Peripartum Cardiomyopathy: Clinical Characteristics, Risk factors, Treatment Patterns and Clinical Outcomes. A Single Center Registry

Abstract

Introduction: Patients admitted with heart failure have a guarded prognosis and a wide range of mortality risk. Identifying that risk as early as possible will help to optimize use of resources and streamline care decisions; for example, triggering a referral to palliative care services. Our organization is a tertiary medical center located in a rural area of the northeastern United States. We compared two established mortality prediction models by applying them to our population. Purpose: The purpose of this investigation was to compare the accuracy of the ADHERE criteria and the Lim criteria (University of Hawaii) in predicting inpatient mortality. Methods: The patient population included all patients admitted to our facility who subsequently had a discharge diagnosis of Heart Failure (HF) for the period of April 2011 to September 2015. Data gathered from the Electronic Medical Record (EMR) included demographics (age, gender, race, ethnicity), admission and discharge date/time, length of stay, type and result of lab tests in the first 24 hours, and vital signs. We calculated the ADHERE score with BUN, Creatinine, and blood pressure. Using the methodology in Lim, we counted the number of abnormal (outside of the normal hospital range) lab tests and grouped them into three categories: one to six, seven to twelve, and thirteen or greater. We used univariate analyses, conducted logistic regressions, and compared the proportions of inpatient deaths in the highest risk group of each model. Results: 1553 patients were admitted to our facility with HF during our review period: 1381 had sufficient data to be scored under the ADHERE criteria and 1548 under the Lim criteria. We created logistic regression models based on the ADHERE and the Lim criteria to predict inpatient mortality. For both the ADHERE and the Lim models, results from univariate analyses indicated that the respective ADHERE/Lim criteria and patient age were significant. Specifically, age group 80 and greater was significant OR 4.4 95% C.I. (1.6, 15.0) in the ADHERE model and OR 6.1 95% C.I. (2.2, 21.3) in the Lim Model. Results from the ADHERE and Lim criteria are found in Table. Conclusions: The Lim criteria seem to provide the better predictor of inpatient mortality, though either model would be acceptable for inpatient mortality prediction. The model is neither over-nor under-fitted as demonstrated by the AUC. As these data are easily extracted from an EMR, they can be used to provide an alert to providers as to patients at greatest risk of mortality. Since our population is demographically homogeneous, these results may not be generalizable to other organizations.