NICE trumpets new standard for health AI evaluations

With artificial intelligence technologies for healthcare on the rise, UK health technology assessment (HTA) agency NICE has published an economic evaluation standard that it says will help inform coverage and reimbursement decisions.

CHEERS-AI – or Consolidated Health Economic Evaluation Reporting Standards for Interventions That Use Artificial Intelligence – is an attempt to arrive at a checklist of factors with potential implications for cost-effectiveness that should be considered when HTAs appraise health AIs.

It comes as AI is gaining increased traction in healthcare, with applications ranging from diagnosing patients, handling and transcribing medical data, and improving administrative efficiency.

According to NICE, it aims to improve the transparency, reproducibility, and quality of decision-making, plug a gap between the rapid pace of development of AI and HTA practices, and ultimately help patients gain access more quickly to the most promising technologies.

The standard has grown out of the Next Generation Health Technology Assessment (HTx) project, funded by the EU, which aims to modernise the framework deployed by HTA agencies when making access and reimbursement judgements for health technologies.

It is an extension of the 2022 CHEERS standard already used in health technology reviews by NICE and other HTAs, with additions specifically tailored to AI.

That includes assessment of user autonomy – in other words, whether the healthcare worker retains the ultimate care decision – as well as how the AI learns over time and the information on how the AI component was developed, including training data.

At the moment, HTA assessments of health AIs are compromised by poor quality input data, author conflicts of interest, a lack of transparent reporting, and unclear information about AI functionality, according to a NICE blog post.

"We realised that setting out specific reporting standards for economic evaluation studies of AI interventions would help improve transparency in this emerging field and increase their usefulness for health and care system decision-makers," it said. "This is crucial for reimbursement decisions about this new field of healthcare technology."

The checklist has been endorsed by ISPOR, a non-profit organisation that represents health economics and outcomes research (HEOR) professionals, which says it should be used "when reporting an EE of an intervention that uses AI to perform its function."

Photo by BoliviaInteligente on Unsplash