PAULA · INTEL
AI GovernanceRevenue at Risk: $410K–$980K (est)ONC HTI-2 — Phased compliance

EHR-Embedded Predictive AI Now Carries Source-Attribute Disclosure Obligations

ONC's HTI-2 framework extends the HTI-1 Predictive DSI rules: certified Health IT must surface a defined set of source attributes for predictive decision-support interventions and implement intervention risk management practices. For hospitals, that means any AI scoring tool embedded in the EHR — sepsis predictors, readmission scorers, AI-flagged DRG suggestions — has a documentation surface that is now discoverable in payer disputes, OCR §1557 complaints, and state AG inquiries.

Published by the PAULA Intelligence TeamReviewed by a board-certified Physician Advisor
Affected Payers
  • · Medicare FFS
  • · Medicare Advantage
  • · Medicaid
  • · Commercial
Service Lines
  • · EHR / Clinical IT
  • · Coding & CDI
  • · Compliance
  • · Patient Safety
01 — Signal

Why PAULA flagged this

The HTI-1/HTI-2 source-attribute disclosure means every predictive AI in your EHR has a documentable provenance: training data summary, validation, intended use, performance, and known limitations. That provenance is exactly what plaintiffs, OCR, and payers will request when a denial, an adverse event, or a coverage decision is tied to an AI-influenced clinical record.

Recommended Action — Next 60 Days

Inventory every predictive DSI currently active in your EHR. For each, capture the vendor's HTI source-attribute disclosure and your local IRM (intervention risk management) record. Tag any DSI whose disclosure is incomplete — those are the AI tools most likely to surface in a §1557 or False Claims context.

Locked · Full Brief

Scenarios, defenses, P2P scripts, executive lenses, and payer impact unlock with any briefing tier.

02 — Denial Scenario PAULA is Watching

One high-probability pattern

Scenario · 01

Payer Cites an AI-Flagged DRG Suggestion in the Provider's Own EHR to Argue Coding Inflation

Payer Argument

During a DRG audit, the payer obtains EHR audit-trail evidence that the provider's CDI workflow surfaced an AI-suggested higher-weighted DRG before the coder finalized the claim. Payer argues the AI suggestion drove inflated coding without independent clinician validation, citing HTI-1 disclosure of the tool's intended-use scope as evidence the tool was used outside its bounds.

Defense

Maintain an IRM record for every coding/CDI AI tool: intended-use boundary, validation evidence, and required human-review attestation per case. CLIP should require coder sign-off language that the DRG selection is based on the documented clinical record and not the AI suggestion alone. Pull the HTI source-attribute disclosure for the tool and confirm the documented use case falls within intended scope.

P2P Framing

The DRG selection in this case was made by a certified coder reviewing the complete clinical record. The AI suggestion is a flag, not the determination. Per our HTI-1 source-attribute documentation, the tool is intended for [scope] and the coder applied independent judgment as recorded in the audit trail. Please identify the specific documentation element you believe is insufficient.

03 — Decision Layer

Two executive lenses

CMIO / Clinical IT

HTI-1/HTI-2 effectively requires a predictive-AI inventory with provenance. Build it once, govern it quarterly, and tie each entry to a named clinical owner. Tools without a clean source-attribute disclosure should be paused until the vendor produces one — the regulatory and litigation exposure of running an undocumented predictive DSI now outweighs the workflow value.

Compliance & Legal

The HTI disclosure surface is discoverable. Treat the predictive-DSI inventory and the IRM logs as litigation-ready documents from day one: version-controlled, dated, with named approvers. When an OCR §1557 or state AG inquiry lands, the absence of an IRM record is itself the finding.

Operationalize this brief

Choose your briefing tier.

One-time purchase · Delivered as a downloadable executive PDF with citations and CLIP-ready operational templates.

Assess your exposure
Source: ONC Health Data, Technology, and Interoperability rulemaking (HTI-1 / HTI-2). Predictive DSI source-attribute and IRM provisions apply to certified Health IT. Source Confidence: HIGH for HTI-1 text; MEDIUM for HTI-2 operationalization.
PAULA · Clinefficiency Pro — Real-time decision intelligence for UM, denial risk, and regulatory change.