- Coherent integrated care board (ICB) level digital procurement strategies, which include AI, will ensure scalability across health systems.
- Similarly, ICBs should aim to create unified information governance rules to help streamline information management and reduce administrative burdens for both NHS employees and private sector providers.
- To further facilitate successful AI implementation in the NHS, multidisciplinary implementation teams should be formed within individual trusts.
A new report from Public Policy Projects (PPP) calls for ICBs to introduce robust assessments of procurement requirements for AI imaging adoption in the NHS. Such assessments, the report argues, would help ensure that costs are more evenly distributed across the departments that will benefit from procuring AI technology. PPP also calls for wider procurement strategies to become more integrated in order to reflect the multifaceted, cross-departmental impact of new technologies.
On 11 October, 2023, Public Policy Projects (PPP) hosted a roundtable in partnership with Siemens Healthineers, Deloitte, and Mills and Reeve, entitled Implementing AI in Imaging Diagnostics. The roundtable was chaired by Dr Rizwan Malik, Consultant Radiologist at Bolton NHS Foundation Trust, and explored the challenges and opportunities relating to AI adoption in imaging diagnostics in the NHS. Along with our partners and the chair, this roundtable was attended by NHS and other experts in radiology, system transformation, innovation, and AI.
The report also highlights how information governance (IG) strategies should be joined up across NHS health systems. The IG landscape across the NHS is currently highly fragmented, with different NHS organisations often having different understandings of what is or is not legal. Rather than a single, or even a small handful of IG rules and guidelines, NHS employees and commercial organisations alike are in effect working 10s, 100s, even thousands of assorted IG rules. A unified IG approach across the NHS, or at the very least, across individual integrated care system (ICS), is highly desirable to help streamline information management and reduce administrative burdens.
The report also argues that individuals in trusts across multiple departments should work together to form multidisciplinary (MDT) AI implementation teams. Where AI strategies don’t yet exist, proactive AI MDTs can also play an instrumental role in shaping the future direction of AI in the NHS. Further, by setting up MDTs for AI in advance of new technology adoption, systems can align new acquisitions to trust, ICS, and NHS priorities.
- Industry advocates and early AI clinical adopters should assure sceptical colleagues that AI in imaging diagnostics is a supplemental tool meant to assist clinicians, not replace them. Demonstrating AI’s utility either through a pilot programme or other clinical case studies can help show the capabilities of imaging AI. In radiology in particular, there is an acute, long-term workforce shortage, so AI will provide an invaluable capacity boost to ease workforce pressures.
- AI vendors should strive to provide clearly defined case studies for the use of their technology in a clinical setting. New technologies, including AI, should solve specific problems, and vendors should work collaboratively with clinicians to identify which problems their technology mitigates. Clearly defined, specific use-cases will help drive clinical trust in AI’s abilities.
- To help ensure their products are clinically applicable, AI vendors should strive to have clinical input at every stage of the product lifestyle where practical, up to and including product implementation. This includes properly compensating clinicians for their advice and input. This clinical input should come in addition to, not in place of, vital advice provided by Clinical Safety Officers.
- NHS trusts, hospitals and ICSs should set up MDTs to help successfully deploy and scale diagnostic imaging AI. These MDTs should be formed in advance of new technology adoption to ensure workplace readiness.
- The NHS should recognise innovation adoption as valid and appropriate work in a clinical job plan. Without explicit allowance for innovation adoption within regular job specifications, only the most passionate employees will push for new tech adoption as doing so in the current environment generally means working overtime or otherwise outside of day-to-day clinical duties.
- It is essential that NHS IT teams are consulted from the very beginning of AI adoption processes. AI is highly taxing on IT systems, and IT experts have the requisite domain expertise to advise on the necessary technical steps for adoption and enabling technical integration with individual trust’s EPRs.
- The Department of Health and Social Care, NICE, Medicines and Healthcare products Regulatory Agency, and other regulatory agencies should create a strategy for the use of synthetic data to train AI models. Clear rules and guidelines on its use will help build trust in AI and will give industry confidence about which datasets can or cannot be used to train AI models.
- Medical indemnity for AI technologies must be clarified. Currently, risk is typically borne by the Medical Director or Clinical Director at the trust level, but as AI becomes more seamlessly integrated into clinical practice the line between clinical negligence and device failure becomes more difficult to establish. In addition to clarifying the responsibility of clinicians using AI, the legal responsibility of AI suppliers for their technology must be made clear.
- ICBs should create clearly defined system-level digital procurement strategies which include AI. ICB level strategies will ensure scalability of technology across the system, and coherence in procurement. Similarly, ICSs should aim to create unified information governance (IG) rules, as current IG practices are highly fragmented leading to inconsistent practices and significant confusion.