Nuril Farid Abshori, Syanindita Puspa Wardhani, Muhammad Nauval Dwi Afandi, Fitri Amalia, Aldivo Pradana, Anisatul Hamidah, Achmad Zainudin Arif, Iwal Reza Ahdi, Djanggan Sargowo, Muhammad Habiburrahman
Heart failure (HF) often coexists with renal dysfunction, which complicates the management and prolongs the length of stay (LOS) of patients in hospital. It is associated with an increased risk of in-hospital complications, higher readmission rates and increased healthcare costs, placing a significant burden on both patients and healthcare systems. The present study evaluated whether renal biomarkers or echocardiographic indices can more effectively predict LOS in hospitalized patients with HF and compared the performance of predictive models derived from these features. In the present retrospective observational single-center study, 112 adult patients admitted with HF between January, 2024 and May, 2025 were analyzed. LOS was evaluated both as a continuous variable and as a binary outcome (≤3 days vs. >3 days). Group comparisons employed the Mann-Whitney U and Chi-squared tests. Predictive models were constructed using logistic regression, linear regression and random forest algorithms. Three feature sets were assessed: i) Renal biomarkers [serum urea, creatinine and estimated glomerular filtration rate (eGFR)]; ii) echocardiographic left ventricular ejection fraction (LVEF); and iii) a combined set including renal biomarkers, LVEF, age, sex and diabetes status. The median LOS was 3 days (range, 1-10 days). Patients with prolonged periods of hospitalization (>3 days) exhibited higher urea and creatinine levels, and a lower eGFR (all P≤0.003), whereas age, sex, diabetes and LVEF exhibited no significant association. Among the predictive models, the random forest model using the combined feature set achieved the highest discrimination (AUC, 0.74; 95% CI, 0.67-0.85) and regression performance (R2=0.26). Models incorporating LVEF alone demonstrated limited predictive capacity, and the combined model did not substantially outperform renal biomarkers alone. Renal dysfunction independently predicts prolonged LOS in HF and outperforms LVEF. Integrating renal biomarkers into admission assessments can improve early risk stratification and guide clinical management by incorporating routinely available measures, such as serum creatinine and eGFR into standardized admission protocols, enabling the early identification of high-risk patients. © 2026, Spandidos Publications. All rights reserved.Copyright © 2026 Abshori et al. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) License.
Faculty of Medicine and Health Sciences, Maulana Malik Ibrahim Islamic State University Malang, East Java 65144, Malang, Indonesia; Faculty of Medicine, Universitas Brawijaya, East Java 65145, Malang, Indonesia; Faculty of Medicine, Universitas Muhammadiyah Surabaya, East Java, Surabaya, 60113, Indonesia; Department of Internal Medicine, Faculty of Medicine and Health Sciences, Maulana Malik Ibrahim Islamic State University Malang, East Java 65144, Malang, Indonesia; Karsa Husada Hospital, East Java, Batu, 65311, Indonesia; Department of Cardiology, Faculty of Medicine, Brawijaya University, East Java 65145, Malang, Indonesia; Faculty of Medicine, Imperial College London, London, W12 0NN, United Kingdom; Faculty of Medicine, Universitas Indonesia, DKI Jakarta, Jakarta, 10430, Indonesia
