Optimisation of blood matching and estimation of alloimmunisation risk in transfusion-dependent patients

Blood transfusion is a life-saving treatment for people with sickle cell disease. Recent advances in genotyping pave the way for extended red blood cell matching. Hence, we formulate sequential allocation decisions as a Markov decision process and study penalty-based policies for matching, which consider up to 17 blood group antigens, including policies that look ahead to future patient appointments.
Presently, blood is matched manually for transfusion using incomplete red cell blood type information to minimise the immunological incompatibility between donor and patient. We develop a Bayesian framework to model alloimmunisation events in the context of missing data. We integrate hospital record data from heavily transfused patients with blood service data from donors and infer missing donor blood types from self-reported demographic information, using a model for the distribution of blood types stratified by latent genetic variables.