Artificial Intelligence-Enabled Electrocardiograms and Pregnancy-Related Cardiomyopathy

BACKGROUND: Pregnancy-related cardiomyopathy (PRCM) is a leading cause of maternal death in the United States. PRCM symptoms often overlap with those of normal pregnancy and the lack of validated screening tools limits early recognition and contributes to diagnostic delays. Multiple artificial intelligence-enabled electrocardiogram (AI-ECG) models have been developed and validated for PRCM detection. OBJECTIVES: We sought to investigate the pooled diagnostic performance of these models. METHODS: A systematic search of PubMed, Scopus, and Cochrane was performed to identify studies investigating the use of AI-ECG models for PRCM detection. Studies reporting complete 2×2 confusion matrix data for the AI-ECG model were included for analysis. Univariate and Reitsma bivariate random effects model meta-analysis was performed. Pooled performance estimates with 95% CIs are presented in forest plots and a summary receiver operating characteristics curve. RESULTS: Our search identified 716 studies (PubMed: 84, Scopus: 544, Cochrane: 88) from which 6 studies involving 2,928 patients were included in the meta-analysis. Univariate random effects model meta-analysis revealed a pooled sensitivity of 0.808 (0.656-0.902), specificity of 0.944 (0.889-0.973), and diagnostic OR of 54.017 (37.918-76.952). Following Reitsma bivariate random effects model meta-analysis, the AI-ECG models yielded a pooled sensitivity of 0.775 (0.623-0.878); specificity of 0.936 (0.871-0.969); and area under the curve of 0.932 (0.813-0.953) for PRCM detection. CONCLUSIONS: AI-ECG models demonstrate good performance for PRCM detection and may represent a promising adjunct for earlier identification of PRCM. Additional studies are needed to evaluate their impact on clinical outcomes in pragmatic settings.