2015 Informs Annual Meeting

SB33

INFORMS Philadelphia – 2015

SB34 34-Room 411, Marriott Smart Medical Prognosis and Decision Making via Data Mining Sponsor: Health Applications Sponsored Session Chair: Danica Xiao, PhD Candidate, University of Washington, Seattle, 3900 Northeast Stevens Way, Seattle, WA, 98195, United States of America, xiaoc@uw.edu Co-Chair: Shouyi Wang, Assistant Professor, University of Texas at Arlington, 3105 Birch Ave, Grapevine, TX, 76051, United States of America, shouyiw@uta.edu 1 - Does Specialization of Health Care Services Increase Operational Efficiency? Saied Samiedaluie, Postdoctoral Fellow, University of British Columbia, E204 - 4500 Oak Street, BC Women’s Hospital, Women’s Health Rese, Vancouver, BC, V6H 3N1, Canada, saied.samiedaluie@gmail.com, Vedat Verter We study a health care network configuration problem considering two scenarios: specialization versus generalization. We characterize the settings in which each scenario is preferred in terms of minimizing the patient admission refusal rate. Our results show that the decision of system configuration for a multi-hospital network requires careful consideration of patient mix among arrivals, relative length of stay of patients, and distribution of patient load between hospitals. 2 - A Model to Predict Depression among Diabetes Patients with Application in Screening Policymaking Haomiao Jin, University of Southern California, 3715 McClintock Ave, GER 240, Los Angeles, CA, United States of America, haomiaoj@usc.edu, Shinyi Wu About 30% of diabetes patients are suffering from depression, but nearly half of them are undiagnosed. Universal screening improves depression diagnosis rates but is labor-intensive. A machine learning model is developed to predict depression among diabetes patients. The model is applied in a screening policy to help healthcare providers to better prioritize the use of their resources and time and increase efficiency in managing their patient population with depression. 3 - Data Mining Techniques Applied to the Study of Canines with Disease Zhenpeng Miao, Saint Joseph’s University, 5600 City Ave, This paper, the third in a series, aims at providing models effective in predicting the degree of pain and discomfort in canines suffering from osteoarthritis, sarcoma, dermatitis and side effects of radiation treatments. The R programing language and SAS JMP are used to clean data and execute multivariate analyses to predict changes in different activity levels. The predictive models provide information that can assist in effective diagnosis and medication of suffering dogs. 4 - Understanding Linear and Non Linear Brain Dynamics During Manual Lifting Tasks Awad Aljuaid, PhD Student, UCF, University of Central Florida 4000 Cent, Department of Industrial Engineering, Orlando, FL, 32816-2993, United States of America, amjuaid@knights.ucf.edu, Waldemar Karwowski, Petros Xanthopoulos The aim of this study is to test the change on different EEG measures during various psychophysical lifting frequencies. High-density wireless dry cell EEG device have been used to record brain signals. Twenty healthy males participated in this experiment performing two physical lifting sessions psychophysical weight lifting (low, medium, and high) and strength measurements (Isometric and isokinetic). EEG recording at different brains locations are analyzed with linear and non-linear methods. Philadelphia, PA, 19131, United States of America, mmkuchi0@gmail.com, Yingdao Qu, Virginia Miori

3 - A Structural Approach to Community Detection in Complex Networks

Song Chew, Associate Professor, Southern Illinois University Edwardsville, Southern Illinois University-Edwardsville, Edwardsville, IL, 62026, United States of America, schew@siue.edu We in this study develop a novel measure of community structure that gauges the strengths and weaknesses of a proposed community structure against an ideal. In addition, we present an algorithm that may, as it maximizes our measure, return several alternative community structures for consideration. We provide several examples to demonstrate use of our measure, and to illustrate applications of our algorithm as well. SB33 33-Room 410, Marriott Applications of Markov Models to Medical Decision Making Problems Sponsor: Health Applications Sponsored Session Chair: M. Reza Skandari, University of British Columbia, Vancouver, Vancouver, Canada, reza.skandari@sauder.ubc.ca 1 - Evaluation of Breast Cancer Mammography Screening Policies Considering Adherence Behavior Maboubeh Madadi, University of Arkansas, mmadadi@uark.edu, Shengfan Zhang, Louise Henderso The efficacy of mammography screening guidelines is highly associated with women’s compliance with these recommendations. Currently, none of the existing policies take women’s behavior into consideration. In this study, we develop a randomized partially observable Markov chain model to evaluate a wide range of screening mammography policies, incorporating heterogeneity in women’s adherence behaviors. 2 - Policy Approximation for Optimal Treatment Planning Wesley Marrero, University of Michigan, 500 South State Street, Ann Arbor, MI, 48109, United States of America, wmarrero@umich.edu, Mariel Lavieri, Jeremy B. Sussman, Greggory J. Schell, Rodney A. Hayward Markov decision process (MDP) models are powerful tools which enable the derivation of optimal treatment policies, but may incur long computational times and decision rules which are challenging to interpret by physicians. To reduce complexity and enhance interpretability, we study how Poisson regression may be used to approximate optimal hypertension treatment policies derived by a MDP for maximizing a patient’s expected discounted quality-adjusted life years. 3 - Optimal Decision Making in a Markov Model with Parameter Uncertainty: The Case of CKD M. Reza Skandari, University of British Columbia-Vancouver, Vancouver, BC, Canada, reza.skandari@sauder.ubc.ca, Steven Shechter, Nadia Zalunardo We investigate a Markov decision process whose unknown transition parameters are revealed partially through state observation. Decisions are made as the state evolves. We use the model to study the optimal time to start preparing a type of vascular access for chronic kidney disease patients who will need dialysis. 4 - Reinforcement Learning Algorithm for Blood Glucose Control in Diabetic Patients Mahsa Oroojeni Mohammad Javad, Northeastern University, 334 Snell Engineering, Northeastern Univ, Boston, United States of America, oroojeni.m@husky.neu.edu, Stephen Agboola, Kamal Jethwani, Ibrahim Zeid, Sagar Kamarthi In this paper a reinforcement learning algorithm is proposed for regulating the blood glucose level of Type I diabetic patients. In the proposed reinforcement learning algorithm body weight and A1C level define the state of a diabetic patient. For the agent, insulin dose levels constitute the actions. As a result of a patient’s treatment, after each time step t, the patient receives a numerical reward depending on the response of the patient’s health condition.

76