Construction and validation of a regularized logistic regression model for predicting late recurrence after ablation in persistent atrial fibrillation: a clinical parameter approach.

Background: Patients with persistent atrial fibrillation (PsAF) exhibit a high recurrence rate following catheter ablation, and there is a lack of individualized prediction tools based on clinical parameters. Objective: To develop and validate a regularized logistic regression model for predicting long-term recurrence risk (>3 months post-procedure) in PsAF patients following catheter ablation. Methods: This study enrolled 300 consecutive PsAF patients undergoing their first radiofrequency catheter ablation (RFCA) from January 2023 to December 2024 in our cardiology department. The first 250 patients were assigned to the training set for model development and internal validation, while the remaining 50 patients served as an independent external validation cohort. Demographic, echocardiographic, laboratory, and procedural parameters were collected. Feature selection was performed using LASSO regression, followed by the construction of a multivariate logistic regression prediction model and a corresponding nomogram. Model performance was evaluated using ROC curves, C-index, and calibration curves. Internal validation was conducted via bootstrap resampling, and external validation was performed on the 50-patient cohort. Results: LASSO regression identified the following predictors: sex, AF duration, preoperative serum calcium, D-dimer, LDL-c, TSH, additional ablation sites, termination mode, early recurrence of atrial fibrillation (ERAF), and creatinine difference. Multivariate analysis revealed pharmacologic termination (OR = 0.248, Conclusion: The regularized logistic regression model incorporating clinical parameters accurately predicts long-term recurrence risk post-ablation in PsAF patients, outperforming traditional risk scoring systems and showing significant potential for clinical application.