2015 Informs Annual Meeting

WB34

INFORMS Philadelphia – 2015

3 - Using Linear Programming Based Exploratory Techniques in Gene Expression Consensus Clustering Victoria Ellison, North Carolina State University, Campus Box 7913, 2500 Stinson Drive, Raleigh, NC, 27695, United States of America, vmelliso@ncsu.edu, Yahya Fathi, Amy Langville We propose a divisive hierarchical consensus clustering algorithm (DHCCA) by modifying a BILP formulation of the Median Partition problem and applying several proposed parametric programming algorithms based on the optimal partition parametric programming algorithm. We prove that executing our DHCCA, under certain assumptions, is equivalent to solving the minimum ratio cut problem. This equivalency can be useful in creating fast heuristics to the Median Partition problem. 4 - Using Data Mining to Predict Drug Courts Outcome Hamed Majidi Zolbanin, Oklahoma State University, 309 S. West Drug court is an alternative for traditional criminal courts that attempts to shift from a punitive to a therapeutic jurisprudence. Under this new philosophy, the eligible offenders are held as individuals in need of rehabilitative treatments. The initiative is proved to be effective in lowering the costs and breaking the cycle of narcotics use. To better manage the resources and maximize the benefits, this study develops a model to predict who will or will not graduate from these courts. WB33 33-Room 410, Marriott Methods and Applications in Disease Detection and Treatment Sponsor: Health Applications Sponsored Session Chair: Shan Liu, Assistant Professor, University of Washington, Seattle, WA, United States of America, liushan@uw.edu 1 - Sequencing Chemotherapy Agents for Metastatic Colorectal Cancer Patients Lakovos Toumazis, PhD Candidate, Department of Industrial and Systems Engineering, University at Buffalo, SUNY, Buffalo, United States of America, iakovost@buffalo.edu, Artemis Toumazi, Loukia Karacosta, Daniel A. Goldstein, Changhyun Kwon, Murat Kurt Colorectal cancer is the third most lethal cancer in the US affecting both genders. Despite advancements in chemotherapy treatment, long-term survival for the advanced stage of the disease remains poor. With the goal of improving the effectiveness of chemotherapy treatment for metastatic colorectal cancer patients we developed a Markov decision process model that jointly optimize the duration and sequence of the available drugs. The obtained optimal policy improves survival by at least 6 months. 2 - Large-scale Personalized Health Surveillance by Selective Sensing Ying Lin, University of Washington, Box 352650, Seattle, WA, 98195-2650, United States of America, linyeliana.ie@gmail.com, Shan Liu, Shuai Huang Detecting subjects who have been on the trajectory towards disease onset holds promises for preventative healthcare. Development of personalized health surveillance is enabled by sensing and information technologies. To scale up personalized surveillance, we developed a selective sensing method that integrates degradation modeling, prognosis, and optimization, which can cost- effectively monitor a large number of individuals by exploiting the similarities of their disease trajectories. 3 - Cost Effectiveness of Expanding Anti-retroviral Therapy to Untreated Subpopulations in Botswana Thomas Keller, University of Minnesota, 111 Church St. SE, St, Unit 6, Stillwater, Ok, 74075, United States of America, hamed.majidi@gmail.com, Durand Crosby, Dursun Delen

4 - Minimizing Overdiagnosis in Cancer Screening Maboubeh Madadi, University of Arkansas, mmadadi@uark.edu, Shengfan Zhang, Edward Pohl, Chase Rainwater Overdiagnosis is defined as the diagnosis of screen-detected cancers that would not have presented clinically in a woman’s lifetime in the absence of screening. Overdiagnosis is known to be the most important disadvantage of cancer screening and it can adversely affect people’s lives. In this study a mathematical programming model is developed to find the optimal cancer screening policy with respect to overdiagnosis risk, while maintaining the lifetime cancer mortality risk at a low threshold. WB34 34-Room 411, Marriott Medical Decision Making Sponsor: Health Applications Sponsored Session Chair: Pooyan Kazemian, PhD Candidate, University of Michigan- Ann Arbor, 1205 Beal Ave., Ann Arbor, MI, 48105, United States of America, pooyan@umich.edu 1 - Why is Screening so Common for Some Diseases when Evidence is so Uncertain? Ozge Karanfil, PhD Candidate, MIT Sloan School of Management, 100 Main Street, E62-379, Cambridge, MA, 02142, United States of America, karanfil@mit.edu, John D. Sterman Practice guidelines for routine screening such as mammography or PSA testing have changed significantly over time. Evidence-based guidelines are often not followed by clinicians and patients, with significant over or under screening. In this study we describe a dynamic model to explain changes in policy action thresholds. We use quantitative and qualitative data to document evidence of gaps between guidelines and practice. Qualitative data includes interviews with health/medical professionals. 2 - Planning for HIV Screening, Testing, and Care at the Veteran’s Health Administration Management, B410 Gold Hall, UCLA Anderson, Los Angeles, CA, 90024, United States of America, krajaram@anderson.ucla.edu, Matthew Goetz, Uday Karmarkar, Sandeep Rath, Sarang Deo CDC has recommended a routine screening policy for HIV. We modeled a QALY maximizing nonlinear mixed integer program incorporating system dynamics and disease progression and found that routine screening may not be always feasible. We applied this model to the Greater Los Angeles station of the Veterans Health Administration and used it to develop and evaluate managerially relevant policies within existent capacity and budgetary constraints to improve upon the current screening policy. 3 - Value of Patient-centric Treatment Policies for Pelvic Organ Prolapse in Women Yueran Zhuo, Ph.D. Candidate, University of Massachusetts Amherst, Isenberg School of Management, Amherst, MA, 01003, United States of America, yzhuo@som.umass.edu, Senay Solak Pelvic organ prolapse (POP) is a common condition that impacts many women’s health and quality of life. The selection of a treatment option for POP depends on several factors, but personal preferences of the patient play a more significant role than most other similar conditions. We specifically take this aspect of POP into account and identify patient-centric treatment recommendations for this syndrome. The value of these policies is then assessed through comparisons with physician decisions. 4 - On Low-cost In-home Sensor Placement for Personalized Tracking of Activity of Older Adults Alexander Nikolaev, Assistant Professor, University at Buffalo (SUNY), 312 Bell Hall, Buffalo, NY, 14260-2050, United States of America, anikolae@buffalo.edu, Ann Bisantz, Siddhartha Nambiar, Melissa Green, Lora Cavuoto Activity-tracking sensors can now be used for personalized care; yet, challenges with usability and cost may prevent their adoption by the elderly. We assess how low-cost in-home sensor systems help infer physical activity levels of tenants. The American Time Use Survey and real apartment layouts are used to study optimal sensor count, placement, and activity prediction error levels. We describe sensor placement strategies and apartment layouts best suited for sensor technology use. Kumar Rajaram, Professor of Decisions, Operations, and Technology Management, UCLA Anderson School of

Minneapolis, MN, 55455, United States of America, kelle665@umn.edu, Gregory Bisson, Diana Negoescu, Daniel Winetsky

In 2002 the Botswana government initiated Africa’s first national anti-retroviral program, which has enrolled over 200,000 Batswana HIV-infected patients since its inception. We develop a dynamic compartmental model to evaluate the cost effectiveness of expanding this program to serve more Batswana citizens as well as migrant workers, who constitute a significant proportion of HIV-infected patients.

411