Engineering solutions for the transplant supply gap: social network analysis, artificial intelligence, and optimization in living kidney donation.

Kidney transplantation is the gold standard for treating end-stage renal disease, yet over 90,000 patients remain on the transplant waitlist. This dissertation introduces engineering-driven solutions to help reduce the transplant supply gap by addressing three challenges: illicit trafficking, donor recruitment, and evaluation inefficiencies. First, we model illicit organ trafficking networks using social network analysis. We demonstrate that targeting transplant clinics alone is insufficient to disrupt operations. Instead, disrupting the network requires detaining brokers who organize behind-the-scenes logistics—offering a more effective strategy for intervention. Second, we seek to identify a latent population of potential living donors who face barriers such as limited information, health concerns, or financial worries. Using deep learning models on online discussions and expert-informed schema design, we classify potential donor profiles and the factors influencing their decisions. To improve model accuracy and adaptability, we introduce Combinatorial Promptimization, a novel automatic prompt engineering technique that outperforms Google DeepMind’s PromptBreeder on the GSM8K benchmark. Third, we address inefficiencies in the donor evaluation process, where 8–86% of potential donors drop out. We propose an optimization model for the Single-Dependence Sequential Testing Problem, minimizing the total expected cost and time required for medical testing. Together, these methods show how tools from network science, AI, and operations research can strengthen legitimate transplant systems, support potential donors, and accelerate evaluations—helping to increase the number of living kidney donations.