Predictive algorithms have quietly become part of routine care delivery — sepsis alerts, readmission-risk scores, no-show predictions — often without clinicians knowing exactly how those numbers were generated or how reliable they are. On April 11, 2023, the Office of the National Coordinator for Health Information Technology (ONC) released a proposed rule aimed squarely at that gap. Published in the Federal Register on April 18, 2023 (88 FR 23746) as “Health Data, Technology, and Interoperability: Certification Program Updates, Algorithm Transparency, and Information Sharing,” and referred to throughout the industry by its shorthand, the HTI-1 rule, the proposal would extend transparency and risk-management expectations to predictive decision support built into certified electronic health record (EHR) technology, while also updating data-sharing standards and information-blocking regulations.

It is important to be precise about where this stands: as of this writing, HTI-1 is a proposed rule, not a final one. ONC accepted public comments through June 20, 2023, and under the federal rulemaking process a proposed rule can be revised, narrowed, delayed, or in some respects dropped before anything becomes a binding legal requirement. Nothing described below is currently in effect as a result of this specific NPRM (Notice of Proposed Rulemaking); it describes what ONC has proposed, not what health IT developers or providers are currently required to do. Readers making compliance or purchasing decisions should consult the primary source materials linked throughout, along with qualified regulatory counsel, since this is a regulatory summary, not legal advice.

What HTI-1 Is and Where It Sits in the Cures Act Framework

HTI-1 is ONC’s latest rulemaking implementing provisions of the 21st Century Cures Act, the 2016 law that directed federal regulators to promote health information interoperability and combat information blocking. ONC has used a series of numbered “HTI” rules to build out the technical and policy infrastructure the Cures Act called for, and HTI-1 is the first in that series, following on from the 2020 ONC Cures Act Final Rule that first established information-blocking regulations and the current ONC Health IT Certification Program structure.

The proposed rule is broad, touching several largely distinct areas of health IT policy under one regulatory vehicle:

  • New transparency requirements for predictive algorithms embedded in certified health IT
  • A revised, expanded baseline for the data classes certified systems must be able to exchange
  • Modifications to information-blocking exceptions
  • A new “Insights Condition” reporting requirement for certified health IT developers
  • A shift toward more frequent, incremental certification criteria updates rather than periodic large “editions”

Each of these pieces would, if finalized, primarily bind health IT developers whose products are certified under the ONC Health IT Certification Program — the certification that most hospital and ambulatory EHR vendors maintain because federal programs and many state and payer requirements reference it.

Proposed Transparency Rules for Predictive Decision Support

The most closely watched piece of HTI-1 is its proposal to overhaul how certified health IT handles clinical decision support. ONC proposed retiring the older “clinical decision support (CDS)” certification criterion in favor of a new, broader category called “decision support interventions,” or DSI, and within that category, ONC drew a specific distinction for predictive DSI — decision support whose logic is derived, in whole or in part, from trained models or statistical/algorithmic methods applied to data, as opposed to a fixed set of rules that clinical experts author directly.

For predictive DSI specifically, ONC’s proposal would require certified health IT to make a defined set of information — often referred to informally as “source attributes” — available to users. As proposed, that information is intended to help clinicians, informaticists, and health system leaders judge whether a given predictive tool is what ONC’s materials describe using the shorthand FAVES: fair, appropriate, valid, effective, and safe. Proposed categories of source attribute information include things such as:

  • A description of the intervention’s purpose and intended use
  • Information about the data used to develop and/or validate the model
  • Details on how the intervention was evaluated for validity and performance
  • Ongoing risk-management and monitoring practices the developer applies
  • Whether and how bias was assessed

ONC’s proposal would also require developers to attest to having risk-management practices in place — encompassing risk analysis, risk mitigation, and governance — for predictive DSIs included in their certified products, rather than leaving that entirely to individual health systems to assess after the fact. The intent, as described in the proposal, is not for ONC to certify that any specific algorithm is accurate or unbiased, but to standardize what information is disclosed so that the organizations deploying these tools can make that judgment for themselves.

Why Predictive Algorithms Drew Regulatory Attention

The proposal reflects a broader concern that has built over several years: as more health systems adopt AI- and machine learning-driven tools embedded in the EHR, there has been no consistent, government-recognized baseline for what information a hospital, clinic, or clinician is entitled to see about how those tools were built, tested, and maintained. Absent that baseline, the same predictive tool might be extensively documented by one vendor and treated as a “black box” by another, with users given no consistent way to compare them or watch for the kind of demographic bias that has drawn scrutiny in health care algorithms generally.

USCDI Version 3: A Larger Baseline Data Set

Separately from the algorithm-transparency provisions, HTI-1 proposed to update the United States Core Data for Interoperability (USCDI) — the standardized set of data classes and elements that certified health IT must be able to exchange — from its current baseline to USCDI version 3. As proposed, ONC would set an expiration date of January 1, 2025, for the older USCDI v1 baseline, meaning Health IT Modules certified to data-exchange criteria would need to demonstrate they could handle the USCDI v3 data set by that date.

USCDI v3 is a meaningfully larger data set than earlier versions, adding data classes such as:

  • Sexual orientation and gender identity
  • Functional status
  • Disability status
  • Mental/cognitive status
  • Social determinants of health (SDOH) elements such as housing, food, and transportation needs

The practical effect, as proposed, is that certified EHRs and related health IT would need to be able to structure, store, and exchange a wider range of data relevant to health equity and whole-person care than prior USCDI baselines required — data many health systems have historically captured inconsistently or only in free text.

Proposed Information-Blocking Exception Updates

HTI-1 also proposed several changes to the information-blocking regulations that ONC finalized in 2020 under the Cures Act. Information blocking refers to practices by health IT developers, health information networks, or health care providers that are likely to interfere with the access, exchange, or use of electronic health information (EHI), unless an exception applies. As proposed, HTI-1 would:

  • Revise the infeasibility exception to require a clearer, more direct causal connection between an uncontrollable event and an actor’s inability to fulfill a request for EHI, rather than allowing more generalized justifications
  • Rename and expand what is currently called the “Content and Manner Exception” into a broader Manner Exception, including a new condition intended to give actors more certainty that fulfilling requests through participation in a Qualified Health Information Network under the Trusted Exchange Framework and Common Agreement (TEFCA) can satisfy the exception
  • Clarify the definition of “health IT offering” to exclude certain customary provider activities, such as subsidizing electronic health record access for smaller or safety-net-affiliated practices

ONC’s stated goal for these changes is to reduce ambiguity that the agency says has made some actors reluctant to share data even in cases the exceptions were meant to cover, while also creating a clearer on-ramp toward TEFCA participation as that nationwide exchange framework matures.

A New “Insights Condition” for Certified Developers

Another major piece of the proposal is what ONC calls the Insights Condition — a new Condition and Maintenance of Certification requirement, authorized under the Cures Act’s EHR Reporting Program provisions, under which certified health IT developers would have to report standardized metrics about how their certified products are actually used in the field. As proposed, this reporting would cover areas such as patients’ electronic access to their health information, clinical data exchange, and public health data exchange, giving ONC (and, in aggregate or de-identified form, the public) a clearer picture of real-world interoperability performance rather than relying solely on certification testing done at a point in time.

Certification Program Structure and Other Proposed Changes

Beyond the headline items, HTI-1 proposed a number of more technical certification program updates, including:

  • Moving away from periodic, large “edition”-style certification updates toward a model of more frequent, incremental criteria updates
  • Revised certification criteria touching electronic case reporting (eCR) to public health agencies
  • Updates to “patient demographics and observations” certification criteria, aligned with the broader USCDI v3 data expansion
  • Proposed minimum refresh-token durations and other API-access technical requirements tied to the existing SMART App Launch framework

Collectively, ONC framed these as modernization steps meant to keep the certification program’s technical requirements from falling behind fast-moving standards work happening in the FHIR and broader health IT standards community.

What Happens Next

Because HTI-1 is a proposed rule, its next formal step after the June 20, 2023 comment deadline was for ONC to review public comments and decide how to proceed — potentially finalizing provisions as proposed, modifying them, dropping some entirely, or in principle pursuing further rulemaking. Readers should treat any description of “final” HTI-1 requirements circulating elsewhere as referring to a later stage of this process, not to the April 2023 proposal discussed here. For the authoritative, current text of the proposal, ONC’s own docket and fact materials — and the Federal Register notice itself — remain the primary sources of record.

Frequently Asked Questions

What is the HTI-1 rule?

HTI-1 is a proposed rule from ONC, published in the Federal Register on April 18, 2023, titled “Health Data, Technology, and Interoperability: Certification Program Updates, Algorithm Transparency, and Information Sharing.” It proposes AI/predictive-algorithm transparency requirements, a new USCDI v3 data baseline, information-blocking exception changes, and a new developer reporting condition.

Has the HTI-1 rule been finalized?

As of this writing, no. HTI-1 was published as a proposed rule (an NPRM) with a public comment period running through June 20, 2023. Proposed rules describe what an agency intends to require; they are not binding until a final rule is issued, and provisions can change materially in the process.

What is a “predictive DSI” under HTI-1?

Predictive DSI, or predictive decision support intervention, is a category ONC proposed for clinical decision support tools whose outputs are derived from trained models or algorithmic/statistical methods applied to data, as distinguished from decision support built on fixed, expert-authored rules.

Why does HTI-1 propose adding USCDI version 3?

ONC proposed USCDI v3 as a broader baseline data set for certified health IT, adding classes like sexual orientation and gender identity, functional status, disability status, and social determinants of health, to support more complete data exchange around health equity and whole-person care than earlier USCDI versions required.

Where can I read the official HTI-1 proposed rule text?

The proposed rule was published in the Federal Register on April 18, 2023, with supporting fact sheets and background materials published on healthit.gov.