Abstract 15353: Identifying Different Clinical Phenotypes in Peripartum Cardiomyopathy by Latent Class Analysis

Introduction: Peripartum cardiomyopathy (PPCM) is a heterogeneous disorder whose pathophysiology remains poorly understood. To better understand the heterogeneity of PPCM, we used latent class analysis (LCA) to explore the different clinical phenotypes in a large sample of PPCM patients. Methods: Using the National Inpatient Sample (NIS) database, we identified 917 patients admitted to hospitals in 2017-2019 with a primary diagnosis of PPCM. We performed LCA based on 8 risk factors: age (< 30 vs > 30), race (African American (AA) vs non-AA), hypertension in pregnancy (HIP) (preeclampsia, gestational hypertension, or chronic hypertension), diabetes mellitus (DM), obesity, tobacco use, cannabis use, and psychiatric disorders (anxiety, depressive, bipolar, or schizophrenia spectrum disorders). We then compared in-hospital outcomes among the different clusters identified by LCA. Data were analyzed using Stata version 17 for descriptive analysis and Mplus version 8.8 for LCA. Results: Four patient clusters were identified. C1 (n=471, 51.4%) had no AA or cannabis users and members had the lowest probability of having HIP (0.387), DM (0.27), and obesity (0.17). C2 (n=328, 35.8%) were all AA who did not use cannabis and had the lowest probability of using tobacco (0.076), and having psychiatric disorders (0.084). C3 (n=31, 3.4%) was the youngest group all of whom used tobacco and cannabis and had the largest probability of having psychiatric disorders (0.484). On the other hand, C4 (n=87, 9.5%) was the oldest group none of whom used cannabis but members had the highest probability being obese (0.664), having HIP (0.825), and DM (0.642). Among all clusters, C3 had the lowest in-hospital mortality (0.0%) and hospital charges ($57,357). C4 had the highest in-hospital mortality 1.1% and longest hospital length of stay (5.4 ± 7.0 days). Conclusions: Using LCA, we identified four clinically distinct classes PPCM with varying in-hospital outcomes. These may reflect different additional underlying mechanisms in PPCM. Our findings may help identify treatment targets and patient selection for future clinical trials.