Aug 2026· Annals of Noninvasive Electrocardiology· Vol 31· 0 citations· 15 references
Medicine
TL;DR
CIED‐detected AF burden is strongly associated with progression to persistent AF, and ML‐based analysis of 6‐month device data enables accurate, point‐in‐time risk stratification to support earlier and more targeted clinical management.
Abstract
ABSTRACT Background Cardiac implantable electronic devices (CIEDs) frequently detect brief, often subclinical atrial fibrillation (AF), but their value for predicting progression to persistent AF remains uncertain. Statistical and machine learning (ML) approaches may enable dynamic risk stratification using this longitudinal device data. Objective To develop a risk stratification model using clinical and CIED‐derived AF burden measured over a rolling 6‐month window to predict progression to persistent AF. Methods We analyzed continuous CIED data from 1985 patients without prior persistent AF implanted between 2016 and 2024 at a tertiary medical center. AF burden and clinical variables were summarized using overlapping 6‐month rolling windows to estimate 1‐year risk of persistent AF. Associations were evaluated using Kaplan–Meier and Cox proportional hazards models. A gradient‐boosted decision tree model (XGBoost) was used to predict progression. Results During a mean follow‐up of 1192 days, 874 patients (44%) developed paroxysmal AF, of whom 257 (29%) progressed to persistent AF after a mean of 813 days. Patients with no AF or < 1 h/day of AF in the prior 6 months had > 97% 1‐year freedom from persistent AF, whereas those with > 8 h/day had a 63% progression rate. Higher AF burden was strongly associated with progression (maximum HR 8.66, p < 0.001). The ML model demonstrated high predictive performance (sensitivity 99.4%, specificity 95.7%). Conclusion CIED‐detected AF burden is strongly associated with progression to persistent AF. ML‐based analysis of 6‐month device data enables accurate, point‐in‐time risk stratification to support earlier and more targeted clinical management.
This data-driven, interpretable XGBoost model enables individualized AF risk assessment in middle-aged and older CHD patients, offering a practical tool for early identification and targeted intervention in clinical practice.
Feng Chen, Qin Fu, Ling Li et al.· Frontiers in Cardiovascular...· 0 citations
Background Postoperative atrial fibrillation is a common complication after cardiac surgery, associated with both short- and long-term adverse effects. Persistent forms of postoperative atrial fibrillation lasting until hospital discharge are less frequent but clinically significant. Despite its impact, reliable prediction remains challenging. This study aimed to develop and evaluate machine learning models that integrate electrocardiographic and clinical variables to predict atrial fibrillation at discharge. Methods In this retrospective single-center study, 1,905 patients undergoing cardiac surgery were analyzed. Perioperative 12-lead electrocardiographic parameters and clinical variables were preselected using univariable logistic regression (p < 0.1). Various machine learning models, including neural networks, support vector machines, k-nearest neighbors' random forests, and Bayesian classifiers were trained, using 10-fold cross-validation and subsequently evaluated on an independent test set. A genetic algorithm was applied for variable selection, and a reject option was implemented to withhold uncertain predictions. Results Of 1,905 patients 2.2% were discharged in atrial fibrillation. The final neural network model yielded a 69.9% coverage rate, with 99.2% of all predictions being correct (95% CI: 98.6–99.6). Key predictors included age, CHADS2-VASc score, left ventricular ejection fraction, EuroSCORE II, cardiopulmonary bypass time, prior cardiogenic shock, perioperative atrioventricular block, and mitral valve surgery. Conclusion A machine learning model integrating perioperative electrocardiographic and clinical variables demonstrated robust performance in predicting the absence of atrial fibrillation at discharge, accurately identifying more than two thirds of patients. For the remaining patients, management continued to rely on clinical judgment. This approach offers a valuable tool for improved risk stratification and may facilitate the implementation of more targeted prophylactic strategies following cardiac surgery.
S. Leiler, W. Hitzl, Andre Bauer et al.· Frontiers in Cardiovascular...· 0 citations
BACKGROUND AND AIMS
Accurate stroke-risk stratification is central to anticoagulation decision-making in patients with atrial fibrillation (AF), but conventional scores may not fully capture risk heterogeneity. We aimed to develop and externally validate an interpretable weighted score using a time-to-event framework.
METHODS
GLORIA-AF Phase II/III data were used to evaluate 17 baseline predictors using LASSO-penalized Cox regression with stability selection; coefficients were converted into integer weights. Performance was assessed using discrimination, calibration, integrated discrimination improvement (IDI), continuous net reclassification improvement (NRI), and decision-curve analysis. External validation was performed in EORP-AF and APHRS-AF registries.
RESULTS
Among 20,517 patients included in the derivation cohort (mean [SD] age, 69.9 [10.3] years; 9,196 women [44.8%]), 487 (2.4%) had stroke, and 17,397 (84.8%) were receiving anticoagulation at baseline. Ten selected predictors formed a 0-23-point score. The derived score achieved a C-index of 0.661 (95% CI, 0.636-0.685), higher than CHA2DS2-VA (0.626; P < 0.001), CHA2DS2-VASc (0.615; P < 0.001), and the unweighted score (P = 0.016), with no significant difference from the full Cox or machine learning models. In external validation (8,309 patients; 147 strokes), the C-index was 0.652 (95% CI, 0.614-0.690) versus 0.616 for CHA2DS2-VA (95%CI, 0.598-0.634; P < 0.001). IDI/NRI, calibration, and decision-curve analyses supported improved risk differentiation, close calibration, and generally greater net benefit than CHA2DS2-VA. The score retained higher discrimination than CHA2DS2-VA among patients without baseline anticoagulation (P < 0.001).
CONCLUSION
The GLORIA-AF Stroke Weighted Risk Score provides risk refinement beyond CHA2DS2-VA while retaining discrimination consistent with more complex models. External validation supports its transportability and potential adjunctive role in guideline-directed thromboembolic risk assessment.
Atrial fibrillation (AF) is frequently asymptomatic and often remains undetected until complications arise. Although artificial intelligence (AI)-enabled electrocardiography (ECG) can predict incident AF from sinus rhythm ECGs, its influence on physician risk assessment in simulated clinical settings remains uncertain. We developed a deep learning model to predict multi-day AF risk using non-AF 12-lead ECGs. For ECG-labeled outcomes, the model achieved an AUROC of 0.79 in the internal EUMC cohort and 0.74 in the external BIDMC cohort. For Holter-labeled outcomes, AUROC values reached 0.87 in the internal EUMC subset and 0.75 in the prospective PROVISION-AF cohort. To assess decision-support utility, a multinational survey of 70 physicians evaluated how AI-derived risk estimates influenced physician risk assessment and follow-up decisions in structured simulated cases. AI assistance significantly improved physicians' AF risk discrimination (AUROC 0.573 to 0.650) and negative predictive value (0.764 to 0.839), with significant net reclassification improvement for non-AF cases (NRI 0.127, p < 0.001). Performance gains were most notable among non-electrophysiologist cardiologists. In conclusion, AI-derived risk estimates improved physician risk discrimination in a structured simulated survey, particularly in non-specialist settings, supporting their potential role as a digital decision-support tool. Further real-world implementation studies are needed to determine whether these effects translate into improved clinical outcomes or healthcare efficiency.
Yeji Kim, Bogeun Kim, J. Yoon et al.· npj Digital Medicine· 0 citations
Background Atrial fibrillation (AF), one of the most common cardiac arrhythmias worldwide, carries a high risk of severe complications. Patients diagnosed with paroxysmal atrial fibrillation (PAF) may progress to persistent atrial fibrillation (PerAF) following a period of time. This study aimed to identify metabolites associated with PerAF using machine learning (ML) and predict the probability of PAF progressing to PerAF, enabling the adjustment of subsequent treatments in clinical practice. Methods AF patients without any anticoagulant therapy within the last 7 days were enrolled between July 2020 and July 2022 at two hospitals in China. Targeted metabolomic profiling was performed on the participating patients’ plasma. Differential metabolites and clinical features were identified through univariate and multivariate analyses. Patients were divided into discovery and validation cohorts (70%:30%). Four ML models (logistic regression, random forest, XGBoost, and LightGBM) were developed for PerAF prediction. Model predictive performance was measured mainly using the area under the curve (AUC). Results One hundred patients (65 PerAF, 35 PAF) were enrolled. Eight metabolites (kynurenine, N-A cetylaspartic acid, glyceric acid, adipic acid, citramalic acid, malic acid, isocitric acid, and oxoglutaric acid) and two clinical features (NT-proBNP and uric acid) were significantly associated with PerAF. Pathway enrichment analysis highlighted alterations in the citrate cycle and glyoxylate/dicarboxylate metabolism. XGBoost was chosen for establishing the final model since its predictive performance outperformed that of the other algorithms. The model based on clinical parameters, metabolites, and demographics achieved the highest AUC in both the discovery cohort (0.751 (95% CI [0.631~0.867])) and validation cohort (0.985 (95% CI [0.940~1.000])). A simplified model with three features (NT-proBNP, citramalic acid, and uric acid) retained robust performance. Conclusions This study identified eight PerAF-related metabolites via targeted metabolomics and ML, and developed accurate predictive models (including a simplified, clinically feasible model) with favorable predictive performance for PerAF risk stratification. Future directions should include large-scale multi-center external validation, comprehensive adjustment for potential confounding factors, and the application of multiple metabolic platforms to deeply explore AF-related metabolic alterations.
Background Prediction models for mortality risk in patients with chronic heart failure (CHF) have traditionally relied on static admission data, which restricts capturing disease dynamics. Longitudinal follow-up data were used to develop a dynamic model to improve accuracy and provide evidence for tailored interventions. Methods We enrolled 1,333 CHF patients from 3 Shanxi centres. Data included CHF patient-reported outcome (PRO) measures (CHF-PROM), lifestyle, medications, and prognosis. Endpoint: all-cause mortality. Using sequential data, we developed Gated Recurrent Units (GRU), Long Short-Term Memory (LSTM), Multi-layer Perceptron (MLP), and Logistic Regression (LR) for 3-year risk. Performance assessed by area under the receiver operating characteristic curve (AUC), accuracy, true negative rate (TNR), true positive rate (TPR), Brier score, and F1-score. Temporal Shapley Additive exPlanations (TimeSHAP) provided interpretability, and a web tool built. Results Among models tested, the GRU model demonstrated strongest predictive accuracy, with performance steadily increasing as follow-up progressed. By 24 months, the GRU-based model attained its peak predictive performance, yielding an AUC of 0.765 (95% confidence interval [CI]: 0.761–0.768), an F1-score of 0.537 (95% CI: 0.531–0.542), and a Brier score of 0.208 (95% CI: 0.199–0.216). TimeSHAP indicated that physical condition, appetite, sleep, physical independence, and anxiety within the CHF-PROM, together with age and New York Heart Association Functional Classification functional class, were key predictors of 3-year all-cause mortality in patients with CHF. Conclusion PRO data from multiple follow-ups, combined with a model constructed using GRU, provides promising tool for predicting mortality risk in patients with chronic heart failure (CHF). The self-developed web-based decision support system allows users to calculate risk scores simply by entering patient information. Trial Registration Study registered with the China Clinical Trial Registry [identifier: ChiCTR2100043337]. Experimental registration date is February 11, 2021.
Yujia Zhang, Mengyi Dou, Fengqin Ding et al.· Vascular Health and Risk Man...· 0 citations