INFORMS 2021 Program Book

INFORMS Anaheim 2021

TD20

3 - Integrated Conditional Estimation-Optimization Meng Qi, University of California, Berkeley, Berkeley, CA, 60614, United States, Paul Grigas, Zuo-Jun Max Shen Many real-world optimization problems have an objective function defined by a random parameter whose probability distribution depends on a contextual feature vector. In contrast to the standard way of first estimating the uncertainty then optimizing the objective based on estimation, we propose an Integrated Conditional Estimation-Optimization (ICEO) method that estimates the underlying conditional distribution of the random parameter while considering the structure of the optimization problem. This allows us to fundamentally capture the correlation between the contextual features, especially when the distribution of contextual features varies. We prove the asymptotic consistency and finite performance guarantees in the form of generalization bounds. We also provide computational methods and numerical studies. TD22 CC Room 204B In Person: Empirical and Behavioral Research in Healthcare General Session Chair: Hyun Seok (Huck) Lee, Korea University Business School, Seoul, 97333-3235 Co-Chair: Junghee Lee, Tulane University, New Orleans, LA, 70118- 5669, United States 1 - Learning in Drug Shortages Hyun Seok (Huck) Lee, Korea University Business School, Seoul, 97333-3235, Korea, Republic of, Junghee Lee, In Joon Noh In this study, we investigate whether pharmaceutical manufacturing plants learn from their own drug shortage instances. Specifically, we examine if more drug shortages recovery at a plant lead to quicker recovery from its subsequent shortages. We also investigate factors that might affect this learning. Our findings will have policy implications for the FDA and will also contribute to the academic literature on learning. TD24 CC Room 205A In Person: Supply Chain & Operational Risk Management General Session Chair: Na Rea Cho, University of Alabama, Tuscaloosa, AL, 35405, United States 1 - Concurrent Sourcing under Supply and Demand Uncertainty Bryant Cassidey, University of Alabama, Tuscaloosa, AL, United States, Nickolas K. Freeman, Sharif Melouk A central question in the supply chain strategy literature related to Supply Chain Risk Management (SCRM) asks how a firm should delineate its boundary with respect to goods composing a product or service it offers. We investigate the setting in which a firm may choose to make and buy components used to manufacture finished goods (concurrent sourcing), and determine the optimal decision strategy under uncertain supply and demand. For simplicity, we assume supply is uncertain in an all-or-nothing manner: either the supply is totally disrupted or not. We show that the optimal strategy follows a threshold structure defined by problem parameters. We also investigate the effect of concurrent sourcing on the supplier’s optimal pricing strategy. Our analysis highlights the conditions under which a manufacturer and a supply chain system benefit the most from concurrent sourcing. 2 - Managing Residential Energy Storage In this paper, we examine a homeowner’s battery management policy when they have access to intermittent renewable energy and are connected to a grid with feed-in tariffs and time-of-use electricity prices. The cost-minimizing battery operating policy depends on the configuration of electricity prices in the market. When electricity sellback prices are lower than purchase prices from the grid, we show that a simple heuristic with a charge up to level in the off-peak period and a discharge down to level in the peak period performs extremely well relative to the optimal solution. We compare the performance and emissions reductions resulting from this recommended policy against other commonly utilized heuristics (full charge/discharge, no battery, do nothing). Na Rea Cho, University of Alabama, Tuscaloosa, AL, 35405, United States, Karthik Murali, Youngsoo Kim, Mesut Yavuz

TD20 CC Room 203B In Person: Pharmaceutical Supply Chains General Session Chair: Minje Park, Boston University, Boston, MA, 02215-1704, United States 1 - Changing Standards and Drug Shortages in the Pharmaceutical Industry Ivan Lugovoi, The Ohio State University, Columbus, OH, United States Matching supply and demand is a fundamental task of supply chain management. Failure to supply a product is painful for consumers, but particularly so in the pharmaceutical industry, where the product is often necessary for the treatment of life-threatening diseases. Drug shortages, therefore, pose significant public health threats. One important manufacturing quality issue is a drug’s non- compliance with quality standards. Our research examines changes in such quality standards. A change of a quality standard can lead to a compulsory change in the manufacturing or testing technology, and such technological change can, in turn, lead to manufacturing quality issues or the decision to completely cease production. As a result, the total manufacturing capacity for a drug in a market can be adversely impacted by a quality standard change. 2 - The Impacts of a Non-profit Organization on Drug Shortages Junghee Lee, University of Notre Dame, Notre Dame, IN, 70118- 5669, United States, Hyoduk Shin, Daewon Sun The ongoing shortage of pharmaceutical drugs critically threatens public health. To mitigate the drug shortages, philanthropies and hospital systems founded a non-profit organization that “better” sources and even manufacturers essential medicines. We investigate how the advent of the non-profit entity reshapes the competition and impacts the performance of each entity in a pharmaceutical drug supply chain. TD21 CC Room 204A In Person: Robust and Stochastic Decision-making under Uncertainty General Session Chair: Meng Qi, University of California, Berkeley, Chicago, IL, 60614, United States 1 - The Power of Adaptivity for Stochastic Submodular Cover Rohan Ghuge, University of Michigan, Ann Arbor, MI, 48105- 2542, United States, Anupam Gupta, Viswanath Nagarajan In the stochastic submodular cover problem, the goal is to select asubset of stochastic items of minimum expected cost to cover a submodular function. Solutions in this setting correspond to a sequential decision process that selects items one by one “adaptively” (depending on prior observations). While such adaptive solutions achieve the bestobjective, the inherently sequential nature makes them undesirable in many applications. We ask: how well can solutions with only a few adaptive rounds approximate fully-adaptive solutions? We consider both cases where the stochastic items are independent, and where they are correlated. For both situations, we obtain nearly tight answers, establishing smooth tradeoffs between the number of adaptive rounds and the solution quality, relative to fully adaptive solutions. 2 - Models and Methods for Ambulance Dispatch Anton J. Kleywegt, ISyE Georgia Tech, School of Ind and Systems Eng, Atlanta, GA, 30332-0205, United States, Vincent Guigues Ambulances are controlled by dispatch decisions. First, when an emergency call arrives, a decision is made which ambulance to dispatch, or whether to place the call in queue to wait for a later dispatch. A number of factors should be taken into account when making this dispatch decision, including the following: (a) Ambulance type and crew. Ambulance services operate different types of ambulances, and ambulance crews have different levels of skills. (b) Ambulance location. At any point in time, different ambulances are at different locations and in different states of readiness. (c) Future coverage. The ambulances that remain available should provide good coverage for future emergencies. Second, when an ambulance completes a task and becomes available, a decision is made regarding the call in queue that the ambulance should serve next, or the location where the ambulance should go and wait. Future coverage also plays a role in this decision. We present a number of stochastic optimization models and methods to support ambulance dispatch decisions.

132