Enhancing kidney organ offer decision making through use of artificial intelligence

Research question Is use of an Artificial Intelligence (AI) driven Clinical Decision Support (CDS) tool at the time of deceased donor kidney offering safe, feasible and effective? Background At any time, there are around 6,000 patients on the UK kidney transplant waiting list with an average wait of 2-3 years. The shortage of organs means that some patients become unfit for surgery or die whilst waiting. Because of this, clinicians often consider organ offers from less-than-optimal donors with existing comorbidities, where the transplant outcome is less certain. Uncertainty around the outcome from a transplant leads to variability in decision-making between centres and clinicians. Our teams have been developing AI-based models that use donor, recipient and organ information to predict the outcome if an offer is accepted and transplanted or declined and the patient waits for another offer. These models provide accurate predictions whilst helping the user understand the reasons for individual predictions. These are useful to aid clinician decision making, and in communicating risk to patients during informed consent. Aims/ objectives This research aims to refine our existing models and integrate them into a web-based tool suitable for clinical use. We will then evaluate safety, feasibility, efficacy and cost effectiveness of these tools in a real-world clinical setting through a randomised clinical trial. Methods Work package (WP) 1 will refine our existing AI models by assessing and mitigating bias, estimating uncertainty and incorporating organ image analysis. This will result in stable, accurate and explainable multimodal models suitable for clinical evaluation. WP2 will integrate these models into a web-based CDS tool, working alongside WP4 to co-design a user interface. Regulatory approval for the tool as a medical device will be sought. WP3 will evaluate the feasibility, safety and efficacy of the CDS tool. An IDEAL stage 2A study will evaluate use retrospectively in a simulated setting using historic organ offers, followed by a prospective, randomised, IDEAL stage 2b/3 clinical trial in three UK transplant centres. The study will randomise 658 kidney offers to use the CDS tool, or standard clinical practice. We will assess impact of the tool on offer acceptance, organ utilisation and transplant outcomes. WP4 will run in parallel, employing qualitative and human factors techniques to co-design the user interface, evaluate usability, and to identify barriers and enablers to implementation. WP5 will undertake a health economic assessment of the tool alongside WP3. Timelines Total study duration is 60 months: WP1 – months 1-18 WP2 – months 10-26 WP3 – months 16-60 WP4 – months 19-60 WP5 – months 37-60 Impact and dissemination Our CDS tool has the potential to reduce variability in offer decision making, reducing inequalities in access to transplantation. It may also improve overall organ utilisation, reducing the burden of dialysis on patients and the NHS. Outputs from this work will have value to the transplant community, clinical AI community, patient groups and policy makers. Our dissemination strategy includes a variety of academic, clinical, patient and mainstream media channels to reach all these groups for maximum impact.