2016 INFORMS Annual Meeting Program

WC90

INFORMS Nashville – 2016

WC90 Broadway D-Omni Health Care, Modeling XV Contributed Session

4 - The Impact Of Data Quality: A Study On The Coast Guard’s Data Nelson Christie, Rutgers University, Princeton, NJ, 08540, United States, christie.l.nelson.phd@gmail.com We report our findings with the US Coast Guard on a project designed to identifying errors in a large operational database. We combined interview results with statistical algorithms to identify a large number of errors in the data. We then examine the impact that data quality has on operational planning.

Chair: Shanshan Wang, PhD Candidate, Beijing Institute of Technology, 5 southstreet Zhongguancun, Haidian District, Beijing, 100081, China, shshwang_bit@163.com 1 - Safety Stock For Blood Products With Short Shelf Life Christine Pitocco, Research Professor, Stony Brook University, 202 Harriman Hall, Harriman Hall Room 202, Stony Brook, NY, 11794- 3775, United States, christine.pitocco@stonybrook.edu, Katsunobu Sasanuma Poorly managed inventory of apheresis platelets in a blood bank can result in a loss of revenue and safety issues for patients in need. A safety stock of platelets must be available, but higher levels of safety stock may cause wastage if not utilized. We discuss how the safety stock level should change according to the change in demand and shelf life. We propose an optimal inventory control policy based on a simulation of blood bank operations. 2 - Facility Location Problem For Stochastic Mixed-integer Programming In Healthcare Mengnan Chen, University of Central Florida, 12800 Pegasus Drive, PO Box 162993, Orlando, FL, 32816-2993, United States, cmn891127@knights.ucf.edu, Qipeng Zheng This paper considers a facility location problem with patients’ appointment and physician scheduling. We model this problem as a two-stage optimization problem. In the first stage, depending on the patients’ choices, which is relative to their characteristics and physicians/clinics’ attributes, physicians will be scheduled to the different clinics. In the second stage, the central hospital will match patients’ choices and physicians’ scheduling. Using discrete choice model, we estimate the probability for patient’s choice. Let the scenario is the different combination of patients’ choices, then we can develop a stochastic mixed-integer programming to solve the facility location problem. 3 - Test Modality Capacity Simulation: A Nuclear Medicine Radiology Assessment Haris Ackerman, Management Engineer, Virtua Health, 303 Lippincott Drive, Marlton, NJ, 08053, United States, hackerman@virtua.org, Mojisola Otegbeye, Hala Sweidan Significant delays in the nuclear medicine radiology department of a 433 bed acute-care hospital increases patient length of stay resulting in patient dissatisfaction and reduced reimbursement rates. Simulation modeling deployed to show a budget neutral increase in daily stress test fulfillment rate from 80% to 99.9% while maintaining current staffing roster by utilizing optimal staff scheduling patterns. 4 - Outpatient Appointment Scheduling And Sequencing Model With Uncertain Service Time And Correlation Shanshan Wang, PhD Candidate, Beijing Institute of Technology, 5 Southstreet Zhongguancun, Haidian District, Beijing, 100081, China, shshwang_bit@163.com, Jinlin Li, Chun Peng As the window of hospital, outpatient appointment scheduling and sequencing plays a critical role in the allocation of healthcare resources. We take different jobs and uncertain service time into consideration. Based on support and moment of service time distribution, we employ mean absolute deviation to capture its correlation, propose distributionally robust models, and can be reformulated them as tractable counterparts. Numerical results show that when sequence is fixed, it’s optimal to allocate time allowances with a decreasing pattern. When considering “New” and “Repeat” patients, optimal outpatient sequence of repeat patients is in the front of new patients.

WC89 Broadway C-Omni

Large-Scale Optimization in Transportation Sponsored: TSL, Intelligent Transportation Systems (ITS) Sponsored Session Chair: Velibor Misic, Massachusetts Institute of Technology, Massachusetts Avenue, Cambridge, MA, 02139, United States, vvmisic@mit.edu 1 - Planning Optimization For Integrated Transportation Systems Bradley Sturt, Massachusetts Institute of Technology, Massachusetts Avenue, Cambridge, MA, 02139, United States, bsturt@mit.edu, Dimitris Bertsimas, Sebastien Martin, Yee Sian Ng, Julia Yan Passengers move through large cities via various public transportation systems, such as subway and bus systems. City operators need to decide how to schedule the trains and buses throughout the day. Prior work has addressed making the decisions for each transportation system in isolation, which may result in a suboptimal citywide transportation system. This work proposes an optimization approach for holistically and cooperatively optimizing the decisions for decision makers for the subway, bus systems and the city. 2 - From Physical Properties Of Transportation Flows To Demand Predictions: An Optimization Approach Julia Y. Yan, Massachusetts Insitute of Technology, Massachusetts Avenue, Cambridge, MA, 02139, United States, jyyan@mit.edu, Dimitris Bertsimas Transportation system management requires accurate demand data. The main data sources are often aggregated datasets such as entry/exit data, and one must recover the original demand. Such problems are generally underspecified. We present an optimization framework to recover origin-destination matrices under minimal assumptions, enforcing reasonable physical constraints such as flow conservation, smoothness, and sparsity. We evaluate this on real-world datasets and show 6-7% improvement in R2 over a baseline. 3 - Online Taxi Routing In New York City Sebastien Martin, Massachusetts Institute of Technology, Massachusetts Avenue, Cambridge, MA, 02139, United States, semartin@mit.edu, Dimitris Bertsimas, Patrick Jaillet Taxi dispatching used to have little room for optimization. However, more and more customers request cabs from their cellphone. This gives transportation network companies prior information that can be leveraged to achieve a better efficiency. Large-scale taxi routing has usually been done with simple rules or heuristics. Our work proposes ways to scale optimization-based online routing algorithms to the largest instances of vehicle routing with real data. We use historical taxi trip data in New York City to dispatch in real time thousands of taxis and serve tens of thousands of customers. 4 - A Modern Optimization Approach To The Airlift Planning Problem For The United States Transportation Command (USTRANSCOM) Velibor Misic, Massachusetts Institute of Technology, Massachusetts Avenue, Cambridge, MA, 02139, United States, vvmisic@mit.edu, Dimitris Bertsimas, Allison An Chang, Nishanth Mundru USTRANSCOM plans missions globally, the majority traveling by air. These missions are challenging to plan due to their combinatorial nature and complex constraints. We propose a novel solution approach that combines local search, mixed-integer optimization and column generation, and show that it provides high quality solutions. This material is based upon work supported by USTRANSCOM under Air Force Contract No. FA8721-05-C-0002. Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of USTRANSCOM.

455