Abstract 4371511: High Sensitivity and Specificity of Electrocardiogram-Based AI Models for Diagnosing Peripartum Cardiomyopathy: A Systematic Review and Meta-Analys

Introduction: Artificial intelligence electrocardiograms are considered efficient for estimating ejection fraction in heart failure patients. However, whether this method is a potential tool to diagnose peripartum cardiomyopathy has not been thoroughly explored. Research Questions: Are electrocardiogram-Base AI models able to predict peripartum cardiomyopathy? Aims: We conducted a meta-analysis and systematic review to evaluate the accuracy of electrocardiogram-base artificial intelligence models to predict peripartum cardiomyopathy. Methods: We searched PubMed, Embase and Cochrane. We computed true positives, true negatives, false positives and false negatives events to estimate pooled sensitivity, specificity and area under the curve under random mode. We used R 4.3.1 to perform statistics. Results: We identified 3 studies of data from 4 different datasets, including 425 patients evaluated for peripartum cardiomyopathy. The mean age ranged from 29 to 33 years. Multiparity ranged from 20.58% to 36.94%. Black population ranged from 58.13% to 62.6% Chronic hypertension ranged from 1.56% to 9.5%. Gestational hypertension ranged from 28% to 32.8% .The AI enabled electrocardiogram data yielded areas under the receiver operator of 0.900, sensitivity of 0.841(0.749-0.903), and specificity of 0.840(0.714-0.917) to predict for peripartum cardiomyopathy. Conclusions: In this systematic review and meta-analysis, the use of electrocardiogram-based artificial intelligence models demonstrated high sensitivity and specificity for the diagnosis of peripartum cardiomyopathy.