Abstract 4139284: Most Important Predictors of 30-Day Death in Patients with Heart Failure Presenting with Acute Dyspnea

Background: Heart failure (HF) manifests with burdensome symptoms leading to frequent emergency department (ED) visits and high rates of poor outcomes and mortality. Dyspnea is a nonspecific chief complaint, making risk stratification in this group very challenging. The purpose of this study was to determine patient characteristics related to poor outcomes in patients with HF seen in the ED for dyspnea. Methods: This was a secondary analysis of a retrospective observational cohort study of patients with known HF seen in the ED for a chief complaint of dyspnea between January 2018 and December 2022 at any one of 40 UPMC-affiliated hospitals in Pennsylvania. Trained independent chart reviewers collected study variables using a structured data collection form. Variables with <10% prevalence or >50% missingness were omitted. Multiple imputation was used to handle sparsely missing data. Primary outcome was all-cause death within 30-days of ED admission calculated from date of ED admission using death dates available in charts and manual obituary search. We used LASSO regression with 10-fold cross-validation to obtain the most significant predictors from 150 data points available early in the ED course (i.e., past medical history, initial vitals, demographics, ED treatments). Nine important variables were selected to build a logistic regression model to derive unbiased coefficients. Analysis was performed using Stata version 18. Results: The sample included 296 patients (mean age 73±14 years, 53% female, 13% Black, mean EF 50±13%) of whom 22 (7.4%) died within 30 days of ED admission. LASSO regression selected nine variables: age, systolic blood pressure, history of Afib/flutter, CAD, MI, stents, sleep apnea (OSA), fluid bolus in ED, and home aspirin. In the subsequent logistic regression using those variables, fluid bolus (OR=3.4, 95% CI 1.08-10.67, p=0.036) and history of sleep apnea (OR=0.09, 95% CI 0.011-0.76, p=0.027) remained significant. Conclusions: Our analysis suggest that patients who received a fluid bolus experienced higher odds of 30-day mortality and patients with a history of OSA experienced lower odds of mortality. The complex relationship of these variables with 30-day death should be investigated further.