Public reference · Decision brief
Illustrative public reference—not client work or a results claim.
Is a digitally derived measure fit for the decision it must support?
A source-linked structure for deciding whether to advance, conditionally advance, pause, or stop a proposed measure for a defined clinical-investigation use.
Decision
Advance only when the evidence matches the context of use.
Start with one sentence: For [population and setting], use [technology and measure] to assess [concept of interest] as [endpoint role] for [specified decision]. Every gate below tests that sentence. A strong sensor or algorithm does not, on its own, validate the resulting measure or make it meaningful for patients.
Decision rule
Advance when all critical gates have sufficient evidence for the proposed use. Conditionally advance only when named gaps can be closed before the measure affects a consequential endpoint or claim. Pause when evidence is incomplete but recoverable. Stop when the construct, population, or intended decision is mismatched.
| Gate | Question the evidence must answer | Minimum record | Do not infer |
|---|---|---|---|
| 1. Meaning | Is the concept of interest clinically relevant and meaningful to the intended population? | Patient, caregiver, and clinician input; construct definition; target population and setting. | That continuous collection automatically makes a measure meaningful. |
| 2. Context | What exact role will the measure play in the investigation and decision? | Context of use; endpoint role; estimand or decision relationship; timing; success criterion. | That evidence for one use transfers to another. |
| 3. Technology | Does the digital health technology measure what it is intended to measure under expected conditions? | Verification; hardware and software version; calibration; operating range; data provenance. | That device availability equals measurement validity. |
| 4. Measure | Are analytical and clinical validity adequate for the population, setting, and intended inference? | Reference method; prespecified analyses; measurement error; reliability; sensitivity to change; uncertainty. | That correlation alone establishes validity or clinical usefulness. |
| 5. Use | Can intended participants and staff use the technology as required? | Usability, accessibility, training, adherence, missingness, burden, and subgroup performance. | That a laboratory result will reproduce in remote or real-world conditions. |
| 6. Change | Will updates, drift, failure, or missing data be detected before they distort an endpoint? | Version control; change assessment; error monitoring; missing-data strategy; escalation and revalidation plan. | That performance remains fixed after deployment. |
Evidence packet
- One-page context-of-use statement.
- Conceptual model connecting patient experience, concept, measure, endpoint, and decision.
- Technology verification and version record.
- Analytical and clinical validation plan with uncertainty.
- Usability, accessibility, burden, adherence, and subgroup evidence.
- Data-provenance, missingness, error, security, privacy, and change-control plan.
Decision record
- Decision owner
- One accountable role—not a committee name.
- Disposition
- Advance, conditionally advance, pause, or stop.
- Evidence relied on
- Record IDs, versions, dates, and source links.
- Conditions
- Specific gaps to close before the next use.
- Review trigger
- Date, new evidence, protocol change, technology update, or safety signal.
Source facts and limits
What the current FDA source adds
FDA’s August 2026 paper highlights considerations drawn from existing guidances: patient-meaningful and clinically relevant concepts, justification of measure selection, verification and validation, usability and human factors, sources of error, and evidence proportionate to intended use. The completed FDA/Duke-Margolis workshop materials add a structured set of statistical and measurement questions, including data provenance and missing data.
This reference is a screening and decision structure. It is not FDA guidance, regulatory advice, a validated instrument, or an endpoint-qualification determination. The workshop materials do not necessarily represent official FDA or HHS views or endorsement; any detailed synthesis requires speaker-level attribution.