Abstract 14649: Artificial Intelligence Model Predicts Sudden Cardiac Arrest Manifesting With Pulseless Electrical Activity versus Ventricular Fibrillation

Introduction: There is no specific treatment for sudden cardiac arrest (SCA) manifesting as pulseless electrical activity (PEA) and survival rates are low; unlike ventricular fibrillation (VF) which is treatable by defibrillation. The ability to distinguish individuals presenting with PEA-SCA versus VF-SCA has potential mechanistic and therapeutic importance, but access to the initial SCA rhythm has been a limiting factor.

Objectives: We trained and tested an AI model to differentiate PEA-SCA vs. VF-SCA in a novel setting that allows for determination of the initial rhythm.

Methods: A subgroup of SCAs are witnessed by emergency medical services (EMS) personnel with zero response time, thereby providing the true initial rhythm. The internal cohort consisted of 421 EMS-witnessed out-of-hospital SCAs with PEA or VF as the initial rhythm in the Portland, OR metro area. Using demographics and clinical variables, we trained and tested an extreme gradient boosting model to distinguish PEA-SCA vs. VF-SCA. External validation was performed in EMS-witnessed SCAs from Ventura, CA.

Results: In the internal cohort, the AI model achieved an area under the operating curve (AUC) of 0.68 (95% CI 0.61-0.76). Model performance was similar in the external cohort, achieving an AUC of 0.72 (0.59-0.84). Anemia, older age, increased weight, and dyspnea as a warning symptom were the most important features of PEA-SCA; younger age, chest pain as a warning symptom and established coronary artery disease were important features associated with VF.

Conclusions: The AI model identified novel features of PEA-SCA and could distinguish EMS-witnessed individuals presenting with PEA-SCA vs. VF-SCA.