Optimization Days 2024

HEC Montréal, Québec, Canada, 6 — 8 May 2024

Schedule Authors My Schedule

WA7 - Healtchare

May 8, 2024 10:30 AM – 12:10 PM

Location: Quebecor (yellow)

Chaired by Duygu Tas Kuten

3 Presentations

  • 10:30 AM - 10:55 AM

    Improving four clinical pathways in respiratory diseases using process mining. An application to a Quebec-specialized hospital.

    • Luca Murazzano, presenter, Université Laval
    • Paolo Landa, Université Laval
    • Jean-Baptiste Gartner, Université Laval
    • Mohamed Hakim Raki, Université Laval
    • André Côté, Université Laval

    Quebec has an increasingly aging population with a growing number of long-term and chronic conditions. Within these chronic conditions, pulmonary and respiratory diseases impact 8% of the overall Quebec population. It is important to meet the service demand effectively and efficiently for this population, analyzing the existing care processes and ensuring the right service configuration. The objective of this study consists of understanding and optimizing the organizational costs and performance of the treatments provided to the patients of the Cardiology and Respiratory University Hospital in Quebec, using a process mining approach. The application of this method will identify how the hospital organization can best deploy resources to meet the needs of the patients. Hospital data from 2018 to 2022 was collected from four macro clinical activities: Emergency Department, Ambulatory, Hospitalisation, and Medical Imaging. Data were processed through a subsequent cleaning and refining process to obtain the raw clinical pathways of the patients. Then a subsequent pass through a process mining tool enabled the identification of macro and micro trajectories for the different disease types. The results made it possible to identify critical points within the clinical pathways and where action is needed to make the care pathway more efficient.

  • 10:55 AM - 11:20 AM

    Prédiction de la congestion à l'urgence à travers les délais logistiques hospitaliers

    • Nadia Lahrichi, Polytechnique Montréal
    • Louis-Martin Rousseau, Polytechnique Montréal
    • Ilitea Kina, presenter, Polytechnique Montréal

    La congestion des services d'urgence est une menace nationale depuis quelques années, exacerbée par une pénurie de personnel soignant accrue depuis la pandémie. Alors que ce contexte appelle à des efforts d'optimisation, peu d'études ont tenté de prédire la congestion en tant que telle. En utilisant le score de congestion EDWIN, validé préalablement dans la littérature, sur les données du Centre Hospitalier de l'Université de Montréal de juillet 2018 à décembre 2021, nous avons étudié la congestion par l'analyse des séries temporelles et entraîné des modèles de régression logistique, des arbres de classification, des forêts aléatoires, des arbres boostés par le gradient et des arbres de classification optimaux pour la prédire la congestion dans 24 heures et dans 48 heures. Bien que les meilleurs modèles de prédiction soient les forêts aléatoires (précision de 0,902 à 24 heures) et les arbres boostés (précision de 0,903 à 24 heures), l'utilisation d'une fonction d'autocorrélation pour établir des délais individualisés pour chaque variable a augmenté la précision de 4,5 à 6 % dans les meilleurs modèles. Les variables les plus importantes dans tous les modèles étaient les scores de congestion antérieurs il y a 12, 24 et 48 heures. La congestion des urgences est donc un phénomène prévisible et un problème qui s'entretient dans le temps.

  • 11:20 AM - 11:45 AM

    Stochastic Operating Room Allocation Problem

    • Duygu Tas, presenter, Sabancı University
    • Raf Jans, HEC Montréal

    In this study, we consider a multi-period operating room allocation problem with stochastic surgery durations (ORAP-SSD). In real-life applications, uncertainty in the surgery durations may lead to postponing the starting times of surgeries and using operation rooms (ORs) beyond their capacities. Thus, we propose a model that considers not only the fixed costs of opening ORs but also the penalty costs proportional to the deviation from the OR capacity (expected overtime) and to the deviation from the starting times of surgeries (expected waiting time). We first assume that ready times of patients are provided by the surgeons (preferred starting times). We further extend the model by considering not only the scheduled starting times but also the patient ready times as decision variables. Moreover, for a given allocation of ORs, we develop a procedure to exactly compute the expected overtime and the expected waiting time where surgery durations follow a given probability distribution. To construct solutions for the ORAP-SSD, we then propose a formulation based on a two-stage stochastic programming model with a set of sample scenarios generated for the surgery duration realizations.

Back