Abstract 335: Predicting Pulse Status From the ECG Signal Without Interrupting CPR

• Published in: Circulation (New York, N.Y.) Vol. 140; no. Suppl_2
• Main Authors: Kwok, Heemun, Bhandari, Shiv, Blackwood, Jennifer E, Coult, Jason, Kudenchuk, Peter, Rea, Thomas
• Format: Journal Article
• Language: English
• Published: 19.11.2019
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.140.suppl_2.335

Abstract only Objective: Currently, cardiac arrest resuscitation requires interruptions in CPR every 2 minutes to assess cardiac rhythm and pulse. A method which analyzed the ECG during CPR to predict whether or not an organized rhythm generated a pulse could help to direct care and limit CPR interruptions. We evaluated a real-time method to predict pulse status from organized rhythm ECG segments with and without CPR. Methods: The study cohort received attempted resuscitation by a metropolitan EMS system following out-of-hospital ventricular fibrillation arrest. Two-minute rhythm/pulse checks on the continuous defibrillator recordings were annotated for CPR, rhythm, and pulse status using the ECG, impedance, and accelerometer signals, the audio recording, and EMS record. Pulse was defined by the presence of a palpable pulse by EMS. Paired ECG segments with and without CPR were extracted at each rhythm/pulse check. Using organized rhythm segments from one-third of cases for training, we developed four ECG features using wavelet analysis (median power values in three frequency bands and QRS rate) and a logistic model to predict pulse status. Predictive performances of each ECG feature and the logistic model were measured by AUC in the remaining validation cases with and without CPR. Results: There were 238 cases and 911 paired segments with a median of 3 (IQR 2,5) paired segments per case. Among 319 organized rhythm segments in the validation set, AUC for pulse prediction ranged from 0.67 to 0.79 for the individual ECG features (Figure). The logistic model was more predictive than any individual feature (AUC 0.84, 95% CI 0.80-0.89, p < 0.05 for each comparison). The model predicted pulse similarly regardless of CPR activity (p = 0.2). Conclusion: ECG features extracted by wavelet analysis were predictive of pulse status among organized rhythm segments with and without ongoing CPR. Further study is required to understand how pulse prediction could guide rescuer actions in real-time.