Comprehensive Summary
This study is concerned with using artificial intelligence(AI) to drive allocation of kidney transplants as a treatment for end stage renal disease(ESRD). This was a systematic review that compiled existing literature about AI driven kidney allocation. The PICOS(Population, Intervention, Comparator, Outcomes, Study, Design) was used to define the scope of the review. The study found that AI models outperform traditional risk assessment models like the Kidney Donor Risk Index (KDRI) and Estimated Post-Transplant Survival (EPTS) in predicting both graft and patient survival with C-indices ranging from approximately 0.65 to 0.72. However, the study stated that this evidence is not universal and that the gains may often be driven by the number and type of variable rather than the pure superiority of the AI models. Although the AI models have been shown to show better discriminative ability compared to traditional risk scoring models, they have not actually been used to drive actual donor matching decisions.
Outcomes and Implications
This study addresses the long standing problem of organ donation allocation by considering the use of AI to solve it. The study also takes into account challenges involving ethics, fairness, and transparency that would need to be faced in the implementation of such AI algorithms. Also the study described how most of the models used were mostly used for risk assessment and outcome forecasting, rather than actual allocation of kidneys which limits their clinical utility.