Informs Annual Meeting 2017

Annual

Meeting program book

October 22 − 25

T E C H N I C A L S E S S I O N S

Sunday, 8:00 - 9:30AM

How to Navigate the Technical Sessions

SA02

There are four primary resources to help you understand and navigate the Technical Sessions: • This Technical Session listing, which provides the most detailed information. The listing is presented chronologically by day/time, showing each session and the papers/abstracts/authors within each session.

310B Decision Analytic Approaches to Green

Infrastructure Solutions Sponsored: Decision Analysis Sponsored Session Chair: Fengwei Hung, Johns Hopkins University, Baltimore, MD, 21218, United States, hfengwe1@jhu.edu Co-Chair: Melissa A. Kenney, University of Maryland, College Park, Snehasis Mukhopadhyay, IUPUI, Indianapolis, IN, United States, smukhopa@iupui.edu, Andrew Hoblitzell, Meghna Babbar-Sebens WRESTORE is a stakeholder-driven design system that optimizes best management practices for the Eagle Creek Watershed in Indiana. The system provides resources on conservation programs and enables stakeholders to visualize the impacts of alternative practices on the watershed. The system utilizes fuzzy logic, neural networks, and deep learning for user modeling, and uses an interactive genetic algorithm with mixed initiative interaction for exploratory search and exploitative optimization. In this talk, we will provide a discussion of the current implementation of the system, as well as a brief discussion of ongoing work and future plans for the system. 2 - Understanding Residential Low Impact Development at the Home Owner Level Domenico Amodeo, George Washington University, Washington, DC, 20052, United States, dcamodeo@email.gwu.edu Environmental engineers and economist have studied the patterns of adoption for green technology. Many of these studies rely on regression and aggregated data. We developed a zero-truncated negative binomial regression model using parcel level data. However, we argue that regression is an inadequate method for predicting home owner adoption of LID. A graph based method is presented to identify configurations of LID emerging from residential property owners who are presented with a municipal incentive to adopt. We argue that understanding the configurations that emerge may provide insight into homeowners motivations and the effectiveness of incentive programs. 3 - A Two-stage Stochastic Programming Model for Adaptive Stormwater Management with Green Infrastructure Fengwei Hung, Johns Hopkins University, 3400 N Charles St. Ames Hall 313, Baltimore, MD, 21218, United States, hfengwe1@jhu.edu Adaptive management (AM) can be viewed as a multi-stage planning problem with the emphasis on learning by doing and uncertainty. The proposed model considers the decision maker’s risk attitudes and the potential learning in latter stages. Risk is modeled in the form of conditional value of risk (CVaR) and learning is defined as the process of updating random distributions which are assumed to depend on the investment now (1st stage decisions). This method can explore the risk-expected value tradeoffs accounting for what we may be able to learn from doing. Finally, we show an example to demonstrate how this model works. 4 - Optimal Planning of Green Infrastructure Placement in an Urban Watershed under Precipitation Uncertainty Masoud Barah, University of Tennessee, Knoxville, TN, United States, mbarah@utk.edu, Anahita Khojandi, Jon Hathaway, Xueping Li Urbanization, infrastructure degradation, and climate change have overwhelmed most stormwater management systems across the nation or rendered them ineffective. Green Infrastructures (GIs) are low-cost strategies that can contribute to stormwater management. We develop a stochastic programming model to determine the optimal placement of GIs across a set of candidate locations in a watershed to minimize the excess runoff in medium-term precipitation uncertainties. We calibrate the model using rainfall projections and stormwater system’s hydrologic responses to them. We provide optimal GI placement in a watershed and perform sensitivity and robustness analyses to provide insights. MD, 20740, United States, kenney@umd.edu 1 - WRESTORE Design System: Optimizes Best Management Practices

The Session Codes

TA01

Room number. Room locations are also indicated in the listing for each session.

Time Block. Matches the time blocks shown in the Program Schedule.

The day of the week

Time Blocks

Sunday - Monday 8:00am - 9:30am 10:00am - 10:50am 11:00am – 12:30pm 1:30pm – 3:00pm 3:10pm - 4:00pm 4:30pm – 6:00pm Tuesday 7:30am - 9:00am 9:40am - 10:30am 10:30am - 12:00pm 12:05pm - 1:35pm 2:00pm - 3:30pm 3:40pm - 4:30pm 4:35pm – 6:05pm Wednesday 7:30am - 9:00am 9:40am - 10:30am 10:30am - 12:00pm 12:30pm - 2:00pm

Rooms and Locations /Tracks All tracks / technical sessions will be held in the George R. Brown Convention Center. Roo¡m numbers are shown on the Quick Reference and in the Technical session listings. Monday and Tuesday Plenary talks will be held in Hilton- Ballroom of Americas, Level 2

1

SA03

INFORMS Houston – 2017

SA03

4 - Estimating Latent Asset-pricing Factors from Large-dimensional Data

310C An Introduction to Two-Stage Stochastic Mixed- IntegerProgramming Invited: Tutorial Invited Session Chair: Jiming Peng, University of Houston, Houston, TX, United States, jopeng@Central.uh.edu Co-Chair: Rajan Batta, University at Buffalo (SUNY), 410 Bell Hall, University at Buffalo (SUNY), Buffalo, NY, 14260, United States, batta@buffalo.edu 1 - An Introduction to Two-Stage Stochastic Mixed-Integer Programming Simge Kucukyavuz, University of Washington, Box 352650, Industrial & Systems Engineering, Seattle, WA, 98195, United States, simge@uw.edu, Suvrajeet Sen This paper provides an introduction to algorithms for two-stage stochastic mixed integer programs. Our focus is on methods which decompose the problem by scenarios representing randomness in the problem data. The design of these algorithms depend on where the uncertainty appears (right-hand-side, recourse matrix and/or technology matrix) and where the continuous and discrete decision variables are (first-stage and/or second-stage). In addition we provide computational evidence that, similar to other classes of stochastic programming problems, decomposition methods can provide desirable theoretical properties (such as finite convergence) as well as enhanced computational performance when compared to solving a deterministic equivalent formulation using an advanced commercial MIP solver. SA03A Grand Ballroom A Applied Probability in Finance Sponsored: Applied Probability Sponsored Session Chair: Markus Pelger, Stanford University, Stanford, CA, 94305, United States, mpelger@stanford.edu 1 - Determinant of Clearinghouse Margins Agostino Capponi, Columbia University, 500 W. 120th Street, New York, NY, 10027, United States, ac3827@columbia.edu We analyze regulatory credit default swap (CDS) data to investigate drivers of clearinghouse margins, exploiting explicitly linked portfolio exposures and collateralizing assets. We document several stylized facts, including heterogeneity of clearing member portfolios, significant time variation in margin levels, and clustering of portfolio return on margins around the mean. Our results show that the advent of margin spirals relies on commonality in exposures, as shocks to common credit exposures can be destabilizing, whereas idiosyncratic shocks are stabilized by clearinghouse margining rules. 2 - Law of the Few: Economics of the Tipping Point Nan Chen, Chinese University of Hong Kong, William M.W. Mong Large regime switch due to social interaction, known as the tipping point, is of great interest in sociology and economics. Two empirical features related to this phenomenon are local conformity/global diversity and punctuated equilibrium effect. The former refers to that a significant conformity can be found within a given community but in other separate communities the same issue is approached by different ways. The latter feature means that the dynamics of a given community tend to have a long period of the dominance of one opinion, punctuated by bursts of opinion shifts. In this paper we propose a simple stochastic model to incorporate the above two features of an interacting population. 3 - Adaptive Learning and Stability of Equilibria in a Dynamic Model with Informed Investors Yiwen Shen, Columbia Business School, New York, NY, United States, YShen21@gsb.columbia.edu, Paul Glasserman, Harry Mamaysky We propose a dynamic model of securities trading between overlapping generations of informed and uninformed investors. We assume becoming informed is costly, and the quality of learnable information improves with the ratio of informed investors. Investors form beliefs about how price volatility depends on the quality of learnable information. We show the beliefs could be self-fulfilling in a low volatility market. With such beliefs, our economy is characterized by high and low information equilibria, both of which are locally stable but with periodic transitions between each other. We discuss the model’s implications for trading volume, investors’ welfare, and boom-bust cycles. Engineering Bldg, Rm 609, Shatin N.T., Hong Kong, nchen@se.cuhk.edu.hk, Steven Kou, Yan Wang

Markus Pelger, Stanford University, 312 Huang Engineering Center, 475 Via Ortega, Stanford, CA, 94305, United States, mpelger@stanford.edu, Martin Lettau We develop an estimator for latent factors in a large-dimensional panel of financial data that can explain expected excess returns. Statistical factor analysis based on Principal Component Analysis (PCA) has problems identifying weak factors that are important for asset pricing. Our estimator searches for high Sharpe-ratio factors that can explain both the expected return and covariance structure. We derive the statistical properties of the new estimator and show that it can find asset-pricing factors, which cannot be detected with PCA, even if a large amount of data is available. Our factors accommodate a large set of anomalies better than notable four- and five-factor alternative models. SA03B Grand Ballroom B Dynamic Pricing Sponsored: Revenue Management & Pricing Sponsored Session Chair: Hamid Nazerzadeh, USC Marshall School of Business, Los Angeles, CA, 90089, United States, hamidnz@marshall.usc.edu Co-Chair: Negin Golrezaei, University of Southern California, Pasadena, CA, 91106, United States, golrezae@usc.edu 1 - Dynamic Pricing and Information Aggregation in Sports Betting Markets Adam Schultz, University of Chicago-Booth School of Business, 5532 South Kenwood Ave, Apt 208, Chicago, IL, 60637, United States, adam.schultz@chicagobooth.edu, John R. Birge, N. Bora Keskin In this paper, we explore how market makers use dynamic pricing as a mechanism to aggregate information in a biased prediction market. We collect a novel data set of time series data to study a sports betting market, including Twitter data to control for breaking news events that lead to information changes in the market. After investigating how market makers adjust prices in this context, we present an approach to estimate potential bias in bettors’ beliefs about game outcomes. This model allows us to perform a counterfactual analysis in which we characterize the optimal point spread for the market maker for each game. We use this model to analyze market makers’ policies and expected profit performance. 2 - On the Efficacy of Static Prices for Revenue Management in the Face of Strategic Customers Yiwei Chen, Singapore University of Technology and Design, Singapore, Singapore, yiwei_chen@sutd.edu.sg, Vivek Farias We consider a RM problem wherein a monopolist seller seeks to maximize revenues from selling a fixed inventory of a product to customers who arrive over time. Customers are forward looking. We allow for multi-dimensional customer types. We allow for a customer’s disutility from waiting to be positively correlated with his valuation. We show that static prices proposed by Gallego and van Ryzin [1994] that study a myopic customer RM problem continue to remain asymptotically optimal in the regime where inventory and demand grow large. Irrespective of regime, the optimal static price captures at least 63.2% of the seller’s revenue under an optimal dynamic mechanism. 3 - Multidimensional Binary Search for Contextual Decision-making Ilan Lobel, NY, United States, ilobel@stern.nyu.edu, Renato Paes Leme, Adrian Vladu We consider a multidimensional search problem that is motivated by questions in contextual decision-making such as dynamic pricing. Nature selects a state from a d-dimensional unit ball and then generates a sequence of d-dimensional directions. After receiving a direction, we have to guess the value of the dot product between the state and the direction. Our goal is to minimize the number of times when our guess is more than away from the true answer. We construct a polynomial time algorithm that we call Projected Volume achieving regret O(d log d), which is optimal up to a log d factor. The algorithm combines a volume cutting strategy with a new geometric technique that we call cylindrification.

2

SA04

INFORMS Houston – 2017

SA04

4 - Dynamic Pricing in High-dimensions Adel Javanmard, USC Marshall School of Business, 300A Bridge Hall, University of Southern California, Los Angeles, CA, 90089, United States, ajavanma@marshall.usc.edu, Hamid Nazerzadeh We study the pricing problem where a firm sells a large number of products, described via a wide range of features, to customers that arrive over time. The firm is oblivious to market values of products. However, it can gain information about valuations based on previous binary feedbacks on whether products were sold at the posted prices. This setting is motivated in part by the prevalence of online marketplaces that allow for real-time pricing. We propose a dynamic policy (RMLP) that leverages the sparsity structure of the demand space and provably achieves an asymptotically optimal revenue. Customer and Server Behavior in Queues Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Mirko Kremer, Frankfurt School of Finance and Management gGmbh, Frankfurt, 60314, Germany, m.kremer@fs.de 1 - Mismanaging the Quality-speed Tradeoff in Congested Environments Mirko Kremer, Frankfurt School of Finance and Management, Frankfurt, 60314, Germany, m.kremer@fs.de, Francis de Véricourt The tradeoff between diagnostic accuracy and speed permeates many manufac- turing and service settings. We present the results from a set of controlled labora- tory experiments designed to test the predictions of a formal sequential testing model that captures this tradeoff in a setting where the gathering of additional information (i.e., diagnostic testing) is likely to improve diagnostic judgments, but may also increase congestion in the system. We find that decision makers are insufficiently sensitive to congestion, with an aversion to stopping a diagnostic process in the face of increasing system congestion. On the other hand, decision makers are overly sensitive to diagnostic signals, with a tendency to stop a diag- nostic process immediately after the first test result, even at low congestion levels that render additional testing inexpensive. As a result of these behavioral pat- terns, besides substantial heterogeneity in how they trade off quality and speed, the majority of decision makers manage their system with both lower-than- optimal diagnostic accuracy and higher-than-optimal congestion cost. 2 - The Effects of Discrete Work Shifts on Non-terminating Queues Robert Batt, University of Wisconsin-Madison, Madison, WI, 53706, United States, bob.batt@wisc.edu, Diwas S.KC, Bradley R. Staats, Brian Patterson We study an emergency department to see how the finite-length shift structure impacts productivity and quality. Using simulation, we show that policies thatprohibit starting new patients near the end of the shift can lead to improved system throughput. 3 - An Empirical Study of Customer Spillover Learning about Service Quality Andres I. Musalem, Universidad de Chile, Beauchef 851, Santiago, 8370456, Chile, amusalem@duke.edu We study social learning in a multi-arm bandit setting where each of a sequence of arriving customers chooses between alternative service providers with uncertain quality, observing only the outcomes from other customers’ service experiences. In a series of experiments, we test the key theoretical prediction that the social learning process results in insufficient exploration, relative to a ,,centralized“ benchmark regime where customers learn about the providers’ quality from their own (previous) choices. SA03C Grand Ballroom C

320A Sustainable and Socially Responsible Operations Sponsored: Manufacturing & Service Oper Mgmt, Sustainable Operations Sponsored Session Chair: Can Zhang, Georgia Institute of Technology, Atlanta, GA, 30309, United States, czhang2012@gatech.edu Co-Chair: Atalay Atasu, Georgia Institute of Technology, Atlanta, GA, 30308, United States, atalay.atasu@scheller.gatech.edu 1 - Agricultural Support Prices in Developing Economies: Operational Analysis and its use in Policy Making Tharanga Kumudini Rajapakshe, University of Florida, 10287, SW. 37th Place, Gainesville, FL, 32608, United States, tharanga@ufl.edu In this paper, we offer analytically-supported insights on several fundamental aspects of the Guarenteed Support Price scheme by analyzing a Stackelberg game between a homogenous population of small farmers and a social planner. To understand the welfare implications of the GSP scheme on its main stakeholders, we characterizethe equilibrium operational decisions of the farmers and the equilibrium value of the GSP. 2 - The Value of Strategic Farmers, Social Entrepreneurs and For-profit Firms in Crop Planting Decisions Wenbin Wang, Shanghai University of Finance and Economics, Shanghai, China, wang.wenbin@mail.shufe.edu.cn, Ming Hu, Yan Liu We study how farmers of heterogeneous production costs make crop-planting decisions over time to maximize their incomes. We consider both strategic farmers, who rationally anticipate the near-future prices and naive farmers, who shortsightedly react to recent crop prices. We find self-interested behavior of strategic farmers alone can reduce price volatility and benefit all farmers. In the absence of strategic farmers, an optimally-designed pre-season procurement contract, offered by a social entrepreneur or for-profit firm, also brings benefit to farmers as well as to the firm itself. However, a poorly designed contract may distort the market and drive non-contract farmers out of business. 3 - Inter-dependency among Mitigation, Preparedness, and Response Decisions in Disaster Management Cycle Shabnam Rezapour, Florida International University, Miami, FL, United States, srezapou@fiu.edu, Reza Zanjirani Farahani, Alfonso J. Pedraza Martinez This research explores the interdependency between the stages of mitigation (fortifying stocks), preparedness (locating stocks), and response (distributing stocks timely) in disaster management cycle. Using a stochastic model, the focus is on minimizing the total logistics costs. We show that controlled decentralization, centralization, and decentralization are the best locating strategy to minimize logistics costs for expensive goods, for cheap goods, and in relief networks with high transportation cost. Stock fortification (in mitigation stage) makes it possible to have short response intervals (in response stage) and fulfill demand with less number of stocks (in preparedness stage). 4 - Perishable Inventory Sharing in a Two-location System Can Zhang, Georgia Institute of Technology, 499 Northside Circle NW, Apt 315, Atlanta, GA, 30309, United States, czhang2012@gatech.edu, Turgay Ayer, Chelsea C. White Motivated by a platelet inventory management problem in a local hospital network, we study an inventory sharing problem for perishable products in a two-location system. We derive structural properties of the optimal policy and present managerial insights that are significantly different from those in existing studies for traditional non-perishable products.

3

SA06

INFORMS Houston – 2017

SA06

mechanisms that prevent fraud. We propose a systems approach for Medicaid fraud inspection and prevention that maximizes the risk of committing fraud for providers serving Medicaid population. 2 - The Lifecycle of Maturity for Organizations to Fight Fraud Arnie Greenland, Robert H. Smith School of Business, 4342 Van Munching Hall, College Park, MD, 20742, United States, agreenland@rhsmith.umd.edu Organizations that are faced with the daunting task of detecting, reducing and eventually eliminating fraud differ greatly in their capabilities to succeed in that endeavor. In this paper, the author will propose a natural lifecycle of development that organizations traverse in developing the data and capabilities to improve their ability to effectively fight fraud; and he will offer some examples, appropriately masked. While the paper will focus primarily on the data and organizational requirements, the author will also discuss the key analytical tools, capabilities and skill sets required as part of the organizational development process. 3 - What Indicators Matter for Fraud Identification in Healthcare and to what Extend do they Matter Didem Egemen, PhD Candidate, Integrity Management Services Inc., Alexandria, VA, United States, didem@gwu.edu, Paulo Macedo, Sewit Araia Detecting healthcare fraud is not a straightforward process since it requires outlier detection methods for large number of indicator. In this study, we recommend combining different methods, which are Mahalanobis distance, Singular Value Decomposition, and One Class Classification. Our goal is to obtain outliers within healthcare providers by taking into account the different information provided by these three methods and try to reduce the number of false positives. For this purpose, the observations flagged differently by these methods are compared by using Multivariate Lorenz curves and the equivalence of these Lorenz curves are tested by using the Bootstrapping method. 322B Smart and Connected Health Sponsored: Health Applications Sponsored Session Chair: Muge Capan, Christiana Care Health System, Wilmington, DE, 19801-3082, United States, mcapan@ncsu.edu 1 - Data-Driven Approach to Early Warning System Implementation Stephen Hoover, MS, Christiana Care Health System, Value Institute, Newark, DE, United States, Stephen.Hoover@ChristianaCare.org Early warning systems are designed to alert healthcare providers to changes in patient status. Despite good intentions, poor implementations can lead to an increase in alert burden and alarm fatigue. We will discuss the prospective analysis of the impact of early warning systems’ on alerts prior to implementation from both an analytical modeling and information technology perspective. Post- implementation results will also be presented. 2 - Signaling Sepsis and Forecasting the Unexpected Transfer to Upgraded Resources in Sepsis Muge Capan, Associate Director of Health Systems Optimization, Newark, DE, Muge.Capan@ChristianaCare.org, Kristen Miller Signaling Sepsis, funded by the National Library of Medicine, creates an evidence- based framework to optimize sepsis clinical decision support. One of its quality improvement projects, FUTURE (Forecasting the Unexpected Transfer to Upgraded Resources in sepsis), uses gamification to elicit predicted patient outcomes from healthcare teams, thus enabling patient-level feedback, raising sepsis awareness, and identifying clinicians who best predict patient decompensation. 3 - Computer-aided Multimodal Health Monitoring System with Wearable Microfluidic Sweat Sensors Daehan Won, Binghamton University, Vestal, NY, United States, dhwon@binghamton.edu Quantitative, simple, and low-cost analysis of biofluids is essential to prevent and diagnosis variety of health conditions along with the understanding metabolisms of human physiology. Therefore, we introduce a new health-related information monitoring framework; various chemicals in human sweat will be measured and analyzed by a newly designed wearable sensors and the mobile application respectively. Our unified framework will hold great potentials for future applications. SA08

320C Design and Evaluation of Differentiated Care Interventions Sponsored: Health Applications Sponsored Session Chair: Diana Maria Negoescu, University of Minnesota, Minneapolis, MN, 55455, United States, negoescu@umn.edu 1 - Machine Learning Discovery of Longitudinal Patterns of Depression and Suicidal Ideation Jue Gong, University of Washington, Industrial and Systems Engineering, MEB B14 BOX 352650, Seattle, WA, 98195, United States, gongjue@uw.edu, Gregory E. Simon, Shan Liu Depression is often accompanied by thoughts of self-harm which are a strong predictor of subsequent suicide attempt and suicide death. We used artificial neural networks to extract latent structures from the trajectory data of depressive symptoms in an on-going treatment population from electronic health record. We discovered correlations between patients’ depression symptoms (PHQ-8) and suicidal ideation (the 9th question of PHQ) through cross-correlation analysis. This work paves the way toward the development of personalized depression monitoring and suicide prevention strategies. 2 - Optimal Learning of the Tumor Response Parameters Distribution in Spatiotemporally Integrated Radiotherapy Ali Ajdari, University of Washington, Seattle, WA, United States, ajdari@uw.edu, Archis Ghate One of the key assumptions in the linear-quadratic formulation of the radiotherapy problem is that the distribution of uncertainty in tumor-response parameters is known to the treatment planner. Recent advances in imaging techniques however, have brought forth the possibility of relaxing this assumption and using tumor-images acquired over the treatment course to learn patient-specific parameter distributions. Therefore, the goal in this study is to learn the tumor’s response while optimally dosing a patient over the treatment course. 3 - Response-adaptive Design of Dose-finding Trials In Clinical trials human participants are assigned to one or more interventions to evaluate the effects of those interventions on health-related outcomes. In a standard dose-finding trial, patients are randomly assigned to predetermined doses such that the number of patients assigned to each dose is roughly equal. Such a design may be inefficient. A better strategy is adaptive design in which modifications are made to dose assignments, while the study is in progress. We consider a response-adaptive dose-finding clinical trial where the decision maker is interested in identifying the optimal dose, and allocation of future patients to doses is based on knowledge gained from previous outcomes 4 - Personalized Prediction of Glaucoma Progression under Different Target Eye Pressure Levels Amir Nasrollahzadeh, Clemson University, Clemson, SC, United States, snasrol@g.clemson.edu, Amin Khademi

Pooyan Kazemian, Harvard University, 50 Staniford Street, 9th Floor, Room 930B, Boston, MA, 02114, United States, pooyan.kazemian@mgh.harvard.edu, Mariel Sofia Lavieri, Mark P. Van Oyen, Joshua Stein

Using perimetric and tonometric data from two randomized clinical trials, we developed and validated Kalman filter models for fast-, slow-, and non- progressing patients with glaucoma. Our models identify patient type and generate personalized and dynamically-updated forecasts of glaucoma progression under different target eye pressure levels. This approach can help eye doctors determine an appropriate treatment regimen and tailor care to individual patients.

SA07

322A Medicaid Fraud Detection and Prevention Sponsored: Health Applications Sponsored Session

Chair: Elham Torabi, James Madison University, torabiex@jmu.edu 1 - Medicaid Fraud Detection and Prevention: A Systems Approach Elham Torabi, Assistant Professor, James Madison University, 1933 Buttonwood Ct., Harrisonburg, VA, 22802, United States, torabiex@jmu.edu, Verbus Counts Medicaid fraud and abuse result in significant waste of resources. Any dollar saved in the Medicaid program can be used in improving access for those who need care. Auditing Medicaid claims and investigating fraudulent activities is time and money consuming. Therefore, what is crucially needed more than ever is

4

SA11

INFORMS Houston – 2017

SA10

4 - Patient-centered and Personalised Scheduling of Colonoscopy Appointment Karmel Shehadeh, University of Michigan, Ann Arbor, MI, 48109, United States, ksheha@umich.edu, Amy Cohn We consider the problem of scheduling colonoscopy patients in an endoscopy unit, recognizing the impact of the quality of the pre-procedure bowel preparation that the patient must undergo in the variability of procedure duration. High-quality preparation can lead to short procedure duration while poor preparation can make the procedure much longer. Combined with patient no-show and arrivals uncertainties, this duration structure contributes to a schedule with many outliers. We use simulation and stochastic programming to analyze and improve the scheduling of colonoscopy patients. Attention is paid to the competing schedule metrics, including patient waiting, provider idle and over times 330A Minimizing Food Waste Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Dorothee Honhon, University of Texas at Dallas, Richardson, TX, 75080, United States, dorothee.honhon@utdallas.edu Co-Chair: Xiajun Amy Pan, University of Florida, Gainesville, FL, 32611-7169, United States, amy.pan@warrington.ufl.edu 1 - Package Size and Pricing Decisions with a Bulk Sale Option Ismail Kirci, University of Texas at Dallas, 800 West Campbell Road, Richardson, TX, 75080, United States, Ismail.Kirci@utdallas.edu, Dorothee Honhon, Alp Muharremoglu We investigate the package size and pricing decisions of a retailer selling a perishable product to a population of heterogeneous consumers who differ in their valuation of the product. We also study the pricing decision when the retailer sells the product in bulk and compare bulk sale with package sale in terms of retailer`s profit and consumers` waste. 2 - Food Distribution and Food Waste in Food for Education Programs Debjit Roy, Indian Institute of Management, House Number 308, IIM. Ahmedabad, Vastrapur, Ahmedabad, 560078, India, debzitt@gmail.com, Elena Belavina Mid-day meal schemes are popular government subsidized schemes run in developing economies, which incentivizes children to attend school. What factors can improve the effectiveness of such state-run food distribution schemes? We develop an empirical model to identify the factors and develop a daily demand estimation model for different school categories. 3 - Policies to Minimise Food Waste in Retail Environment Emel Aktas, Cranfield University, School of Management, Bedfordshire, Cranfield, MK43 0AL, United Kingdom, emel.aktas@cranfield.ac.uk, Akunna Oledinma, Hafize Sahin, Zahir Irani, Amir Sharif, Samsul Huda, Zeynep Topaloglu, Mehran Kamrava Food retail is one of the largest sectors both in the UK and the US. IGD predicts the UK food and grocery market to reach £200bn over the next five years. Supermarkets sell 60-90% of all fresh fruits and vegetables, with high levels of imported produce. We focus on fresh fruits and vegetables due to their high value and perishability. We build a causal loop of the fresh fruits and vegetables supply chain in Qatar, considering imported and locally grown produce, the demand, and the waste due to inefficient operational processes and consumer behavior. Using the deSolve library in R, we test inventory control policies based on service levels, lot sizes, and demand across a range of fresh produce categories. 4 - Shelf Space Optimization of Perishable Products Arzum E. Akkas, Boston University, 100 Memorial Drive, 8-11C, Cambridge, MA, 02142, United States, aakkas@bu.edu In practice, shelf lives within a product category can vary significantly due to product characteristics such as sugar content and pungency levels (e.g, 3 months versus 6 months); yet, shelf life information is not considered when configuring planograms. This paper examines the shelf space allocation problem from the perspective of the relationship between shelf life and expiration. SA09

330B Operational Issues in Organ Transplants Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Baris Ata, University of Chicago, Chicago, IL, 60637, United States, Baris.Ata@chicagobooth.edu Co-Chair: Ali Cem Randa, The University of Chicago-Booth School of Business, 5807 South Woodlawn Avenue, Chicago, IL, 60637, United States, randa@chicagobooth.edu 1 - Optimizing Simultaneous-offering Mechanism of Cadaveric Organs Tinglong Dai, Johns Hopkins University, 100 International Dr, Baltimore, MD, 21202, United States, dai@jhu.edu In this study, we consider an OPO’s problem of determining the optimal batch size of simultaneous offers made to transplantation centers. We model the strategic interaction among transplant centers both within and across batches, generating structural properties and computational insights. 2 - Patient Centered Organ Acceptance Decisions in Kidney Transplant Sanjay Mehrotra, Northwestern University, Dept of I.E./ M.S.C246 Tech Inst., 2145 Sheridan Road, Evanston, IL, 60208-3119, United States, mehrotra@iems.northwestern.edu It is not always clear if an offered kidney should be accepted by a patient. We will present results from a patient centered model for accepting an offered kidney that quantifies its benefits using a decision tree. The parameters of the tree are quantified using appropriate statistical models, which will also be presented. 3 - Personalized Wait Time Estimates for Kidney Transplant Offers Nikolaos Trichakis, Massachusetts Institute of Technology, 100 Main Street, E62-576, Cambridge, MA, 02143, United States, ntrichakis@mit.edu, Chaithanya Bandi, Phebe T. Vayanos We deal with the problem of estimating the wait time until a patient on the kidney national wait list is offered an organ of desired quality, based on the patient’s own characteristics and knowledge of the wait list status. We discuss the unique challenges this estimation problem poses, and argue that existing estimation methods are inadequate. We use a maximum likelihood process to calibrate our method, based only on data that patients observe in practice. Using highly-detailed historical data, we conduct numerical experiments that show our method to significantly outperform existing ones. 4 - An Empirical Analysis of the Effect of Kidney Allocation Policies on Patient Behavior A. Cem Randa, University of Chicago-Booth School of Business, 5807 S Woodlawn Avenue, Chicago, IL, 60637, United States, randa@chicagobooth.edu, Baris Ata Organ Procurement and Transplantation Network (OPTN) allocates deceased donor kidneys to the patients based on an additive point system. The patients accumulate points for waiting time and other factors, and organs are offered to patients according to their points. The patients carry no obligation to accept any organ offers. OPTN continuously updates the weight of factors that are affecting contributing to point system, in order to have a more effective allocation policy. However, this effort excludes the patient behavior. We will develop an empirical model to capture the affect of patient behavior in this system and evaluate different counterfactual policies empirically.

SA11

332A Behavioral Operations in Services Sponsored: Manufacturing & Service Oper Mgmt,

Service Operations Sponsored Session Chair: Mor Armony, New York University, New York, NY, 10012, United States, marmony@stern.nyu.edu 1 - Experiment of Hospital Unit Admission Decision Behavior under Congestion Song-Hee Kim, University of Southern California, Bridge Hall 307A, Los Angeles, CA, 90089, United States, songheek@marshall.usc.edu, Jordan Tong Hospital inpatient units have limited capacity. We explore how physicians make admission decisions in order to understand how to improve this important decision-making process. Specifically, in a controlled laboratory experiment setting, we observe and compare admission decision-making behaviors based on current unit occupancy and severity of arriving patient conditions.

5

SA12

INFORMS Houston – 2017

2 - Allocating Inpatient Beds to Off-service Patients: Tradeoffs and Consequences

4 - Managing High Worker Turnover: Productivity & Peer Effects on Modern Production Lines Ken Moon, The Wharton School, Philadelphia, PA, 19104, United States, kenmoon@wharton.upenn.edu We study the drivers and consequences of high worker turnover in modern, large-scale production processes. Leveraging a uniquely rich dataset drawn from China-based final assembly, testing, and packaging (FATP) facilities producing millions of units of consumer electronic goods per week, we study the role of peer effects in worker turnover and quantify the effects of such turnover on production outcomes. We prescribe policies to both manage turnover and mitigate its effects: in practice, managers lack guidance on the optimal mix of control versus mitigation and often suspect that they may undervalue worker experience. 332C Health Care, Strategy and Policy Contributed Session Chair: Ashlea Milburn, University of Arkansas, Fayetteville, AR, United States, ashlea.bennett@uark.edu 1 - The Impact of Bundled Payment Policy On Health Care Operations Yiming Fan, University of Science and Technology of China, Hefei, 230026, China, jinkui0612@163.com Healthcare reimbursements in China have been traditionally based on a fee-for- service (FFS) scheme, leading to the rapid increase in healthcare costs, since FFS provides incentives for high volume of care, rather than efficient care. During the healthcare reform, a new reimbursement scheme, called bundled payment (BP), is introduced in order to remove such incentives. In this paper, we aim to empirically study the impact of BP on the operational performances of a healthcare system using a data set from Anhui New Rural Cooperative Medical System. 2 - Comparative Two Types of the Value Co-creation Model of Long-term Care for the Elder in Taiwan Ching-Fang Wu, PhD Student, Nation Cheng Kung University, No.1, University Road, Tainan City, Taiwan (R.O.C), 701, Taiwan, r48021037@mail.ncku.edu.tw, Shih-Chieh Fang, Ching-Ying Huang The purpose of this research is to examine the elderly customers’ satisfaction and well-being from customer Effort in Value Co-creation Activities point of view. Besides, we compare two different models of long-term care and try to find which factors affect customer satisfaction and quality of life. This study is a quantitative study conducted by questionnaires. There are two types of Subjects for the questionnaires, one is the elder who lives in the day-care center, the other one is in the hospital-based institution. Three parts of the questionnaires which are demographic and living characteristics, value co-creation activities, satisfaction, and life satisfaction scale. 3 - Multiple Chronic Condition Patterns in Baby Boomers, Generation X, and Early Millennials Ajit Appari, University of Texas Health Science Center at Houston, 1200 Pressler Street, RAS.W-310, School of Public Health, Houston, TX, 77030, United States, ajit.appari@uth.tmc.edu, Maria Ukhanova, Robert O. Morgan Multiple chronic conditions (MCC) in working-age adults pose enormous burden to the American economy. Critical gaps exist in our understanding of MCC patterns and their geographic variation among working-age adults. In this study we analyzed claims data on 451,694 individuals, continuously enrolled with a leading insurer in Texas during 2008-2013. Using Exploratory Factor Analysis, we determined five major patterns- Metabolic, Cardiovascular, Musculoskeletal, Liver-Immune Disorder, and Neurologic-Pain. Next, logistic regression was used to identify community socioeconomic factors (zip code) that correlate with MCC risk, adjusting for selection bias using Heckman’s approach. 4 - A Study of the 2008 Subprime Mortgage Crisis’ Impact on Taiwanese Pregnant Women and Their Children Development by an Epigenetic Approach Jonathan Tong, PhD Program Student, National Taiwan University, 7F., No. 28, Ln.325, Zhungjing Rd., Taipei, 110, Taiwan, d03323003@ntu.edu.tw Previous studies show that embryo will change functionally relevant to cellular and physiological phenotypic traits without involving a change in the nucleotide sequence, nor in DNA sequence, when pregnant women endure external/environmental impacts. This study will investigate Taiwanese children born in the period of economic crisis 2008 based an epigenetic approach and see if such changes also include some critical social cognitive capabilities. SA13

Hummy Song, The Wharton School, University of Pennsylvania, 560 Jon M. Huntsman Hall, 3730 Walnut Street, Philadelphia, PA, 19104, United States, hummy@wharton.upenn.edu, Anita L. Tucker, Ryan Graue, Sarah Moravick, Julius J. Yang Given a highly variable patient census at the service level yet a fixed allocation of inpatient beds to services, a significant portion of admitted patients become “off- service” patients. These patients are physically located in a bed that belongs to a different service (e.g., general surgery) while still being cared for by a physician of the service (e.g., cardiac medicine). We examine the tradeoffs and consequences of assigning incoming patients to an off-service bed as opposed to an on-service bed. 3 - The Influence of Customer Emotions in Chat Service Operations Galit Yom-Tov, Technion - Israel Institute of Technology, Techbion City, Haifa, 32000, Israel, gality@technion.ac.il, Anat Rafaeli, Daniel Altman, Monika Westphal, Michael Natapov, Neta Barkay We develop a sentiment analysis tool for objective assessment of customer emotions in interactions between service agents and customers. We then explore customer sentiment of real customer-agent interactions in three companies. Our analyses shed light on the presence of positive and negative emotion in customer service, and identify patterns of customer emotion over the course of an interaction. The analyses quantify the link between customer emotions, service quality measures, and agent behaviour. We discuss implications of using automated sentiment analysis tools for the management of service centers. 4 - Strategic Capacity Management in Outpatient Care with No-shows and Rescheduling Yunchao Xu, New York University, 44 W 4th Street, KMC#8-152, New York, NY, 10012, United States, yxu4@stern.nyu.edu, Mor Armony, Nan Liu An important factor that determines whether a patient will show up or not for his scheduled appointment is how easy to reschedule if he chooses not to show up. In this talk, we discuss how an outpatient care provider should manage her practice by considering such patient strategic behavior. While this work is motivated by healthcare, our models and insights can also be applied to appointment-based services in other settings. 332B Empirical Studies of Supply Chains and Production Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Ken Moon, The Wharton School, Philadelphia, PA, 19104, United States, kenmoon@wharton.upenn.edu 1 - Impact of Order-fulfillment Service Levels on Customer Value Nitish Jain, London Business School, Sussex Place, Regent’s Park, London, NW1 4SA, United Kingdom, njain@london.edu, Karan Girotra, Andres I. Musalem Faster fulfillment of an order is one of the key service components for online retailers. Based on a detailed transactional data on fulfillment performance (actual and promised) and on subsequent purchase patterns, we impute the economic Ioannis Stamatopoulos, University of Texas at Austin, McCombs School of Business, 2110 Speedway B6000, Austin, TX, 78705, United States, yannis.stamos@mccombs.utexas.edu, Soheil Ghili We leverage detailed data from an international retailer who adopted the electronic shelf label technology to some - but not all - of its stores to identify the underlying operational costs of pricing. We impose a simple model for pricing on that data and use the adoption shock to separate the effects of menu costs on prices from those of changes in demand trends. Our approach can be thought of as a structural implementation of the difference-in-difference identification strategy. 3 - Does Ration Gaming Drive the Bullwhip Effect? Robert Louis Bray, 830 Hinman Ave, 2s, Evanston, IL, 60202, United States, robertlbray@gmail.com We model a single-supplier, 73-store supply chain as a dynamic discrete choice problem. We estimate the model with transaction-level data, spanning 3,251 products and 1,370 days. We find two phenomena: ration gaming (strategic inventory hoarding) and the bullwhip effect (the amplification of demand variability along the supply chain). We estimate that the bullwhip effect would be 12% smaller in the counterfactual scenario without ration gaming incentives, confirming the long-standing hypothesis that ration gaming causes the bullwhip effect. value of faster and timely fulfillment to online retailers. 2 - Estimating the Operational Costs of Pricing SA12

6

SA15

INFORMS Houston – 2017

5 - A Study of Co-occurrence of Multiple Chronic Conditions using Online Healthcare Forums Juxihong Julaiti, PhD Student, Penn State University, 445 Waupelani Drive, Apt J1, State College, PA, 16801-4445, United States, jpj5196@psu.edu In recent years participation in online forums to discuss health related issues is increasing. We study the on-line forums to study the relationship between multiple chronic conditions and resources allocation. We use linear, logistic regression, random forest to derive the relationships. The results indicate that: 1) patients with cancer and depression are more likely to participate in online forums; 2) patients with costly chronic conditions are more likely to use online forums; 3) whether a patient has cardiovascular disease, arthritis are significant variable to predict if the patient will have asthma etc. 6 - Dispensing Medical Countermeasures in Public Health Emergencies via Home Health Agencies and Points of Distribution Ashlea Bennett Milburn, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, ashlea.bennett@uark.edu, Anna Hudgeons, Charleen McNeill A major concern regarding emergency preparedness at the state government level involves the handling and dispensing of the Strategic National Stockpile of medicinal supplies. This research utilizes simulation modeling to determine if using home health agencies (HHAs) in addition to Points of Distribution (PODs) to dispense SNS supplies can provide aid in a more timely fashion than using PODs alone. The comparative effectiveness of the two dispensing methods is demonstrated via a case study of pandemic influenza in Northwest Arkansas. 332D Analytics and Incentives for Efficient Healthcare Delivery Sponsored: Manufacturing & Service Oper Mgmt, Healthcare Operations Sponsored Session Chair: Hamsa Sridhar Bastani, Stanford University, Stanford, CA, 94305, United States, hsridhar@stanford.edu 1 - Yardstick Competition for Hospital Queues Nicos Savva, London Business School, London, United Kingdom, nsavva@london.edu, Tolga Tezcan, Ozlem Yildiz In this paper, we first show that the hospital reimbursement system currently used in practice does not incentivize hospitals to reduce waiting times. We then propose a modification which can achieve socially optimum investment without placing an onerous informational burden on hospital payers. 2 - Misaligned Incentives in Kidney Exchange Itai Ashlagi, Stanford University, Department of Management Sci. and Engr., Stanford, CA, 94305, United States, iashlagi@stanford.edu Growth of live-donor kidney exchange has stagnated although an increasing number of transplant centers are facilitating exchanges through large national exchanges, in isolation or in small consortiums with other centers. Using a novel dataset we document significant fragmentation and heterogeneity in participation by centers, an adversely selected and hard to match national exchange pool, and inefficient transplants outside the largest exchange. Using a simple supply and demand model we estimate the inefficiency and in the current market and calculate a rewards scheme that would be implemented by the optimal mechanism. 3 - Exploiting the Natural Exploration in Contextual Bandits Hamsa Sridhar Bastani, IBM Thomas J. Watson Research Center, Yorktown Heights, NY, United States, hamsab@wharton.upenn.edu, Mohsen Bayati, Khashayar Khosravi The contextual bandit literature focuses on an exploration-exploitation tradeoff because exploration-free greedy policies may yield poor performance in general. However, greedy policies are desirable when experimentation is costly or unethical (e.g., clinical settings). We posit a sufficient set of additional assumptions under which a greedy policy is asymptotically optimal. Next, we present Greedy-First, a new algorithm that uses observed data to determine whether to follow a greedy policy or to explore. This algorithm is asymptotically optimal without our additional assumptions, and significantly reduces experimentation in simulations. SA14

4 - Dynamic Bandit Approach for Personalized Treatment Yonatan Mintz, UC Berkeley, San Francisco, CA, United States, ymintz@berkeley.edu, Anil Aswani, Philip Kaminsky, Elena Flowers, Yoshimi Fukuoka Personalized fitness tracking devices enable the implementation of data driven exercise programs intended to treat sedentary lifestyles. Despite the prevalence of these devices, many individuals have a hard time adhering to the recommended programs since they may be incompatible with an individual’s schedule or provide ineffective motivation. In this talk, we address this problem by leveraging the data and infrastructure of mobile fitness tracking devices to personalize exercise programs for participating individuals. We develop a multi-armed bandit approach to adaptively learn each participant’s exercise preferences to personalize their exercise programs and increase adherence. 332E Demand Learning Sponsored: Manufacturing & Service Oper Mgmt, Supply Chain Sponsored Session Chair: Yiangos Papanastasiou, University of California Berkeley, Berkeley, CA, 94720, United Kingdom, yiangos@haas.berkeley.edu 1 - Dynamic Selling Mechanisms for Product Differentiation and Learning N. Bora Keskin, Duke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC, 27708-0120, United States, bora.keskin@duke.edu, John R.Birge We consider a firm that designs a menu of vertically differentiated products for a population of customers with heterogeneous quality sensitivities. The firm faces an uncertainty about production costs. We characterize the structure of the firm’s optimal dynamic learning policy and construct simple and practically implementable policies that are near-optimal. 2 - Bayesian Dynamic Learning and Pricing with Strategic Customers Xi Chen, New York University, 44 W. 4th St, NYU. KMC Room 8-50, New York, NY, 10012, United States, xchen3@stern.nyu.edu, Zizhuo Wang In this talk, we study such a learning problem when the customer is aware of the seller’s policy, and thus may behave strategically when making a purchase decision. We propose a randomized Bayesian policy (RBP), which updates the posterior belief of the customer in each period with a certain probability. We show that the seller can learn the customer type exponentially fast with the RBP even if the customer is strategic, and the regret is bounded by a constant. We also propose policies that achieve asymptotically optimal regrets when only a finite number of price changes is allowed. 3 - Dynamic Pricing with Online Product Reviews Dongwook Shin, Hong Kong University of Science and Technology, Kowloon, Hong Kong, dwshin@ust.hk Dongwook Shin, Columbia University, New York, NY, 10027, United States, dwshin@ust.hk, Assaf Zeevi We investigate how the presence of product reviews affects a dynamic-pricing monopolist who strives to maximize its total expected revenue. To formulate the problem in tractable form, we first study a fluid model, which is a good approximation when the volume of sales is large. In the context of the fluid formulation, the optimal pricing policy can be characterized in a closed-form, which is leveraged to design pricing policies that perform well with respect to the underlying revenue maximization problem. Lastly, we discuss the impact of product reviews on learning when the monopolist operates without knowing the demand function a priori. 4 - A Tutorial on Thompson Sampling Daniel Russo, Northwestern University, Chicago, IL, United States, dan.joseph.russo@gmail.com Thompson sampling is an algorithm for online decision problems where actions are taken sequentially in a manner that must balance between exploiting what is known to maximize immediate performance and investing to accumulate new information that may improve future performance. The algorithm addresses a broad range of problems in a computationally efficient manner and is therefore enjoying wide use. I will present selected examples and insights from a recent tutorial on Thompson sampling. The talk will focus on applications in revenue management, and on practical challenges like nonstationarity, prior specification, and the need for approximate posterior inference. SA15

7