2016 INFORMS Annual Meeting Program

MC04

INFORMS Nashville – 2016

MC02 101B-MCC Data Mining Innovations in Healthcare Sponsored: Data Mining Sponsored Session

models and decision tools to find effective appointment scheduling strategies with the objective of maximizing the likelihood that a majority of patients can be seen within a time threshold. 3 - Managing Access To Primary Care Through Advanced Scheduling

Sina Faridimehr, PhD Student, Wayne State University, 4815 Fourth Street, Detroit, MI, 48202, United States, sina.faridimehr@wayne.edu, Ratna Babu Chinnam, Saravanan Venkatachalam

Chair: Michael Lash, University of Iowa, S210 John Pappajohn Business Building, Iowa City, IA, 52242-1000, United States, michael-lash@uiowa.edu 1 - Multi-site Evidence-based Best Practice Discovery – Finding Factors That Influence Treatment Outcome Eva Lee, Georgia Tech, evakylee@isye.gatech.edu Cody Wang This work is joint with Care Coordination Institute. The work focuses on establishing inter-operability among electronic-medical-records(EMRs) from 737 providers for large-scale data mining to identify discriminatory characteristics that can predict the quality of treatment outcome. We demonstrate the system usability by analyzing Type II diabetic patients. DAMIP establishes a classification rule on a training set that results in greater than 80% predictive accuracy on a blind set of patients. This facilitates evidence-based treatment and optimization of site performance through best practice dissemination and knowledge transfer. 2 - Scalable Support Vector Machines For Massive Healthcare Datasets Talayeh Razzaghi, Clemson University, trazzag@g.clemson.edu Solving the optimization model of support vector machines is often an expensive computational task for massive healthcare training sets. We propose an efficient, effective, multilevel algorithmic framework that scales to very large data sets. Our multilevel framework substantially improves the computational time without loosing the quality of classifiers for balanced and imbalanced datasets. 3 - Exploring Feasibility To Early Detect Alzheimer’s Disease (AD) And Dementia Progression Using Big Data + Machine Learning Approach Chih-Lin Chi, University of Minnesota, Minneapolis, MN, United States, cchi@umn.edu, Wenjun Zeng, Wonsuk Oh, Soo Borson We aim to develop a 3-step data strategy to develop a model to predict long-term cognitive changes for AD. We show results from the first step: exploring feasibility of long-term prediction by optimizing time-varying risk factors. The developed model predicts cognitive scores and annual changes in up to 8 years when most subjects were cognitive normal or mild cognitive impairment. This model demonstrates accurate prediction of how dementia progress for cognitive stable, mild deterioration, moderate deterioration, and sharp deterioration subgroups. The presentation will also discuss the next steps that aim at converting the data- research results into informatics tools. MC03 101C-MCC Health Care, Modeling I Contributed Session Chair: Sina Faridimehr, PhD Student, Wayne State University, 4815 Fourth Street, Detroit, MI, 48202, United States, sina.faridimehr@wayne.edu 1 - Measure And Predict Medication Adherence Behavior Using Administrative Data Shan Xie, Purdue University, 315 N Grant Street, W Lafayette, IN, 47907, United States, xie34@purdue.edu, Yuehwern Yih For patients with diabetes, poor adherence to medication has been associated with suboptimal glycemic control, increased health care costs and adverse health outcomes. Thus, improving medication adherence is important to realize the full benefit of medication therapies. The existing measures of medication adherence based on administrative data only provide an aggregate number, which lack the ability to distinguish between different adherence behaviors. This study will develop an analytic framework to quantify and predict medication adherence patterns, and provide useful information to efficiently target patients at high risk

In this research, we develop a scheduling framework that employs stochastic programming to improve access to care within primary care clinics. The model leverages correlations between scheduling practice, continuity of care, appointment utilization and access performance. Results from testing the models at VA facilities are promising. MC04 101D-MCC Optimal Procurement, Tariff, and Cybersecurity in Smart Grid Sponsored: Energy, Natural Res & the Environment, Energy I Electricity Sponsored Session Chair: Lawrence V Snyder, Lehigh University, 200 West Packer Avenue, Bethlehem, PA, 18015, United States, lvs2@lehigh.edu 1 - Optimal Day-ahead Power Procurement With Renewable Energy, Storage, And Demand Response Soongeol Kwon, Texas A&M University, College Station, TX, United States, soongeol@tamu.edu, Natarajan Gautam, Lewis Ntaimo Motivation of this research stems from pressing issues related to reducing energy cost, specifically focused on demand-side. From energy consumers perspective, there exist opportunities to reduce energy cost by adjusting purchase and consumption of energy in responding to time-varying electricity prices while utilizing renewable energy with energy storage. Considering this scenario, the main research objective is to develop a decision model to determine optimal day- ahead purchase commitment while considering real-time adjustments in response to variability and uncertainty in actual power demand, renewable supply, and electricity price. 2 - A Game Theoretic Analysis Of Electricity Time-of-use (TOU) Tariff For Residential Customers Dong Gu Choi, Pohang University of Science and Technology, Pohang, Korea, Republic of, dgchoi@postech.ac.kr, Valerie Thomas We properly formulate a game-theoretic model for analyzing not only the optimal behaviors of both an electric utility and residential customers but also their monetary gains or losses under a TOU tariff. With two heterogeneous customer types in terms of consumption pattern, we identify that a win-win situation is not possible. Also, we emphasize our analytic results by describing a numerical example, and we discuss the implications of our results for electric utilities and regulatory agencies. 3 - Risk Assessment And Network Optimization For Smart Grid Cybersecurity Mesh communication networks are widely used to facilitate communication to and between smart grid sensors such as advanced meters and demand response devices. Despite this wide use, a comprehensive risk assessment of cyber attacks on distributed networks has not been fully explored. In this work, we propose a framework for connectivity analysis of smart grid sensor networks to ensure robust communication when a given number of communication nodes are compromised. We also propose an efficient algorithm to construct a graph to meet given connectivity criteria by augmentation of communication links or strong authentication on certain nodes. 4 - Electricity Market Clearing With Enhanced Dispatch Of Wind Producers: Market Design And Environmental Implications Ali Daraeepour, Duke University, Durham, NC, 27708, United States, a.daraeepour@duke.edu, Dalia Patino-Echeverri This study explores the market design, operational, and environmental effects of the stochastic electricity market clearing. We propose a framework that allows a robust assessment of the relative advantages of the stochastic market clearing with respect to the conventional deterministic mechanism under wind production uncertainty. Using a stylized version of PJM, the two mechanisms are compared in terms of air emissions, wind integration, prices and supply-side revenue adequacy, and out-of-market adjustments. Lawrence V Snyder, Lehigh University, lvs2@lehigh.edu, Jiyun Yao, Parv Venkitasubramaniam, Shalinee Kishore, Rick Blum

and customize adherence improvement interventions. 2 - Improving Access To Healthcare By Minimizing Appointment Delays

Ashley N. Anhalt, PhD Student, University of Pittsburgh, 525 S. Aiken Avenue, APT #3, Pittsburgh, PA, 15232, United States, ana88@pitt.edu, Jeffrey P. Kharoufeh Providing patients with timely access to healthcare is an important issue for major healthcare providers. One major problem is that patients are often unable to schedule appointments with specialists in a reasonable time. We present queueing

183