A method to predict ventricular fibrillation shock outcome during chest compressions

Area of Science:

• Cardiology
• Biomedical Engineering
• Emergency Medicine

Background:

• Out-of-hospital cardiac arrest due to ventricular fibrillation (VF) is a major cause of mortality.
• Quantitative electrocardiogram (ECG) analysis of VF can predict patient outcomes.
• Cardiopulmonary resuscitation (CPR) artifact in ECGs complicates continuous analysis during resuscitation.

Purpose of the Study:

• To develop a method for predicting VF shock outcomes during CPR.
• To enable real-time, patient-specific guidance of resuscitation therapy.

Main Methods:

• An algorithm was developed using 4577 VF segments from 1151 cardiac arrest patients.
• Variable-frequency notch filters were employed to mitigate CPR artifact in ECGs.
• Support vector machines and logistic regression combined 10 ECG features and 3 patient characteristics.

Main Results:

• The algorithm achieved an area under the receiver operating characteristic curve (AUC) of 0.74 for predicting defibrillation success during CPR.
• AUC for predicting functional survival during CPR was 0.75.
• Performance during CPR was comparable to that without CPR (AUC 0.77 for defibrillation success, 0.76 for functional survival).

Conclusions:

• A novel algorithm accurately predicts defibrillation success and functional survival during ongoing CPR in VF arrest.
• This represents a potential proof-of-concept for real-time resuscitation therapy guidance.