Takotsubo Cardiomyopathy in HFrEF and Thrombocytopenia: Multi-Center Machine Learning Supported Analysis of Mortality and Costs among 6+ Million Hospitalizations

Abstract

Introduction Distinguishing Takotsubo Cardiomyopathy (TTC) from ischemic heart disease in patients presenting with acute on chronic heart failure is particularly challenging in patients with thrombocytopenia due to the need for coronary angiography. This is the first known nationally representative study on TTC predictors and its relation to mortality and total costs in HFrEF and thrombocytopenic patient admissions. Methods The 2016 data from the nation's top all-payer hospitalization dataset (the National Inpatient Sample) was used for backward propagation neural network machine learning augmenting multivariable regression fully adjusting for age, race, income, primary malignancy, metastases, and mortality risk among chronic thrombocytopenia (platelets Results Among 6,039,969 inpatient admissions, TTC compared to non-TTC patients were more likely to have HFrEF (8.27% versus 2.84%, p Conclusions This study provides the first large multi-center robust analysis suggesting HFrEF, thrombocytopenia, and sociodemographics as TTC predictors. It is possible patients with thrombocytopenia and HFrEF are more likely to be labeled as TTC rather than have invasive testing to rule out ischemic heart disease. The subset of patient admissions in which TTC patients also received PCI resulting in improved mortality and total costs suggests that ischemic heart disease rather than TTC was the culprit. In addition, the thrombocytopenic patients with acute on chronic heart failure that received PCI and were labeled as TTC also had decreased total costs. This study may help provide cardiologists clinically relevant predictors of TTC and its outcomes to risk stratify patients. In addition, it suggests that despite thrombocytopenia providers should not be quick to label patients with TTC but rather consider invasive workup with possible PCI that may decrease costs of inpatient admissions.