Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WB58

3 - Optimization-based Subclassification for Causal Inference Shasha Han, NUS Business School, National University of Singapore, NUS Business School, Biz 2 Building B1, 1 Business Link, Singapore, Singapore, Joel Goh We are interested in estimating the individualized and population treatment effects of several drugs respectively. To obtain unbiased estimates, one often need to classify the entire population to several subclasses, such that the covariates between the contrasted treatments are distributed as similar as possible. For the purpose, the study propose an optimal subclassification method. 4 - Optimal Stopping of Response-adaptive Clinical Trials Amir Ali Nasrollahzadeh, Research Assistant, Clemson University, 278 Freeman Hall, Clemson University, Clemson, SC, 29634, United States, Amin Khademi One advantage of response-adaptive clinical trials is the ability to modify its key elements such as allocation policy, sample size, or stopping decisions while the trial is still in progress. In this study, we consider optimal stopping of Phase-II clinical trials where the objective is to find a ``target dose’’ in a safe dosage range. Response-adaptivity of the design suggests that the decision maker must choose between continuation of the trial with more sampling, termination and moving to a confirmatory phase, and abandoning the trial whenever new observations of patients’ responses become available. n WB58 West Bldg 101C Health Care System Optimization Sponsored: Health Applications Sponsored Session Chair: Naoru Koizumi, George Mason University, Arlington, VA, 22201, United States 1 - Optimal Pooling Strategies for Nucleic Acid Testing of Donated Blood Considering Viralload Growth Curves and Donor Characteristics Hadi El-Amine, George Mason University, Fairfax, VA, 22030, United States, Ebru Korular Bish, Douglas R. Bish Blood product safety, in terms of being free of transfusion-transmittable infections (TTIs), is crucial. Nucleic Acid Testing (NAT) technology enables earlier detection of infections but is more expensive, hence, most blood centers administer NAT to pools of blood samples from multiple donors. Since some donor characteristics are uncertain, we develop a chance-constrained model that determines the optimal NAT pool sizes for various TTIs, considering both non-universal (where first-time donors undergo more extensive screening), and universal (i.e., common testing for all donors) strategies, so as to minimize the TTI risk, while remaining within the testing budget with a high probability. 2 - Interactive Maps for UNOS Data Visualization Monica Gentili, University of Louisville, J.B. Speed School of Engineering, Louisville, KY, 40292, United States, Naoru Koizumi, Andrew Rivard In this talk we will show a web application to visualize UNOS data on heart transplants by means of interactive maps. The available maps will allow the user to visualize different statistics related to the data for different geographic granularity level (transplant center, Organ Procurement Organization, and UNOS region). The statistics will include (but are not limited to) (i) the likelihood of receiving a quality donor heart, (ii) the rate of donor heart offers which were declined (iii) pre- and post-transplant mortality rate. Further developments of the website will be discussed. 3 - Simulating the Outcome of Blood Incompatible Deceased Donor Kidney Transplantation - Time to Remove the Barrier Mehdi Nayebpour, George Mason University, Founders Hall, Fairfax Dr, Arlington, VA, 22201, United States, Keith Melancon, Naoru Koizumi, Karthika Mahendran ABO incompatible kidney transplantation could be accomplished with good out- comes. Most ABO-i kidney transplants have been accomplished using live donors after preparing recipients weeks before transplantation. We report the first case of successful directed deceased donor ABO-i kidney transplantation other than A2 to B. We desensitized the patient within 24 hours of receiving the organ offer and before performing the transplantation. If our current allocation system could allow for deceased donor ABO-i transplantation, we could significantly curtail kidney discards. We simulated this policy by using KPSAM and witnessed lower wait times and better outcomes.

n WB59 West Bldg 102A Joint Session HAS/DM: Healthcare Analytics Sponsored: Health Applications Sponsored Session Chair: Krista Foster, University of Pittsburgh, Pittsburgh, PA, 15217, United States Co-Chair: Zlatana Dobrilova Nenova, University of Denver, Englewood, CO, 80113, United States 1 - Assessing Multi-modality Breast Cancer Screening Strategies for High-risk Populations Caglar Caglayan, PhD Candidate, Georgia Institute of Technology, 755 Ferst Drive NW, ISyE Main Building #321, Atlanta, GA, 30332, United States, Turgay Ayer, Donatus U. Ekwueme Women with BRCA 1/2 gene mutations and family history of breast or ovarian cancer are at higher risk for breast cancer. For these women, the use of ultrasound (US) and MRI might address some of the limitations of mammography, the standard modality for average-risk women. Currently, there is no consensus on the optimal use of these technologies. We propose a population- dynamics based optimization approach for the multi-modality breast cancer screening problem for the high-risk population. We develop a Markov model to capture the disease progression in high-risk women, and formulate a mixed integer linear program to identify optimal structured strategies that are practical for implementation. 2 - Approximate Dynamic Programming for a Dynamic Appointment Scheduling Problem Zlatana Dobrilova Nenova, University of Denver, Denver, CO, 80113, United States, Dan Zhang, Manuel Laguna We consider a dynamic appointment scheduling problem. There are multiple patient classes with associated arrival rates, cancellation rates, and no-show probabilities. Patients arrive and make appointment requests over time. The provider can either assign the patient to an appointment slot or reject patients. Overbooking is allowed. The objective is to balance patients’ rejection and waiting costs, and the doctor’s idle- and overtime costs. Patients may cancel appointments or no-show. The problem is formulated as a finite-horizon Markov decision process and solved using an approximate policy iteration algorithm. We test the solution method on problem instances constructed from clinic data. 3 - Matching Emergency Physicians and Apps to Meet Patient Demand: Quasi Real-time Scheduling for Improved Patient Satisfaction and Resource Deployment Krista Foster, University of Pittsburgh, Pittsburgh, PA, 15217, United States, Jennifer S. Shang We study the current roles of advanced practice providers in emergency departments throughout a large multi-facility network and identify conditions under which the addition of advanced practice providers would benefit hospitals (reduce costs) and patients (reduce length of stay). We propose a model to optimally staff hospitals with emergency physicians and advanced practice providers to meet patient demand. 4 - A Spatial Model of the Opioid Epidemic in California Sema Nur Kaynar Keles, UCLA Anderson School of Management, Los Angeles, CA, United States, Elisa Frances Long With more than 67,000 overdose deaths in 2017 in the United Statesùincluding nearly 5,000 deaths in Californiaùadditional efforts to mitigate the opioid crisis are urgently needed. We propose a spatial “epidemic model of opioid use to forecast new opioid prescription rates, opioid addiction, cessation, and overdose death rates in California. Such a model could be used to prioritize key populations and allocate limited resources (eg, naloxone) most efficiently.

462