Procedural sedation is often performed by non-anesthesiologists in various settings and can lead to respiratory depression. A tool that enables early detection of respiratory compromise could not only enhance patient safety during procedural sedation, but also reduce the risk of medical liability. In this study, we aimed to develop a machine learning model that integrates detailed body composition data from patients undergoing liposuction to enhance the prediction of respiratory depression during procedural sedation. Features from bioelectrical impedance analysis, 3D body scanning, and manual measurements were extracted and used to train machine learning models. SHAP analysis, an explainable AI approach, was conducted to visually interpret feature importance. The XGBoost model, particularly when incorporating 3D body scanning data, demonstrated superior predictive performance, achieving an AUROC of 0.856 and a sensitivity of 0.805. The main predictors identified were upper abdominal volume, BMI, and age, highlighting the importance of the acquisition of detailed body composition data for assessing respiratory risks during sedation. The developed model effectively predicts the risk of respiratory depression in patients undergoing liposuction, offering a potential for personalized sedation protocols.