Risk Prediction for Peripartum Cardiomyopathy in Delivering Mothers: A Validated Risk Model

Background Peripartum cardiomyopathy (PPCM) causes significant morbidity and mortality in childbearing women. Delays in diagnosis lead to worse outcomes; however, no validated risk prediction model exists. We sought to validate a previously described model and identify novel risk factors for PPCM presenting at the time of delivery. Methods and Results Administrative hospital records from 5,277,932 patients from 8 states were screened for PPCM, identified by International Classification of Disease-9 Clinical Modification codes (674.5x) at the time of delivery. Demographics, comorbidities, procedures, and outcomes were quantified. Performance of a previously published regression model alone and with the addition of novel PPCM-associated characteristics was assessed using receiver operating characteristic area under the curve (AUC) analysis. Novel risk factors were identified using multivariate logistic regression and the likelihood ratio test. In total, 1186 women with PPCM were studied, including 535 of 4,003,912 delivering mothers (0.013%) in the derivation set compared with 651 of 5,277,932 (0.012%) in the validation set. The previously published risk prediction model performed well in both the derivation (area under the curve 0.822) and validation datasets (area under the curve 0.802). Novel PPCM-associated characteristics in the combined cohort included diabetes mellitus (odds ratio [OR] of PPCM 1.93, 95% confidence interval [CI] 1.23–3.02, P = .004), mood disorders (OR 1.74, 95% CI 1.22–2.47, P = .002), obesity (OR 1.92, 95% CI 1.45–2.55, P < .001), and Medicaid insurance (OR 1.54, 95% CI 1.22-1.96, P < .001). Conclusions This is the first validated risk prediction model to identify women at increased risk for PPCM at the time of delivery. Diabetes mellitus, obesity, mood disorders, and lower socioeconomic status are risk factors associated with PPCM. This model may be useful for identifying women at risk and preventing delays in diagnosis.