Prediction of shock-refractory ventricular fibrillation in patients with out-of-hospital cardiac arrest: external validation of an ECG-based approach.

BACKGROUND: Shock-refractory ventricular fibrillation (VF) patients can be defined as those requiring at least three defibrillation attempts. Patients with refractory VF may benefit from personalized resuscitation treatments. We sought to externally validate a previously described electrocardiogram (ECG) feature calculation and analysis strategy for predicting refractory VF in an Asian out-of-hospital cardiac arrest (OHCA) cohort. METHODS: We conducted a retrospective cohort study using the Seoul OHCA registry (2020-2023). We included patients with an initial shockable rhythm and 3-s ECG segments before and after the initial shock. Refractory VF was defined by patients requiring ≥3 total shocks (primary definition) or ≥3 consecutive shocks (alternative definition). We trained a logistic model to predict refractory VF based on 25 previously-described ECG scalogram features calculated from ECG segments before and after the first shock. Performance was quantified by the area under the receiver operating characteristic curve (AUC) evaluated via leave-one-out cross-validation (LOOCV), and was compared against the original study's reported AUC. RESULTS: Of the 428 included OHCA patients with initial shockable rhythm, 202 (47.2%) received ≥3 shocks during prehospital resuscitation. The logistic combination of 25 ECG features demonstrated good performance, with LOOCV AUCs of 0.782 [95% CI 0.736-0.828] for predicting ≥3 total shocks and 0.805 [95% CI 0.762-0.858] for ≥3 consecutive shocks. These AUCs were not significantly different versus the original study's AUC of 0.846 (P = 0.060 and P = 0.22 for respective differences). CONCLUSION: A strategy using ECG features surrounding initial shock showed robust performance for predicting refractory VF in a distinct patient population.