Findings
Digitally derived measures generated from digital health technologies, including AI-enabled tools, enable continuous remote health tracking outside traditional clinical settings. These measures can serve as clinical outcome assessments, biomarkers, or multicomponent endpoints. To be fit for purpose, digitally derived measures require rigorous sensor verification alongside analytical and clinical validation. Identifying potential sources of measurement error—such as sensor noise, user interactions, software updates, and environmental factors—is critical to maintaining validity throughout a clinical investigation.

Recommendations
Developers and sponsors should engage patients, caregivers, and clinicians early to ensure measures capture meaningful aspects of health and reflect relevant clinical experiences. Conceptual models should be used to clearly define the concept of interest and establish rationale relative to target patient populations. Sponsors must ensure device instructions are clear through human factors and usability studies, evaluating user burden, adherence, and calibration requirements. Verification data should be established or leveraged from third parties, and appropriate analytical validation must be conducted in target populations before clinical validation.

Regulatory Considerations
The FDA applies a risk-based approach to evidence requirements, tailoring validation expectations to the measure’s intended context of use and role in clinical investigations. For artificial intelligence and machine learning models, developers should consider credibility assessments aligned with proposed agency frameworks. Sponsors are strongly encouraged to engage early with the relevant FDA review division before incorporating digitally derived measures into medical product development protocols.