See what Trussed catches that your current tool misses, live in your stack

    No migration, no commitment, just a direct comparison in your environment.

    Set up a technical evaluation
    ONC HTI-1 Rule

    What Is a Decision Support Intervention (DSI)? ONC Guide

    A Decision Support Intervention (DSI) is ONC's HTI-1 term for any technology within certified health IT that produces a prediction, classification, recommendation, evaluation, or analysis to inform a clinical or operational decision. HTI-1 divides DSIs into evidence-based DSIs, which reference clinical guidelines or literature, and predictive DSIs, which use algorithms or AI/ML models. Predictive DSIs carry additional transparency obligations, including disclosure of defined source attributes to end users.

    DSI Under HTI-1 at a Glance

    DSI Definition

    Replaces the older CDS term, covering clinical and operational decision support in certified health IT.

    Evidence-Based vs Predictive

    Guideline-driven logic is separated from algorithm or AI/ML-driven predictive DSIs.

    Source Attributes

    Defined disclosures certified developers must make available for predictive DSIs.

    Governance Obligation

    Certification compliance sits with developers; usage risk oversight sits with deploying organizations.

    How ONC Defines a Decision Support Intervention

    A Decision Support Intervention is the term ONC's HTI-1 rule uses for any technology within certified health IT that produces a prediction, classification, recommendation, evaluation, or analysis intended to inform a clinical or operational decision. It replaces the earlier Clinical Decision Support terminology, and it intentionally covers a broader scope: decision support used for population health and operational purposes falls under the DSI definition, not only direct clinical care at the point of service.

    Evidence-Based DSI vs Predictive DSI

    HTI-1 separates DSIs into two categories. Evidence-based DSIs reference established clinical guidelines or published literature to generate their output. Predictive DSIs instead rely on algorithms or AI/ML models. This distinction carries direct compliance consequences: predictive DSIs are subject to defined source attribute disclosure requirements that evidence-based DSIs are not, since evidence-based logic is grounded in guidelines rather than trained models.

    AspectEvidence-Based DSIPredictive DSI
    Underlying basisClinical guidelines or published literatureAlgorithms or AI/ML models
    Source attribute disclosureNot subject to this requirementMust disclose defined source attributes to end users

    Operational and Governance Implications

    Responsibility for DSI compliance is split between two parties. Certification compliance for source attribute disclosure rests with the certified health IT developer. The deploying provider organization retains responsibility for oversight of how predictive DSI outputs are actually used and interpreted in practice. For teams implementing certified modules, this split has direct architectural consequences for how predictive DSI logic is built, tracked, and audited.

    Architectural Implications for Predictive DSIs

    Predictive DSIs introduce technical requirements that reach beyond the recommendation logic itself.

    1. 1

      Metadata storage and versioning

      Modules need a mechanism to store, version, and surface source attribute metadata alongside each predictive DSI output.

    2. 2

      Third-party model traceability

      When a predictive DSI relies on an externally hosted or third-party model, the system must still trace and expose the required source attribute data.

    3. 3

      Logical separation of DSI types

      Predictive DSI logic should be architecturally separable from evidence-based DSI logic so the correct transparency obligations apply to each.

    4. 4

      Audit logging

      DSI outputs and their associated source attributes should be logged to support compliance verification and incident review.

    5. 5

      Access control differentiation

      Runtime access controls should distinguish who can view, modify, or override predictive DSI recommendations within the module.

    Procurement Questions for Predictive DSI Compliance

    Governance and procurement teams evaluating certified health IT should be able to get clear answers to the following before adoption.

    • Which decision support features in this certified health IT module are classified as predictive DSIs versus evidence-based DSIs?
    • Can the vendor provide complete source attribute documentation for each predictive DSI, including its development and validation basis?
    • How and when are source attribute disclosures updated when a predictive model is retrained or replaced?
    • What mechanisms let end users access source attribute information at the point of care or operational use?
    • How does the vendor's certification documentation map to the specific ONC certification criterion governing decision support interventions?

    Frequently Asked Questions

    Is a DSI the same as traditional Clinical Decision Support (CDS)?

    Not exactly. DSI is the term ONC's HTI-1 rule uses in place of CDS, and it covers a broader scope, including decision support used for population health and operational purposes, not only direct clinical care within certified health IT.

    Does every predictive DSI need to disclose source attributes?

    Under HTI-1, predictive DSIs in certified health IT are subject to defined source attribute disclosure requirements. Evidence-based DSIs are treated differently, since they rely on clinical guidelines rather than algorithmic models.

    Who is responsible for source attribute compliance, the vendor or the provider organization?

    Certification compliance for source attribute disclosure rests with the certified health IT developer. The deploying provider organization retains responsibility for oversight of how predictive DSI outputs are used and interpreted.

    Extending Oversight to AI-Driven Decision Support

    Predictive DSIs introduce runtime behavior that certification alone does not fully govern. Understanding how source attribute transparency fits into your broader AI oversight approach is a practical next step for governance leaders working with certified health IT.

    Explore Runtime Governance