Feature Selection and Tree-based Models to Predict MultidrugResistance

• Ó. Escudero Arnanz ^[1] ; I. Mora Jiménez ^[1] ; S. Martínez Agúero ^[1] ; J. Álvarez Rodríguez ^[2] ; C. Soguero Ruiz ^[1]
  1. [1] Rey Juan Carlos University, Fuenlabrada, Madrid, Spain
  2. [2] Intensive Care Department, University Hospital of Fuenlabrada, Madrid, Spain
• Localización: XXXVIII Congreso Anual de la Sociedad Española de Ingeniería Biomédica. CASEIB 2020: Libro de actas / Roberto Hornero Sánchez (ed. lit.), Jesús Poza Crespo (ed. lit.), Carlos Gómez Peña (ed. lit.), María García Gadañón (ed. lit.), 2020, ISBN 978-84-09-25491-0, págs. 464-467
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • Antimicrobial resistance is one of the greatest threats to the health system. Owing to the critical health condition of patients admitted in the Intensive Care Unit (ICU), time is critical to identify bacteria and their resistance to antibiotics. In this paper, the acquisition of the first multidrug-resistance (MDR) has been temporarily characterized by considering data from the University Hospital of Fuenlabrada for 16 years. We have considered different feature selection strategies and data-driven methods inspired in trees, to predict whether a patient will acquire the first MDR or not in the next 24 hours. The obtained results are quite promising, providing a 97% success rate for non-MDR and 83% for MDR cases. They can be considered as a first step to help the clinician in decision making, for example, to isolate the potential MDR patient to avoid the spread of multiresistant strains in the ICU.