Predictive Modeling of Weight Loss and Metabolic Health Outcomes: A Retrospective Predictive Modeling Study

体重减轻与代谢健康结果的预测建模:一项回顾性预测建模研究

阅读:1

Abstract

BACKGROUND: Obesity is a chronic, complicated, and progressive disease that significantly affects mortality, quality of life, and overall health in nearly 13% of the adult population worldwide. Thus, solutions like a hypocaloric diet with a Mediterranean diet pattern aim to control this and other metabolic problems. OBJECTIVES: This study developed and measured the performance of different machine learning (ML) models designed to predict body weight loss and/or metabolic syndrome (MetS) change after a 3-month hypocaloric diet with a Mediterranean pattern in obesity-diagnosed patients. METHODS: The data set was provided by a clinical trial of 893 obese patients. Five machine learning architectures were implemented: Logistic Regression, Decision Tree Classifier, Random Forest Classifier, eXtreme Gradient Boosting Classifier (XGBoost), and Support Vector Classifier. Performance metrics such as accuracy, precision, recall, F1-score, and ROC curve were used to assess the prediction models. The influence of the predictors was also evaluated in every case. RESULTS: For body weight loss prediction, Stacking and Random Forest models outperformed the other models, with accuracies of 81.37% and 76.44% and AUC of 86.79% (95% CI: 82.9%-90.4%) and 86.25% (95% CI: 82.3%-89.9%), respectively. For MetS change, Stacking had the best performance, with an accuracy of 85.90% and an AUC of 83.65% (95% CI: 76.9%-89.8%). For the prediction model of body weight loss and MetS change, Stacking was the best algorithm, with an accuracy of 94.74% and an AUC of 95.35% (95% CI: 88.7%-99.9%). Furthermore, variables associated with metabolic and inflammatory markers exhibited stronger correlations with the outcomes. CONCLUSION: Machine learning models, especially ensemble methods like Stacking and XGBoost, effectively predict body weight loss and MetS improvement in obese patients following a Mediterranean diet. Key predictors include age, insulin resistance markers, and inflammatory biomarkers. Integrating these predictive tools can significantly enhance personalized dietary interventions, optimizing treatment outcomes in clinical practice.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。