Studies have shown 70% of the patients who require Stem Cell Transplants have to rely on Unrelated Donors for a successful treatment. Public registries, such as BeTheMatch maintained by the National Marrow Donor Program, are responsible for meeting this demand. Maintaining a large volunteer registry is a complex and expensive process. We propose data-analytic modeling for three specific problems that can aid in unrelated donor search and registry management. I. Donor Selection: Donor selection for Stem Cell Transplant often requires physicians to manually select 3-5 donors from a long list of genetically compatible donors with varying degrees of match probabilities as identified by Human Leukocyte Antigen (HLA) matching algorithms. The decision process is based upon non-HLA donor attributes, and is very time consuming. We provide a binary classification model that is trained on historical past donor selection data to help make future donor selections faster and consistent. II. Donor Availability: Donors can decline a sample donation request for a number of reasons, which adversely affects the time taken to complete a transplant. Past responses show that only half the requests receive a positive response. We propose a binary classification model for predicting donor availability based on donor demographic information and responses to outreach programs. III. Donor Utility and Recruitment: Power law like distribution of HLA types implies that a large number of registered donors are never utilized. We provide a mathematical framework to combine the Donor Selection and Availability models with donors’ HLA type to determine donor utilization, that can be used to identify donors for future cost management efforts in the registry, such as advanced typing. Studies indicate that a large number of patients also don’t find a match due to lack of diversity in the registry. We provide recommendations for targeted donor recruitment to enhance diversity based on donors’ geographic information.
University of Minnesota Ph.D. dissertation. January 2018. Major: Biomedical Informatics and Computational Biology. Advisor: Vladimir Cherkassky. 1 computer file (PDF); vii, 73 pages.
Applications of Data Analytic Modeling for Efficient Stem-Cell Transplants and Donor Registry Management.
Retrieved from the University of Minnesota Digital Conservancy,
Content distributed via the University of Minnesota's Digital Conservancy may be subject to additional license and use restrictions applied by the depositor.