Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SC64

2 - A Markov Decision Process to Identify Optimal Policies for Stopping a Trial of Labor Karen T. Hicklin, University of North Carolina at Chapel Hill, B-24 Hanes Hall, Chapel Hill, NC, 27599-3260, United States, Julie Simmons Ivy For first time moms the decision to have a cesarean delivery (C-section) can lead to future complications in subsequent pregnancies such as uterine rupture or repeat C-sections. In addition to the health risk associated with C-sections, there is general consensus that the C-section rate in the United States is too high and not associated with a decrease in maternal and neonatal morbidity or mortality. We model the mode of delivery decision using a Markov decision process (MDP). This MDP evaluates when a C-section is optimal as a function of total time spent in labor and the associated rate of complications. 3 - Analysis of Aggregate Data using Phase-type Distribution for Reliability Estimation Samira Karimi, University of Arkansas, Fayetteville, AR, United States, Haitao Liao One of the most accessible component reliability data in industry is aggregate lifetime data from fielded systems. While Gamma and Inverse Gaussian have been used in reliability estimation using aggregate data, Phase-type (PH) distribution has not been studied in the related literature. In this work, an expectation- maximization (EM) algorithm is proposed to estimate the parameters of PH distribution based on aggregate data. In addition, a Bayesian alternative is also studied and confidence region for model parameters are provided. A numerical study shows the strength of using PH distribution as an alternative for handling aggregate lifetime data. 4 - Modeling Patient Flow Information with Covariates using Phase-type Distributions Wanlu Gu, University of Arizona, Tucson, AZ, 85719, United States, Neng Fan The hospital length-of-stay (LOS) measures the time from admission to discharge. It demonstrates patient flow and plays an important role in health care quality improvement. In this paper, we fit the Coxian phase-type (PH) distributions to the patient flow information collected in Banner University Medical Center Tucson and assess the effects of covariates, including age, gender, admission type etc. The resulting estimated PH distributions can classify patients into different LOS groups and the pattern under each group is identified. The estimated coefficients and the statistical significance of covariate effects will help decision making in healthcare service and sources assignment. n SC64 West Bldg 104A Joint Session DM/Practice Curated: Data Science and Analytics in Health Care III Sponsored: Data Mining Sponsored Session Chair: Hamidreza Ahady Dolatsara, Auburn University, Auburn, AL 1 - Neural Fuzzy-based Unscented Kalman Filter Model for Atrial Fibrillation Onset Prediction Trung Le, North Dakota State University, Fargo, ND, 58102, United States Atrial fibrillation is the most common arrhythmia, which increases the risk of stroke by 5 times and potentially leads to embolism. Assessing the risk of developing PAF is important to avoid the risk of death; previous work has only focused on addressing the challenge for predicting the onset of paroxysmal atrial fibrillation (PAF) from the morphological-temporal features of surface electrocardiogram by utilizing machine learning-based techniques. In this paper, we propose a method based on a combination of Kalman filter algorithm and a Neural Fuzzy network to predict PAF onset for 70 patients suffering from PAF. 2 - Quantification of Stroke Risk in Patients with Atrial Fibrillation and Obstructive Sleep Apnea Rupesh Kumar Agrawal, Research Assistant, Oklahoma State University, Tulsa, OK, 74106, United States, Daniel Tran, Matt Wilkett, Dursun Delen, Bruce Benjamin Extant research work has established strong correlation in patients with Atrial Fibrillation (A-Fib), the Obstructive Sleep Apnea (OSA) and the increased risk of stroke. CHADS2 and CHA2DS2-VASc scores are used to calculate the risk of stroke based on circulatory and respiratory mechanism. This is a pilot study to quantify stroke risk in patients with A-Fib, OSA, to show that the stroke risk is higher in patients, independent of treatment with anticoagulants used to treat patients with A-Fib. Furthermore, our goal is also to develop human subject clinical trial, based on the knowledge gained from the analytics study to develop a score for OSA as part of the criteria for calculating a risk of stroke

3 - Quantifying the Motivation of Physicians’ Continuous Charitable Clinics Choice Online Han Yang, Beijing Institute of Technology, No.5 Zhongguancun South Street, Haidian Dist, Beijing, 100081, China, Zhijun Yan, Lun Li Online healthcare communities is a novel channel for physicians to share healthcare knowledge with patients.The services that physicians provide in OHCs consist of paid health services and charitable clinics services, and we have limited knowledge on the motivation of physicians’ continuous charitable clinics choice online. We develop a logistic model to examine three types of motivation of physicians’ continuous charitable clinics choice, including economic , social image and altruistic motivation. Based on the data from Guahao.com, we find that different types of motivation has different effects on physicians’ continuous charitable clinic behavior based on different situations. n SC65 West Bldg 104B Big Data, Economic Modeling, and Reliability Sponsored: Data Mining Sponsored Session Chair: Kezban Yagci Sokat, Northwestern University, IEMS, 2145 Sheridan Rd, Evanston, IL, 60208, United States 1 - Influencing Donation Choice in Social Networks after Disasters Trilce Encarnacion, Rensselaer Polytechnic Institute, 110 8th Street, Jonsson Engineering Center # 4049, Troy, NY, 12180, United States The goal of this research is to develop analytical methods to effectively influence donation choice in a social network setting. To do so, several objectives need to be accomplished: a) Develop a network influence model formulation to represent the social network interactions b) Incorporate behavior research regarding individuals attitudes and socio-economic characteristics in relationship to donation behavior; c) Develop a diffusion model to characterize influence propagation through the social network after a disaster occurs; and, d) Select an appropriate influencing strategy to effect change in donation choice. 2 - Predicting Major Stock Indices using Macroeconomic Indicators Lin Lu, Auburn University, Auburn, AL, 36830, United States, Waldyn Martinez, Bin Weng, Fadel Mounir Megahed This paper proposes a two-stage approach to investigate whether macroeconomics variables alone can be used to accurately predict the one-month ahead price for major U.S. stock and sector indices. Stage 1 evaluates the hypothesis that the price for different indices is driven by different economic indicators. Stage 2 uses a hybrid approach of the recurring neural network and the ensembles models to examine the secondary hypothesis that residuals from the time-series model are not random and can be explained by incorporating macroeconomic indicators through the ensemble approach. Results showed that the macroeconomic indicators are leading predictors of the price of 13 U.S. sector indices. 3 - Reliability Estimation for Balanced Systems with Multi-dimensional Distributed Units Elsayed A. Elsayed, Distinguished Professor, Rutgers University, Department of Industrial and Systems Eng, 96 Frelinghuysen Road, Piscataway, NJ, 08854, United States, Jingbo Guo Balanced systems with multi-dimensional distributed units are emerging in a diverse range of industries. For example, Unmanned Aerial Vehicle (UAV) with multi-layer of rotary wings and spherical UAVs are used in many applications due to their stability and flexibility. A minimum number of units is required to function properly in a balanced arrangement for the system to operate properly. Reliability estimation for such complex systems is studied under different conditions. 4 - A Note on Pricing Model of ESG Investing Hiroshi Ishijima, Chuo University, 42-8 Ichigaya-honmura-cho, Shinjuku, Tokyo, 1628473, Japan, Akira Maeda Modern investors have been paying more attention to ESG investing than ever which is defined as the consideration of environmental, social and governance factors as well as the conventional finance factor regarded as the discounted sum of future cash flows. Along these lines, we would like to develop a rational pricing model of ESG investing. We then conduct a preliminary empirical analysis in the Japanese stock market to show whether ESG investing is profitable enough to compensate for fiduciary duties as stated in the Principles for Responsible Investment. 5 - Modelling Human Trafficking Kezban Yagci Sokat, Northwestern University, IEMS, 2145 Sheridan Rd, Evanston, IL, 60208, United States, Nezih Altay Human trafficking has become a serious concern for society and the global economy. While there has been a lot of attention in this topic in the social contexts, there is little in the humanitarian operations community. We develop a model for human trafficking and study the impact of relevant law.

84