HTI-1 Falls Short with DSI and Predictive Decision Support Proposals

By David Bucciferro, Chair, EHR Association

In other installments of this five-part blog series on the EHR Association’s issues with ONC’s Health Data, Technology, and Interoperability: Certification Program Updates, Algorithm Transparency, and Information Sharing Proposed Rule (HTI-1), we discussed our overarching concerns, as well as concerns with the Insights Condition program, transition to USCDI v3, and Patient Requested Restrictions. In this installment, we examine the considerable issues we have identified with provisions related to Decision Support Interventions (DSI) and Predictive Models.

According to ONC, the existing scope and structure of the Health IT Certification Program are to enhance transparency around predictive decision support, with requirements to make transparent information regarding the authorship, bibliographic, and other kinds of “source attribute” information for evidence-based decision support and linked referential interventions. Noting that AI/ML in healthcare “is often best considered a form of decision support or ‘augmented intelligence’,” ONC says its goal with the proposed rule related to DSI and predictive models is to update the existing decision support criterion to directly include predictive decision support, inclusive of ML technologies.

From our perspective, while ONC is clearly attempting to leverage its certification program to help the FDA increase transparency related to decision alerts, there are other policy paths that could be taken that would not place the burden on health IT developers. Instead, regulations imposing transparency requirements should be applied by the FDA or HHS more broadly to those creating the decision alerts rather than the developers of EHRs, which serve primarily as alert delivery mechanisms. 

Duplicate Efforts and Other Concerns

What’s more, the burden inherent to the proposals about DSIs is significant and would result in duplicative work across software developers who rely on decision support content sourced from the same vendors. It also disregards the fact that many healthcare organizations create their own alerts; vendors have no oversight or responsibility for those decisions or the content underlying them. As such, given the magnitude of the proposed changes, we feel that the Dec. 31, 2024, compliance timeline is unrealistic.

We also have recommendations regarding specific proposed changes. For starters, we don’t support the proposal to replace “clinical decision support” (CDI) with DSI in the certification criterion, as “intervention” has other meanings in the healthcare space. We also recommend narrowing the definition of “predictive decision support intervention,” as some interventions are not conducive to source attributes or feedback gathering. We would also like to see greater clarity around how ONC defines “predictive.” 

Further, ONC’s proposal to require user-accessible source attribute information on all decision support interventions available within the EHR is unreasonably broad, considering the scope of what could fit within the definition of a DSI. Thus, we would like to see the definition of those DSIs impacted by the proposed rule narrowed to focus on interruptive alerts that are created by the developer.

Regarding the proposal to require health IT modules to communicate when an attribute is missing, there are many client- or third-party-created DSI about which the developer will not be able to provide source attribute information.

Finally, in terms of enabling feedback from users regarding DSI information displayed through the health IT module – which will be made available for export in a computable format – we recommend limiting this to interruptive alerts, as passive alerts can’t have associated user actions. Nor do we want to inadvertently encourage interruptive alerts with a negative effect on usability. In fact, absent standards and a target audience, we feel that this recommendation will have questionable use or value.

Multiple Concerns

Other blogs in this series delve into our concerns with three additional provisions of HTI-1: the Insights Condition program, the transition to USCDI v3, and Patient Requested Restrictions. 

By sharing the issues the EHR Association has identified in our review of HTI-1, our hope is that ONC will give full consideration to our concerns – which we share with other impacted stakeholders – as it moves forward with finalizing the rule to ensure its end goals can be achieved without unrealistic deadlines and without placing an undue burden on healthcare providers and health IT developers. 

Leave a comment

3 Comments

  1. HTI-1 Patient Requested Restrictions Risks Negative Unintended Consequences | EHRA Blog
  2. HTI-1’s Insufficient USCDI v3 Transition Timeframe | EHRA Blog
  3. HTI-1’s EHR Reporting Provision Puts EHR Developers in the Middle | EHRA Blog

Share your thoughts on this topic!

This site uses Akismet to reduce spam. Learn how your comment data is processed.

  • Categories

  • Follow EHRA on Twitter

  • Enter your email address to follow this blog and receive notifications of new posts by email.

    Join 198 other subscribers
  • Contact Us

    Kasey Nicholoff
    staff @ ehra.org

    Amanda Patanow
    Communications and Media
    ehracomms @ npccs.com