Validating a sensor-based measure means comparing it with reference measurements, in the right people, at the right times. Capture holds the reference assessments, participant-reported context and device summary metrics in one study with aligned timestamps.
Free sandbox · No credit card · 21 CFR Part 11 aligned
| Verification | Analytical | Clinical | |
|---|---|---|---|
| Sensor output accuracy | |||
| Algorithm vs reference test | |||
| Relates to clinical state | |||
| Study with participants |
Key points
Study design
A digital biomarker validation study usually has three streams of data. First, reference measurements: a timed walk in clinic, a clinician-rated scale, polysomnography, a spirometry result. Second, the digital measure: summary metrics derived by the algorithm for defined time windows. Third, context: when the device was worn, what the participant was doing, how they felt. Validity is established by comparing the first two, with the third explaining the gaps.
Capture handles the reference measurements as site forms (performance tests are PerfO site forms), participant context as recurring questionnaires on their phones, and device summary metrics as structured numeric fields per session or day. Every record is time-stamped and attributed, so alignment in analysis is straightforward.
High-frequency raw sensor data usually stays in the device platform or a dedicated data lake. The study database should hold the metrics, time windows and identifiers needed for the analysis, and a clear link to where raw data lives. See EDC for wearable device studies.
Clinic 4-m walk
1.12 m/s
Sensor gait speed
1.08 m/s
Wear time
94%
Reference measures
Reference tests are run by site staff and recorded as site forms with range checks, units and assessor attribution. Participant-reported context comes from phone questionnaires with completion windows, so each session's data is complete.
Patient-reported
Participant, own phone
Observer-reported
Caregiver
Clinician-reported
Site staff, clinic tablet
Performance outcome
Site-run test, e.g. chair rise
Build a validation study skeleton in the free sandbox.
Study build
What the digital measure is supposed to capture, in which population.
Accepted reference for each aspect being validated.
How sessions are defined and matched.
Minimum wear time and data quality for a valid session.
Participant diaries for activities and symptoms.
See the wearable data consent template.
Examples
| Digital measure | Possible reference | Context to capture |
|---|---|---|
| Gait speed from a wearable | Timed walk test in clinic | Walking aids, pain, falls |
| Daily step count | Observed walking, activity questionnaire | Illness days, weather, device removal |
| Sleep duration and efficiency | Polysomnography or sleep diary | Bedtime routine, naps, medication |
| Tremor severity | Clinician-rated tremor scale | Medication timing, activities during recording |
| Cough frequency | Manually counted recordings | Environment, other sounds, symptoms |
| Heart rate variability | Clinical ECG | Posture, activity, caffeine, stress |
The right reference depends on the concept of interest and the intended context of use.
Wear adherence
A digital measure is only as good as the hours it covers. Participants take devices off to charge them, forget them after showering, or stop wearing them when they feel unwell, and those gaps are rarely random. Define in advance what counts as a valid day or session, monitor wear time during the study, and ask participants about the reasons for gaps in their context diary, so the analysis can handle missing data honestly.
Short context questionnaires on the participant's own phone, with a completion window and reminders, are the simplest way to capture that information. See the wearable data sharing consent template for how to explain wear expectations up front.
A digital biomarker is a characteristic measured by a digital device that indicates a biological or clinical state. A digital endpoint is such a measure used to assess treatment effect in a trial.
It depends on the analysis, such as agreement with the reference or sensitivity to change. Plan it with a statistician, as for any study.
A framework separating verification, analytical validation and clinical validation of digital measures from sensor-based technologies.
Capture stores structured summary metrics, reference assessments and participant context. Raw high-frequency data usually stays in the device platform.
As site forms completed by the assessor, with units, range checks and attribution.
Yes. FDA has issued guidance on digital health technologies for remote data acquisition in clinical investigations.
Yes, in the free sandbox.
Keep exploring
Reference, device and context data in one study. Free sandbox.