4 Presentations

• 
  03:30 PM - 03:55 PM

  Queueing Causal Models for Comparative Analytics in Service Systems

  • Opher Baron, University of Toronto
  • Dmitry Krass, University of Toronto
  • Mark van der Laan, University of California, Berkeley
  • Arik Senderovich, York University
  • Zhenghang Xu, presenter, University of Toronto

  Simulation is a powerful tool for the prescriptive analysis of queueing models. With ample data and expert knowledge of the underlying system structure, a good model can be constructed and used to predict impact of various interventions. However, such manual construction is both time- and skilldemanding. Moreover it is somewhat subjective – if the expert failed to note an important feature of the system (e.g. different customer types receiving different service priorities), the model will not be accurate. As an alternative, we propose a data-driven representation of system building blocks, justified by the G-computation results from causal inference literature. We describe the queueing data generation process with structural equations and apply machine learning models to fit the equations to the data. Through numerical experiments, we show that this approach can replace the explicit queueing dynamics of the simulator. Our model is shown to capture the intervention effect in M/G/c queues with independent hyper-exponential service time and first-come-first-serve queueing discipline.
• 
  03:55 PM - 04:20 PM

  Co-exposure management in service systems

  • Binyamin Oz, presenter, Hebrew University of Jerusalem
  • Yael Perlman, Bar-Ilan University

  We consider multi-class service systems where customers (or jobs) are affected by being exposed to other customers while waiting for service. We study the steady-state performance with respect to some co-exposure measures under different scheduling policies.
• 
  04:20 PM - 04:45 PM

  Adaptive Server Behavior to Schedule Deviations and Its Consequences: Evidence from Operating Rooms

  • Yiwen Jin, presenter, UBC Sauder School of Business
  • Yichuan Ding, McGill University
  • Steven Shechter, University of British Columbia
  • Jugpal Arneja, UBC Department of Surgery

  We study how clinical teams adaptively respond to real-time deviations from the planned operating room (OR) schedules and the associated consequences of these responses. Specifically, we investigate whether clinical personnel will adjust their service speed when they are ahead of or behind the original schedule and whether this affects patient readmission and reoperation rates. We empirically explore these questions using a unique surgery data set that includes actual and scheduled surgery time stamps. Using Arellano-Bond GMM estimation we find that surgical teams prolong the subsequent surgery by 10.5% on average when facing one standard deviation (SD) early start and expedite by 5.6% when facing one SD delayed start. In the turnover stage of a shift, the turnover duration is expedited when facing delay (and is prolonged when ahead of schedule). Overall the cleaning teams respond more intensely than the surgical teams, whereas the slowing down effect when being ahead is stronger than the speedup effect for both. We then leverage the deviation from the scheduled start as instruments. We present a causal study that a faster-than-scheduled procedure duration erodes surgical quality by increasing 30-day readmission and reoperation probabilities.
• 
  04:45 PM - 05:10 PM

  Improving the Uninsured ESRD Patients’ Experience: The Case of Compassionate Dialysis

  • Olga Bountali, presenter, Rotman School of Management
  • Sila Cetinkaya, Southern Methodist University
  • Farnaz Nourbakhsh, Southern Methodist University

  Uninsured patients suffering from chronic diseases may have access to medical treatment under federal law, EMTALA, only after being evaluated as in ‘emergent, life-threatening condition’. In the case of End Stage Renal Disease (ESRD) patients who do not have access to regular treatment, the practice of offering dialysis conditional on a screening assessment in the emergency room (ER) is known as “compassionate dialysis”. However, the screening assessment itself may lead to severe congestion in the ER and significant treatment delays for the patients. In this paper, we explore the impact of the current protocol at play and various mitigation strategies to compare the value of patient-centric outcomes relative to the current practice.