2016 INFORMS Annual Meeting Program

WB21

INFORMS Nashville – 2016

WB21 107A-MCC Emerging Methods for Healthcare Analytics and Visualization Sponsored: Health Applications Sponsored Session Chair: Rahul C Basole, Georgia Institute of Technology, 85 Fifth Street NW, Atlanta, GA, 30332, United States, basole@gatech.edu 1 - A Semi-supervised Learning Approach To Enhance Health Care Community-based Question Answering Papis Wongchaisuwat, Northwestern University, Evanston, IL, United States, papiswongchaisuwat2013@u.northwestern.edu, Diego Klabjan, Siddhartha R Jonnalagadda Community-based Question Answering (CQA) sites play an important role in addressing health information need. We developed an algorithm to automatically answer health-related questions based on past questions and answers. Our algorithm uses information retrieval techniques to identify candidate answers from resolved QA. In order to rank these candidates, we implemented a semi- supervised learning algorithm that extracts the best answer to a question. On our dataset, the semi-supervised learning algorithm has an accuracy of 86.2% while UMLS-based (health-related) features used in the model enhance the algorithm’s performance by approximately 8%. 2 - Visual Analytics For Population Level Health Analysis Rahul C Basole, Georgia Institute of Technology, 85 Fifth Street, Atlanta, GA, 30332, United States, basole@gatech.edu, Mark L. Braunstein, Hyunwoo Park, Dhruv Mutturaju, Myung Choi, Richard Starr We present the design, implementation, and use cases of a FHIR-centric population health analysis and visualization platform. 3 - Double Sided Network Externalities In Healthcare Information Exchanges Emre Muzaffer Demirezen, School of Management, Binghamton University, School of Management Binghamton University, State University of New York AA278, Binghamton, NY, 13902, United States, edemirezen@binghamton.edu, Subodha Kumar, Arun Sen Based on our interactions with different healthcare information exchange (HIE) providers, we develop models to study participation levels and sustainability of HIEs. We examine how heterogeneity among healthcare practitioners (HPs) affects participation of HPs in HIEs. We find that, under certain conditions, low- gain HPs choose not to join HIEs. Hence, we explore several measures that can encourage more participation in HIEs and find that it might be beneficial to: (i) establish a second HIE in the region, (ii) propose more value to the low-gain HPs, or (iii) offer or incentivize value-added services. We present several other interesting and useful results. WB22 107B-MCC Empirical Analysis of Resource Utilization Sponsored: Health Applications Sponsored Session Chair: David Anderson, CUNY Baruch, 55 Lexington Ave, New York, NY, 10010, United States, davidryberganderson@gmail.com 1 - Comprehensive Prediction Models For Colorectal Cancer Mortality David Anderson, CUNY Baruch, David.Anderson@Baruch.Cuny.Edu Having accurate, unbiased prognosis information can help patientsand providers make better decisions about what course of treatment to take.Using a comprehensive dataset of all colorectal cancer patients in Califoria, wegenerate predictive models that estimate short-term and medium-term survivalprobabilities for patients based on their clinical and demographic information.Our study addresses some of the contradictions in the literature about survival rates and signicantly improves predictive power over the performance of anymodel in previously published papers.

2 - Slow First, Fast Later: Temporal Speed-up In Service Episodes Of Finite Duration Aditya Jain, CUNY Baruch, Aditya.Jain@baruch.cuny.edu, Sarang Deo Many services comprise repeated episodes of finite duration wherein customers must be served before the end of that episode leading to non-stationary operational dynamics. We hypothesize and empirically validate (using data from a high volume tertiary care outpatient department) the presence of a ``slow first, fast later’’ work speed pattern in such environments. This pattern allows sufficient build-up of inventory earlier for more efficient utilization of faster work speed later. As a natural corollary of this pattern, we also find that greater anticipated workload, which causes faster inventory build-up, leads to a greater increase in work speed earlier during the service episode. 3 - The Impact Of Reminder Calls For A Pediatrics Practice Kiatikun Louis Luangkesorn, Assistant Professor, University of Pittsburgh, 1028 Benedum Hall, 3700 Ohara St, Pittsburgh, PA, 15261, United States, lluangkesorn@pitt.edu, Tricia Pil Primary care practices often use reminder phone calls to reduce missed appointments. The same factors that make reminder calls useful can also be used for improving patient engagement in the form of well child visits and vaccinations. However, studying the impact in a clinical setting can be difficult because it may not be practical or ethical to conduct random control trials on patients. We present two studies, one in the context of a series of interventions to inprove human papillomavirus (HPV) vaccination and to increase the fraction of patients who meet the recommendation of annual well child visits within a multi- practice pediatrics network. 108-MCC Optimal Treatment & Screening of Chronic Care Patients Sponsored: Health Applications Sponsored Session Chair: Huaiyang Zhong, Stanford University, 475 Via Ortega, Stanford, CA, 94305, United States, hzhong34@stanford.edu 1 - Optimal Statin Therapy Plan For Diabetic Patients Saeideh Mirghorbani, University of Alabama, Tuscaloosa, AL, United States, smirghorbani@crimson.ua.edu The importance of cardiovascular risks in diabetic patients has been emphasized because of their high cardiovascular mortality rate. Statins are a class of medicines used to lower blood cholesterol levels and mitigate the risk of heart problems. In this research, we address the optimal time to initiate and terminate statin therapy considering patient adherence as well as the influence of statin side-effects. We develop a finite horizon, discounted Markov decision process in which patients transition through health states. The objective is to maximize the expected quality-adjusted life years. 2 - Fairness In Down Syndrome Screening Jia Yan, Georgia Institute of Technology, Atlanta, GA, United States, jyan40@gatech.edu, Turgay Ayer, Pinar Keskinocak Detection and false positive rates of prenatal screening for Down syndrome depend on the selected risk cutoff values. In current practice, one-size-fit-all type cut off values are being used, which may lead to suboptimal outcomes and unfairness among different age groups. We first propose a Monte Carlo simulation model to capture prevalence and test outcomes in the population. Then, we combine this simulation model with an optimization modeling framework to identify the optimal age-specific risk cutoff values by taking fairness among different age groups into account. Our findings indicate that age-specific cutoff values significantly improve health outcomes and fairness. 3 - Improving The HIV Care Cascade Via Mental Health Interventions Huaiyang Zhong, Stanford University, hzhong34@stanford.edu The UNAIDS’ 90-90-90 targets having 90% of HIV-infected people aware of their status, 90% of diagnosed HIV-positives on antiretroviral therapy (ART), and 90% of those on ART virally suppressed. In most sub-Saharan African countries, these percentages are far lower. Developing effective and cost-effective approaches to improve the HIV care cascade is critical. We focus on one potential opportunity for improving the HIV care cascade: the provision of antidepressant therapy (ADT) for HIV-infected individuals with depression. WB23

402