Adherence Is a Measurement Problem

Part of Research Nurse (RN) Training: Participant Education, Adherence and Retention.

About 2 minutes. Free, in your language, no account needed.

Why this matters

If you do not know how much drug a participant actually took, you cannot interpret what happened to them.

Adherence is not a virtue to be assessed. It is a variable to be measured, and measuring it badly is worse than not measuring it.

How it is measured

Returned product counts. Tablets brought back, counted against tablets dispensed.

Diaries. What the participant recorded at the time.

Device data, where the product records its own use.

What they tell you, which is often the most accurate and the least trusted.

Each has weaknesses. Counts are defeated by tablets thrown away. Diaries get filled in at the last minute. Self-report is shaped by how you ask.

Ask in a way that permits the truth

"Have you been taking it as prescribed?" invites yes.

"Most people miss a few. How many do you think you have missed in the last two weeks?" invites a number. The second question produces better data and does not imply fault.

Normalize the gap and you will hear about it.

Record what you find, not what you hope

If the count says four tablets unaccounted for, the record says four. Not zero because the participant explained it.

Record the count and the explanation separately. "Four not returned; participant reports leaving them at a relative's house" is complete. Either half alone is not.

Bring back the packaging

Participants forget. Remind them at the previous visit and again by phone.

An unreturned bottle is a missing measurement for that whole period, and it cannot be reconstructed.

In practice

A participant returns with more tablets than expected, meaning they missed several doses.

You do not tell them it is fine. You do not tell them off. You count, record, ask what happened, record that too, and tell the investigator whether it crosses any protocol threshold.

Missed doses are a finding, not a failure of the participant.

What this looks like when it goes wrong

A site records everyone as fully adherent because counts are roughly right and nobody wants to flag a participant.

The drug appears less effective than expected. Nobody can tell whether that is the drug or the dosing, because the adherence variable contains no variation and therefore no information.

Key takeaways