysis.

Main Results:

• 1.72% of 8214 patients died in-hospital.
• Sepsis, Elixhauser Comorbidity Index, and cardiac arrest were significant predictors.
• The Naive Bayes model demonstrated superior predictive performance (AUC=0.898) and net benefit.

Conclusions:

• Machine learning models effectively predict in-hospital mortality in body contouring patients.
• The Naive Bayes model offers a reliable tool for risk stratification.