When the Device and the Participant Disagree
Part of Decentralized Trials (DCT) Training: Connected Devices and Wearables.
About 2 minutes. Free, in your language, no account needed.
Both are data, and neither is automatically right
A participant reports sleeping badly and the wearable shows eight hours. A participant says they walked to the shops and the step count shows four hundred steps.
The instinct is to trust the device because it produces a number. That instinct is wrong often enough to be dangerous.
Consider the mundane explanations first
The device was not worn. It was worn on the other wrist. It was charging. Somebody else wore it. The participant carried the phone and left the watch at home.
Most disagreements resolve here, and they resolve fastest if you ask before you theorize.
Ask without implying doubt
"Help me understand the day" works. "The device says something different" puts the participant on the defensive and makes them less likely to tell you the useful thing.
They usually know what happened.
Record both values
Do not reconcile them into one. The participant-reported value and the device value are two observations, and the analysis may want either.
Adding your own conclusion as a third value is the thing to avoid.
A disagreement can be the clinical signal
A participant who reports feeling fine while the device shows a sustained resting heart rate rise is a safety question, not a data quality question.
Do not close it as a discrepancy. Escalate it.
Systematic disagreement is a study problem
One participant's mismatch is a participant. Twenty participants whose sleep data contradicts their diaries is the device, the algorithm, or the instructions.
Report the pattern, not just the instances. Nobody else has your view of the individual cases.
Write what you did
The query, the participant's explanation, whether the device was checked, and what was concluded. An unexplained discrepancy in the record is a question left open for an analyst who has less information than you do.