Informs Annual Meeting 2017

MB05

INFORMS Houston – 2017

4 - Food Waste in Grocery Retail Elena Belavina, University of Chicago Booth School of Business, 5807 S.Woodlawn Ave, Chicago, IL, 60637, United States, elena.belavina@chicagobooth.edu Food waste is a major contributor to carbon emissions (as big as road transport). Identifying and influencing market conditions that can decrease food waste is thus important to combat global warming. We build and calibrate a stylized two- echelon perishable-inventory model to capture grocery purchases and expiration at competing stores and households in a market. We examine how the equilibrium waste in this model changes with store density. Our analysis shows that higher density reduces food waste up to a threshold density; it leads to higher food waste beyond this threshold. Put differently, in so far as food waste is concerned, there exists an optimal store density. 320B Analytics in Medical Decision Making and Population Health Sponsored: Health Applications Sponsored Session Chair: Turgay Ayer, Georgia Institute of Technology, Atlanta, GA, 30332, United States, ayer@isye.gatech.edu Co-Chair: Anthony Bonifonte, Georgia Institute of Technology, Austell, GA, 30106, United States, ABonifon@gatech.edu 1 - Analytics in Blood Pressure Management: From Data to Decisions Anthony Bonifonte, Georgia Institute of Technology, 5940 Water Oaks Dr, Austell, GA, 30106, United States, ABonifon@gatech.edu, Turgay Ayer, Ben A.Haaland, Ben A.Haaland Antihypertensive drug treatment can control elevated blood pressure (BP) and reduce the risk of cardiovascular disease. We propose data-driven models that capture heterogeneity in BP progression and combine stochastic optimization and statistics to identify the optimal thresholds for initiating treatment and for increasing the treatment dosage. We analytically characterize the expected value and variance of the hazard ratio, which enables us to easily compute the optimal treatment decisions, and capture different attitudes towards risk (e.g. risk neutral or risk averse). Our findings may help guide future RCT design for hypertension treatment. 2 - Prioritizing Hepatitis C Treatment in U.S. Prisons Turgay Ayer, Georgia Institute of Technology, School of Industrial and Systems Engineering, Groseclose 417, Atlanta, GA, 30332, United States, ayer@isye.gatech.edu, Can Zhang, Anthony Bonifonte, Jagpreet Chhatwal, Anne Spaulding Prison systems offer a unique opportunity to control the hepatitis C virus (HCV) epidemic. New HCV treatments are very effective, but providing treatment to all inmates is prohibitively expensive. We propose a restless bandit modeling framework to support HCV treatment prioritization decisions in U.S. prisons. Parameterized by real-world data, we derive a closed-form index policy and demonstrate the improvement in our policy over current practice and benchmark policies. Our results shed light on several controversial health policy issues in HCV treatment prioritization in the prison setting and have important policy implications, such as needing to consider more than health state of inmates. 3 - Dynamically Consistent Home Health Care Routing and Scheduling Seyma Guven-Kocak, Georgia Institute of Technology, Atlanta, GA, United States, seymaguven@gatech.edu, Pinar Keskinocak, Alejandro Toriello, Aliza R.Heching This work addresses a real-world home health care routing and scheduling problem (HHCRSP) faced by a home care agency in the U.S. In home health care scheduling, there is a desire to retain consistency with respect to the aides servicing each patient, which is referred to as continuity of care. In order to handle continuity of care, we propose a dynamic approach and introduce the dynamically consistent home health care routing and scheduling problem (D- Con-HHCRSP). We present solution methods to address HHCRSP and D-Con-HHCRSP, where the goal is to be able to quantify and control the deviation from the existing schedule in place, so that some of the existing assignments may be retained in the new schedule. 4 - Short-term Mortality Prediction in a Cohort of Cancer Patients Ying Zhuo, Massachusetts Institute of Technology, 70 Lincoln Street, Unit L612, Boston, MA, 02111-2670, United States, MB05

identified 23,983 patients who initiated 46,646 anti-cancer regimens. We fitted state-of-the-art machine learning models to predict 60-, 90- and 180-day mortality and achieved out-of-sample AUCs of 0.90, 0.89, 0.87, respectively, significantly higher than existing models.

MB06

320C Health Policy Modeling, Analysis, and Insight Sponsored: Health Applications Sponsored Session Chair: Pooyan Kazemian, PhD, Harvard Medical School, Boston, MA, 02114, United States, pooyan.kazemian@mgh.harvard.edu 1 - Flexible FDA Approval Policies Taylor Corcoran, University of California-Los Angeles, 110 Westwood Plaza, Los Angeles, CA, 90024, United States, taylor.corcoran.1@anderson.ucla.edu, Elisa Frances Long, Fernanda Bravo Current U.S. FDA approval standards require pharmaceutical companies to demonstrate drug efficacy at a 0.05 significance level using clinical trial outcomes. This policy ignores heterogeneity in disease characteristics, such as the severity and prevalence, the level of research and development taking place, and the number of alternative drug treatments available for patients. We develop a queueing framework to analyze the drug approval process that incorporates such factors. We derive optimal approval policies and conduct a numerical study for three diseases (HIV, hypertension, and breast cancer). 2 - Same-day Follow-up Appointments, Spoilage, and Continuity of Care: An Empirical Investigation Xiaoxu Tang, University of Minnesota, Minneapolis, MN, United States, tangx238@umn.edu, Diwakar Gupta, Yichuan Ding We use a proprietary dataset (2M+ records) to study follow-up appointments in outpatient clinics. We propose methodologies to tag episodes of care and measure continuity of care and appointment-slot spoilage (late cancellation or no-show). We find that same-day follow-up (SDFU) appointments (those booked on the day the prior appointment occurred) have greater continuity of care and higher spoilage rate while controlling for patient, physician, and clinic characteristics. Counterfactual analysis is performed to evaluate SDFU appointments as a strategy for managing follow-ups. 3 - Robust Optimization Framework to Account for Prediction Errors for Cancer Diagnosis Selin Merdan, University of Michigan, 1640 McIntyre Street, Ann Arbor, MI, 48105, United States, smerdan@umich.edu, Brian T.Denton In the early detection and diagnosis of disease, multiple diagnostic tests are often available in discriminating between diseased and nondiseased individuals; however, how best to use these tests to render a diagnosis is challenging because there is often a tradeoff between the benefits of an accurate diagnosis of the anticipated disease and harms and costs associated with the diagnostic tests themselves. The motivation for our work in this research is to combine multiple diagnostic tests into an optimal composite diagnostic testing, which will provide a good balance between unnecessary testing and missed cases with higher sensitivity and specificity in the context of prostate cancer detection. 4 - Optimizing Monitoring Policies under Unknown Patient Adherence Zhenhuan Zhang, University of Minnesota, 425 13th Ave SE Apt.1001, Minneapolis, MN, 55414, United States, zhan4490@umn.edu, Diana Maria Negoescu In most instances of disease management, adherence to treatment is crucial to treatment success. However, measurements of patient adherence are often noisy and adherence behavior is unknown to physicians prior to treatment initiation. We optimize HIV Viral Load (VL) monitoring policies in this setting, and assess the total costs and quality-adjusted life years achieved by different policies. 5 - Estimating the Economic Cost of Physician Burnout in the United States Shasha Han, NUS.Business School, Biz 2 Building B1, 1 Business Link, PhD office, NUS.business school, Singapore, Singapore, shashahan@u.nus.edu, Tait D. Shanafelt, Christine A. Sinsky, Karim Awad, Joel Goh Importance: Even though physician burnout is known to be associated with negative clinical and organizational outcomes, the costs of burnout are poorly understood. Consequently, leaders of healthcare institutions cannot properly assess the benefits of initiatives to remediate physician burnout.Objective: To estimate the annual cost associated with physician burnout in the United States and to develop a methodology for individual healthcare institutions to estimate their institution-level costs associated with physician burnout.

zhuo@mit.edu, Dimitris Bertsimas, Colin F. Pawlowski, John M. Silberholz, Alexander M. Weinstein, Eddy Chen, Aymen Elfiky

As cancer patients often receive aggressive and potentially unnecessary care at the end of life, there is significant need to accurately predict mortality. We obtained 11 years of electronic health records (EHR) from a large cancer center and engineered 401 features: demographics, genomics, vitals, lab tests, etc. We

166