Machine learning–based prediction of in-hospital postoperative ischemic stroke following off-pump coronary artery bypass grafting

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Coronary artery disease undergoing off-pump coronary artery bypass grafting (OPCABG) (n=11855). Intervention: XGBoost machine learning risk prediction model (POISM) vs. Support vector machine, random forest, logistic regression, and k-nearest neighbour (KNN) models. Primary outcome: In-hospital postoperative ischemic stroke prediction (AUC in internal validation) (AUC 0.863, 95% CI 0.793-0.933).

Q: What does this research mean for the field?

An XGBoost machine learning model predicts in-hospital postoperative ischemic stroke following off-pump coronary artery bypass grafting with strong accuracy (AUC of 0.863 internally and 0.786 externally). Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Zhengdong Kong; Wen-Long Zou; Guang-Pu Fan; Wenqing Zhang; Chun Liu; Hui Xue; Chaobin Wang; Lu-Feng Zhang; Dan Miao; Xiaoqing Li; Bin Cai; Rui-Yuan Guan; Xiu-Bin Yang; Qing-Hua Cui; Guang-Zhi Liu

doi:10.21203/rs.3.rs-10789974/v1

October 14, 2026Open Access

Machine learning–based prediction of in-hospital postoperative ischemic stroke following off-pump coronary artery bypass grafting

View Full Paper

Ask AI

Bookmark

Share

Key result

XGBoost model predicts ischemic stroke after off-pump CABG with ~0.86 AUC.

AUC 0.863
95% CI 0.793-0.933
n=11,855

Why the study?

This study aimed to develop and validate a machine learning-based model to assess the risk of in-hospital postoperative ischemic stroke in hospitalised patients with CAD undergoing off-pump coronary artery bypass grafting.

Does a machine learning-based model (XGBoost) accurately predict in-hospital postoperative ischemic stroke in hospitalized patients with coronary artery disease undergoing off-pump coronary artery bypass grafting?

Key Points

To develop and validate a machine learning-based model that predicts the risk of in-hospital ischemic stroke in patients with coronary artery disease undergoing off-pump coronary artery bypass grafting.
Extracted electronic health record data from 11,855 patients undergoing off-pump coronary artery bypass grafting across two centers: Beijing Anzhen Hospital (training set: n=8,147; internal validation set: n=2,037) and Peking University People’s Hospital (external validation set: n=1,671).
Evaluated five machine learning models (support vector machine, random forest, logistic regression, k-nearest neighbour, and XGBoost) using perioperative clinical characteristics, interpreting variable importance with Shapley Additive exPlanation (SHAP).
In-hospital ischemic stroke occurred in 127 patients (1.58%) in the training cohort and in 24 patients (1.44%) in the external validation cohort.
The perioperative XGBoost model achieved the best predictive performance, yielding an AUC of 0.863 (95% CI, 0.793–0.933) in internal validation, 0.786 (95% CI, 0.68–0.891) in external validation, and 0.827 (95% CI, 0.765–0.890) in combined validation.
An open-access web-based clinical decision-support tool, the Post-OPCABG IS Model (POISM), was deployed to facilitate rapid risk estimation for clinical use.

Study Design

Type

Cohort (n=11,855)

Multicenter

Yes

Structured PICO

Does a machine learning-based model (XGBoost) accurately predict in-hospital postoperative ischemic stroke in hospitalized patients with coronary artery disease undergoing off-pump coronary artery bypass grafting?

P

Population

11,855 hospitalized adult patients (median age 63 years, 20.9% female) with coronary artery disease undergoing isolated off-pump coronary artery bypass grafting across two centers in China, followed during their index hospitalization (median 13 days).

E

Exposure

Machine learning-based risk prediction models (specifically XGBoost and the resulting web-based Post-OPCABG IS Model [POISM] tool) utilizing 10 key pre-, intra-, and postoperative predictors (prolonged intraoperative hypotension <65 mmHg, history of stroke, plasma fibrinogen, fasting blood glucose, type 2 diabetes mellitus, RDW-CV, severe carotid artery stenosis ≥70%, female sex, postoperative atrial fibrillation, and fibrinogen degradation products)

C

Comparator

Other candidate machine learning algorithms (support vector machine, random forest, logistic regression, and k-nearest neighbour [KNN])

O

Outcome

In-hospital postoperative ischemic stroke (IS), defined as a sudden onset of global or focal neurological deficits lasting for >24 hours with confirmation via computed tomography and/or magnetic resonance imaging, independently confirmed by two neurologistshard clinical

Main Result

Effect estimate: AUC 0.863 (95% CI 0.793-0.933)

Limitations

Low incidence of postoperative ischemic stroke limited the number of positive cases and may compromise statistical power
Retrospective study design susceptible to incomplete data recording and missing values
Primary outcome was restricted to in-hospital postoperative ischemic stroke rather than 30-day stroke due to lack of longitudinal follow-up data
Data were sourced exclusively from the Beijing region, potentially introducing selection bias and limiting generalizability

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May aid perioperative stroke risk stratification after off-pump CABG; leaves open need for prospective validation before clinical adoption.

Abstract

Background This study aimed to develop and validate a machine learning (ML)-based model to assess the risk of ischemic stroke (IS) in hospitalised patients with coronary artery disease (CAD) after off-pump coronary artery bypass grafting (OPCABG). Methods Data were collected from the electronic health records of 25,488 hospitalised patients with CAD receiving coronary artery bypass grafting at the Beijing Anzhen Hospital of Capital Medical University (BAH-CMU cohort, training and internal validation sets) between January 1, 2019, and December 31, 2021, and at the Peking University People’s Hospital (PKUPH cohort, external validation set) between May 1, 2016, and June 30, 2024. Patient and clinical characteristics available pre-, intra-, and postoperatively were used as inputs for five candidate ML models: support vector machine, random forest, logistic regression, k-nearest neighbour (KNN), and XGBoost. Model performance was compared using the area under the receiver operating characteristic curve (AUC). Shapley Additive exPlanation (SHAP) method was utilised to visualise and interpret the complex relationships between the variables and outcomes. Post-OPCABG IS Model (POISM), a web-based tool, was further designed utilising ML algorithms. Results A total of 11,855 patients (2452 [20.9%] female) were included in this study. The best-performing model, using perioperative data, achieved the following AUCs: internal validation (XGBoost), 0.863 (95% CI, 0.793–0.933); external validation (XGBoost), 0.786 (95% CI, 0.68–0.891); and combined validation (XGBoost), 0.827 (95% CI, 0.765–0.890). The SHAP analysis illustrated significant clinical contributors associated with risk of postoperative IS. Based on this model, we created an easy-to-use website tool, POISM, which clinicians can easily utilise for clinical decision-making in patients at risk of IS. Conclusions This XGBoost model showed strong predictive performance on both the training and test datasets. Moreover, the user-friendly web tool, POISM, can quickly predict the risk of postoperative IS in hospitalised patients with CAD undergoing OPCABG, thereby supporting clinical decision-making and reducing adverse outcomes.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Kong et al. (2026) conducted a cohort in Coronary artery disease undergoing off-pump coronary artery bypass grafting (OPCABG) (n=11,855). XGBoost machine learning risk prediction model (POISM) vs. Support vector machine, random forest, logistic regression, and k-nearest neighbour (KNN) models was evaluated on In-hospital postoperative ischemic stroke prediction (AUC in internal validation) (AUC 0.863, 95% CI 0.793-0.933). The XGBoost machine learning model predicted in-hospital postoperative ischemic stroke after off-pump CABG with an AUC of 0.863 (95% CI 0.793-0.933) internally and 0.786 externally.

synapsesocial.com/papers/6ae682a3e432e7a5c4ef35e5 https://doi.org/10.21203/rs.3.rs-10789974/v1

View Full Paper

Ask AI

Bookmark

Share

Machine learning–based prediction of in-hospital postoperative ischemic stroke following off-pump coronary artery bypass grafting

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Machine learning–based prediction of in-hospital postoperative ischemic stroke following off-pump coronary artery bypass grafting

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading