> ## Content Index
> Fetch the complete content index at: https://www.coios.me/llms.txt
> Use this file to discover other available public pages before exploring further.

# Machine learning beat risk scores for predicting AF recurrence after catheter ablation
- URL: https://www.coios.me/machine-learning-beat-risk-scores-for-predicting-af/
- Published: 2026-10-17T08:57:09.000Z
- Updated: 2026-10-17T09:52:09.000Z
- Description: Across 11 studies, machine learning pooled a C-statistic of 0.80 for AF recurrence after ablation, against 0.58 for traditional scores.
- Author: Daniel Ryan
- Tags: Data, models and systems, AI in medicine, #peer-reviewed, #confirms, #item, #source-2026-10, #score-55

This systematic review compared machine learning models with traditional risk scores for predicting atrial fibrillation recurrence after catheter ablation. Four databases were searched to June 2026; eleven studies covering 17,150 patients were pooled by random-effects meta-analysis of the C-statistic. Machine learning gave a pooled discrimination of 0.80 (95% CI 0.72-0.86) against 0.58 (95% CI 0.53-0.62) for traditional scores. Multimodal fusion models discriminated best, then neural networks; tree-based models did not significantly beat the traditional scores. Calibration was largely unreported and no model was externally validated. A PROBAST appraisal found the included studies methodologically weak.

*Why it is interesting: Recurrence after ablation is common and the existing scores discriminate little better than chance. The pooled gap is large, but no model has been externally validated.*

Source

[Current Problems in Cardiology, 15 October 2026](https://doi.org/10.1016/j.cpcardiol.2026.103484?ref=coios.me)

Paper

Machine-learning versus traditional scoring methods for prediction of atrial fibrillation recurrence after catheter ablation: A systematic review and meta-analysis

Authors

Koh T, Jeong S, Dababneh E, Lal V, Hughes I, Nayyar S

DOI

10.1016/j.cpcardiol.2026.103484

Type

Peer-reviewed article

Design

Systematic review and random-effects meta-analysis of 11 head-to-head studies, n=17,150, databases searched to June 2026, PROBAST quality appraisal

Verdict

Confirms prior evidence

Driver

[AI in medicine](https://www.coios.me/d-ai-medicine/)