INFORMS 2021 Program Book

INFORMS Anaheim 2021

TB23

2 - Designing Physician Payments for Diagnostic Accuracy under Limited Outcome Visibility Elodie Adida, University of California-Riverside, Riverside, CA, 92521-9800, United States, Tinglong Dai The prevailing payment system in the U.S. does not sufficiently incentivize physicians to exert diagnostic effort. Meanwhile, misdiagnosis remains frequent and hard to keep track of. In this paper, we develop a model to analyze the effect of a fee-for-service payment system on diagnostic accuracy, physician effort, and social welfare for a given condition. We consider a partially altruistic physician who may (1) exert costly but non-reimbursable effort and/or (2) order a reimbursable diagnostic test that is costly to the patient. Exerting effort generates an imperfect signal, whereas performing a test detects the patient’s true condition. We also analyze a diagnosis-based payment system and study how to best approach the socially optimal outcome. 3 - The Role of Schedule Volatility in Employee Turnover: The Case of Home Health Care Alon Bergman, Postdoctoral Scholar, University of Pennsylvania, Philadelphia, PA, United States, Guy David, Hummy Song High rates of employee turnover increase employee search costs and training costs, and can contribute to the loss of organizational knowledge and productivity. In the case of licensed nurses, high job turnover is likely to raise healthcare costs and reduce its quality. This paper identifies schedule volatility, an operationally measurable element, to be a key determinant of nursing turnover in home health care. Using administrative data from a large home health agency, we define and document different measures of worker schedule volatility, and recover causal estimates for the effect of schedule volatility on nurses’ voluntary separation (quitting) using an IV approach. We then consider several counterfactual scheduling policies. Through simulations, we calculate the counterfactual schedule volatility resulting from each policy and its effect on nursing turnover. TB23 CC Room 204C In Person: Theoretical and Empirical Models in Service Operations General Session Chair: Lina Wang, Arizona State University, Scottsdale, AZ, 85257, United States 1 - The Impact of Piracy on Movie Distribution Franco Berbeglia, Professor, Purdue University, West Lafayette, IN, 15232-2962, United States, Timothy Derdenger, Michael D. Smith, Rahul Telang We develop and estimate a dynamic discrete choice model that embeds piracy downloads as a substitute for legal distribution channels. In general, high quality pirated movies become available as soon as there is a home video released in some market, which then impacts demand in all other markets, as piracy belongs to a unique global market. The current problem in the industry is that home video releases are optimized locally, without considering the impacts created by piracy on other markets. Even though piracy may not be eliminated; it can be controlled by choosing the release timing of foreign releases more wisely. Through counterfactuals, our model can predict the impact on studio revenues of delaying piracy, which may inform business practitioners about the trade-off they face when optimizing local home video releases. 2 - Delegation with Technology Migration: An Empirical Analysis of Mobile Virtual Network Operators Fan Zou, University of South Carolina, West Columbia, SC, 29169, United States, Yan Dong, Kejia Hu, Sriram Venkataraman This study examines the impact of mobile virtual network operators (MVNOs) on the performance of mobile network operators (MNOs) in the presence of overlapping generations of wireless mobile technologies (2G and 3G). MVNOs distribute MNOs’ mobile services to customers without owning any spectrum or network infrastructures. Some MVNOs are wholly owned by MNOs with a revenue-sharing mechanism (branded MVNOs), while others operate through wholesale agreements with MNOs (third-party MVNOs). By focusing on the impact of MVNOs on MNOs’ performance, we investigate governance issues, e.g., delegation vs. ownership, in value chains with overlapping generations of technologies.

TB20 CC Room 203B In Person: Data-driven Modeling for Disease Management General Session Chair: Gizem Nemutlu, Brandeis University, Waltham, MA, 01803- 3872, United States 1 - Determining the Optimal Covid-19 Testing Centre Locations and Capacities Considering the Disease Dynamics and Target Populations Esma Akgun, University of Waterloo, Waterloo, ON, N2L 6P1, Canada, Sibel Alumur Alev, F. Safa Erenay Testing individuals at risk to identify COVID-19 infections and isolating them help control and mitigate the pandemic. However, during the peaks, the existing testing capacity may need to be expanded. We develop a location and capacity allocation model integrated with an SEIR model to determine the optimal locations of new pop-up testing centers, capacities of the existing centres, and the assignments of demand regions to the testing centres considering time-variant testing demand due to ever-changing disease prevalence. The objective function is to minimize the total distance traveled subject to budget and capacity constraints. We applied the model to the case of Ontario, Canada using real data. 2 - Limits of Capacity Flexibility: Impact of Hallway Placement on Patient Flow and Quality of Care in the Emergency Department Arshya Feizi, PhD Candidate, Boston University, Boston, MA, 02134, United States, William Baker A common practice in busy emergency departments (EDs) is to admit patients from the waiting area to hallway beds as the regular beds fill up. Using data from a large ED, we first perform a causal analysis to quantify the impact of hallway placement on wait times and quality of care as defined by disposition time, ED length of stay (LOS) and likelihood of adverse outcomes. Next, we perform a counterfactual analysis using a data-driven simulation of the ED to find better hallway usage policies. We find that a pooling policy, where hallway beds are used only if all regular beds are full, has the greatest impact on reducing wait times, albeit at the cost of higher hallway utilization. Also, too little or too much wait tolerance for rooming patients may result in over- or under-utilization of the hallway space, both of which are detrimental to ED average throughput times and wait times. 3 - Liver Cancer Surveillance in the Era of New Hepatitis C Antiviral Treatments: A Value of Information Analysis Gizem S. Nemutlu, Brandeis University, Waltham, MA, 01803- 3872, United States, Jagpreet Chhatwal The treatment landscape for chronic hepatitis C has changed with the use of direct-acting antivirals; 95-100% of individuals with hepatitis C can now be cured. However, the risk of liver cancer is not eliminated for individuals with advanced liver disease. The value of routine cancer surveillance in this population is widely debated; long-term data on liver cancer incidence is lacking. Our objective was to evaluate the value for future research on liver cancer surveillance in hepatitis C cured individuals using a validated microsimulation model. We estimated the cost-effectiveness of routine surveillance and performed a value of information analysis informed by the population-level expected value of perfect information to determine the value of future research. Our analysis showed that the routine surveillance was cost-effective only in individuals with cirrhosis. TB22 CC Room 204B In Person: Healthcare: Incentives and Operations General Session Chair: Alon Bergman, University of Pennsylvania, Wynnewood, PA, 19096, United States 1 - Scheduling Smarter: Scheduling Decision Impact on Nurse-Aide Turnover Kevin Mayo, Indiana University, Bloomington, IN, 47408, United States, Eric Michael Webb, George Ball, Kurt M. Bretthauer High turnover rates in long-term nursing facilities exacerbates the current and worsening shortage of caregivers. Part-time Certified Nursing Assistants (CNAs) provide a significant amount of patient care in these facilities and have high turnover rates, potentially harming health outcomes and increasing cost of care. We empirically analyze the effect of scheduling decisions on part-time CNA turnover. Using novel data for 6,221 part-time CNAs at 157 facilities over a 26- month period, we identify three scheduling levers that can reduce turnover: reducing co-worker variability, reducing variation in weekly scheduling and increasing hours worked which follows a nuanced non-linear relationship. These findings suggest that smart managers will benefit from identifying quality workers and assigning them more consistent schedules as part of a team.

105