ONC Drops 14 EHR Certification Criteria and the AI Model Card Rule. Information Blocking Changes Wait.
The HTI-5 final rule, effective Dec. 28, removes 14 of 60 EHR certification criteria, not the 34 proposed, and drops AI source-attribute requirements.
The Office of the National Coordinator for Health Information Technology (ONC) has removed 14 of the 60 criteria in its health IT certification program and revised seven, in a final rule published in the Federal Register on Oct. 28 that takes effect Dec. 28.
The rule, known as HTI-5 and titled "ONC Deregulatory Actions To Unleash Prosperity," finalizes part of a December 2025 proposal. Certification matters to physicians because the certified EHR technology definitions that CMS uses for the Medicare Promoting Interoperability Program and the MIPS Promoting Interoperability category are built from these criteria. ONC's own 2024 survey data, cited in the rule, show 91 percent of physicians report having a certified EHR.
How much of the proposal survived?
Less than half of the removals. By The Script Pad's count from the rule's summary, ONC proposed in December 2025 to remove 34 of the 60 criteria and revise seven. The final rule removes 14, revises the same seven and leaves 39 unchanged, so 20 proposed removals, about three in five, did not make it into this rule.
Removed criteria include the already-expired clinical decision support criterion, family health history, the two security tags criteria, care plan, the quality measure filter, patient health information capture and automated numerator recording. Some removals tied to CMS program definitions take effect Jan. 1, 2027. ONC kept the implantable device list criterion to support a new Promoting Interoperability measure on unique device identifiers, did not finalize removal of the privacy and security criteria, and kept public health reporting criteria, including cancer registry and electronic case reporting. The electronic prior authorization and real-time prescription benefit criteria added last year are untouched. Planned removal of Direct messaging and C-CDA document exchange requirements was pushed to Jan. 1, 2030.
What happens to AI transparency?
It leaves certification. The rule removes the "source attribute" requirements, often called model cards, from the decision support interventions criterion, along with its intervention risk management requirements. The rule refers to these tools as "super intelligence" (SI) enabled, a term it draws from a Sept. 29 executive order. Commenters warned the change could reduce transparency for underrepresented patient groups. ONC responded that it was "not aware of any publicly available evidence" that the transparency requirements improved patient care, and encouraged users to discuss quality and performance with developers before deploying predictive tools. ONC also removed the criterion's reference to drug-drug and drug-allergy contraindication checking, saying retained requirements for medication and allergy data still support those checks.
The certification label will now say less about how a predictive tool was built and tested. Practices that want that information will have to ask for it.
What about information blocking?
Not yet. The final rule covers only the certification program and the Trusted Exchange Framework. ONC says comments on proposals it did not address "remain under review and consideration" and may appear in later rules. Healthcare Dive reported that ONC postponed its information-blocking proposals, and quoted national coordinator Dr. Tom Keane saying HTI-5 takes "decisive action to significantly lower the administrative compliance burden for developers." ONC estimates savings to developers with a present value of $1.55 billion, and puts their implementation and compliance cost at $0.
What practices should do now
Nothing changes in the exam room on Dec. 28, and ONC says it expects several removed functions, such as patient health information capture, to remain in products. But certification no longer guarantees them. Practices renewing EHR contracts should list the functions they rely on, such as family history capture or patient-entered data, and get them in writing. Groups using predictive decision support should ask vendors for the training, validation and performance information that model cards used to carry, and should watch CMS's 2027 rules for matching changes to the certified EHR definition. This site has covered FDA's final guidance on prescription drug apps and the push for electronic prior authorization in self-funded plans.
Reported, fact-checked and edited by The Script Pad's editorial staff. See How we use AI.
This article is for professional education and does not replace clinical judgment. Treatment decisions should be based on the individual patient and current guidelines.
