Thesis Title: Analytics-based approaches to improve patient outcomes in healthcare

Advisor:

Dr. Pinar Keskinocak, School of Industrial and Systems Engineering, Georgia Tech

Thesis Committee:

Dr. David Goldsman, School of Industrial and Systems Engineering, Georgia Tech

Dr. Brian Gurbaxani, School of Industrial and Systems Engineering, Georgia Tech

Dr. Raymond Lynch, Division of Transplantation, Department of Surgery, Emory University School of Medicine

Dr. Joel Sokol, School of Industrial and Systems Engineering, Georgia Tech

Date and Time: Friday, June 24th, 2022, at 10 am (EDT)

Meeting Link: https://gatech.zoom.us/j/99610038299

Abstract:

Patient outcomes in healthcare are tremendously affected by the efficiency of the care-providing system. This thesis focuses on the application of analytics-based approaches to improve patient outcomes in two healthcare settings: rehabilitation scheduling and deceased-donor kidney transplantation.

​

Appointment scheduling in rehabilitation centers is a complex process that requires patients to be scheduled for a combination of services over multiple days with different providers. In Chapter 2 of the thesis, we develop a mixed-integer programming model and a heuristic to generate appointment schedules that provide timely and ideal care to the patients. We also develop a decision support tool for scheduling the appointments and discuss its implementation at Shepherd Center, a rehabilitation hospital.

​

Kidney transplantation is the preferred option for patients with chronic kidney disease due to lower long-term costs and better quality of life resulting from transplantation. However, in the United States, there is a large gap between the supply and demand of organs. Owing to this organ scarcity, many patients on the transplant waitlist either die or become too sick to receive a transplant. Thus, there is an urgent need to develop and evaluate allocation policies and acceptance practices that improve patient health outcomes and reduce organ wastage. In Chapter 3, we discuss the development and validation of a discrete-event simulation model that mimics the current deceased-donor kidney allocation process. In Chapter 4, we develop an alternative kidney allocation policy that is based on geographical supply-demand trends and present a quantitative comparison to the current policy via the simulation model. Our findings indicate that the alternative allocation policy results in better system- and patient-level outcomes – such as higher number of transplants, lower average wait time, and higher kidney utilization – than the current system. In Chapter 5, we develop an ensemble survival model to predict a patient's post-transplant survival probability using donor and patient characteristics. We then propose an organ offer accept/decline practice, in which a patient accepts a kidney offer if the estimated post-transplant survival probability is above a certain threshold. Using the simulation model, we investigate the potential benefits of using this proposed data-driven predictive method in kidney offer accept/decline decisions.