Supplementary Material for: Predicting Incident Atrial Fibrillation After Stroke: A Scoping Review of Clinical Scores, Biomarkers, and AI-enhanced Strategies

Post-stroke atrial fibrillation (AFib) is a frequent yet undetected complication, particularly in resource-limited settings, where systematic screening remains challenging. Timely identification is essential for guiding anticoagulation strategies and reducing recurrent stroke risk. This scoping review synthesizes evidence on predictive strategies integrating artificial intelligence, circulating biomarkers, and advanced rhythm-monitoring modalities in adults with ischemic stroke or transient ischemic attack without known AFib. Predictive variables from conventional clinical scores and modern AI-based models were harmonized into a unified framework, highlighting incremental contributions from natriuretic peptides, imaging radiomics, and electronic health record–derived laboratory parameters. A novel analytical construct—area under the curve (AUC)–cost–feasibility mapping—was introduced to compare diagnostic strategies, including risk scores, handheld and patch electrocardiography, smartwatch-based photoplethysmography (with ECG confirmation required for diagnosis), and implantable loop recorders, with explicit consideration of scalability in low- and middle-income countries. Based on this synthesis, a tiered diagnostic pathway is proposed, combining clinical risk stratification with biomarker-guided triage (particularly NT-proBNP and MR-proANP) to inform allocation of extended monitoring resources, thereby optimizing diagnostic yield and cost-effectiveness. Persistent knowledge gaps include the absence of standardized biomarker thresholds, limited head-to-head evaluations of AI-enabled workflow in post-stroke populations, insufficient external validation in diverse populations, and a lack of prospective cost-effectiveness analyses. By integrating predictive domains, quantifying performance–cost trade-offs, and outlining an implementation-oriented, risk-stratified strategy, this review aims to inform AFib screening after stroke from theoretical innovation toward context-adapted clinical application, offering a structured framework to guide both research and practice in diverse healthcare environments.

卒中后心房颤动（AFib）是一类常见却未被检出的并发症，在资源受限地区尤为高发，此类地区的系统性筛查仍存在诸多挑战。及时识别该病症对于指导抗凝治疗策略、降低卒中复发风险至关重要。本范围综述（scoping review）整合了针对无已知AFib的缺血性卒中或短暂性脑缺血发作成人患者的、整合人工智能、循环生物标志物与先进节律监测手段的预测策略相关证据。研究将常规临床评分与现代人工智能模型中的预测变量整合为统一分析框架，着重阐明了利钠肽、影像组学以及电子健康档案衍生实验室检测指标的增量贡献。本研究提出了一种新型分析框架——曲线下面积（AUC）-成本-可行性映射模型，用于对比各类诊断策略，包括风险评分、手持与贴片式心电图、基于智能手表的光电容积描记术（诊断需经心电图确认）以及植入式循环记录仪，并明确考量了该框架在中低收入国家的可推广性。基于上述证据整合，本研究提出了分层诊断路径：将临床风险分层与生物标志物指导下的分诊（尤其是N末端B型利钠肽原与中间区域心房利钠肽原）相结合，以指导延长监测资源的分配，从而优化诊断检出率与成本效益。目前尚存的知识空白包括：缺乏标准化的生物标志物临界值、针对卒中后人群人工智能赋能工作流程的头对头对比评估有限、不同人群中的外部验证不足，以及前瞻性成本效益分析的缺失。本综述通过整合预测维度、量化性能与成本的权衡关系，并提出面向实施的风险分层策略，旨在推动卒中后AFib筛查从理论创新转向适配场景的临床应用，为不同医疗环境下的相关研究与实践提供结构化指导框架。