INFORMS 2021 Program Book

INFORMS Anaheim 2021

TD19

2 - Processing Large Scale DEA: The State-of-the-art Jose H. Dula, University of Alabama, School of Business,, Tuscaloosa, AL, 35487, United States, Dimitris K. Despotis, Gregory Koronakos Dula’s 2012 algorithm, BuildHull, was the fastest way to process DEA when it came out. Since then there has been an interest in analyzing and testing the algorithm’s performance. We report on a study comparing BuildHull with a recent competing approach for DEA using a well-structured data suite which includes massive data sets and allows insights into the effects of dimensionality, cardinality, and extreme-efficient DMU density. 3 - Dynamic L1 Regression Botan Citil, University of Alabama, Tuscaloosa, AL, United States, Jose Dula The objective of this project is to apply L1 regression to streaming data. L1 regression is considered to be more robust than ordinary least squares and is indicated when the data contain outliers such demand spikes, etc. We report on results that enhance and accelerate resolving the special LP formulation for this problem. We present numerical results from our tests. TD17 CC Room 202A In Person: Decision Analysis General Session Chair: Tim Marler, PhD, RAND Corporation, Santa Monica, CA, United States 1 - Presenter Jack Soll, Duke University, Fuqua School Of Bus., Durham, NC, 27708-9972, United States I will serve as a discussant for the papers presented in this session. 2 - Analysis of Cyber-attacks and Cost-effective Methods of Cybersecurity Einstina Wang, Oxford Academy, Cypress, CA, United States, Gabriel Han, Hannah Jang, Won Jang Increasing cybersecurity risks can cause data breaches that expose personal and sensitive information, damaging the reputation of targeted companies and hurting their clients. We test various preventative security methods, including employee education, firewalls, encryption, and software updates, with consideration of costs to determine practical solutions to mitigate the risks of cybersecurity data breaches for companies of various sizes. We determine that the most important action a company could take would be through behavioral changes such as securing passwords and using multi-factor authentication. 3 - The Role of Overconfidence on Personal Attitude Toward Covid-19 and Risk Mitigating Factors Dominik Piehlmaier, Assistant Professor, University of Sussex, Brighton, United Kingdom The experimental study sheds light on the impact of overconfidence on a person’s attitude toward COVID-19 as well as the likelihood of wearing face masks, getting vaccinated, utilizing contact tracing apps, and following mandatory quarantine rules by conducting a randomised controlled trial data from 600 UK panellists. Building on the theory of correlation neglect, we show that respondents who are overconfident in their knowledge about infectious diseases illustrate a laxer attitude toward the current outbreak. The study provides evidence to help inform public health officials to focus on a subpopulation who would benefit from a nudge to follow official COVID-19 guidance and regulations. 4 - A Game Theory Approach for Engineering Optimization and Decision-making Tim Marler, RAND Corporation Groups of decision-makers, interacting in a design process, can be modeled using game theory, which in turn can be solved as a multi-objective optimization problem. From this perspective, decision-makers rarely cooperate completely in a theoretical sense; rather, the exchange of information is iterative. Ultimately, this can result in a non-optimal solution or design. Given multiple decision-makers, each managing a separate objective function and controlling unique variables, this paper presents a new algorithm for modeling design process as a non-cooperative game theoretic scenario. This algorithm is then used in the context of a broader novel multi-objective optimization approach for resolving such non-cooperative situations, thus yielding a Pareto optimal solution. This approach provides not only a mathematical method for extending a Nash equilibrium point (non- cooperative solution) towards the Pareto optimal set, but also a means for modeling how decision-makers actually interact. This, in turn provides significant insight into engineering project management and decision-making. The proposed approach is demonstrated with two illustrative design problems.

TD19 CC Room 203A In Person: Information, Technology, and Analytics in Healthcare General Session Chair: Mehmet U.S. Ayvaci, The University of Texas at Dallas, Richardson, TX, 75080-3021, United States 1 - Does EMR Adoption by Nursing Homes Decrease Hospitalization Costs ? Atiye Cansu Erol, University of Pennsylvania, Philadelphia, PA, United States, Lorin Hitt, Prasanna Tambe Electronic Medical Records (EMR) have the potential to decrease medical expenditures by increasing communication between healthcare providers and by reducing unnecessary tests and medical errors. Using a three−year panel of Medicaid spending for long−term care patients in nursing homes, we find evidence that EMR adoption by nursing homes reduces hospitalization costs for residents by 3.5−14 percent. We also find a further reduction of hospitalization costs for residents when hospitals and nursing homes both adopt EMR: an average savings of 13 percent of Medicaid expenditure and as much as 35 percent for system−member hospitals. Given the interdependent nature of healthcare delivery for long−term care patients or patients with chronic conditions, our findings underscore the importance of looking outside the adopting institution when accounting for health IT value. 2 - Nursing Home Staff Networks and Covid-19 Elisa F. Long, UCLA Anderson School of Management, Gold Hall 110 Westwood Plz # B-508, Los Angeles, CA, 90024-5055, United States, Keith Chen, Judith Chevalier Skilled nursing homes (SNFs) accounted for a disproportionate share of COVID- 19 fatalities worldwide, with outbreaks persisting despite the March 2020 nationwide ban on visitors. Using device-level geolocation data for 50 million smartphones, we analyze SNF connections via shared staff and observe 500,000 individuals entering at least one SNF, with 5.1% entering two or more facilities. Nursing homes share connections with 7.1 other facilities, on average. Network measures of connectivity, including node degree, strength and Eigenvector centrality, are highly predictive of COVID-19 cases, whereas traditional regulatory quality metrics are unimportant in predicting outbreak size. 3 - A Multi-treatment Forest Approach for Analyzing the Heterogeneous Effects of Team Familiarity Minmin Zhang, The University of Texas at Dallas, Richardson, TX, 75252, United States, Wallace J. Hopp, Guihua Wang, Michael Mathis We study the heterogeneous effects of team familiarity on surgery duration. We develop a multi-treatment forest consisting of multiple tree models that divide patients into different subgroups based on their features and estimate the effects of familiarity within each subgroup. The results show that the effects of familiarity are different for different types of patients. Our results can help hospital administrators to improve operational efficiency by matching patients with surgery team members using patient-specific information. 4 - A Tool to Inform Global Hepatitis C Elimination in Developing Countries Huaiyang Zhong, Harvard Medical School, Massachusetts General Hospital, Boston, MA, United States, Madeline Adee, Lindsey Hiebert, John Ward, Jagpreet Chhatwal The World Health Organization (WHO) recently launched a global campaign for eliminating hepatitis C virus (HCV) as a public health threat by the year 2030. However, most countries do not have a national strategy for HCV screening and treatment that can lead to HCV elimination. We developed a microsimulation model to assess various combinations of screening and treatment strategies, and built an online, publicly accessible tool to help policy makers identify a path to HCV elimination.

131