Artificial Intelligence (AI)

AI development stages

Describe the entry point (or the minimum requirement on eligibility), the exit point, as well as the activities, for each phase of the AI development stage:

1. 1. Proof-of-concept
   2. Development and clinical evaluation of the prototype
   3. Real-world testing of the innovation in health or social care settings
   4. Health system adoption

The researcher should consider and give details of the innovation and its AI component(s), including the following.

• The need
  • What is the health problem (unmet clinical need) that the technology is trying to solve?
  • In what way does the technology support the priorities of the health and social care systems?
  • Who are the intended users?
• The innovation
  • Give a clear description of the proposed innovation.
  • What is the rationale for choosing AI to solve the problem?
  • What AI approaches are being used and why?
  • Is there clear proof-of-concept, that is, are you able to demonstrate that the approach is technically feasible and has practical potential?
  • What are the functionality, structure, intended use, and limitations of the technology?
  • What are the benefits and concerns of the technology (safety, risks, ethics, and distortions to decision-making)?
• Feasibility and acceptability of the technology
  • Is the AI product ethically permissible, justifiable, worthy of public trust, fair and non-discriminatory?
  • Has the technology been shown to be trusted/accepted by diverse end-users and clinicians, including different demographics and professional backgrounds (e.g., patient and public involvement and engagement, and clinical acceptability studies)?
  • What will be your data protection procedure to mitigate the risks of patients, care professionals, or service users being identified?

• Model validation
  • What training datasets will be used (source, access, size, security, diversity)?
  • Will independent datasets be used to validate the technology?
  • How representative are the training and validation datasets of “real-world” data?
  • How will the robustness of the model be assessed and improved?

• Regulatory and adoption
  • Has the technology been regulatory assessed/approved?
  • Does the product have or need a CE marking and risk classification?
  • How and by whom will the technology be adopted into the clinical pathway and what is the pathway to adoption?

Examples of NIHR-funded projects involving AI are available at:  https://www.nihr.ac.uk/documents/ai-in-health-and-care-award-funded-projects-2021/27866

• A buyers guide to AI in health and care
• A guide to good practice for digital and data-driven health technologies. Department of Health and Social Care
• AI Code of Conduct
• Artificial Intelligence Funding
• Artificial Intelligence in Health and Care Award – AI definition
• Artificial intelligence: How to get it right. Putting policy into practice for safe data-driven innovation in health and care
• Invention for Innovation - Product Development Award Supporting Information for Applicants
• NHS Digital Standards
• NICE Evidence Standards Framework
• Recommendations for diversity, inclusivity, and generalisability in artificial intelligence health technologies and health datasets
• Regulatory Horizons Council: The regulation of Artificial Intelligence as a Medical Device
• The National AI strategy
• The National Strategy for AI in Health and Social Care
• The NHS Long Term plan
• Understanding artificial intelligence ethics and safety
• Understanding regulations of AI and digital technology in health and social care