Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC58

Modeling of behavior are essential to early detection of drug abuse and optimization of behavioral interventions from clinicians and pharmacists. A longitudinal drug dispensing dataset is used to study drug use behavior. We develop a semi-Markov mixture modeling approach to model heterogeneous drug use behavior. We identify a weighted combination of canonical semi-Markov models for each individual. We use the frailty term embedded in the semi-Markov models to characterize the heterogeneity among individual behaviors. 2 - A Multi-attribute Optimization Approach for Post Acute Care System Management Hossein Badri, Wayne State University, Detroit, MI, United States, Ashkan Hassani, Maryam Khatami, Mark Lawley, Kai Yang In this study, a stochastic multi-attribute decision making approach is developed for the Post Acute Care System (PACS) management. In this problem, besides the cost metrics, there are some service coverage requirements that make the decision making complex. The proposed approach provides the decision making procedure for healthcare providers who are willing to subcontract post-acute care services. The proposed approach is implemented for PACS management problem in the city of Houston, TX. 3 - Using Analytics to Understand and Model the Problems of Seniors Blocking Hospital Beds Michael Carter, ON, Canada, Pavel Shmatnik The aging demographic created by the “Baby Boomers” is creating some serious capacity issues for downstream sub-acute facilities like nursing homes, retirement homes, rehab and home and community services. The problem is bound to get worse over the next 10-20 years. Today, over 16% of hospital beds are occupied by someone who is supposed to be somewhere else. This is a fairly common problem internationally and Health Authorities are scrambling for intelligent solutions that do not simply involve more money. In this paper, we will present some of our preliminary analysis on root causes and modelling alternatives. 4 - Midterm Nurse Scheduling with Specialized Constraints and Preference Considerations Jia Guo, PhD Student, The University of Texas at Austin, Austin, TX, 78731, United States, Jonathan F. Bard A nurse scheduling IP model is developed to minimize the sum of weighted unsatisfied demand and individual preference violations. Constraints account for vacations, birthdays, maximum number of consecutive working days and days off, and minimum number of rest hours. The problem is solved in two phases using a column generation algorithm. In the first phase, overtime is not allowed. In the second phase, we take the first phase solution and add overtime to cover unsatisfied demand. Results are presented for instances with up to 30 nurses. n TC60 West Bldg 102B Health Applications Sponsored: Health Applications Sponsored Session Chair: Bryan A. Norman, Texas Tech University, Lubbock, TX, 79409- 3061, United States 1 - Creating a Robust Hospital Pharmacy Delivery System Anna Svirsko, University of Pittsburgh, 4200 Fifth Avenue, Pittsburgh, PA, 15260, United States, Hamdy Salman, Bryan A. Norman Large hospital systems route medications from the hospital pharmacy to the inpatient units daily. The medication demand on each unit varies based on the patients in the unit and their medical conditions. First, we create a linear program to establish delivery pathways and then formulate the robust counterpart. Using historic data, we consider the deterministic problem where our unit demand is the average medication demand compared to the robust formulation where the demand is given by an uncertainty set based on the medication demand across each day for a year. We compare the deterministic and robust solution to note the difference between the two formulations. Yi-Chin Kato-Lin, PhD, Hofstra University, Hempstead, NY, 11549, United States, Rema Padman, Vibhanshu Abhishek, Julie Downs A mobile app for healthy eating was designed for our study participants’ use. Users’ behaviors in interacting with peers on a closed social platform were observed. Matching the app usage data with survey data allows us to understand who participate in what activities, or why they were not participating. 2 - Social Platforms on Healthy Eating Mobile App: Who Participate in What Activities, and Why Not?

n TC58 West Bldg 101C Modeling and Optimization in Organ Allocation Sponsored: Health Applications Sponsored Session Chair: Murat Kurt, Bristol-Myers Squibb, 3401 Princeton Pike, Sanjay Mehrotra, Northwestern University, Dept of I. E. / M. S. C246 Tech Inst, 2145 Sheridan Road, Evanston, IL, 60208-3119, United States, Vikram Kilamb, UT Dallas, Dallas, TX, Kevin Bui We will present a distributionally robust model for designing a organ allocation mechanism to achieve equitable sharing of deceased donor organs. The model will be tested using data from liver allocation, using historical supply-demand information as nominal distribution. Parametric sensitivity analysis using the distributional robust modeling will provide information on the stability of the design. 2 - A Sub-distribution Hazard Model to Predict the Likelihood of Deceased Donor Renal Transplant Under New Kidney Allocation System Yeongin Kim,, Mehmet Ayvaci, Bekir Tanriover As the rules for kidney allocation have changed recently, it is inevitable to develop a new model to predict time-to-transplant. We study adult waitlisted kidney patients in the USA after the implementation of the new Kidney Allocation Systems (KAS) using the national registry of kidney patients employing for sub-distribution hazard regression. The sub-distribution hazard regression prediction on waitlist outcomes in the KAS era carries clinically relevant information. Preparation of a risk calculator based on this model is feasible and would be a useful online tool for transplant health care providers to counsel kidney transplant candidates. 3 - Head of Line Matching Policies for Allocation of Deceased Donor Kidneys Yichuan Ding, University of British Columbia, 2053 Main Mall, Vancouver, BC, V6T1Z2, Canada, Baris Ata, Stefanos Zenios We seek the deceased-donor kidney allocation policy that achieves the best efficiency-equity tradeoff, taking into account patients’ strategic choices. We developed an analytical framework based on the fluid model that can analyze a broad class of ranking policies, which provides approximations to the previously and currently used policies in practice. The analysis characterizes the optimal scoring formula for given efficiency-fairness tradeoffs. Our results show that the policies that incorporate characteristics of both the donor and the recipient can improve the quality-adjusted-life-years up to $3.1\%$ based on the fluid model Lawrence Township, NJ, 08648, United States 1 - Developing an Organ Allocation Design using Distributional Robustness Fatemeh Karami, University of Louisville, JB Speed Bldg. 211, e, Louisville, KY, United States, Mehdi Nayebpour, Monica Gentili, Naoru Koizumi, Andrew Rivard Heart failure is a growing health problem affecting nearly 6 million in the United States, with approximately 200,000 patients suffering from a progressed to end- stage or Stage D heart disease. Heart transplantation is the definitive therapy for the end-stage heart disease patients. However, the geographic disparity in access to heart transplant in the United States is substantial. This research analyzes the geographic disparity in the heart allocation system and explores different heart allocation boundaries derived from our mathematical optimization models to decrease disparity. n TC59 West Bldg 102A Aged Care Analytics: Models, Methods and Applications: Part II Sponsored: Health Applications Sponsored Session Chair: Nan Kong, Purdue University, West Lafayette, IN, 47906-2032, United States Co-Chair: Mingyang Li, Tampa, FL, 33647, United States 1 - Heterogeneity Modeling of Drug Use Behavior via Semi Markov Mixture Modeling Zhouyang Lou, Purdue University, West Lafayette, IN, 47906, United States, Nan Kong, Christopher Callahan, Wanzhu Tu, Noll Campbell, Qing Tang Drug use disorder harms individual health as well as the welfare of others. predictions, and also eliminates kidney wastage. 4 - Effective Adult Heart Allocation Policies: An Optimization Approach

328