EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Machine Learning Versus Traditional Risk Scores for Predicting In-Hospital Mortality in Acute Heart Failure

Ameer Hamza, Sana Yousaf, Hamza Akram, Muhammad Kashif

DEVELOPMENTAL MEDICO-LIFE-SCIENCES · 30.06.2026 · 3 (6) · 32-38

DOI: 10.69750/dmls.03.06.0217 ↗

Kurz gefasst

Fachartikel in der Zeitschrift DEVELOPMENTAL MEDICO-LIFE-SCIENCES, online erschienen am 30.06.2026. Verfasst von Ameer Hamza und 3 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

However, early prediction of in-hospital mortality in acute heart failure needs to be accurate for timely escalation of care. Traditional risk scores are helpful but are not necessarily able to detect the complex nonlinear interactions between clinical and laboratory variables.

Objective

To assess machine-learning algorithms in comparison to traditional risk assessments for predicting in-hospital mortality in patients admitted with acute heart failure.

Methods

This prospective observational study was conducted from January 2025 to August 2025, including 120 consecutive individuals identified with acute heart failure. Traditional risk evaluation was performed with ADHERE, GWTG-HF, and OPTIMIZE-HF metrics. This study used logistic regression, support vector machines, Random Forest, LightGBM, XGBoost, and artificial neural networks as machine learning algorithms. The efficacy of the models was evaluated by AUC, accuracy, sensitivity, specificity, predictive values, and F1-score.

Results

18 patients (15.0%) died in hospital. Non-survivors were older and had lower blood pressure, oxygen saturation, higher creatinine, hyponatremia, elevated lactate, reduced left ventricular ejection fraction, cardiogenic shock, and greater inotropic requirement. GWTG-HF had the highest AUC value when compared to traditional scores (0.78). Among the three models, XGBoost outperformed the others with the largest AUC of 0.91, followed by LightGBM and Random Forest, with values of 88.3%, 83.3% and 89.2% respectively. Systolic blood pressure, serum creatinine, lactate, oxygen saturation, and left ventricular ejection fraction were the best predictors.

Conclusion

The machine-learning models, especially XGBoost, predicted in-hospital mortality in acute heart failure better than traditional risk scores.

Zitieren

APA

Hamza, A., Yousaf, S., Akram, H., & Kashif, M. (2026). Machine Learning Versus Traditional Risk Scores for Predicting In-Hospital Mortality in Acute Heart Failure. DEVELOPMENTAL MEDICO-LIFE-SCIENCES, 3(6), 32-38. https://doi.org/10.69750/dmls.03.06.0217

Vancouver

Hamza A, Yousaf S, Akram H, Kashif M. Machine Learning Versus Traditional Risk Scores for Predicting In-Hospital Mortality in Acute Heart Failure. DEVELOPMENTAL MEDICO-LIFE-SCIENCES. 2026;3(6):32-38. doi:10.69750/dmls.03.06.0217

BibTeX
@article{Hamza2026,
  author = {Hamza, Ameer and Yousaf, Sana and Akram, Hamza and Kashif, Muhammad},
  title = {Machine Learning Versus Traditional Risk Scores for Predicting In-Hospital Mortality in Acute Heart Failure},
  journal = {DEVELOPMENTAL MEDICO-LIFE-SCIENCES},
  year = {2026},
  volume = {3},
  number = {6},
  pages = {32--38},
  doi = {10.69750/dmls.03.06.0217}
}
Kostenlos starten →