Accurate prediction of bleeding risk after coronary artery bypass grafting with dual antiplatelet therapy: A machine learning model vs. the PRECISE-DAPT score

Yi Yang; Yuqing Yan; Zhou Zhou; Jifan Zhang; Haolong Han; Weihui Zhang; Xia Wang; Chen Chen; Weihong Ge; Jun Pan; Jianjun Zou; Hang Xu

doi:10.1016/j.ijcard.2024.132925

Accurate prediction of bleeding risk after coronary artery bypass grafting with dual antiplatelet therapy: A machine learning model vs. the PRECISE-DAPT score

Int J Cardiol. 2024 Dec 22:132925. doi: 10.1016/j.ijcard.2024.132925. Online ahead of print.

Authors

Yi Yang¹, Yuqing Yan², Zhou Zhou³, Jifan Zhang⁴, Haolong Han⁵, Weihui Zhang¹, Xia Wang¹, Chen Chen³, Weihong Ge⁶, Jun Pan⁷, Jianjun Zou⁸, Hang Xu⁹

Affiliations

¹ Department of Pharmacy, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing 210008, China; School of Basic Medicine and Clinical Pharmacy, China Pharmaceutical University, Nanjing 210009, China.
² School of Basic Medicine and Clinical Pharmacy, China Pharmaceutical University, Nanjing 210009, China; Department of Pharmacy, Nanjing First Hospital, Nanjing Medical University, Nanjing 210001, China.
³ Department of Pharmacy, Nanjing First Hospital, Nanjing Medical University, Nanjing 210001, China.
⁴ Nanjing Foreign Language School, Nanjing 210008, China.
⁵ Department of Pharmacy, Nanjing Drum Tower Hospital Clinical College of Nanjing University of Chinese Medicine, Nanjing 210008, China.
⁶ Department of Pharmacy, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing 210008, China.
⁷ Department of Cardiothoracic Surgery, The Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing 210008, China. Electronic address: pj791028@njglyy.com.
⁸ Department of Pharmacy, Nanjing First Hospital, Nanjing Medical University, Nanjing 210001, China. Electronic address: zoujianjun100@126.com.
⁹ Department of Pharmacy, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing 210008, China; School of Basic Medicine and Clinical Pharmacy, China Pharmaceutical University, Nanjing 210009, China. Electronic address: njglyyxh@163.com.

PMID: 39719216
DOI: 10.1016/j.ijcard.2024.132925

Abstract

Background: Dual antiplatelet therapy (DAPT) after coronary artery bypass grafting (CABG), although might be protective for ischemic events, can lead to varying degrees of bleeding, resulting in serious clinical events, including death. This study aims to develop accurate and scalable predictive tools for early identification of bleeding risks during the DAPT period post-CABG, comparing them with the PRECISE-DAPT score..

Methods: Clinical data were collected from patients who underwent isolated CABG at Nanjing Drum Tower Hospital between June 2021 and December 2023. The dataset was split into derivation and validation cohorts at a 7:3 ratio. Machine learning models were developed to predict bleeding within six months post-CABG in DAPT patients and tested in a temporal external validation cohort. The SHapley Additive exPlanations method visualized variable importance regarding outcomes. The performance of the PRECISE-DAPT score was also validated in this cohort.

Results: Among 561 enrolled patients, 165 (29.4 %) experienced bleeding events, with 49 (8.7 %) cases being significant. In the validation cohort, eXtreme gradient boosting (XGB) achieved the highest area under the receiver operating characteristic curve (0.915) and precision-recall curve (0.692). Compared to PRECISE-DAPT, XGB showed no difference in AUROC (p = 0.808) but had a higher AUPRC (p = 0.009). In the temporal external validation cohort, the XGB model has an AUROC of 0.926 and an AUPRC of 0.703. We developed a dynamic high-accuracy bleeding risk calculator based on the XGB model and created a mobile-friendly QR code for easy access to this tool.

Conclusion: Bleeding risk during DAPT in post-CABG patients can be reliably predicted using selected baseline features. The XGB model outperforms the Precise-Dapt model, showing better precision and recall.

Keywords: Coronary artery bypass grafting; Dual antiplatelet therapy; PRECISE-DAPT score; Risk prediction; Ticagrelor.