Development And Validation Of A Predictive Model For 90 Day Readmission Risk In Mild Heart Failure Exacerbation Patients Discharged From The ED

Document Type

Conference Proceeding

Publication Date

1-1-2026

Publication Title

J Card Fail

Keywords

adult, aged, Caucasian, chronic kidney failure, cohort analysis, comorbidity, conference abstract, controlled study, drug therapy, female, follow up, heart failure, heart infarction, high risk patient, hospital readmission, human, least absolute shrinkage and selection operator, major clinical study, male, medical history, mobile application, predictive model, probability, retrospective study, risk factor, scoring system, very elderly

Abstract

Background: Heart failure (HF) frequently leads to hospital readmissions, especially in mild exacerbation HF (MEHF) patients. This study identifies 90-day readmission risk factors and develops a predictive model. Methods: We retrospectively analyzed data from MEHF patients discharged from the Henry Ford Jackson Hospital ED and referred to the HF clinic. Demographic factors, medical history, and clinical markers were collected to build a predictive model using linear and lasso regression. A point-based scoring system was created based on lasso coefficients: strong predictors (≥0.4) received 2 points, moderate (≥0.1) 1 point, and weak predictors (<0.1) 0 points. Scores were normalized (0-100) to represent 90-day readmission probability. A web-based application was developed for real-time risk calculation, and the model underwent internal validation, including AUC analysis, using R version 4.4.1. Results: The study included 98 patients (50 male, 48 female; mean age 69.3 ± 13.8). The cohort was 79% white, 17% black, and 2% other. Of 47 patients followed in the HF clinic, 55.3% were readmitted, while 74% of 50 without follow-up were readmitted. Seventeen baseline characteristics and adherence to four GDMT medications were analyzed. The predictive model identified comorbidities (e.g., chronic kidney disease, prior myocardial infarction), clinical history, and medications as key risk factors. The model was integrated into a mobile app for clinicians. Internal validation showed an AUC/c-statistic of 0.77. Conclusions: This study developed a predictive model for 90-day readmission risk in low-risk HF patients discharged from the ED. The AI-powered risk calculator app helps clinicians identify high-risk patients for better management and outcomes. However, due to the small sample size and lack of external validation, the model's generalizability requires further testing.

Volume

32

Issue

1

First Page

262

Last Page

263

Find It @ Sladen

Share

COinS