Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MC58

2 - Deep-learning-based Approach for Precise Health Cost Prediction Weiguang Wang, University of Maryland, 3330 B. Van Munching

n MC55 North Bldg 232C Navigating NSF: Funding Opportunities at NSF Emerging Topic: NSF Emerging Topic Session Chair: Georgia-Ann Klutke, National Science Foundation, Arlington, VA, 22230, United States 1 - Navigating NSF: Funding Opportunities at NSF Georgia-Ann Klutke, National Science Foundation, National Science Foundation, Arlington, VA, 22230, United States The National Science Foundation (NSF) offers a number of funding opportunities for investigators working in the fields of industrial engineering and operation research, both within the disciplinary programs in Engineering and other directorates, and through cross-cutting initiatives that are foundation-wide. This presentation will describe opportunities that are relevant to the Industrial and Operations Engineering communities, with particular emphasis on the Operations Engineering program in the Division of Civil, Mechanical & Manufacturing Innovation. The talk will also briefly describe guidelines for proposal preparation and NSF’s Intellectual Merit and Broader Impacts criteria. Question-and-answer session will follow the presentation. n MC56 West Bldg 101A HAS Distinguished Scholar Lecture Margaret L. Brandeau Sponsored: Health Applications Sponsored Session Chair: Ebru Korular Bish, Virginia Tech, Blacksburg, VA, 24060, United States Co-Chair: Tinglong Dai, Johns Hopkins University, Baltimore, MD, 21202, United States 1 - What Should We do About the Opioid Epidemic? Models to Support Good Decisions Margaret L. Brandeau, Stanford University, Management Science and Engineering, 475 Via Ortega, Stanford, CA, 94305-4026, United States The US is currently experiencing an epidemic of drug abuse caused by prescription opioids and illegal opioid use, including heroin. In addition to crime and social problems, rising levels of drug abuse have led to a sharp increase in overdose deaths as well as significant outbreaks of infectious diseases such as HIV and hepatitis C. How should we deploy limited public health resources to help solve this complex public health problem? This talk describes models used to support decision making regarding the control of drug abuse — and associated diseases such as HIV and hepatitis C — in the US. We conclude with discussion of key areas for further research. n MC57 West Bldg 101B Joint Session HAS/AI: Deep Learning in Healthcare Sponsored: Health Applications Sponsored Session Chair: Gordon Gao, MD, United States 1 - A Deep Learning Architecture for Psychometric Natural Language Processing Jingjing Li, PhD, University of Virginia, Charlottesville, VA, United States, Ahmed Abbasi, Faizan Ahmad, Hsinchun Chen We propose a novel deep learning architecture -PyNDA- to extract psychometric dimensions from user-generated texts in a timely and unobtrusive manner. PyNDA contains a representation embedding, a demographic embedding, a structural equation model encoder, and a multitask learning mechanism designed to address the unique challenges associated with extracting sophisticated and user-centric psychometric dimensions. Our experiments on three real-world datasets encompassing eleven psychometric dimensions, including trust, anxiety, and literacy, show that PyNDA markedly outperforms traditional feature-based classifiers as well as the state-of-the-art deep learning architectures.

Hall, College Park, MD, 20742, United States, Margret V. Bjarnadóttir, Guodong (Gordon) Gao

Healthcare cost predictions have multiple applications, however precisely predicting healthcare costs at the individual patient level remains a challenge in the machine learning domain. In this study, long short-term memory based recurrent neural network is proposed to incorporate the sequential healthcare cost information for more accurate healthcare cost predictions. We compare the performance of our deep learning approach with multiple traditional machine learning methods. Among all the methods, deep learning shows the best performance. Finally, the performance of deep learning is explained using subgroup analyses and pattern extraction. 3 - Automated Depression Detection Using Multimedia Data Guohou Shan, University of Maryland-Baltimore County, baltimore, MD, 21250, United States, Lina Zhou, Dongsong Zhang Depression is a major mental health problem. Automated depression detection has been explored with social media content; however, it has been rarely studied using multimedia data such as audio and video. This research aims to improve automatic depression detection by incorporating new predictive features and by identifying fine-grained indicative content. Incentives in Healthcare Sponsored: Health Applications Sponsored Session Chair: Burhaneddin Sandikci, University of Chicago, Chicago, IL, 60637, United States 1 - Expanding the Donor Pool: The Use of Marginal Organs for Transplantation Sait Tunc, University of Chicago, McGiffert House 4th Floor, 5751 S. Woodlawn Avenue, Chicago, IL, 60637, United States, Burhaneddin Sandikci, Bekir Tanriover Transplantation emerged as the standard treatment for organ failure. Despite the growing shortage for donor organs as demand outpaces supply, a significant fraction of organs harvested for transplantation are rejected and discarded. As a viable option to alleviate the burden of shortages, we study the use of marginal organs for transplantation through a queueing-theoretic framework. We investigate socially efficient and equilibrium utilizations of donor organs, and discuss incentive mechanisms that help increase the utilization of marginal organs while also improving overall social welfare. 2 - Contracts to Improve Vaccine Effectiveness and Availability Taylor Corcoran, UCLA Anderson School of Management, 110 Westwood Plaza, Los Angeles, CA, 90024, United States, Fernanda Bravo, Elisa Long Two key players in vaccine markets are pharmaceutical companies, who develop and manufacture vaccines, and global health organizations (GHOs), who procure vaccines for low- and middle-income countries. The payment that pharmaceutical companies receive from GHOs depends on the quantity of vaccines produced, but is independent of their effectiveness, which may lead to the development of low quality vaccines that make it more difficult for GHOs to achieve their public health targets. To address this inefficiency, we use a game-theoretic model that incorporates disease dynamics to analyze performance-based contracts which link vaccine price and effectiveness. 3 - Data Driven Incentive Design in the Medicare Shared Savings Program Auyon Siddiq, UCLA Anderson School of Management, Los Angeles, CA, United States, Anil Aswani, Max Shen The Medicare Shared Savings Program (MSSP) was created to control escalating Medicare spending by incentivizing providers to deliver healthcare more efficiently. We formulate the MSSP as a principal-agent model and propose a new type of contract that includes a performance-based subsidy for the provider’s investment. We show that the proposed contract dominates the existing MSSP contract by producing a strictly higher expected payoff for both Medicare and the provider. We then present an estimator based on inverse optimization for estimating the parameters of the principal-agent model and the potential increase in savings under the proposed contract. n MC58 West Bldg 101C

205