Informs Annual Meeting 2017

TD42

INFORMS Houston – 2017

2 - Confirmatory Data Analytics to Reduce Excess Days in Hospitals Shaghayegh Norouzzadeh, Program Manager, Northwell Health, 1 Fairhaven Mall, Apt B83, Mineola, NY, 11501-8517, United States, snorouzzad@northwell.edu, Christina Mouradian The objective of this study is to determine and reduce the frequency and reasons for medically unnecessary hospital days also contributing to excess days. Additional length of stay creates risk for communication errors and hospital acquired adverse events. This study focuses on reducing the turnaround time (TAT) from when patient is medically ready to be discharged to actual discharge. A staged configuration is designed to integrate the qualitative and quantitative data analysis for administrative decision making in two inpatient units. The results indicate that the average TAT and variation improved by minimum of 21% and 26% respectively. Correspondingly, excess days decreased by 45%. 3 - Service and Price Design for Healthcare Information Service Platform Shifu Pan, PhD Candidate, University of Southeast, No. 2, Southeast University Road, Jiangning D, Nanjing, 211189, China, shifupan@outlook.com In order to study the sustainable operation mechanism of health information service platform, we construct a supply chain consisting of medical data providers, health information service platforms and medical information demanders. We design the service content of health information service platform and discuss pricing strategies for health information service platforms that maximize the supply chain revenue. 4 - Weighted Yield Bundle to Quantify Patient Experience in Hospitals Shaghayegh Norouzzadeh, Program Manager, Northwell Health, 1 Fairhaven Mall, Apt B83, Mineola, NY, 11501-8517, United States, snorouzzad@northwell.edu, Nancy Riebling, Ryan Cowan, Christina Mouradian, Eric Chen, George Reeder Strategies to target patient experience typically rely on a random sample of post- discharge satisfaction surveys and Likelihood to Recommend (LTR) scores. This study’s purpose is to enhance the patient experience by reducing the avoidable defects in hospital service delivery. This study presents an innovative two-stage approach. A patient experience bundle© development with a weighted yield model to quantify the qualitative parameters in the process and concurrently measure patient experience, followed by specification of process inefficiencies for improvement. Overall bundle compliance improved on average by 41% with a relative LTR improvement of 5.8%ile. 5 - A Markov Chain Analysis for Medical Resource Allocation for Disease Outbreaks Farbod Farhadi, Roger Williams University, 1 Old Ferry Road, Bristol, RI, 02809, United States, ffarhadi@rwu.edu We encounter many disease outbreaks around the world every year. In year 2016 alone, there were 136 disease outbreaks with variety of impacts recorded by the World Health Organization (WHO). Every disease outbreak affects populations and economies negatively in multiple dimensions. For instance, the Ebola outbreak of West Africa in 2014 caused over 28 thousand suspected cases and more than 14 thousand deaths. In this study, we develop a markovian chain process to propose an efficient allocation of medical resources for disease outbreak containment. 360A Transportation Issues in Smart Cities Sponsored: Public Sector OR Sponsored Session Chair: Leila Hajibabai, Ph.D., Stony Brook University (SUNY), Stony Brook, NY, 11794, United States, leila.hajibabaidizaji@stonybrook.edu 1 - Optimal Control Logic for Intersection Management in an Autonomous Vehicle Environment Amir Mirheli, Stony Brook University, 30 Piedmont Dr. Apt 9, Stony Brook, NY, 11776, United States, amir.mirheli@stonybrook.edu, Leila Hajibabai, Ali Hajbabaie This research presents an optimal control logic to minimize travel delay while upholding safety constraints. The problem is formulated into a dynamic programming model, where vehicular information is obtained from neighboring intersection approaches. A stochastic look-ahead model is developed based on Monte Carlo Tree Search algorithm to determine the optimal vehicle trajectories that prevent movement conflicts. Numerical results confirm that reliable coordination between connected autonomous vehicles can effectively address the consequences of traffic signals. TD42

2 - Real-time Dynamic Traffic Metering in Connected Urban Street Networks Rasool Mohebifard, Washington State University, 1630 NE Valley Road, Apt B204, Pullman, WA, 99163, United States, rasool.mohebifard@wsu.edu, Ali Hajbabaie This study develops a real-time optimization algorithm to determine dynamic metering rates at entry gates to urban street networks. The proposed approach decomposes the problem into smaller sub-problems with the objective of reducing the computational complexity while creating effective coordination among them to improve the quality of the solutions. The proposed approach was tested on a realistic case study network and found dynamic metering rates in real-time that improved network performance. 3 - Dynamic Speed Harmonization in Connected Urban Transportation Networks Mehrdad Tajalli, Washington State University, 210 SE McKenzie St, Pullman, WA, 99163, United States, mehrdad.tajalli@wsu.edu, Ali Hajbabaie This study develops a mathematical linear program for dynamic speed harmonization in connected transportation networks. The solution algorithm decomposes the network to several sub-networks to reduce computational complexity of the problem and improve its runtime. At the same time, the decisions of various sub-problems are coordinated to push the solutions towards optimality. The results show that the proposed algorithm finds real-time solutions in networks of various scales that significantly reduce the number of stops and travel time while increasing the number of completed trips. 4 - Distributed Coordination and Optimization for Optimal Signal Control under Various Connected Vehicle Penetration Rates This research presents distributed coordination and optimization algorithms for real-time optimal signal control with various connected vehicle penetration rates. The algorithm distributes the optimization to the intersection-level to improve the run-time and scalability, and creates effective coordination between neighboring intersections to push solutions towards optimality. Besides, we developed an algorithm to incorporate point detector data and car following models to predict the location of unconnected vehicles. The results indicated significant improvement in network performance under various connected vehicle penetrations rates compared to solutions found by VISTRO. 360B Prevailing Issues in Public Sector OR Sponsored: Public Sector OR Sponsored Session Chair: Ebru Korular Bish, Virginia Tech, Blacksburg, VA, 24060, United States, ebru@vt.edu Co-Chair: Hrayer Aprahamian, ahrayer@vt.edu 1 - Robust Prevalence Rate Estimation under Limited Resources Ngoc Nguyen, Virginia Tech, Blacksburg, VA, 24060, United States, ntn@vt.edu, Ebru Korular Bish Prevalence rate estimation is essential to surveillance studies in public health, and is normally conducted via pooled testing. However, as the optimal pooling design relies on the true and unknown prevalence rate, it is often determined under uncertainty. We take a robust optimization approach to prevalence rate estimation by minimizing the maximal regret in the estimate mean squared error, given an interval in which the true prevalence is presumed to be. We then compare the numerical results from the robust optimization model to those obtained from an expectation-based pooling design model to demonstrate the benefits of robust optimization in pooling design for prevalence estimation. 2 - Resource Allocation Policies for Post-disaster Relief Operations with Explicit Consideration of Human Choice Behavior Xinchang Wang, Assistant Professor, Mississippi State University, 324C McCool Hall, Marketing Department, P.O. Box 9582, Mississippi State, MS, 39762, United States, xwang@business.msstate.edu It has been seen that disaster-affected victims choose to visit a point of distribution (POD) to acquire life-sustaining commodities from a set of alternatives after disasters. We study a variety of post-disaster resource allocation models with victim choice to minimize the total unmet demand. These models are non-differentiable and non-convex due to nonlinear utility functions capturing the reality that no informed victim would visit a POD with no inventories. We fully characterize the optimal policies for these models under a homogeneous case. For a general non-homogeneous case, we provide a partial characterization of the optimal policies and propose a search method to find optimal policies. S.M. Islam, Washington State University, Pullman, WA, United States, smabinal.islam@wsu.edu, Ali Hajbabaie TD43

390