HIMSS Calls for Risk-Based Regulatory Approach to AI in the U.S.

HIMSS is calling for the United States to adopt a risk-based approach for governing the development, deployment and use of artificial intelligence in healthcare space.

HIMSS submitted its recommendations on March 14 in response to a request for information from the White House Office of Technology Policy (OSTP).

HIMSS reinforced the importance of providers having the right information to make informed decisions about the implementation of AI solutions. The letter also recommended mechanisms be in place to mitigate harmful bias and urged the government to take steps to ensure the United States workforce has the expertise to develop innovative AI solutions and deploy those solutions in a safe and effective manner.

Referencing HIMSS Global Policy Principles on Artificial Intelligence and guidance from a diverse array of HIMSS membership, HIMSS stressed the importance of healthcare subject matter expertise as part of the development of the White House AI Action Plan, referenced in President Donald Trump’s “Removing the Barriers to American Leadership in Artificial Intelligence” executive order from Jan. 23.

HIMSS indicated the following considerations should be included in the Action Plan:

• Policy frameworks for the use of AI in the health and human services sector should have guardrails that consider the unique use cases and risks associated with patient safety.
• The Action Plan should endorse a risk-based regulatory approach that weighs the different intended uses, risk levels, types (generative or adaptive) and potential patient impacts of AI when establishing requirements impacting the development and monitoring post-deployment of AI.
• The Action Plan should ensure appropriate feasibility and safety testing during the development cycle of AI to ensure the AI tool will produce comparable and consistent results against intended outcomes in all appropriate settings and not perpetuate or introduce harmful biases.
• HIMSS encourages OSTP to convene clinician leaders to determine what information is needed for providers to make informed decisions for adopting AI tools using a nimble approach, as the requirements for “explainability” will adjust as health systems increase the sophistication of their understanding of AI performance over time.
• The Action Plan should require thoughtful action and monitoring by all parties, from market suppliers to hospitals and providers, to ensure AI solutions are generating appropriate outputs that do not put patients at risk once those tools are deployed in healthcare settings.
• The Action Plan should require that AI data governance and stewardship models are developed and regularly updated to promote the authorized use and disclosure of data.
• The Action Plan should drive consensus-based development and harmonization of standards to ensure AI technologies can appropriately exchange data and learnings across different healthcare systems using different technologies and be safely integrated the exchanged information into existing clinician workflows.
• The Action Plan should ensure that AI data governance and stewardship models are developed and regularly updated to promote the authorized use and disclosure of data.

HIMSS also acknowledged that the workforces of many healthcare organizations do not have the competencies to assess safety, accuracy and appropriate use of AI at the present time. Without those competencies and clear analytics strategy, similar to the requirements for a Stage 7 validation on HIMSS Analytics Maturity Adoption Model, the organization should carefully consider not adopting technologies they cannot safely manage. The Action Plan should support adding the necessary skillsets and knowledge for a clinical and IT workforce focused on testing, monitoring and revalidating AI following deployment in the field.