INFORMS 2021 Program Book

INFORMS Anaheim 2021

WC21

3 - Acuity-based Nurse Assignment in Cancer Center Infusion Departments Bryan A. Norman, Professor, Texas Tech University, Lubbock, TX, United States, Maryam Keshtzari In this study, a mixed-integer mathematical model is proposed to assign nurses to patients based on patient acuity levels to minimize patient wait time and balance the workload across nurses. Moreover, the proposed model considers different levels of flexibility in nurse assignment in order to provide nurse continuity. A heuristic algorithm is developed to investigate the impact of an alternative nurse- patient assignment approach. Numerical examples are presented to compare the performance of the exact and heuristic methods. 4 - A Wearable Gait Recognition System Based on Soft Robotic Sensors Using Deep Learning Methods Samaneh Davarzani, Mississippi State University, Starkville, MS, 39759-1110, United States, Reuben F. Burch, Brian K. Smith Gait recognition systems have gained much attention due to the potential applications in healthcare, sports biomechanics, and the workplace. A new solution to gait recognition tasks can be provided by wearable sensors integrated in clothing and textiles and paired with mobile and computer devices. In this study, a sock prototype designed with embedded soft robotic sensors (SRS) is implemented to measure foot-ankle complex kinematic data during treadmill walking. Various deep learning methods have been employed and compared for modeling SRS data against a motion capture system to determine their ability to provide accurate kinematic data from dynamic movements using SRS measurements. WC21 CC Room 204A In Person: Operations Management General Session Chair: Simeng Shao, University of Southern California, Los Angeles, CA, United States 1 - Correlation Neglect in Supply Chains Anugna Reddy Gondi, Cornell University, Ithaca, NY, United States We study the effect of correlation in two operations contexts: supply risk and demand uncertainty, and hypothesize that individuals do not adequately account for correlation in such contexts. Through controlled lab experiments, we investigate how human subjects make ordering decisions under non-zero correlation (positive or negative) and zero correlation scenarios. Despite it being optimal to set different orders across these two scenarios, we find that participants set orders that are virtually the same across both settings. We also find that the effect of correlation on order bias is robust to controlling for other known biases. 2 - Disclosure-driven Social Engagement in Supply Chains Caleb Kwon, Harvard Business School, Cambridge, MA, United States, Jun Li, Andrew Wu We empirically examine the real effects of corporate social responsibility (CSR) disclosures on actual societal outcomes of disclosing firms and their supply chain connections. To do so, we exploit the passage of the California Transparency in Supply Chains Act (CTSCA) as a regime shift in mandated disclosures and examine the law’s impact on an objective measure of CSR impact, which we measure using actual, media-covered incidents in human trafficking, forced labor, child labor, and human rights abuses. Firm-level responses to the CTSCA are measured by scraping the Web for firm disclosures and by analyzing the disclosure’s contents using NLP machine learning algorithms. Our principal finding is that CSR activity has significant spillover effects which suggests that supply relations could serve as an important link for the propagation of socially impactful actions. 3 - Multi-Product Dynamic Pricing in High-Dimensions with Heterogeneous Price Sensitivity Simeng Shao, University of Southern California, Los Angeles, CA, United States We consider the problem of multi-product dynamic pricing, in a contextual setting, for a seller of differentiated products. In this environment, the customers arrive over time and products are described by high-dimensional feature vectors. Each customer chooses a product according to the Multinomial Logit (MNL) choice model. Our model allows for heterogenous price sensitivities for products. The seller a-priori does not know the parameters of the choice model but can learn them through interactions with the customers. The seller’s goal is to design a pricing policy that maximizes her cumulative revenue. We propose a pricing policy, named M3P, that achieves a T-period regret of O(log(Td) ( √ T+ d log(T))) under heterogenous price sensitivity for products with features of dimension d. We also prove that no policy can achieve worst-case T-regret better than Ω ( √ T).

WC19 CC Room 203A In Person: Bonder Scholar Session Award Session Chair: Arielle Elissa Anderer, The Wharton School, Wynnewood, PA, 19096-2455, United States 1 - Estimating the Value of Incorporating Patient Behavior in Return to Play From Concussion Gian-Gabriel P. Garcia, Harvard University, Cambridge, MA, 48103, United States Concussion, the most common type of traumatic brain injury, is a major public health issue. For patients with sports-related concussion, the timing of return-to- play (RTP) is critical; premature RTP can increase likelihood of catastrophic injuries while delayed RTP can decrease benefits of physical activity. RTP decisions are complicated by the potential for strategic symptom-reporting. We formulate this decision problem as a partially observable stochastic game and analyze the equilibrium and doctor’s best-response RTP strategy. We then use simulation to quantify the value of incorporating patient behavior by comparing this behavior- aware RTP strategy with practice-based RTP policies. 2 - Improving Diabetes Care with Thermal Imaging and Machine Learning Jas Wodnicki, University of Wisconsin-Madison, Madison, WI, United States, Thor Larson Diabetic foot ulcers (DFU) are among the most common and deadly complications of diabetes. Foot ulcers progress rapidly, leading to one million amputations globally which traditional risk assessments fail to prevent. Using thermal imaging, we are quantifying the inflammatory response behind DFU and developing machine learning algorithms for ulceration risk assessment. Our analysis leverages an ensemble of automated image processing methods, with practical use in mind. This technology is tailored to low and middle income countries and the unique cultural and systemic challenges they face. WC20 CC Room 203B In Person: Health Care II Contributed Session Chair: Samaneh Davarzani, Mississippi State University, Starkville, MS, 39759-1110, United States 1 - Generalized Bandits with Learning and Queueing in Split Liver Transplantation Yanhan Tang, Carnegie Mellon University, Pittsburgh, PA, United States, Alan Scheller-Wolf, Sridhar R. Tayur, Andrew A. Li We study liver allocation where surgeons with different abilities learn split liver transplantation. We formulate a multi-armed bandit with embedded learning curves to address the trade-off between discovering talents (exploration) and strengthening extant surgeons’ skills (exploitation). Our QFL-UCB algorithm, enhanced with queueing dynamics, and fairness, has O(log T) regret. Our algorithms could be applied to help evaluate strategies to increase the use of SLT and other technically difficult procedures that require practice. Methodologically, our proposed MAB model and algorithms are generic and have broad applications. 2 - Cooperative Blood Inventory Ledger (CoBIL): A Decentralized Decision Making Framework for Improving Blood Product Management Rishabh Bhandawat, University at Buffalo, Buffalo, NY, United States Existing blood product supply management systems are limited by their segmentation, lack of detailed blood product information, and lack of real-time updating. We propose a novel architecture for blood product information sharing and a pseudo-collaborative decision-making mechanism (CoBIL) to overcome organizational competitive advantage, and to reduce outdates and shortages to benefit donors, patients/hospital, and demand nodes while minimizing operational costs. The work also presents the CoBIL framework along with an inventory routing algorithm to support the adoption of blockchain technology for blood product management.

161