SCRO / Journées de l'optimisation

HEC Montréal, 29-31 mai 2023


HEC Montréal, 29 — 31 mai 2023

Horaire Auteurs Mon horaire

HOMII Healthcare Operations Management II

30 mai 2023 15h30 – 17h10

Salle: Serge-Saucier (bleu)

Présidée par Vahid Roshanaei

4 présentations

  • 15h30 - 15h55

    Mechanisms Driving Service Duration: A Large-Scale Empirical Analysis

    • Hossein Abouee Mehrizi, University of Waterloo
    • Hamid Arzani, prés., University of Toronto
    • Mohammad Hossein Eshraghi, University of Waterloo

    The COVID-19 pandemic has resulted in a significant backlog of medical surgeries worldwide, making it crucial to improve system throughput and reduce these backlogs. A critical factor affecting system throughput is the duration of medical procedures. This paper investigates the determinants of MRI procedure service time and explores how these determinants can be leveraged to increase throughput. We conducted a comprehensive empirical analysis using patient-level MRI services data from 66 hospitals in Ontario, Canada. Our findings show that several covariates, including batching of similar procedures, the time of day when the procedure is performed, and the system's workload, affect service duration. For instance, performing an additional procedure of a similar body type within a batch can reduce the next procedure's duration by approximately 4.2%. We demonstrate through a trace-driven simulation that proper sequencing can significantly increase the daily number of procedures. We also observed that emergent procedures have longer service durations during night shifts, while low-priority procedures experience shorter MRI scan times. Furthermore, our analysis reveals an inverted-U-shaped relationship between service duration and workload, which has been previously observed in the literature for other applications. Our results provide practical insights that hospitals can use to improve MRI procedure throughput.

  • 15h55 - 16h20

    Optimal Use of Home Hemodialysis Using Competitive Incentive Plans

    • Maryam Afzalabadi, prés., Lazaridis School of Business and Economics, Wilfrid Laurier University
    • Mojtaba Araghi, Lazaridis School of Business and Economics, Wilfrid Laurier University
    • Salar Ghamat, Lazaridis School of Business and Economics, Wilfrid Laurier University

    An increasing number of patients with end-stage renal disease receive long-term dialysis every year. Although available evidence suggests that home-based hemodialysis (HHD) may achieve similar clinical outcomes to in-center hemodialysis (ICHD) and are less resource intensive, this treatment modality has been underutilized with dialysis facilities. To increase the utilization rate of HHD, we formulate a target-based incentive model utilizing a comparative approach in the form of exogenous achievement benchmarks alongside the providers’ own improvement. The existing literature on incentive payment models in healthcare has been primarily focused on performance-based and target-based incentive models separately. Furthermore, we propose a novel “competitive” incentive plan in which, instead of exogenous achievement benchmarks, the rank of the providers is the criteria for qualifying for the achievement rewards.
    Our paper is the first in the performance-based payment literature to analytically study incentive models tied to individual performance (improvement) and competition (achievement) simultaneously.

    This approach can be considered a significant development in incentive payment model literature since it can incentivize all the participants to obtain a reward regardless of their initial performance. We obtain the equilibrium solution for target-based and competitive incentive models and analyze the behavior of system equilibria in different environmental and individual settings.

  • 16h20 - 16h45

    Peel Pack Planning Using Clustering and Decomposition Approach

    • Yixuan Wang, prés.,
    • Satyaveer S. Chauhan, Concordia University

    In order to improve the operational efficiency in the Operating Room, hospitals customize surgical trays for each surgical procedure. Since, surgical procedures involve a variety of patients and surgeons, the usage of surgical instruments (in terms of quantity) differs from case to case. This poses a significant challenge as the variability in instrument usage makes it difficult to determine the optimal quantity for each instrument for each procedure. In this study, we address the issue of excessive waste in surgical trays by proposing the implementation of custom peel packs. These peel packs could be used (in place of a new surgical tray) if the surgical tray ran out of the instruments. Our objective is to reduce waste by designing custom peel packs associated with multiple surgical procedures while ensuring that all the necessary instruments are available during the procedure without opening a new main tray. We present one decomposition approach to finding the exact solution and a simplified fast approach based on clustering and mathematical programming to find the near-optimal solution in a fractional time compared to the exact approach. Numerical experiments demonstrate the performance of the presented approaches. The findings indicate that the proposed K-means clustering method is more efficient in configuring peel packs.

  • 16h45 - 17h10

    Multi-Class Advance Patient Scheduling: A Data-Driven Robust Approach

    • Hossein Abouee Mehrizi, University of Waterloo
    • Hamid Arzani, prés., University of Toronto
    • Saeed Ghadimi, University of Waterloo

    In this paper, we study a multi-class advance patient scheduling problem where patients of different classes have different service times and incur different waiting costs to the system. It is known in the literature that dynamic advance patient scheduling is a challenging problem due to the high variability in the daily arrivals and high dimensionality of the problem. To overcome these challenges, we focus on analyzing the regret of any online scheduling policy relative to an offline controller. This helps us to develop a novel dynamic optimization framework where the problem can be approximately decomposed into multiple single-stage stochastic problems. Each single-stage problem determines how patients who arrived in that period should be scheduled over a booking horizon. We then extend the framework to a data-driven setting where the true distribution of demand in each period is unknown. We propose an algorithm to solve the robust model and examine its performance by leveraging the MRI data from hospitals in Ontario. We observe that the proposed approach schedules patients such that the system is efficiently protected against the high variability in the daily and performs well compared to an offline policy that is endowed with the full knowledge of demand.