A machine learning model may help identify patients at risk of recurrent syncope after patent foramen ovale (PFO) closure. New research from Scientific Reports outlines a K-nearest neighbours (KNN) algorithm that demonstrated strong predictive performance in a retrospective study of 284 patients.
PFO, a small opening between the heart's upper chambers that fails to close after birth, has been linked to unexplained syncope. While transcatheter closure can reduce fainting episodes, clinicians previously lacked tools to predict recurrence.
Researchers analyzed 284 patients who underwent PFO closure between 2017 and 2023. Syncope recurred in 41 patients (14.4%). The team evaluated clinical, echocardiographic, laboratory, and procedural variables, narrowing the model to 10 key predictors, including syncope triggers, episode frequency, plateletcrit, occluder type, D-dimer levels, blood pressure status, attack duration, age, platelet distribution width, and diabetes status.
Among 10 algorithms tested, the KNN model achieved an AUC of 0.993, 100% sensitivity, 98.6% specificity, and 98.8% accuracy. SHAP analysis identified syncope inducements, preoperative episode burden, and plateletcrit as the most influential factors.
The authors caution that the model is preliminary. Limitations include binary outcome analysis, a small number of recurrence events, and lack of external validation. The findings are considered hypothesis-generating, and larger prospective studies with time-to-event analysis and external validation are needed before clinical use.