4 présentations

• 
  15h30 - 15h55

  A comprehensive framework for designing, improving, implementing, and assessing a Machine Learning tool for automatizing abstract screening in Systematic Literature Reviews: a case applied to Healthcare Logistics

  • Ana María Anaya Arenas, ESG - UQAM
  • Angel Ruiz, Université Laval
  • Julia Isabel Serrato Fonseca, prés., Université Du Québec À Montréal
  • Julia Isabel Serrato Fonseca, prés., Université Du Québec À Montréal

  When performing a Systematic Literature Review (SLR), the abstract screening process can be a very consuming and laborious task, specially when researchers retrieve a significant number of citations after running queries in scientific databases. We achieved a way to automatize the identification of relevant citations, then we condensed it into a comprehensive framework, and developed an open source GitHub tool, so researchers can reproduce our methodology. Nonetheless, the SLR we are performing is related to Operations Research applied to the benefit of Healthcare Logistics, and we plan to use more Natural Language Processing (NLP) techniques so we can identify the main axis and problems in the domain.
• 
  15h55 - 16h20

  Distance-based critical node detection for strategic vaccination policies

  • Faraz khoshbakhtian, prés., Department of Mechanical & Industrial Engineering, University of Toronto
  • Hamidreza Validi, Department of Industrial, Manufacturing & Systems Engineering, Texas Tech University
  • Mario Ventresca, School of Industrial Engineering, Purdue University
  • Dionne Aleman, University of Toronto
  • Randy Giffen, IBM Canada
  • Proton Rahman, Eastern Health, Newfoundland & Labrador

  Vaccination is a crucial tool to mitigate the spread of infectious diseases, and effective vaccination policies are needed to ensure significant mitigation of disease spread with a limited budget of vaccines. We develop a novel approach to formulate optimal vaccination policies under budget constraints using the distance-based critical node detection problem (DCNDP) approach. DCNDP identifies a subset of nodes in a network whose removal maximally degrades a pre-defined distance-based metric of connectivity. We introduce a lightweight integer programming (IP) model for 2-hop DCNDP and implement a DCNDP pipeline that combines exact IP formulation with divide-and-conquer partitioning and heuristics to find near-optimal solutions in networks with hundreds of thousands of nodes. Using a contact network of Newfoundland & Labrador with 500,000 individuals and a budget of 20% vaccine coverage, our approach divides the network into 30 partitions and by spending just 1 hour on each partition it decreases the overall 2-hop connectivity of the network by 84%. We further enhance our approach by integrating granular agent-based pandemic modelling and machine learning to derive effective vaccination policies. Our framework can be adapted to different regions and can be used as a decision-making tool to enhance public health outcomes.
• 
  16h20 - 16h45

  Clinical Trajectories for Short and Long Term Capacity Management in Healthcare

  • Eduardo Redondo, prés., Cirrelt
  • Angel Ruiz, Université Laval
  • Valérie Bélanger, CIRRELT, HEC Montréal

  Capacity management is critical in healthcare as it impacts patient outcomes, staff workload, and hospital costs. Our study proposes an approach for capacity management in healthcare that leverages the use of clinical trajectories. Patient with similar needs are identified, grouped and represented as archetypes, and a trajectory reflecting their expected resource utilization over time is associated with them. A generator of patients allowed us to emulate the uncertainty of heterogeneous patients' needs and their arrivals. For the short-term simulation techniques were used to enable rapid experimentation and assessment of different scenarios in a controlled environment without incurring additional costs. For long-term capacity management, we used stochastic optimization techniques to find an optimal staffing level over a given period that meet demand while minimizing a cost function. The uncertainty within patient arrival and resource utilization/allocation processes was represented by stochastic parameters in the long-term optimization model and the internal competition for resources in the short-term simulations. Our results (1) demonstrated that considering clinical trajectories can improve capacity planning accuracy while enabling proactive resource allocation that leads to better patient outcomes, and (2) highlights the benefits of considering actors' heterogeneity and demand uncertainty, where demand consists of chained resource consumption over time.
• 
  16h45 - 17h10

  Cluster-based Trajectory Analytics of Functional Decline and Recovery in Older Adults

  • Ghazal Khalili, prés., DeGroote School of Business, McMaster University
  • Manaf Zargoush, DeGroote School of Business, McMaster University
  • Kai Huang, DeGroote School of Business, McMaster University

  The ability to perform activities of daily living (ADL) independently is a crucial measure of an individual's health status, and its loss can significantly impact their well-being. This study aims to analyze the trajectories of functional decline and recovery among patients and identify typical patterns in ADL performance. Understanding these trajectories can help develop personalized interventions and improve individuals' health outcomes, leading to reduced direct and indirect costs for their families and society as a whole.
  While various methods have been employed to study ADL trajectories, this study is unique in its application of clustering and sequence analysis approaches to investigate typical functional decline and recovery sequences. A Markov model is incorporated to address computational challenges in the clustering phase. The study's results have significant implications for healthcare professionals and policymakers, as they can use them to develop targeted and effective interventions that support patients in maintaining their independence and improving their quality of life.
  Overall, this study contributes to the field of healthcare analytics by using clustering and advanced sequence analysis methods to understand the trajectories of functional decline and recovery among older adults. The findings can help healthcare providers make more informed decisions and optimize patient care.