ID: 4349277 BEYOND THE HOSPITAL: REAL-WORLD SURVIVAL BENEFIT OF REMOTE MONITORING IN CHRONIC HEART FAILURE

Objective: Use Holter data to predict diagnosis and arrhythmia risk and Long QT Syndrome (LQTS).Method: A single-center retrospective case-control study was conducted comparing congenital LQTS patients and normal controls who underwent a 24-hour Holter between 01/22-05/ 24.After identifying patients with cardiac events, two neural network models were constructed: 1) LQTS diagnostic prediction model, 2) cardiac event (resuscitated sudden cardiac arrest, ventricular tachycardia/fibrillation) prediction model in the LQTS cohort.Holter QTc values (minimum, maximum, average, and % time QTc.460 ms) were input variables.Models were trained on 80% of the patients and tested on the remaining 20%.Results: We identified 78 LQTS patients (mean age: 1267 yrs, 35% male) and 158 normal patients (mean age: 1364 yrs, 30% male with 193 and 169 Holter records respectively.Five patients experienced cardiac events with all having LQTS2 diagnosis and 15 Holter records among them.LQTS Diagnosis Prediction Model: Using average, maximum, and % time QTc.460 ms as input variables, this model achieved an AUC of 0.91 in predicting LQTS diagnosis with generalizability of AUC 0.94 on test set.Maximal QTc was the most influential input with AUC decreasing by 0.19 with scrambled values.Cardiac Event Prediction Model: Using minimal, average, and maximum QTc as inputs, this model demonstrated an AUC of 0.91 with generalizability AUC of 0.94 on test set.Maximal QTc was the most significant feature with an AUC decreasing to 0.38 with permutated values.Conclusion: Holter monitoring QTc data has value in aiding in the diagnosis of LQTS and greater value in identifying LQTS patients at-risk for cardiac events.Future studies enrolling prospective patients across sites is warranted.