2015 Informs Annual Meeting

SD38

INFORMS Philadelphia – 2015

SD37 37-Room 414, Marriott Health Care Modeling and Optimization IV Contributed Session Chair: Xiang Zhong, University of Wisconsin, 5019 Old Middleton Rd, Madison, WI, 53706, United States of America, oliver040525@gmail.com 1 - A Predictive Readmissions Model for Coronary Bypass Artery Grafting Patients Jingyun Li, California State University Stanislaus, 7740 McCallum Short-term hospital readmissions due to CABG surgery is a burgeoning problem. Drawing on archival data of CABG patients from 27 hospitals in North Texas, during a three-year period, our model predicts the 30-day readmission propensity of CABG patients, as well as their frequency, and time to readmission. 2 - Innovation in Healthcare Management using Data-driven Clinical Pathways Yiye Zhang, Carnegie Mellon University, 4800 Forbes Ave, Pittsburgh, PA, 15213, United States of America, yiyez@andrew.cmu.edu, Rema Padman This paper investigates how service innovations in the management of healthcare delivery can be facilitated through the development of data-driven clinical pathways. We propose a clinical pathway learning algorithm that models the association between treatments and patients health conditions as a hidden Markov model, and also makes predictions for patients’ future states. We customize clinical pathways by patient and treatment types using hierarchical clustering and frequent sequence mining. 3 - Analysis of the Impact of Electronic Visits on Patient Care Delivery Xiang Zhong, University of Wisconsin, 5019 Old Middleton Rd, Madison, WI, 53706, United States of America, oliver040525@gmail.com, Jingshan Li, Philip Bain, Albert Musa To improve care access, many healthcare organizations have introduced electronic visits to provide patient-physician communication. In this study, we introduce an analytical model to study the care delivery with e-visits. Analytical formulas to evaluate the mean and variance of patient length of stay during access to care are derived. The impact of e-visits on patient access to other care delivery venues is investigated. Scheduling and control policies to improve care access are discussed. SD38 38-Room 415, Marriott Big Data III Contributed Session Chair: Munir Majdalawieh, Associate Professor, Zayed University, Academic City, Dubai, United Arab Emirates, munir.majdalawieh@zu.ac.ae 1 - A Simulation-Optimization Method for Quantitative Aggregation of Prior Statistical Findings Mohammad Jalali, Massachusetts Institute of Technology, Cambridge, MA, 02141, United States of America, jalali@mit.edu We introduce a simulation-optimization method for quantitative aggregation of prior statistical findings. The method uses only available statistical results from prior studies to estimate a meta-model that is consistent with those original findings. As an empirical demonstration, we aggregate prior studies of the determinants of basal metabolic rate. Our model proves more accurate than existing models in the literature and the models by World Health Organization and Institute of Medicine. 2 - Guaranteed Matrix Completion via Non-convex Factorization Ruoyu Sun, Stanford University, Menlo Park, CA, 94025, United States of America, sundirac@gmail.com, Zhi-Quan Luo Matrix factorization is very popular for large-scale matrix completion. However, due to the non-convexity, there is a limited theoretical understanding of this approach. We show that under similar conditions to those in previous works, many standard optimization algorithms converge to the global optima of the non- convex factorization based formulation, thus recovering the true low-rank matrix.Our result is the first one that provides exact recovery guarantee for many standard algorithms. Blvd., Dallas, TX, 75252, United States of America, jli9@csustan.edu, Steves Ring, Indranil Bardhan

4 - Optimal Post-donation Blood Screening under Prevalence Rate Uncertainty Hadi El-amine, PhD Student, Virginia Tech, 250 Durham Hall Perry St., Blacksburg, VA, 24061-0118, United States of America, hadi@vt.edu, Ebru Bish, Douglas Bish Blood product safety, in terms of being free of transfusion-transmittable infections, is crucial. Under prevalence rate uncertainty, various objective functions, including minimization of a mean-variance objective and minimization of the maximum regret, were considered in order to determine a “robust” post- donation blood screening strategy that minimizes the risk of releasing an infected unit of blood into the blood supply. Efficient and exact algorithms are provided.

SD36 36-Room 413, Marriott Public and Nonprofit Sector Applications Sponsor: Public Sector OR Sponsored Session

Chair: Ece Zeliha Demirci, PhD Candidate, Bilkent University, Department of Industrial Engineering, Bilkent University, Ankara, 06800, Turkey, edemirci@bilkent.edu.tr 1 - Designing Intervention for Public-interest Goods Ece Zeliha Demirci, PhD Candidate, Bilkent University, Department of Industrial Engineering, Bilkent University, Ankara, 06800, Turkey, edemirci@bilkent.edu.tr, Nesim K. Erkip We study intervention design problem for public-interest goods with two intervention tools: investment on demand increasing strategies and subsidies. We consider a setting composed of a retailer whose demand is exponentially distributed and a central authority with fixed budget. We characterize the optimal solution structure and enrich our findings with detailed analysis of results. 2 - A Two-stage Model for Dynamic Staff-job Assignments in the Non-for-profit Sector We design mechanisms for large-scale assignment problems that appear in public sector applications by identifying complete, stable, and fair staff-job matchings over time, even when preference lists are truncated. To address these issues we consider dynamic and multi-stage negotiation policies using stochastic optimization. Equilibrium concepts and heuristics are proposed to approximate the proposed problem to optimality. 3 - Modelling and Analysis of New Zealand (NZ) Legislation Network Neda Sakhaee, PhD Candidate, University of Auckland, Room Network representation of legal documents is a novel approach to study this complex system. The result is a huge citation network of legal documents (as nodes) and links between them (as edges). We present this network for NZ acts using a data set of more than 700 year old acts from NZ legislation website. We study the structure of the network, measures, clusters and time evolution. Then, we present correlation studies between the clusters and government policies considering longitudinal changes 4 - Optimizing Government Resource Allocation to Increase Community Resilience Saba Pourreza, PhD Candidate, University of North texas, 1307 West Highland Street, Denton, TX, 76201, United States of America, saba.pourrezajourshari@unt.edu, Brian Sauser This study constructs an optimization model that considers two decision variables job creation, goods and service production. The aim of the model is to enhance the community impact of small medium size businesses (SMB) when a disruption hits. Tina Rezvanian, PhD Student, Northeastern University, Huntington Ave, Boston, 02115, United States of America, rezvanian.t@husky.neu.edu, Ozlem Ergun 576, 38 Princes Street, Auckland, 1010, New Zealand, nsak206@aucklanduni.ac.nz, Golbon Zakeri, Mark Wilson

131