
Artificial Intelligence to aid Decision making in Kidney Transplantation
(AID-KT)
Researchers from the CET have been awarded an NIHR Programme Grant for Applied Research (PGfAR) to support a 5-year programme of research exploring the use of AI-driven Clinical Decision Support (CDS) tools for offers of kidneys for transplantation.
The project builds on previous work undertaken in collaboration with the AI for Digital Health group at the University of Oxford, led by Tingting Zhu. During this work, we used 20 years of UK Transplant Registry data to train explainable machine learning models that predict patient outcomes when organ offers are accepted or declined.
The AID-KT project aims to further refine these models, assessing risk of bias and quantifying uncertainty. We will then integrate these models into a web-based CDS tool and their impact in real-world clinical practice.
Work Packages
The project has 5 work packages, that are tightly integrated with one another. The overall aim is to deliver and test a clinically-viable clinical decision support tool.
Work Package 1 - Model Development
In this work package, we will build on our previous models to include aspects such as:
- Evaluation of bias and bias mitigation
- Uncertainty quantification
- Use of post-retrieval organ image analysis to refine predictions
We will also externally validate our models using non-UK transplant data.
Work Package 2 - Integration and regulatory
AI-driven clinical decision support tools are considered as "Software as a Medical Device" (SAMD), and therefore need to meet the relevant regulatory requirements for medical devices. This work package will focus on developing a web-based user interface for the CDS tool in collaboration with end-users (clinicians and patients) integrating the tool into clinical workflows, and creating the regulatory submission to the MHRA for clinical investigation.
Work Package 3 - Clinical Evaluation
Clinical evaluation of the CDS tool will follow the IDEAL framework. Clinical studies will comprise a retrospective IDEAL 2a evaluation using historical organ offers in a simulated setting, and a prospective IDEAL 2b/3 randomised clinical trial.
IDEAL 2A retrospective study
In this phase of the clinical evaluation, we will evaluate use of the CDS tool in a simulated setting using historical organ offers. This will allow us to test the integration and user interface, evaluate the potential for clinical impact, and assess elements such as learning curve and effect of experience on clinical decision making.
IDEAL 2B/3 prospective study
This study will take the form of a randomised, prospective clinical trial in three UK transplant centres (Oxford, Coventry and Newcastle). Organ offers will be randomised to have the CDS tool available or not, and the impact of CDS use on clinical decision making and transplant outcomes will be evaluated.
The clinical trial will be jointly run by the Aberdeen Centre for Healthcare Randomised Trials (CHaRT) and the Oxford Surgical Interventional Trials Unit (SITU).
Work Package 4 - Qualitative and Human Factors
This work package will run alongside WP2 and WP3. We will use a mixed-methods approach with a combination of quantitative survey data and qualitative work, involving transplant surgeons, nephrologists, transplant coordinators and patients from the three participating centres. Evaluation will follow the DECIDE-AI guidelines, assessing interface design, usability, user trust and barriers to implementation.
Work package elements include:
- Stakeholder engagement workshops - with both clinical staff and patients
- User interface co-design workshops
- User interface evaluation
- Evaluation of facilitators and barriers to use
- Evaluation of user trust
- Learning curve evaluation
We will also undertake an evaluation of the DECIDE-AI reporting guidelines for early-stage clinical evaluation of AI-driven CDS tools, in collaboration with the DECIDE-AI team.
Work Package 5 - Health Economic Analysis
A health economic work package will run alongside the prospective clinical study in WP3. A full economic evaluation will be conducted comparing the use of the web-based CDS tool with standard care (i.e., standard NHSBT offer data and clinician experience only) for the decision on acceptance of an organ offer for individuals on the national kidney transplant waiting list. A decision analytic model (e.g., Markov cohort model, individual sampling model) will be developed to assess the relative efficiency for the compared strategies. The model structure will be informed by the literature and through discussions with the clinical collaborators and the project management team.
Collaborators
This project involves a large group of collaborators from across the UK.
- Nuffield Department of Surgical Sciences, University of Oxford
- Simon Knight (Chief Investigator)
- James Hunter
- Peter McCulloch
- Baptiste Vasey
- Jessica Scaife
- Reshma Rana Magar
- AI for Digital Health, University of Oxford
- Centre for Healthcare Randomised Trials (CHaRT), University of Aberdeen
- University of Newcastle
- University of Nottingham
- Patient Representatives
- Judy Marks
- Fez Awan
Funding

This work is funded by a grant from the National Institite of Health Research (NIHR) Programme Grants for Applied Research (PGfAR) scheme (grant number NIHR 208885).
AID-KT Publications
A selection of papers related to this project are listed below: