Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MB61

3 - Bundled Payments, Fee-for-service, and Competition: Implications on Quality and System Performance Zheng Han, University of Kansas, Lawrence, KS, 66049, United States, Mazhar Arikan, Suman Mallik We study the quality competition between two hospitals where one under the fee-for-service (FFS), while the other under the bundled payment (BP). The demand, the costs, and the probability of successfully treating a patient depends on the hospital’s chosen quality. Under such a setting, we develop a game theoretic model to answer the following questions. Is BP (FFS) payment scheme always associated with high (low) equilibrium quality? What factors affect the equilibrium outcomes and how? What insights can a policymaker (i.e., an insurer) obtain from the equilibrium quality outcomes? 4 - Scheduling Smarter and Working Harder: The Key to Reducing Turnover Kevin Mayo, Indiana University, 1275 E. 10th St, Bloomington, IN, 47405, United States, Eric Webb, Kurt M. Bretthauer, George Ball Turnover rates among nurse aides in skilled nursing facilities is extremely high and likely to get worse as demand outpaces supply. Such turnover rates have significant negative effects on patient health outcomes and various costs associated with turnover. We examine the scheduling characteristics of 6,634 part time nurse aides and 5,305 turnovers to determine how scheduling policies affect turnover. We identify the impacts of the amount, variation, and type of scheduling that all significantly influence the likelihood of turnover, and the differential effects in high or low workload environments. Using these insights, managers can better schedule their nurse aides to reduce turnover. n MB61 West Bldg 102C Stochastic Models in Healthcare Sponsored: Health Applications Sponsored Session Chair: Oguzhan Alagoz, University of Wisconsin-Madison, Madison, WI, 53706, United States Co-Chair: Ali Hjaar, Wisconsin-Madison, Wisconsin-Madison, WI, United States 1 - Optimizing Hospital Resources to Improve Care Delivery -An Application to Bed Capacity Eva Lee, Georgia Tech, Industrial & Systems Engineering, Ctr for Operations Research in Medicine, Atlanta, GA, 30332-0205, United States We consider the problem of partitioning clinical services in hospitals into groups with the goal of efficiently allocating existing inpatient beds. We derive a 2-stage approach stochastic approach to address the 3-fold problem: 1) how many groups of services to form; 2) how many beds to allocate to each group; and 3) how to partition services among the groups. Three full-scale examples willl be presented to demonstrate the flexibility and diverse application of our framework with managerial insights for different utility optimization goals and queueing systems. 2 - Optimizing Breast Cancer Screening using Partially Observable Markov Decision Processes Ali Hajjar, University of Wisconsin-Madison, 1513 University Avenue, 3233 Mechanical Engineering Building, Madison, WI, 53706, United States, Oguzhan Alagoz Breast cancer, the leading cause of cancer death for women, can be detected at earlier stages through mammography screening. Therefore, We formulate a finite- horizon, partially observable Markov decision process (POMDP) model for this problem. 3 - Optimal Defibrillator Deployment versus Actual Deployment Timothy Chan, University of Toronto, Mechanical and Industrial Engineering, 5 Kings College Road, Toronto, ON, M5S 3G8, Canada, Christopher Sun Public defibrillators, which are located throughout cities worldwide, can be used to resuscitate cardiac arrest victims by bystanders with no training. However, location decisions to date are not data-driven. In this talk, we present a head-to- head comparison of optimal defibrillator locations against actual defibrillator locations using nine years of real data. At every decision epoch, the optimization model can only make decisions based on past data. On out-of-sample, future cardiac arrests, the optimization model improves spatiotemporal coverage, a measure of spatial proximity to cardiac arrests and temporal accessibility of the defibrillator, by 50-100% compared to actual location decisions.

4 - A Data-driven Stochastic Programming Approach to the Outpatient Colonoscopy Scheduling Problem Karmel Shehadeh, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48105, United States, Amy Cohn We present a data-driven stochastic programming approach to optimize scheduling templates for stochastic colonoscopy procedures. Particular attention is paid to the underlying impact of pre-colonoscopy bowel preparation on variability in colonoscopy duration and the competing schedule metrics, including patient delays, clinic overtime, and colonoscopy procedure outcomes. A case study based on an outpatient procedure center (OPC) at a large medical center is used to draw some useful managerial insights for OPC managers. n MB62 West Bldg 103A Decision Making and Data Mining Sponsored: Data Mining Sponsored Session Chair: Michael Lash, University of Iowa, Iowa City, IA 1 - Hybrid Decision Making: When Interpretable Models Collaborate with Black-box Models Interpretable machine learning has received increasing interest especially in domains where humans are involved in the decision-making process. However, the possible loss of the task performance for gaining interpretability is often inevitable. We propose a novel framework for building a Hybrid Decision Model that integrates an interpretable model with any black-box model to make better decisions. We design a principled objective function that considers predictive accuracy, model interpretability, and data explainability. Experiments show that hybrid models do not necessarily trade accuracy for explainability and provide higher flexibility in model designing. 2 - ELM-SOM: A Continuous Self-Organizing Map for Visualization Renjie Hu, University of Iowa, 1505 W. Benton Street, Iowa City, IA, 52246, United States, Venous Roshdibenam, Hans J. Johnson, Emil Eirola, Anton Akusok, Yoan Miche, Kaj-Mikael Bj÷rk, Amaury Lendasse This paper presents a novel dimensionality reduction technique: ELM-SOM. This technique preserves the intrinsic quality of Self-Organizing Maps (SOM): it is nonlinear and suitable for big data. It also brings continuity to the projection using two Extreme Learning Machine (ELM) models, the first one to perform the dimensionality reduction and the second one to perform the reconstruction. ELM-SOM is tested successfully on six diverse datasets. Regarding reconstruction error, ELM-SOM is comparable to SOM while bringing continuity. 3 - Text Mining of Online Reviews using Deep Learning Techniques Asil Oztekin, University of Massachusetts Lowell, 333 1st Street, Unit 210, Lowell, MA, 01850-2580, United States In this study, we used deep learning techniques to analyze user-generated content, particularly online text reviews of travelers who describe their experience of airports and suggest a recommendation for other travelers. By performing aspect-oriented sentiment analysis on the user reviews, we develop a holistic predictive approach. The study reveals that aspect-oriented sentiment scores significantly improve the predictive power for the recommendation decision of the user. Recurrent Neural Network based deep learning approach performed superior to other machine learning models in predicting customer recommendations. This research has methodological, application-based, and managerial implications. Tong Wang, University of Iowa, Pappajohn Business Build, 21 East Market Street, Iowa City, IA, 52245, United States

170