CliniGA CRO

CliniGA CRO Blog

Wearable Updates: Linking Algorithm Versions to Measurements

Link wearable measurements to software and algorithm versions when assessing the effect of an update on study data.

Representative research photograph: Wearable Updates: Linking Algorithm Versions to Measurements
Photo: Linda Bartlett (Photographer) · License: Public domain · Cropped and resized.
Published
Updated

A wearable firmware update may generate different outputs from the same underlying movement. Recording only the device model may not explain which algorithm or configuration produced a measurement used in a clinical study.

Preserve the meaning of data

Link relevant device, application and processing versions to data provenance. Assess whether a change affects presentation alone or the production of measurements. Manufacturer release notes inform this assessment but do not automatically establish suitability for the study’s intended use. Plan separately for automatic updates that the research team cannot directly control.

Worked example

For example, an update changes how brief interruptions contribute to daily activity totals. Differences before and after the update might appear to reflect participant behaviour. The team links version-transition dates to participant records, refers comparability questions to relevant scientific experts, and separately checks the consistency of the technical transfer.

Review checks

  • Check that device and processing versions accompany relevant measurements.
  • Compare output definitions and known limitations when a version changes.
  • Define how reports distinguish data generated under mixed versions.

Acceptance and boundaries

Acceptance means the technical context of a measurement can be reconstructed and the change impact has a reasoned assessment. An update does not automatically invalidate data, but absence of impact should not be assumed. Document scientific suitability separately from transfer correctness. Preserve relationships between original outputs, any reprocessing and the rules used to create revised results.

Frequently asked questions

Is manufacturer approval sufficient?
Manufacturer information matters, but the effect on the specific study measurement and intended use needs separate assessment.
Can historical data be processed with a newer algorithm?
It may be considered for a defined purpose, while preserving traceability between raw data, earlier outputs and revised outputs.

Sources

  1. FDA — Electronic systems, records and signatures in clinical investigations
  2. MHRA — Guidance on GxP data integrity