INFORMS 2021 Program Book

INFORMS Anaheim 2021

WC22

2 - Online Policies for Efficient Volunteer Crowdsourcing Scott Rodilitz, Yale, New Haven, CT, 06511-2572, United States Nonprofit crowdsourcing platforms encourage volunteers to complete tasks by using nudging mechanisms to notify a subset of volunteers with the hope that at least one of them responds positively. However, since excessive notifications may reduce volunteer engagement, the platform faces a trade-off between notifying more volunteers for the current task and saving them for future ones. Motivated by these applications, we introduce the online volunteer notification problem and develop an online randomized policy that achieves constant-factor guarantees. Further we demonstrate the effectiveness of our policy by testing it on data from a volunteer-based food recovery platform. 3 - Operational Issues in Large Jail and Judiciary Systems Charlie Hannigan, University of Chicago-Booth School of Business, Chicago, IL, 60616-1740, United States People are often held in jail excessively long pre-trial. In fact, these pre-trial case lengths are sometimes longer than their eventual sentence. When this happens, the detainees are colloquially called “overstays.” In 2016, total excess time for overstays released from the jail we study was over 244 years. Overly long case lengths and overstays arise, in part, due to detainees who intentionally delay their cases. We employ machine learning techniques and structural estimation to identify and understand intentional delaying behavior in the large jail system we study. We also propose policy interventions to reduce excessive case lengths and overstays, some tailored for intentionally delaying detainees and some for non- delaying detainees. 4 - School Choice In Chile Boris Espstein, Columbia University, New York, NY, United States, Jose Correa, Rafael Epstein, Juan Escobar, Ignacio Rios, Bastian Bahamondes, Carlos Bonet, Natalie Epstein, Nicolas Aramayo, Martin Castillo, Andres Cristi, Felipe Subiabre Centralized school admission mechanisms are an attractive way of improving social welfare and fairness in large educational systems. In this paper we report the design and implementation of the newly established school choice system in Chile, where over 274,000 students applied to more than 6,400 schools. This is a simultaneous nationwide system that includes all 14 school grade levels, making it one of the largest school choice problems worldwide. One of our primary goals is to favor the assignment of siblings to the same school. By adapting the standard notions of stability, we show that a stable assignment may not exist. Hence, we propose a heuristic approach that elicits preferences and breaks ties between students in the same priority group at the family level. In terms of implementation, we adapt the Deferred Acceptance algorithm as in other systems around the world. WC24 CC Room 205A In Person: Data-Driven Supply Chain Management General Session Chair: Reza Yousefi Maragheh, Walmart Global Tech 1 - Near-optimal Primal-dual Algorithms for Quantity-based Network Revenue Management Zijie Zhou, PhD student, Massachusetts Institute of Technology (MIT), Cambridge, MA, United States We study the canonical quantity-based network revenue management problem. The exact solution to the problem by dynamic programming is computationally intractable. Existing works in the literature make use of the solution to the deterministic linear program to design asymptotically optimal algorithms. Those algorithms rely on repeatedly solving DLPs to achieve near-optimal regret bounds. It is, however, time-consuming to repeatedly compute the DLP solutions in real- time.In this paper, we propose innovative algorithms that are easy to implement and do not require solving any DLPs. Our algorithm achieves a regret bound of $O(\log k)$, where $k$ is the system size. To the best of our knowledge, this is the first NRM algorithm that (i) has an $o(\sqrt{k})$ asymptotic regret bound, and (ii) does not require solving any DLPs. 2 - Debiasing In-sample Policy Performance in the Small-data, Large-scale Regime Michael Huang, University of Southern California, Los Angeles, CA, 90007-2488, United States, Vishal Gupta, Paat Rusmevichientong Motivated by the poor performance of cross-validation when data are scarce, we propose a novel method called the Variance Gradient Correction to estimate the out-of-sample performance of a policy by debiasing the in-sample performance. Unlike cross-validation, our method does not sacrifice data set for a test set. We prove it is approximately unbiased, and, in many settings, its estimation error tends to zero in probability uniformly over a policy class. Through a empirical study on dispatching EMS services, we show our proposed approach outperforms state-of-the-art approaches in estimating out-of-sample policy performance.

WC22 CC Room 204B In Person: Healthcare Policy and Delivery Innovations General Session Chair: Shima Nassiri, University of Michigan, Ann Arbor, MI, 48109- 1234, United States 1 - Generic Drug Treatment Effectiveness: An Empirical Study Xinyu Shirley Liang, Doctoral Candidate, Ross School of Business, University of Michigan, Ann Arbor, MI, United States, Jun Li, Ravi Anupindi Around 90% of drugs consumed in the US are generics, saving billions in prescription costs. However, the cost-saving benefit can only be realized when the drug efficacy is ensured. We examine generic drug’s effectiveness by exploiting the market entry of generic Lipitor. We find that generic drugs are associated with higher healthcare service utilization and worse clinical outcomes. We also find heterogeneous effects of generic drugs across patients and manufacturers. Our findings highlight the importance of ensuring generic drugs’ quality and accounting for their treatment heterogeneity in prescriptions. 2 - Proximal Policy Optimization for Hospital Inpatient Bed Management Pengyi Shi, Purdue University, Krannert School of Management, Kra, West Lafayette, IN, 47907, United States, Jim Dai, Mark Gluzman, Jingjing Sun When waiting time is excessively long before a bed in the primary ward becomes available, patients may be assigned to beds in a non-primary ward, known as off- service placement or overflow. We model this overflow problem as a queueing-network based Markov Decision Process. We leverage the Proximal policy optimization algorithm to tackle the large action space. Specifically, we adopt randomized routing policies and sequentially assigning beds to each waiting patient. Through a novel state aggregation, we further decompose the value function to facilitate the computation. We perform extensive numerical experiments with real hospital data to demonstrate the effectiveness and scalability of our algorithm. 3 - Comprehensive Primary Care Plus: Financial Incentives and Recommendations Fernanda Bravo, UCLA Anderson School of Management, Los Angeles, CA, 90024-5055, United States, Elodie Adida The Comprehensive Primary Care Plus (CPC+) is a new payment model used by the Centers for Medicare & Medicaid Services (CMS) to reform the way primary care physicians are paid, to better align incentives, and thus to improve the quality and delivery of care. In this talk, we analyze the impact of the CPC+ payment system on the different stakeholders (physician, patient, and payer), how best to design it, and for what type of patient population it is best suited. WC23 CC Room 204C In Person: Service Operations and Societal Impact General Session Chair: Russell Charles Hannigan, University of Chicago-Booth School of Every representative democracy must specify a mechanism under which voters choose their representatives. The most common mechanism in the United States — winner-take-all single-member districts — both enables substantial partisan gerrymandering and constrains `fair’ redistricting, preventing proportional representation in legislatures. We study the design of multi-member districts (MMDs), in which each district elects multiple representatives, potentially through a non-winner-takes-all voting rule. We carry out large-scale analyses for the U.S. House of Representatives under MMDs with different social choice functions, under algorithmically generated maps optimized for either partisan benefit or proportionality. Doing so requires efficiently incorporating predicted partisan outcomes — under various multi-winner social choice functions — into an algorithm that optimizes over an ensemble of maps. We find that with three- member districts using Single Transferable Vote, fairness-minded independent commissions would be able to achieve proportional outcomes in every state up to rounding, and advantage-seeking partisans would have their power to gerrymander significantly curtailed. Simultaneously, such districts would preserve geographic cohesion, an arguably important aspect of representative democracies. In the process, we open up a rich research agenda at the intersection of social choice and computational redistricting. Business, Chicago, IL, 60616-1740, United States 1 - Combatting Gerrymandering with Social Choice: The Design of Multi-member Districts Nikhil Garg, Cornell Tech, New York, NY, 94025-4714, United States

162