Healthcare Modeling and Optimization VIII
By Amira Hijazi
Operations research and analytics are powerful tools that can be used to model, analyze, and improve almost every aspect of life. Today’s session, “Health Care Modeling and Optimization VIII,” included six talks with a goal of using O.R. and analytics to help improve healthcare: starting with Dr. Zlatana Nenova, with a focus on predicting future demand of kidney disease patients. She presents a two-step framework to identify high impact individuals; identification of those patients is very critical for hospitals as it affects future hiring needs.
Moving from hiring needs to patient’s admission policies, Jorge Acuna from the University of South Florida proposes a Markov decision process to optimize the admission of different classes of patients to the inpatient ward. By considering different patient length of stay in the admission policy, Acuna shows the significant improvement over the current policies in overcrowding management.
Speaking of overcrowding, Mohammadreza Torkjazi from West Virginia University analyzes the current medical care policies during mass gatherings. Such policies can be on-site, off-site, or both. Torkjazi presents a mathematical model to study the trade-offs between these policies and when to use each policy.
I was honored to be the fourth presenter in this session. In my presentation, I talked about the current policy of flu vaccine allocation where vaccines are administered continuously through the season. In the current policy, it is not always possible to vaccinate a large population due to insufficient availability of vaccine doses. These observations suggest a new two-phase vaccination policy where we vaccinate at least a minimum statewide proportion at the beginning of the epidemic then wait to realize how the disease is spreading. Based on that, we decide whether to vaccinate more or stop. We use sample average approximation to solve the problem and we found that the proposed policy is better than the current policy. Finally, we take into consideration the risk aversion of healthcare officials.
The fifth presenter, Dr. Soheil Eshghi from Yale University, discusses the effectiveness of different policies used to find and treat infected individuals in outbreaks. He uses data from the Ebola outbreak in Liberia to show that contact tracing policy is not very effective if used alone. However, using contract tracing with enhanced surveillance can help the balance between disease burden and policy expenditure.
Finally, Dr. Jian Hu from the University of Michigan presents his work with Henry Ford Macomb Hospital to improve nurse scheduling by using a two-stage resilient dynamic programming model. The simulation model he used shows a significant improvement in the nurse scheduling and good robustness against uncertainty.