Development of predictive model for predicting postoperative BMI and optimize bariatric surgery: a single center...

Area of Science:

• Bariatric Surgery Outcomes
• Obesity Treatment
• Machine Learning in Healthcare

Background:

• Predicting postoperative body mass index (BMI) trajectories after bariatric surgery is challenging.
• This complexity hinders personalized preoperative obesity treatment strategies.
• Accurate BMI prediction is crucial for effective patient management.

Purpose of the Study:

• To develop and validate machine learning (ML) algorithms for forecasting BMI reduction up to 5 years post-bariatric surgery.
• To create an accessible web-based calculator for healthcare professionals to aid in planning and postoperative care.
• This study is the first to compare multiple ML methods for BMI prediction in this context.

Main Methods:

• Retrospective review of 1104 adult patients undergoing bariatric surgery (RYGB, SG) from 2012-2021.
• Inclusion criteria: preoperative/postoperative data; exclusion: incomplete data, pregnancy, preoperative BMI ≤30 kg/m².
• Data from 883 patients used for model training and 221 for evaluation.

Main Results:

• Developed ML models demonstrated reliable predictive capabilities with low Root Mean Square Error (RMSE).
• RMSE values: 2.17 (next BMI), 1.71 (any future BMI), 3.49 (5-year BMI curve).
• Results were integrated into a web application for enhanced clinical decision-making.

Conclusions:

• Machine learning holds significant potential for improving bariatric surgery outcomes.
• Precise BMI predictions can lead to more personalized intervention strategies.
• ML-driven insights enhance overall healthcare efficiency in obesity management.