Machine Learning-Based Hospital Readmission Prediction: A Comparative Analysis of Speciality-Specific vs. All-Specialities Models.

Journal: Studies in health technology and informatics
Published Date:

Abstract

Hospital readmissions are a major challenge for healthcare systems, leading to increased costs and adverse patient outcomes. Predicting which patients are at risk of readmission is critical for improving care and optimizing resource allocation. This study explores the effectiveness of machine learning models in predicting hospital readmissions, comparing the performance of speciality-specific models with a general, all-specialties model. Using data from 79,886 admissions across different medical specialties we trained a variety of machine learning algorithms. Our results show that while speciality-specific models tend to achieve better performance, the difference is not statistically significant and are more prone to overfitting.

Authors

  • Teresa García-Navarro
    Vicomtech Foundation, Basque Research and Technology Alliance (BRTA), Spain.
  • Jon Kerexeta
    Vicomtech Foundation, Basque Research and Technology Alliance, (BRTA), 20009 Donostia, Spain.
  • Maria Rollan-Martínez-Herrera
    Vicomtech Foundation, Basque Research and Technology Alliance (BRTA), Spain.
  • Paula Laccourreye
    Biodonostia Health Research Institute, 20014 San Sebastián, Spain.
  • Regina Martínez Idarreta
    Clinica Asuncion, Tolosa, Spain.
  • Francisco Martínez
    Department of Applied Mathematics and Statistics, Technological University of Cartagena, Cartagena 30203, Spain.
  • Nekane Larburu
    Vicomtech Foundation, Basque Research and Technology Alliance, (BRTA), 20009 Donostia, Spain.