Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA58

Mehdi Nayebpour, Hadi El-Amine Cold ischemic time and post-operative complications are associated with 30-day mortality of the donor graft. However, long-term patient survival is related primarily to tissue matching, vasculopathy, immunosuppression, and infection. This study examines the preeminent role of tissue matching using an optimization regression algorithm to examine the effect of randomly allocated HLA: A, B, and DR kidney donor on long term patient survival in the largest retrospective study known to date. 2 - Exploring Models for Reducing Geographical Disparity in Organ Allocation Michal Mankowski, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia, Sommer Gentry Reducing the geographic disparity in access to organs is emerging issues in the transplant community. The OPTN/UNOS has recently appointed the Ad Hoc Geography Committee to define guiding principle and review models for the use of geographic constraints in organ allocation.We investigate a potential impact of some of the approved models. 3 - Information Theoretic Learning in Markov Decision Processes Peeyush Kumar, B14, Industrian and Systems Engineering, MEB, University of Washington Seattle, Seattle, WA, 98195-2650, United States I will present my research on Markov decision processes where the decision maker is uncertain about the model of the system. To maximize expected reward over the planning horizon, we must balance the exploration versus exploitation tradeoff: learn the transition probabilities sufficiently well, and utilize this information to quickly zero-in on actions with high rewards. I propose Information Directed Policy Sampling, which is an information theoretic framework that explicitly manages this tradeoff. We obtain a worst-case regret bound for IDPS. The theoretical guarantees are supplemented with numerical results on a sequential auction-design problem, and a response-guided dosing problem. 4 - The Impact of Imperfect Testing on the Performance of Personalized Medical Interventions Hadi El-Amine, George Mason University, 4310 Cotswolds Hill Ln, Fairfax, VA, 22030, United States In this paper, we develop a general framework that analyzes the performance of personalized medical interventions driven by historical patient data. We prove that finding optimal interventions is hard and provide structural properties that provide policy insights. n WA60 West Bldg 102B Joint Session HAS/Practice Curated: New Topics in Healthcare Policy Sponsored: Health Applications Sponsored Session Chair: Soroush Saghafian, Harvard Univeristy, Cambridge, MA, 02138, United States 1 - The Adoption and Diffusion of Medical Technology: Evidence from Cardiac Procedures Ariel Stern, Harvard Business School, Morgan Hall 433, Boston, MA, 02163, United States, Robert Huckman The adoption of new health care technologies is a learning process, with evidence showing that new products and procedures often involve physician learning and tradeoffs in quality and productivity. We consider the early years of uptake for a new cardiac procedure - transcatheter aortic valve replacement (TAVR) - and its implications for physician procedure mix and patient outcomes. Using data on all aortic valve replacement procedures performed in New York State over five years, we evaluate patterns in the uptake of TAVR across physicians and hospitals as well as patterns of access and receipt among (potential) patients. 2 - Development of Imminent Mortality Predictor for Advanced Cancer, a Tool to Predict Short-term Mortality in Hospitalized Advanced Cancer Patients Junchao Ma, Yale School of Management, New Haven, CT, United States, Edieal J. Pinker, Donald Lee End-of-life care for advanced cancer patients is aggressive and costly. Although cancer patients rely on oncologists for information about prognosis to make decisions about end-of-life care, physicians tend to over-estimate life expectancy and inconsistently initiate goals of care discussions. We developed and evaluated a novel prognostic tool, which generates life expectancy probabilities in real time using EHR time series to support oncologists in counseling patients about end-of- life care. We will also discuss use of similar decision-support tools to improve quality of care in intensive care units.

n WA58 West Bldg 101C Joint Session HAS/PSOR/Practice Curated: Emerging Issues in Treatment Planning and Management Sponsored: Health Applications Sponsored Session Chair: Shengfan Zhang, University of Arkansas, Fayetteville, AR, 72701, United States 1 - Opioid Overdose and the Role of Comorbidity in US Emergency Departments Nisha Nataraj, Centers for Disease Control and Prevention, 4770 Buford Highway, MS - F62, Atlanta, GA, 30319, United States, Kun Zhang, Gery Guy, Christina Mikosz The United States is in the midst of an opioid overdose epidemic with 42,249 opioid overdose-related deaths in 2016. Certain medical comorbidities may increase the risk for opioid overdose. We use regression and variable clustering models on national-level hospital discharge data to identify comorbidities most associated with opioid overdose in patients presenting to emergency departments. Results show that the overall number of comorbid chronic conditions is increasing in this population and that comorbid psychiatric, neurologic, and fluid/electrolyte disorders are especially associated with overdose. 2 - A Markov Decision Process Approach to Find the Optimal Time of Operating Lung Volume Reduction Surgery in Patients with Severe Emphysema Maryam Alimohammadi, University of Arkansas, Fayetteville, AR, 72701, United States, Shengfan Zhang, Art Chaovalitwongse Emphysema is a chronic lung disease that can be treated with lung volume reduction surgery (LVRS). Despite the advantages of LVRS in specific patients, it has mortality and morbidity risks and costs more than other treatments which makes it crucial to determine the subgroup of patients that can benefit the most from LVRS and assign them to surgery at the best time. To develop an optimization model that maximizes the quality-adjusted life time of the patients, we used Markov Decision Process (MDP) based on the National Emphysema Treatment Trial (NETT), a comprehensive dataset that collected the data of 1218 patients with severe emphysema who were randomized to undergo surgery or have medical treatment. 3 - Characterizing the Uncertainty Associated with Treatment Outcomes for Tuberculosis Patients Shengfan Zhang, University of Arkansas, 4207 Bell Engineering Center, Department of Industrial Engineering, Fayetteville, AR, 72701, United States The goal of this research is to use data analytics and stochastic modeling approaches to characterize patient recovery pathway from treatment for tuberculosis (TB). This research will use an existing anonymous data that contain information about follow-up test results for TB patients in Moldova upon initiation of treatment. Specifically, we aim to (1) characterize the pattern of recovery as denoted by the smear and culture test results at follow-ups; and (2) predict patient disposition (i.e., recovered to died) based on the recovery pattern. 4 - Enhancing Community Resilience to Combat Crisis of Opioid Addiction MD Noor E. Alam, Assistant Professor, Northeastern University, 334 Snell Engineering Center, 360 Huntington Avenue, Boston, MA 02115, United States, MD Mahmudul Hasan, Gary Young, Alicia Modestino This study focuses on detecting the variation of opioids prescribing pattern by leveraging the Massachusetts All Payer Claim Data (MA APCD) set. Based on the meaningful insights obtained from the data mining, we will devise a set of optimal policies to improve the community resilience to combat the opioid addiction epidemic. n WA59 West Bldg 102A Joint Session HAS/Practice Curated: Equity and Efficiency in Treatment of Chronic Illness Sponsored: Health Applications Sponsored Session Chair: Hadi El-Amine, George Mason University, Fairfax, VA, 22030, United States 1 - Optimization of Human Leukocyte Antigen Based Donor Recipient Matching in Kidney Transplantation Naoru Koizumi, George Mason University, School of Public Policy, 3351 North Fairfax Drive, Arlington, VA, 22201, United States,

432