2015 Informs Annual Meeting

SC44

INFORMS Philadelphia – 2015

SC42 42-Room 102B, CC Joint Session MSOM-Health/HAS: Healthcare Analytics and Optimization Sponsor: Manufacturing & Service Oper Mgmt/ Healthcare & HAS Operations Sponsored Session Chair: Anahita Khojandi, Assistant Professor, University of Tennessee, Knoxville, TN, United States of America, khojandi@utk.edu 1 - Estimating Lipid Management Guidelines’ Risk Value of a Life Year on Treatment Murat Kurt, Merck Research Labs, 351 N. Sumneytown Pike, Statins reduce the risk of heart attack and stroke with adverse side effects, but how to quantify these effects to help physicians make treatment decisions remains to be an open question. We gauge these adverse effects for patients with Type 2 diabetes from a central policy maker’s point of view by formulating a dynamic decision model in which the objective is to minimize the risk of a first major cardiovascular event where time spent on treatment is penalized by a perceived risk increase. We seek penalty factors that make published lipid management guidelines as close as possible to optimal. We present computational results using clinical data and derive insights. 2 - Predicting No-show Behavior of Patients at a Mental Health Clinic Fan Wang, University of Arkansas, 4207 Bell Engineering Center, Mental health clinics have relatively high no-show rates of patients, which reduce provider productivity and clinic efficiency. This study presents two approaches for no-show prediction: a logistic regression model and an artificial neural network. The models are formulated using multiple factors including visit type, ICD-9 classification, insurance type, lead time to visit, no-show history, month, weekday and hour. The predictive performances are evaluated based on AUC values. 3 - Optimizing Dynamic Interventions in Sleep Studies Maryam Zokaeinikoo, Graduate Research Assistant, University of Tennessee, 1700W. Clinch Ave., Apt. 505, Knoxville, TN, 37916, United States of America, mzokaein@vols.utk.edu, Anahita Khojandi, Oleg Shylo We discuss a mathematical framework that takes advantage of the technological advances in wearable neuro-headsets to provide an objective, reliable, inexpensive and scalable approach to sleep interventions. This framework is based on semi-Markov decision models that rely on general signal processing methods for continuous sleep assessment. SC43 43-Room 103A, CC Joint Session RMP/PPSN: Socially Responsible Revenue Models Sponsor: Revenue Management and Pricing & PPSN Sponsored Session Chair: Ioana Popescu, INSEAD, 1 Ayer Rajah Avenue, 138676, Singapore, Singapore, ioana.popescu@insead.edu 1 - Is It Enough? Evidence from a Natural Experiment in India’s Agricultural Markets Kamalini Ramdas, kramdas@london.edu, Nicos Savva, Chris Parker Does access to timely and accurate information provided through ICT applications have additional impact over and above access to mobile phones, in improving market efficiency? Using data from the Reuters Market Light text message service in India that provides daily price information to market participants and a natural experiment where bulk text messages were banned unexpectedly, we find that this information reduces crop price dispersion by about 12%, over and above access to mobile phones. 2 - Certainty Equivalent Planning for Multi-product Batch Differentiation Yang Wang, UC Berkeley, IEOR Dept., Berkeley, CA, 48109, United States of America, yangwang0803@berkeley.edu, Philip Kaminsky, Stefanus Jasin Motivated by a problem in biopharmaceutical manufacturing, we consider a discrete time finite horizon inventory problem where several retailers place orders to meet stochastic demand, and in each period, the sum of order quantities across retailers must be a multiple of a standard batch size. We propose several easy-to- North Wales, PA, 19454, United States of America, murat.kurt7@gmail.com, Niraj Pandey, Mark Karwan Fayetteville, AR, United States of America, fxw005@email.uark.edu, Shengfan Zhang

implement heuristics using certainty equivalence and derive their performance bounds analytically. 3 - Bridging the Gap between for Profit and Social Responsibility Strategies Enno Siemsen, Associate Professor, University of Minnesota, 321 19th Ave S, Minneapolis, MN, 55455, United States of America, siems017@umn.edu, Lisa Jones-christensen, Sridhar Balasubramanian This field experiment compares two different for-profit market entry strategies with a philanthropic strategy in terms of how each influences consumer behavior in base-of-the-pyramid communities. We analyze reactions to a water purification product offered at three price points (moderate discount, deep discount, and free) in rural Malawi. 4 - Revenue Models for Providing Clean Energy at the Bottom of the Pyramid Ioana Popescu, INSEAD, 1 Ayer Rajah Avenue, 138676, Singapore, Singapore, ioana.popescu@insead.edu, Bhavani Shanker Uppari, Serguei Netessine One in every five people does not have access to electricity, relying mostly on kerosene for light. Solar technologies are healthier and offer greater value, yet they require significant one-time investments which are not affordable to people living on $2/day. We develop a consumer behavior model that accounts for income variability and liquidity constraints specific to bottom of the pyramid markets, and investigate alternative revenue models, based on a case study in Rwanda. SC44 44-Room 103B, CC Revenue Optimization and Related Methodologies Sponsor: Revenue Management and Pricing Sponsored Session Chair: Michael Katehakis, Professor And Chair, Rutgers University, 100 Rockafeller Road, Room, Piscataway, NJ, 08854, United States of America, mnk@rutgers.edu 1 - An Inventory System with Multiple Demand Classes Min Wang, Assistant Professor, Drexel University, 3220 Market St, Gerri C. LeBow Hall 740, Philadelphia, PA, 19104, United States of America, mw638@drexel.edu We consider a single-product inventory system with multiple demand classes. Inventories are replenished using a (R, Q) policy and are rationed among demand classes according to a threshold policy. We establish structural results for the key performance measures and develop an efficient algorithm for computing the policy parameters. 2 - Optimal Pricing for a GI/M/k/N Queue with Several Customer Types and Holding Costs This paper deals with optimal pricing for a GI/M/k/N queueing system with several types of customers. A price for a new arrival depends on the number of customers in the system. In addition, the system incurs costs caused customer delays. The holding costs are non-decreasing and convex with respect to the number of customers in the queue. This paper describes average-reward optimal, canonical, bias optimal, and Blackwell optimal policies for this pricing problem. 3 - Efficient Markov Models for Dynamic Pricing Problems Laurens Smit, Leiden University, Niels Bohrweg 2, Leiden, Netherlands, laurens@pipe.nl, Flora Spieksma, Michael Katehakis We model revenue problems as a two dimensional Markov chain, where the arrival rate of customers depend on the charged price. We consider processes that satisfy down entrance state or the restart entrance state classes of quasi skip free processes. We derive explicit solutions and bounds for the steady state probabilities of both processes, and show that these methods work fast and efficiently. In addition we present a procedure to decompose Markov processes into separate thinned processes. 4 - Models and Problems of Dynamic Pricing in the Multi-armed Bandit Framework Wesley Cowan, Rutgers University, 110 Frelinghuysen Rd., Eugene Feinberg, Distinguished Professor, Stony Brook University, Department of Applied Mathematics & Stat, Stony Brook, NY, United States of America, eugene.feinberg@stonybrook.edu, Fenghsu Yang

Piscataway, NJ, United States of America, c.wes.cowan@gmail.com, Michael Katehakis

After a brief review of basic issues of dynamic pricing under partial information on the underlying demand distributions, we provide new models that address some of these issues. A main contribution is a new model and solutions for problems that involve unobserved lost sales.

107