What type of study is this?

This is a Validation Study study.

What question did this study set out to answer?

This study aims to evaluate whether machine learning models using claims data can predict one-year atrial fibrillation ablation outcomes more accurately than traditional risk scores.

January 8, 2026JMIR CardioOpen Access

Machine learning models outperformed traditional risk scores in predicting 1-year AF-free outcomes post-ablation, with AUCs of 0.528-0.544 compared to 0.498-0.505.

View Full Paper

Ask AI

Bookmark

Key result

Authors

YLYijun Liu MOMustapha Oloko-Oba KWKathryn Wood

Discussion

Member takes

Overview

Machine learning models predict ablation outcomes in atrial fibrillation, suggesting improvements over traditional risk scores.

Key Points

This study aims to evaluate whether machine learning models using claims data can predict one-year atrial fibrillation ablation outcomes more accurately than traditional risk scores.
Analyzed claims data from the Merative MarketScan Research Medicare database (2013-2020) for 14,521 patients who underwent AF ablation.
Developed logistic regression and extreme gradient boosting (XGBoost) models using demographic characteristics, comorbidity indices, and ICD diagnostic codes.
Compared model predictions to established risk scores: CHADS 2, CHA 2 DS 2 -VASc, and a modified CAAP-AF.
Assessed the models on subgroups with paroxysmal AF, persistent AF, and both AF and atrial flutter.
AF ablation success rate was 54.01% among patients.
XGBoost outperformed established risk scores with AUCs ranging from 0.528 to 0.529 across AF ablation groups.
While traditional scores like CHA 2 DS 2 -VASc had higher recall (>0.798), XGBoost achieved better precision (0.552-0.556).
In subgroup analyses of ICD-10 patients, models using ICD codes showed improved performance, particularly in paroxysmal AF with the highest AUC of 0.544.

Cite This Study

Liu et al. (2025) studied this question. Machine learning models outperformed traditional risk scores in predicting 1-year AF-free outcomes post-ablation, with AUCs of 0.528-0.544 compared to 0.498-0.505.

synapsesocial.com/papers/69608779fa51ca23bb9809ce https://doi.org/10.2196/77380

View Full Paper

Ask AI

Bookmark

January 8, 2026JMIR CardioOpen Access

Machine learning models outperformed traditional risk scores in predicting 1-year AF-free outcomes post-ablation, with AUCs of 0.528-0.544 compared to 0.498-0.505.

View Full Paper

Ask AI

Bookmark

Key result

Machine learning models outperformed traditional risk scores in predicting 1-year AF-free outcomes post-ablation, with AUCs of 0.528-0.544 compared to 0.498-0.505.

Authors

YLYijun Liu MOMustapha Oloko-Oba KWKathryn Wood

Discussion

Member takes

Overview

Machine learning models predict ablation outcomes in atrial fibrillation, suggesting improvements over traditional risk scores.

Key Points

This study aims to evaluate whether machine learning models using claims data can predict one-year atrial fibrillation ablation outcomes more accurately than traditional risk scores.
Analyzed claims data from the Merative MarketScan Research Medicare database (2013-2020) for 14,521 patients who underwent AF ablation.
Developed logistic regression and extreme gradient boosting (XGBoost) models using demographic characteristics, comorbidity indices, and ICD diagnostic codes.
Compared model predictions to established risk scores: CHADS 2, CHA 2 DS 2 -VASc, and a modified CAAP-AF.
Assessed the models on subgroups with paroxysmal AF, persistent AF, and both AF and atrial flutter.
AF ablation success rate was 54.01% among patients.
XGBoost outperformed established risk scores with AUCs ranging from 0.528 to 0.529 across AF ablation groups.
While traditional scores like CHA 2 DS 2 -VASc had higher recall (>0.798), XGBoost achieved better precision (0.552-0.556).
In subgroup analyses of ICD-10 patients, models using ICD codes showed improved performance, particularly in paroxysmal AF with the highest AUC of 0.544.