Algorithm library

Multi-modality

Activity, cardio, sleep, respiratory, gait

Versioning

Tracked

Every release documented with change history

Validation

V3-aligned

Verification, analytical, clinical

Validated algorithms for digital endpoints

Validated Digital Endpoints
for Clinical Trials

Digital endpoints are more than sensor readings. They are validated, versioned, reproducible measures — built from wearable and connected-device data, aligned to the protocol, and documented end-to-end so they can stand up in regulatory review.

Reproducible · Versioned · Submission-ready

GAIT
HR
SLEEP
SPO₂

Digital Endpoint Workflow

Signal → Algorithm → Endpoint → Evidence

Will my step count from the protocol match what my watch shows me?
Your watch gives a consumer estimate. The clinical endpoint uses a validated algorithm — same input, but interpreted against the protocol.
That's the difference between a feature and a measurement.
Validated, not just collected Signal + algorithm + clinical context

What a Digital Endpoint Really Is

In clinical research, a digital endpoint is a defined, validated measurement derived from wearable or connected-device data — used to answer a clinical question. It is not the raw sensor stream, and it is not the consumer-facing summary the patient sees on their phone.

A digital endpoint is reviewable. A consumer wearable feature is not.

Related pages: Signal Harmonization · Regulatory-Ready Endpoints

Validated digital endpoint algorithms derived from wearable signal data

Why Digital Endpoint Programs Stall

Most digital endpoint programs stall not because the sensor is wrong, but because the algorithm, validation, or provenance layer is missing.

Consumer metrics in clinical roles

Vendor 'step count' or 'sleep score' isn't a validated clinical endpoint, even if it looks like one.

No algorithm versioning

When the math changes mid-study, results drift silently and re-analysis becomes impossible.

Missing validation evidence

Without verification, analytical, and clinical validation, the endpoint can't survive review.

Lost provenance

Values arrive without device, firmware, or algorithm context — leaving statisticians flying blind.

Endpoint-protocol mismatch

Algorithm produces a value, but the protocol defines the endpoint differently — forcing post-hoc work.

Black-box vendor algorithms

If you can't explain how the value was computed, you can't defend it in a submission.

An endpoint that can't be reproduced or defended is a metric, not a measurement.

Delve Digital Endpoints vs Consumer Wearable Metrics

Consumer wearable metrics

  • Vendor-defined, opaque algorithms
  • No version history
  • No validation documentation
  • Provenance not preserved
  • Not aligned to protocol definitions

Delve digital endpoints

  • Documented, transparent algorithms
  • Versioned with documented change history
  • Verification + analytical + clinical validation
  • Provenance preserved with every value
  • Aligned to the protocol's endpoint definition

An endpoint is only useful if it can be reproduced, reviewed, and defended.

What a Strong Digital Endpoint Includes

Endpoints that survive a regulatory review are built around a documented method — not just a number that came out of a sensor.

Strong digital endpoints look the same on day one as they do at submission — that's the whole point.

See related pages: Validation · Endpoint APIs · Analytics

Digital endpoint pipeline from signal to validated clinical measurement

Digital Endpoints Across Therapeutic Areas

Delve's digital endpoint library covers the modalities most clinical programs need, built on the same harmonization and validation infrastructure.

Activity & gait

Step quality, gait speed, daily activity volume, sedentary behavior.

Cardiovascular

HR, HRV, recovery metrics, BP trends, arrhythmia signal detection.

Sleep

Sleep duration, efficiency, staging, sleep continuity over time.

Respiratory

Respiratory rate, SpO₂ trends, breathing-pattern stability.

Metabolic

CGM-derived adherence, time-in-range, weight trends.

Behavioral

Engagement with eDiary, training, and protocol-required tasks.

All endpoints flow through the same harmonization, QC, and API layers — so behavior across modalities is consistent.

FAQ

Can you build endpoints specific to our protocol?

Yes. Studies frequently define their own endpoint logic on top of the underlying validated algorithms. The protocol-specific layer is documented and versioned alongside the base algorithm.

How are algorithms validated?

Each endpoint goes through verification (sensor measures correctly), analytical validation (algorithm computes correctly on labeled data), and clinical validation (metric reflects the intended outcome). All three layers are documented.

What happens when an algorithm needs to change mid-study?

Algorithms are versioned. New versions can be introduced for new cohorts while existing data stays on its original version — preserving reproducibility for analyses already in flight.

Move From Sensor Data to Defensible Digital Endpoints

Delve builds digital endpoints as a fit-for-purpose, validated, versioned, documented layer on top of harmonized wearable and connected-device data.

Book a Digital Endpoint Discussion

Explore Delve wearables capabilities →