AI system information
1. How many artificial intelligence (AI) systems does your trust/health board currently have in development?
2. How many AI systems are currently deployed within your trust/health board?
3. For every AI model currently being used or developed in your trust:
3a. Is the AI model being used operationally, clinically, or for another purpose?
3b. What department(s) is the AI system being used in?
3c. What month and year was the AI system first deployed? (n/a if in development).
3d. Was the AI system created by a commercial entity, university, in-house, or within another NHS trust? Please give the name of the organisation.
3e. What is the AI systems architecture? (e.g. deep neural network, random forest, logistic regression, large language model).
3f. Which coding language and packages are used to deploy the AI system? (e.g. python - scikit-learn, pytorch, tensorflow)
3g. What is the nature of the input of the AI system? (e.g. a medical scan, free text notes, tables of lab results).
3h. What is the nature of the output of the AI system? (e.g. a masked image, a risk score, natural text).
3i. How was the AI system validated in the target population before deployment?
3j. What measures are in place to monitor for degradation in the performance of the AI system post-deployment?
3k. What was the cost to procure the AI model and what is the ongoing cost of use?
3l. Which departmental budget is the cost paid from?
Data management
4. Was any local patient data used for training or fine tuning the AI system?
5. Is any patient data collected specifically for the purpose of training any of the AI systems currently in development or deployment?
6. Do you share, or plan to share, any patient data with third-party developers for AI-related purposes? If yes, please provide details of the data-sharing agreement or relevant policy.
Regulation
7. What is the regulation and certification of the AI model under the European Union Medical Device Directive and/or United Kingdom Medicines and Healthcare products Regulatory Agency (MHRA)?
Patient/public involvement
8. Was any patient/public engagement undertaken before deployment of the AI system?
9. Is there any ongoing patient/public engagement input into the use of AI within your organisation?
Governance
10. Who has responsibility for the AI systems being using?
11. Does your organisation have a governance policy that covers:
11a. Use of AI systems within your organisation?
11b. Ongoing evaluation of an AI model's performance after deployment?
11c. How to monitor for bias in the AI system and how to mitigate against this?
