Multivariable data-driven framework of predictive, preventive, and personalized medicine for long-term atrial fibrillation risk in patients with new-onset obstructive sleep apnea

基于多变量数据驱动的预测、预防和个性化医疗框架,用于评估新发阻塞性睡眠呼吸暂停患者的长期房颤风险

阅读:1

Abstract

INTRODUCTION: Atrial fibrillation (AF) presents a significant challenge in patients with obstructive sleep apnea (OSA), as traditional risk factors often fail to accurately predict individual risk levels. This study exemplifies the paradigm of predictive, preventive, and personalized medicine (PPPM/3PM) by developing a comprehensive predictive nomogram for atrial fibrillation (AF) risk in patients with diabetes mellitus and obstructive sleep apnea (OSA). METHODS: This cohort study included 797 patients with new-onset OSA but without AF from January 2017 to December 2019. A total of 53 baseline clinical features were systematically collected and analyzed. Clinical feature pre-screening was conducted using the Boruta algorithm. The best predictive model was selected from nine machine learning algorithms based on area under the curve (AUC), calibration curve, and decision curve analysis (DCA). The SHapley Additive exPlanations (SHAP) tool was used to explain model decisions. RESULTS: With a median follow-up period of 56.78 months, the development cohort consisted of 494 participants, whereas the time-series independent verification cohort comprised 303 participants. Key predictors identified included age, fasting blood glucose, very low-density lipoprotein cholesterol, triglycerides, triglyceride-glucose index, autonomic nervous system indicators, apnea-hypopnea index, and average blood oxygen saturation. The XGBoost model was chosen for its superior performance, achieving an AUC of 0.922 and a Brier score of 0.071 in the time-series independent verification cohort. Calibration curve and DCA demonstrated that the XGBoost model exhibited the closest alignment with the ideal calibration line and provided the highest net clinical benefit across various threshold probabilities. SHAP analysis revealed significant contributions of individual variables to long-term AF risk. CONCLUSIONS: This study proposes a comprehensive XGBoost-based PPPM/3PM framework that demonstrates superior clinical performance in predicting long-term atrial fibrillation risk among patients with new-onset OSA. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s13167-025-00422-7.

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。