Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD57

4 - Innovation Positioning Under Disruptive Threats Xiaochen Gao, University of California-San Diego, La Jolla, CA, 92093, United States, Viswanathan Krishnan, Sreekumar R. Bhaskaran Typically, while developing a new technology or product, a firm wouldalso have a set of products which it currently makes available toconsumers. As a result, one of the key decisions of a firm after developinga new product is its product positioning strategy. After having investedin and developed a new product, a firm has to decide whether thisproduct should replace the existing line of products or whether boththe new and old products be sold simultaneously to consumers. In thispaper, we develop a framework for studying such an integrated productdevelopment and portfolio design problem. Specifically, we examinethe value of retaining an existing lower quality product after developinga new higher quality version of the same product family. The economicand game-theoretic analysis used to model this problem brings forthseveral interesting insights. One of the key results of this researchis that a product-line based approach to NPD can provide significantstrategic advantages to a firm. In contrast to existing literatureon development intensive products, we show that this strategy allowsa firm to cover multiple market segments and price discriminate betweenconsumers. Interestingly, the value of pursuing a product line basedapproach assumes greater significance under competition. Having multipleproducts in its product-portfolio enables a firm to extract highermargins from its high-end markets without forgoing its presence inlow-end markets. Moreover, when the level of competition is high,cannibalization tensions in maintaining a larger portfolio could alsoserve to deter mutually destructive price competition. n TD56 West Bldg 101A Screening and Treatment Models Sponsored: Health Applications Sponsored Session Chair: Steven Shechter, University of British Columbia, Vancouver, BC, V6T 1Z2, Canada 1 - Screening for Breast Cancer: The Role of Supplemental Tests and Breast Density Information Burhaneddin Sandikci, University of Chicago, Booth School of Business, 5807 South Woodlawn Avenue, Chicago, IL, 60637, United States, Mucahit Cevik, David Schacht The imperfect nature of mammography led to increased consideration of supplemental ultrasound and Magnetic Resonance Imaging (MRI) screening for timely detection of breast cancer, particularly for high-risk women including those with dense breasts. Breast density not only impairs screening accuracy, but also significantly increases risk of developing breast cancer, resulting in disproportionate risk of death from breast cancer for millions of women with dense breasts. We formulate the optimal breast cancer screening problem using a partially observable Markov decision process model. We compare its results to mass screening guidelines and quantify the value of supplemental screening. 2 - Partially Observable Collaborative Model for Optimizing Personalized Treatment Selection Jue Gong, University of Washington, Seattle, WA, 98115, United Statesu, Shan Liu Personalized treatment selection has become an increasingly important topic in health care research. The main challenges include the modeling of individual disease progression dynamics and designing the adaptive treatment selection strategy. We developed the Partially Observable Collaborative Model (POCM), to learn the individual disease progression model under various treatment options when the true health state is hidden to the decision maker. Next, utilizing the learned individual models, a personalized treatment plan can be derived by solving a POMDP. This research helps to advance the development of artificial intelligent decision support tools for chronic disease care. 3 - Analytics in Blood Pressure Management and Control: From Data to Decisions Anthony Bonifonte, Denison University, Granville, OH, 30106, United States, Turgay Ayer, Ben A. Haaland, Peter Wilson Antihypertensive drug treatment can control elevated blood pressure and reduce the risk of cardiovascular disease. We propose data-driven models that 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 initiation and intensification decisions, and capture different attitudes towards risk (e.g. risk neutral or risk averse). Our findings have policy suggestions and may help guide future RCT design.

4 - A High-fidelity Model to Predict Length-of-stay in the NICU Kanix Wang, Research Professional, University of Chicago, Booth School of Business, Chicago, IL, United States, Walid Hussain, Arnab Sur, John R. Birge, Michael Schreiber, Daniel Adelman Effective use of medical staff and clinical resources depends on the accuracy of length-of-stay forecasting to match required resources with patient load. Current forecasts are based on an initial severity ranking and are not updated as patients progress. This problem is particularly acute in the Neonatal Intensive Care Unit (NICU) setting because patient lengths of stay tend to be quite long and highly variable, extending from several days to many months. We develop a dynamic forecasting model, and apply machine learning, to predict remaining length-of- stay depending on a patient’s severity scores history, clinical progression and demographic and environmental covariates. n TD57 West Bldg 101B Amy Cohn Sponsored: Health Applications Sponsored Session Chair: Amy Cohn, University of Michigan, Ann Arbor, MI, 48109, United States 1 - Simulating the Outcome of Make-ahead Drug Policies at an Outpatient Chemotherapy Infusion Center Donald B. Richardson, University of Michigan, 2753 IOE Building, 1205 Beal, Ann Arbor, MI, 48109-2117, United States, Amy Cohn We have developed a discrete-event simulation tool to evaluate a variety of policies for selecting which drugs to pre-mix for outpatient chemotherapy patients. By making patient-specific drugs ahead of appointment time, patient waiting time can be reduced but this comes at the risk of waste cost if the patient must defer treatment once arriving for their appointment and the drug must be discarded. We utilize data from the University of Michigan Rogel Cancer Center to validate our methods. 2 - Evaluating Veteran Access to Eye Care Services Using Facility Location Models Adam VanDeusen, University of Michigan, 1205 Beal Ave, Ann Arbor, MI, 48109, United States, April Maa, Amy Cohn Many patients, including United States veterans, face barriers to accessing appropriate, affordable healthcare. These barriers can be addressed by optimizing clinic locations while delivering care that effectively utilizes providers’ practice responsibilities. We present a model to evaluate veterans’ eye care facility location options with consideration for overall system access. We further present a case study in which trained technicians perform visual disease screenings typically conducted by ophthalmologists. Our work may guide decision-makers in locating and staffing clinics to improve patient access. 3 - A Comparative Analysis of Resident Block Scheduling Models William Pozehl, Ann Arbor, MI, 48104, United States, Amy Cohn Operations research approaches to block scheduling for medical residencies have emerged over the last several years. In particular, mathematical programming models have proven popular and generally successful. Still, though, computational complexity can make solving these models excessively time- consuming. We present a comparison of different modeling approaches and evaluate their performance in relation to a case study for an internal medicine residency program at a large academic hospital. 4 - Stochastic Optimization Models for Childhood Vaccine Distribution Network: A Case Study in Niger Zahra Azadi, Assistant Professor, University of Miami, Miami, FL, United States, Sandra D. Eksioglu The main objective of this research is to increase vaccine coverage in low income countries by improving the performance of the corresponding supply chain. We propose a chance constraint programming model which identifies optimal supply chain designs and management strategies. The model considers the limited shelf life of vaccines, facility and transportation storage capacities, as well as variations in patient arrivals at health clinics. The proposed model is an extension of the supply chain network design model. A sample average approximation (SAA) method is used to solve the problem. We develop a case study for Niger by utilizing GeoNames Geographical Database and Demographic Health Survey.

367