Abstract 16753: Baseline Prediction of Atrial Fibrillation Recurrence After Catheter Ablation: Comparative Analysis of Prognostic Models Using Data Recorded by Implanted Cardiac Monitors

Intro: Tools for baseline prediction of catheter ablation outcome are instrumental to treatment management that best balances positive outcomes against risk of treatment complications. To date, many such predictive models exist in the literature; yet, it is unclear which should be adopted as no study has compared all using a single dataset. Goals: examine 8 published models using a single dataset with expert-annotated data recorded by implanted monitors. Methods: We collected data from a cohort of ~350, acquired from a randomized clinical trial blinded to ablation outcome. Based on a 2020 review, we shortlisted 12 models for baseline prediction of recurrences recorded between days 91-365 post ablation per standard. Models that use postoperative data, thus unsuited for baseline prediction, were excluded. Four models were further dropped from this work as their required variables (e.g. normalized atrial area) are unavailable in our cohort. Results: There was no missing variable; 34% were female and 53% of patients experienced recurrence. Figure reports each model’s performance with metrics as sensitivity (SEN), specificity (SPEC), positive predictive value (PPV), negative predictive value (NPV), area under receiver operating characteristic curve (AUC), positive and negative clinical utility indices (SENxPPV; SPECxNPV, resp.). The digit after each model name indicates the cut-off used (some studies used different cut-offs). Models are presented in rank-order by utility indices. All evaluated scores achieved AUC Conclusions: Our results reinforce statements of previous reviews that many published models are lacking and that there remains a need to develop and validate models for prediction of AFR post ablation.