Clinical AI Guidelines
Good guidelines make clinical AI safer, more transparent, and more reproducible. This page brings together key reporting standards, checklists, and frameworks relevant to the development, evaluation, and deployment of AI in clinical settings. AfCAI will add to this collection as the field develops.
Reporting Checklists
Below are four essential checklists for researchers and clinicians working with AI in health. These are free to use and directly relevant to NHS clinical AI practice.
SPIRIT-AI Checklist
The Standard Protocol Items: Recommendations for Interventional Trials – Artificial Intelligence (SPIRIT-AI) extension. A checklist for writing clinical trial protocols that involve an AI intervention, ensuring trials are prospectively registered with sufficient detail to be reproducible and appraised.
Use when: designing or writing up the protocol for a clinical trial involving an AI system.
CONSORT-AI Checklist
The Consolidated Standards of Reporting Trials — Artificial Intelligence (CONSORT-AI) extension. A checklist for reporting completed randomised controlled trials involving an AI intervention, addressing items specific to AI that are not covered by the standard CONSORT statement.
Use when: writing up the results of a completed RCT in which an AI system was evaluated.
DECIDE-AI Checklist
The Developmental and Exploratory Clinical Investigations of DEcision support systems driven by Artificial Intelligence (DECIDE-AI) checklist. A checklist for reporting early-stage clinical evaluation of AI-based decision support systems — the “first-in-human” equivalent for clinical AI.
Use when: reporting an early, small-scale live clinical evaluation of an AI decision support system.
STARD-AI Checklist
The Standards for Reporting of Diagnostic Accuracy Studies — Artificial Intelligence (STARD-AI) extension. A checklist for reporting diagnostic accuracy studies in which the index test involves an AI system.
Use when: writing up a study evaluating the diagnostic accuracy of an AI system.
Other Key Reporting Standards
The following internationally recognised standards and frameworks are widely used in clinical AI research and deployment. Where AfCAI endorses or has adapted a standard, this is noted.
| Guideline | What it covers | Link |
|---|---|---|
| PROBAST | Risk of bias assessment tool for prediction model studies, including AI-based models. | probast.org |
| NHS AI Lab: AI and Data Regulations | Practical guidance from NHS England on navigating regulation, safety, and governance for AI in health and care. | NHS AI Lab |
| MHRA AI/ML Guidance | UK regulatory guidance on software as a medical device (SaMD) and AI/ML-based medical devices. | MHRA guidance |
AfCAI Guidelines in Development
AfCAI is developing its own guidance for clinical AI practitioners across three areas that align with its Charitable Objects: safe and ethical deployment, professional education standards, and scientific advancement. These will be published here as they are finalised.
If you would like to contribute to or be consulted on AfCAI’s guidelines work, please register your interest via the membership page.