Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SA61

2 - Cost-effectiveness Analysis of Clinical Management Strategies for Undifferentiated Febrile Illness in the Era of Responsible Antibiotic Use Zhenhuan Zhang, University of Minnesota, Minneapolis, MN, 55414, United States, Diana Maria Negoescu, Claudia Munoz-Zanzi Febrile illnesses such as dengue, leptospirosis and scrub typhus have similar symptoms and are often difficult to differentiate without diagnostic tests. If not treated appropriately, patients could experience serious complications. The question of what diagnostic test to use and when to administer antibiotic treatment to avoid misuse of scarce resources and ensure best possible health outcomes remains an open problem. We construct a Markov model of febrile illness progression to assess the cost-effectiveness of fifteen clinical management strategies for diagnosing and treating acute undifferentiated febrile illness in Medical research has found the progression of chronic diseases to transition between different clinical phases (e.g. acute and stable). These trajectory phases are relevant for clinical practice, since they serve as the basis for providing care and nursing. Yet their correct identification is challenging, as symptoms are only stochastically related to them. We formalize a hidden Markov model with latent states matching the trajectory phases as defined by the Corbin-Strauss trajectory framework. A copula approach is implemented to handle multivariate observations (e.g. pain and disability). 4 - A Robust Approach to Study Multiple Treatments: Hierarchical Contrast-specific Propensity Score Shasha Han, NUS Business School, National University of Singapore, NUS Business School, Biz 2 Building B1, 1 Business Link, Singapore, Singapore, Joel Goh, Fanwen Meng, Donald Rubin The worldwide rise in number of patients with diabetes and the consequent secondary complications is afflicting human population globally. To study the multiple medications treatment effect for diabetes, we propose a hierarchical contrast-specific propensity score(CSPS) approach. One merit of the approach is that it is robust to misspecification of the functional form of CSPS. The results from diabetes data in Singapore corroborates our theoretical findings. Due to such robustness, the hierarchical CSPS could be influential in causal inference for multiple treatments. Also, the approach charts one of the paths towards personalized medication in healthcare. n SA61 West Bldg 102C Information Provision in Selling Platforms, Health Care and Public Sectors Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Shouqiang Wang, The University of Texas at Dallas, Richardson, TX, 75080, United States 1 - Conspicuous by Its Absence: Diagnostic Expert Testing under Uncertainty Tinglong Dai, Johns Hopkins University, 100 International Dr, Baltimore, MD, 21202, United States, Shubhranshu Singh In many professional services, diagnostic experts may not be able to immediately reach correct diagnoses for their clients’ conditions, and often resort to diagnostic testing with cost implications (e.g., money, time, discomfort, or side effect). In this paper, we formulate a diagnostic expert’s pathway-selection problem when the peers observe these decisions and form beliefs about the expert’s skill level. We find that the high-type expert’s optimal diagnostic pathway may entail not performing the test even when it generates a positive surplus. We also show under-testing is the unique pattern allowing the high-type expert to credibly signal her type. 2 - Information Design of Time-locked Sales Campaigns for Online Platforms Can Kucukgul, The University of Texas at Dallas, Richardson, TX, 75080, United States, Ozalp Ozer, Shouqiang Wang Many online retailing platforms nowadays offer time-locked sales campaigns as an innovative selling mechanism, whereby third-party sellers sell their products at a typically discounted price for a fixed time horizon of pre-specified length. To incentivize purchases, platforms provide some information on up-to-date sales as campaigns progress, in the hope of influencing an upcoming consumer’s valuation of products. We propose an easy-to-implement and near optimal information provision strategy for such a time-locked sales campaign. Our policy has important managerial implications, and yields significant profit improvement upon some na ve policies currently implemented in practice. Thailand under different attitudes towards antibiotic misuse. 3 - Data Driven Personalized Treatment Planning for Chronic Diseases Christof Naumzik, ETH Zurich, Zurich, Switzerland, Stefan Feuerriegel

3 - Fast and Slow Learning from Reviews Ali Makhdoumi, MIT, 77 Massachusetts Ave., 32D-640, Cambridge, MA, 02139, United States We study the design of rating systems in online platforms. In particular, we consider a model of Bayesian learning from online reviews and characterize the speed of asymptotic learning of the quality of a product. We then characterize the impact of information provision on the speed of learning and identify situations in which providing more information leads to slower learning. 4 - Warning Against Recurring Risks: An Information Design Approach Shouqiang Wang, The University of Texas at Dallas, Naveen Jindal School of Management, 800 W. Campbell Rd, Richardson, TX, 75080, United States, Saed Alizamir, Francis E. De Vericourt Organizations or government agencies typically emit warning messages to alert relevant stake-holders about potential disastrous events of repetitive nature. Nonetheless, when the adverse event does not materialize, these false alarms affect the agencies’ credibility and hence their ability to mobilize timely responses to future threats. Thus, when sounding an alarm, the agencies need to weigh their current ability to elicit immediate actions against the efficacy of their future warning messages. Our research aims to elucidate how to resolve such trade-offs. n SA62 West Bldg 103A Data Mining Applied Paper Finals Sponsored: Data Mining Sponsored Session Chair: Tong Wang, University of Iowa, Iowa City, IA, 52245, United States Co-Chair: Ramin Moghaddass, University of Miami, Miami, FL. 1 - Data Mining Applied Paper Finals n SA63 West Bldg 103B Data Analytic to Facilitate Better Decision Making Sponsored: Data Mining Sponsored Session Chair: Roy Jafari Marandi, Cal Poly, 1 Grand Avenue, San Luis Obispo, CA, 93407, United States, 1 - Decision-driven Data Analytics: Case Study of Breast Cancer Diagnosis Roy Jafari Marandi, Mississippi State University, 260 McCain Engineering Buildin, Mississippi, Mississippi State, MS, 39762, United States This talk focuses on presenting the development and adaptation of data analytic tools that are created or adapted to improve decision-making. The emphasis of this talk is to highlight the efforts and existing research gaps that pertain to improving data analytic techniques to better serve decision-making goals. Breast cancer diagnosis, which is an essential medical decision, will be used as an example and case study to further clarify the presented concepts, techniques, and contributions. 2 - Predicting Personnel Fraud Ekrem Duman, Ozyegin University, Istanbul, Turkey Predictive analytics is an important tool to drive useful business results from data. However, for successful predictive models one should have enough number of examples for the classes to be predicted. When the number of examples is small, building strong predictive models becomes a very challenging task. In this study we pick up one such problem: predicting the bank personnel which might commit fraud. In order to have a strong enough predictive model, we decided to combine the powers of descriptive and predictive modeling techniques where we developed several descriptive models and used them as input of predictive model at the last stage. The results show that our solution approach perform quite well. Tong Wang, University of Iowa, Iowa City, IA, USA., Five finalists of 2018 Data Mining Best Paper Awards (Applied Track) will present their work in this session.

24