INFORMS 2021 Program Book

INFORMS Anaheim 2021

SC21

2 - Health Insurance Plan Selection under Uncertainty using Stochastic Integer Programming Behshad Lahijanian, University of Florida, Herbert Wertheim College Of Engineering, Industria, Gainesville, FL, 32611-6595, United States, Michelle M. Alvarado Selecting a health insurance plan can be complicated for individuals or families. It is due to lots of available plans in the marketplace and how they split the cost in a year. A stochastic program model is developed to determine the health insurance selection model by considering different network types that can help people to understand, compare, and choose the right health insurance plan to suit their individual needs. We present a solving algorithm to minimize the total costs including covered and uncovered expenses in a year. 3 - Capacity Allocation in Cancer Centers Considering Demand Uncertainty Maryam Keshtzari, University of San Diego, San Diego, CA, 79416-1752, United States, Bryan A. Norman Capacity allocation in cancer centers is an important but challenging problem due to the high variability in patient demand and requirements. This study proposes a stochastic chance-constrained model to consider uncertainties in new and returning patient demand. The proposed model finds the optimal specialization mix for oncologists based on demand distribution by cancer type to avoid potential mismatches between oncologists’ specializations and the demand distribution by cancer type. To demonstrate the capability of the proposed model to answer important operational and tactical questions in cancer centers, the model is solved using data collected from our collaborating cancer center. SC21 CC Room 204A In Person: System Modeling to Inform Health Policy Making General Session Chair: Yu-Hsin Chen, Pennsylvania State University, State College, PA, United States 1 - Multi-resource Allocation and Sequence Assignment in Patient Care Management: A Stochastic Programming Approach Xinyu Yao, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, Rema Padman, Karmel S. Shehadeh The emergence of Real-Time Location System (RTLS) technologies has enabled healthcare organizations to collect highly granular location data on people and medical assets in the clinical environment. Using RTLS data from healthcare delivery settings, this study aims to minimize the waiting in the system by developing optimization models and methods to allocate and sequence medical resources for major patient care trajectories. We employ the two-stage stochastic programming model with the Monte Carlo Optimization approach to demonstrate improved patient and resource management strategies in complex care delivery settings. The results indicate that our model can significantly reduce waiting time for patients in the clinic by 60%, on average, with acceptable computational resource requirements and time complexity. 2 - An Analysis of Structured Optimal Policies for Hypertension Treatment Planning: The Tradeoff Between Optimality and Interpretability Wesley J. Marrero, Harvard University, Cambridge, MA, 48105- 2419, United States, Gian-Gabriel P. Garcia, Lauren N. Steimle Markov Decision Process (MDP) models are commonly used tools for optimizing sequential decisions under uncertainty in medical decision making. If the parameters of an MDP satisfy certain assumptions, the optimal policy is guaranteed to be monotone. Unfortunately, these assumptions are not always satisfied. In this research, we define the price of interpretability (PI), which measures the gap between the optimal and an interpretable policy. We assess the PI for the best-performing monotone policy (BMP) and the novel class-ordered monotone policy (CMP), which preserves interpretability along user-defined state and action classes. Within the context of hypertension treatment, we demonstrate that the CMP can be computed faster and achieves greater total quality-adjusted life years across a population of 66.5 million people in the US, compared to the BMP. 3 - Optimal Risk Threshold for High-Risk Screening under Capacity Yu-Hsin Chen, Pennsylvania State University, State College, PA, United States, Qiushi Chen Autism Spectrum Disorder (ASD) is a developmental disorder that affects 1 in 54 children in the US. The recommendation of universal ASD screening for children 18-24 months old has been widely debated, since it may instead increase the diagnosis delay due to unnecessary diagnostic evaluations in extra false-positive cases. In this study, with the risk of ASD identified before the screening, we developed a finite-horizon stochastic MDP model that determined the optimal risk threshold for screening which balanced between early diagnosis rate and diagnosis delay for young children under 30 months old. The simulation

demonstrated that the optimal high-risk screening policy outperformed the universal screening.

SC22 CC Room 204B In Person: Advances in Health Care Policy General Session Chair: Steven Foster, Virginia Tech, Blacksburg, VA, 24060, United States 1 - Outcome-based Pharmaceutical Contracting with Heterogeneous Patient Groups Andrew El Habr, Georgia Tech, Atlanta, GA, 30318-8272, United States, Can Zhang, Turgay Ayer We study under what market conditions and drug characteristics payers and pharmaceutical manufacturers are better off engaging in an outcomes-based contract, an agreement that links payments for drugs to drug effectiveness, over a nominal-pricing contract when there are heterogenous patient groups. One finding is that drugs that are more effective for the larger patient group can be good candidates for outcomes-based contracts. We also find that drugs that are not highly valuable to patients can be good candidates for outcome-based contracts in this setup. 2 - Ed Triage: An Empirical Study of Fast-track Admission and its Implication for Patient Outcomes As an effective way to improve emergency department throughput efficiencies, many hospitals have opened a separated fast-track service line that is dedicated to low acuity patients. However, these hospitals don’t have consistent routing policies and systematic routing criteria. This largely due to the fact that the impact of fast-track routing decisions on patient outcomes for heterogeneous patients hasn’t been well determined. Utilizing a unique data set from three urban hospitals in Canada, we first identified the behavior bias within fast-track routing decision-making process. We came up with an instrumental variable related to this bias to help us quantify the impact of routing decisions on patients with different severity levels whom we classified using a data-driven approach. 3 - COVID Response: Sanitizer Deployment Steven Foster, Clemson University, Clemson, SC, United States, Tyler O’Brien, Emily L. Tucker, Sudeep Hegde COVID-19 has forced universities to create strategies to combat infections in their student populations. To allow students to return to traditional instruction, Clemson University has deployed hand-sanitizing dispensers across campus to reduce viral transmissions, limit outbreaks, and promote adherence to CDC health guidelines. This study integrates optimization modeling and human factors methods to maximize usage of these hand-sanitizing dispensers. The facility location model presented uses door-access data to determine optimal dispenser locations within 37 buildings across campus. Interviews are used to define behavioral uncertainty and stakeholder decision-making to better direct the model to enhance future dispenser allocation. SC23 CC Room 204C In Person: INFORMS Optimization Society Prize Talk Session Award Session Chair: Andrea Lodi, Polytechnique de Montréal, Montreal, QC, H3C 3A7, Canada 1 - Award Presenter Andrea Lodi, Polytechnique de Montréal, Montreal, QC, H3C 3A7, Canada 2 - Award Presenter Shixuan Zhang, Georgia Institute of Technology, Atlanta, GA, 30339, United States 3 - Sufficient Conditions for Exact SDP Reformulations of QCQPs Alex Wang, Carnegie Mellon University, Pittsburgh, 15213, United States Quadratically constrained quadratic programs (QCQPs) are a fundamental class of optimization problems well known to be NP-hard in general. In this talk, we discuss sufficient conditions under which the standard semidefinite program (SDP) relaxation of a QCQP satisfies objective value exactness (the condition that the optimal values of the two programs coincide) or a stronger notion of exactness, convex hull exactness (the condition that the convex hull of the QCQP epigraph coincides with the projected SDP epigraph). We will additionally highlight applications and point to extensions of these results in follow-up work. Shuai Hao, University of Illinois at Urbana-Champaign, Champaign, IL, United States, Yuqian Xu, Zhankun Sun

22