Abstract 4136643: Automated cardiac arrest detection using a photoplethysmography wristband: Evaluation of the time to detection of the circulatory arrest event

Introduction: Out-of-hospital cardiac arrest survival depends on immediate recognition and early cardiopulmonary resuscitation. For victims of unwitnessed cardiac arrest, accounting for 30-50% of cases worldwide, survival chances are dismal as help often comes too late. Automated cardiac arrest detection and alarming would offer a solution to shorten treatment delays. Within the DETECT project, a wristband with the functionality of automated cardiac arrest detection is developed. During the DETECT-1 study, we developed a photoplethysmography (PPG) algorithm to detect circulatory arrest in patients with induced circulatory arrest. Research question: In this sub analysis of the DETECT-1 study, we aim to study the delay between the actual circulatory arrest and the algorithms’ cardiac arrest alarm. Methods: Patients undergoing circulatory arrest induction during transcatheter aortic valve implantation (TAVI), subcutaneous ICD (S-ICD) implantation or ventricular tachycardia (VT) ablation wore a PPG wristband. Continuous ECG and arterial blood pressure data were collected as reference standard. PPG data were used to develop a circulatory arrest detection algorithm in three training iterations and one validation. Time to detection was defined as the difference between the algorithm’s cardiac arrest alarm and the circulatory arrest annotated based on the arterial blood pressure data. Results: The validation set consisted of 86 patients (77 TAVI, 3 S-ICD, 6 VT ablation) with 97 induced circulatory arrest events of which 95 were correctly identified by the algorithm (sensitivity 97% [95% CI 90-100%]). The median time to circulatory arrest detection was -1 second (IQR -2 – 0). In total, 50 (53%) events were detected before the annotated time (median; -1 sec), 33 (35%) events were detected at t=0, and 12 (13%) events were detected after the annotation (median; +2 sec). A total of 65 (68%) induced circulatory arrest events were detected within one second before and after the annotated start time of the circulatory arrest. Return of PPG pulsations were detected in all patients, with a median of 4 seconds (IQR 3 – 7) after pulsations returned (Figure 1 and 2). There were no differences in detection time between the three subgroups. Conclusion: The PPG algorithm detects the induced circulatory arrest events directly when blood pressure lowers. Further research is needed to validate the circulatory arrest detection algorithm and detection interval in spontaneous cardiac arrest.