CIFoundry

ONC HTI-1: what the EHR decision-support rules require, and what HTI-5 changes

HTI-1 is a rule from the Office of the National Coordinator for Health IT (now ASTP/ONC). It sets requirements for certified health IT, mainly EHRs, including how they handle decision support and predictive AI.

It applies to EHR developers who certify their software, not directly to hospitals or to standalone AI vendors. It still shapes what you can ask your EHR vendor. This is general information, not legal advice.

Last reviewed November 2026 · Independent and vendor-neutral

What HTI-1 set up

Decision support interventions (DSI)

  • HTI-1 (final rule published January 2024) replaced the old CDS certification criterion with a "decision support interventions" criterion, 45 CFR 170.315(b)(11). Developers had to meet it by the end of 2024.
  • It separates evidence-based DSIs (rules, guidelines) from predictive DSIs (models that learn from data to produce a prediction, score or recommendation).
  • For predictive DSIs the EHR developer supplies, it required "source attributes": a model-card-style list covering intended use, training data, fairness testing, validation and ongoing maintenance, which users could review.
  • It also required intervention risk management practices for those predictive DSIs, with summaries made public.

What HTI-5 changes

Key points

  • The HTI-5 final rule, published October 28, 2026, takes effect December 28, 2026.
  • It removes the source attribute (model card) requirements and the intervention risk management requirements for predictive DSIs from the certification criterion, and narrows the data the remaining DSI requirements cover.
  • Other obligations remain. Under the Affordable Care Act section 1557 rule (45 CFR 92.210), covered health programs must make reasonable efforts to identify and reduce discrimination risks from patient care decision support tools they use.

What this means for buyers

Ask your EHR vendor

  1. “Which predictive models are built into our EHR, and will you keep publishing the model information you produced under HTI-1?”
  2. “How were these models validated, and on which populations? How often are they re-checked after go-live?”
  3. “Can we switch off or configure each predictive alert, and see its performance at our site?”

Ask standalone AI vendors

  1. “HTI-1 never applied to you directly. Will you give us the same information anyway: intended use, training data, validation, fairness testing and update process?”
  2. “How do you help us meet our own section 1557 duty to review decision support tools for discrimination risk?”
Whatever the tool does, the 12 core questions (evidence, data rights, liability, monitoring, contracts) also apply.

Further reading

More on rules & regulations

Last reviewed November 2026. Educational information, not medical, legal or regulatory advice, and not advice about any specific tool or patient. Rules change; check the linked official sources.

No vendor reviewed, sponsored or edited this guide. Listing is free; no one can pay to change the order.