Tuesday, August 29

Tuesday, August 29

Tuesday, August 29 – Health and Humanitarian Systems

Workshop Chair: Pinar Keskinocak

Time: 9:00 – 10:15
Speaker: Dave Goldsman
Title: Simulating Healthcare Systems
Abstract: Simulation has emerged as a valuable modeling and analysis tool in the healthcare space. This talk describe a variety of healthcare research problems in which simulation has found great applicability. We provide examples ranging from traditional hospital operations to pandemic disease propagation to disease surveillance.

Time: 10:15 – 10:45 – Break

Time: 10:45 – 12:00
Speaker: Joel Sokol
Title: Transplant Analytics
Abstract: The waiting list for organ transplants in the United States continues to get longer and longer, with severe consequences: on average, a person is added to the waiting list every 10 minutes, and an average of about 22 people die per day while waiting for a transplant. In this tutorial, we describe how analytics — statistics, optimization, machine learning, etc. — can be used to address a variety of transplant-system questions, all with the goal of increasing both the number of life-saving transplants and the quality of their medical outcomes.

Time: 12:00 – 13:15 – Lunch

Time: 13:15 – 14:00
Speaker: Turgay Ayer
Title: Prioritizing Hepatitis C Treatment in U.S. Prisons
Abstract: About one out of six inmates in the United States (U.S.) is infected with hepatitis C virus (HCV). HCV prevalence in prison systems is ten times higher than the general population, and hence prison systems offer a unique opportunity to control the HCV epidemic. New HCV treatment drugs are very effective, but providing treatment to all inmates is prohibitively expensive, which precludes universal HCV treatment in prison systems. As such, current practice recommends prioritizing treatment based on clinical and incarceration-related factors, including disease staging, remaining sentence length, and injection drug use (IDU) status. However, there is controversy about how these factors should be incorporated because of the complicated tradeoffs. In this study, we propose a restless bandit modeling framework to support hepatitis C treatment prioritization decisions in U.S. prisons. We first prove indexability for our problem and derive several structural properties of the well-known Whittle’s index, based on which, we derive a closed-form expression of the Whittle’s index for patients with advanced liver disease. From the interpretation of this closed-form expression, we anticipate that the performance of the Whittle’s index would degrade as the treatment capacity increases; and to address this limitation, we propose a capacity-adjusted closed-form index policy. We parameterize and validate our model using real-world data from Georgia state prison system and published studies. We test the performance of our proposed policy using a detailed, clinically-realistic simulation model and show that our proposed policy can significantly improve the overall effectiveness of the hepatitis C treatment programs in prisons compared with the current practice and other benchmark policies, including the commonly used Whittle’s index policy. Our results also shed light on several controversial health policy issues in hepatitis C treatment prioritization in the prison setting and have important policy implications including: 1) prioritization based on only liver health status, a commonly practiced policy, is suboptimal compared with many other policies we consider. Further, considering remaining sentence length of inmates and IDU status in addition to liver health status in prioritization decisions can lead to a significant performance improvement; 2) the decision of whether to prioritize patients with shorter or longer remaining sentence lengths depends on the treatment capacities inside and outside the prison system, and prioritizing patients with shorter remaining sentence lengths may be preferable in some cases, especially if the treatment capacity inside the prison system is not very tight and linkage-to-care level outside prison system is low; and 3) among patients with advanced liver disease, IDUs should not be prioritized unless their reinfection is very well controlled.

Time: 14:00 – 14:45
Speaker: Eva Lee
Title: Saving Lives: Building Capacities, Capabilities, and Real-Time Operations
Abstract: Its a terrifying doomsday scenario: A novel infectious disease is sweeping
through the worlds population, and health officials have only a day or two to stop
its deadly spread. While this may sound like the plot of a movie thriller, health officials
argue that an event of this kind could become a reality sometime in the near
future. According to epidemiology experts, new drug-resistant infectious diseases are
appearing more frequently and are spreading faster than ever before. With the ability
to strike anywhere in the world at any given time, these pathogens are not only a
major threat to public health, but they also place a substantial burden on the global
economy. In order to best address the next outbreak, the U.S. Centers for Disease
Control and Prevention (CDC) partnered with Dr. Eva Lee, a math/OR professor at
Georgia Institute of Technology, to produce a modeling tool to help health personnel with the challenge of population protection in an emergency. The software, known as RealOpt􀀀, has decision support capabilities for modeling and optimizing the public health infrastructure for hazardous emergency response. It is designed for use in biological and radiological preparedness, for disease outbreak planning and response, and for natural disaster planning. RealOpt helps officials plan for dispensing facility locations, to ensure optimal facility staffing and allocation of resources, including routing of the population and dispensing modalities. The program sifts through massive amounts of data to better optimize decision-making during the event of an emergency scenario especially in the case of a deadly outbreak. In this talk, Lee will share her experience in assisting with the U.S. response to the earthquake in Haiti, in Japan for the Tohoku earthquake and tsunami and the Fukushima radiological emergency response, the 2014 Ebola outbreak in West Africa, and the recent Zika virus in the Americans and within United States. She will discuss the system capabilities and the technical challenges.

Time: 14:45 – 15:00 – Break

Time: 15:00 – 15:30
Speaker: Ethan Mark
Title: Using Machine Learning and Simulation to Help Patients Decide Whether to Accept a High Risk Kidney or Remain on the Waiting List.
Abstract:In 2012, over 10% of organ donors in the U.S were labelled “Increased Risk” (IR). With an organ shortage that has been increasing every year, many patients are faced with a decision: should patients offered an IR organ accept it, or remain on the waiting list for non IR organ? Using machine learning and simulation, we built an interactive tool to help patients make this decision. We will discuss our model building process and show some interesting results from our simulation. We find that in a large number of cases, patients would have a higher benefit for accepting the “increased risk” organ. In this talk, we will focus on the Kidney, and organs coming from donors with either HBV, HCV or HIV.

Time: 15:30 – 16:00
Speaker: Seyma Guven-Kocak
Title: Dynamically Consistent Home Health Care Routing and Scheduling
Abstract:This work addresses a real-world home health care routing and scheduling problem (HHCRSP) faced by a home care agency in the United States. In home health care scheduling, there is a desire to retain consistency with respect to the HHA servicing each patient, which is referred to as continuity of care. In order to handle this consistency requirement, we propose a dynamic approach and introduce the dynamically consistent home health care routing and scheduling problem (D-Con-HHCRSP). We present two different constructive methods to solve HHCRSP on a daily basis: an integer programming-based method with approximations and a variant of a petal heuristic. We present adjustments on these methods to address D-Con-HHCRSP, where the goal is to be able to quantify and control the deviation from the existing schedule in place, so that some of the existing assignments may be retained in the new schedule that is produced. We discuss the performance and computational efficiency of these methods.

Time: 16:00 – 16:30
Speaker: Zihao Li
Title: Value of Inventory Information
Abstract: We evaluate the value of dynamic availability of vaccine inventory information during an influenza pandemic, with the goal of efficiently effectively allocating limited vaccine supplies to reduce the number of infections and to improve public health outcomes. We adapt an agent-based simulation model to predict the spread of the disease both geographically and temporally. We study a vaccine inventory allocation strategy considering population and inventory information when uptake rates vary geographically, and compare it to the commonly used pro-rata allocation strategy. The simulation model is run using detailed population and work-related commuting data from the state of Georgia, and is flexible to be run with data from other locations. The proposed vaccine inventory allocation strategy, when compared to the pro-rata allocation strategy, reduces the infection attack rate from 23.4% to 22.4%, decreases the amount of leftover inventory from 827 to 152 thousand, and maintains or increases the percentage of vaccinated population. In addition, the proposed allocation strategy maintains fairness, i.e., the percentage of vaccinated population remain the same or increases in all the locations, compared to the pro-rata strategy. Our results indicate the need for greater vaccine inventory visibility in public health supply chains, especially when the supply is limited and uptake rates vary geographically. Such visibility has a potential to decrease the number of infections, help identify locations with low uptake rates and to motivate public awareness efforts.