Attributable, Legible, Contemporaneous, Original, Accurate

Part of Clinical Data Management and Integrity: Record Integrity and ALCOA+.

About 5 minutes. Free, in your language, no account needed.

Why this matters

Almost every data finding an inspector writes can be expressed in five words. They are not a mnemonic to recite; they are the five distinct ways a record can fail to support what it claims, and each fails differently.

Learning them as a list produces someone who can name them. Learning what each one is actually protecting produces someone who can spot a violation in a record that looks fine.

Attributable

Who did this, and who recorded it.

A value in a record with no identifiable author is a value nobody stands behind. The protection is accountability: if a measurement is questioned, there must be a person who can be asked.

Common failures: a shared login, so the system's record of who entered a value is a department rather than a person. An unsigned paper worksheet. A correction made by someone whose identity the system does not capture. A value entered by one person on behalf of another with nothing recording that.

The shared login is worth naming specifically because it is usually adopted for convenience by people who do not realize it destroys attributability for every record it touches.

Legible

The record can be read, and will still be readable later.

The obvious form is handwriting. The more common modern form is a record that is technically readable and not interpretable: an abbreviation only its author understands, a comment referring to "the earlier discussion," a value with no unit.

Legibility also extends through time. A record stored in a format that requires software no longer available is not legible, whatever it looked like when written.

Contemporaneous

Recorded at the time, or as close to it as the work allows.

This is the one most often violated with entirely good intentions. A coordinator who conducts six visits and writes them up at the end of the day has produced six records that are not contemporaneous, and the memory of the third visit is now reconstructed rather than recorded.

A record written later is not worthless. It is a different kind of evidence, and it must show when it was written and, where relevant, that it was written later. The failure is not lateness but lateness presented as timeliness.

Original

The first place the information was recorded, or a verified true copy of it.

The concept is harder than it sounds in an electronic environment, and it is the one that most often confuses. If a measurement is taken on a device that prints a strip, the strip is original. If the value is transcribed into a form, the form is a copy. If the device writes directly to a system, that record is original and the printout is the copy.

The practical requirement is that the original is retained, or that a copy is verified as true and complete and the verification is documented. Discarding an original after transcribing it, without that verification, removes the only thing that could confirm the transcription.

Accurate

The record says what actually happened.

This is the one everyone starts with and it is doing less work than people assume, because a record can be accurate and still fail the other four. A value that is correct, entered under a shared login, written up the following week from memory, with the original discarded, is accurate and worthless as evidence.

Accuracy is a property of the value. The other four are properties of the record as evidence. That distinction is the single most useful thing in this lesson.

In practice

A monitor reviews a participant's file. The blood pressure in the case report form reads 138/88 and matches the clinical note exactly.

Accurate. She looks further.

The clinical note is initialed with a set of initials belonging to two staff members at the site, and the delegation log does not distinguish them. Attributability fails: the record does not identify who took the measurement.

The note is dated three days after the visit, with no annotation that it was written retrospectively. Contemporaneity fails, and it fails invisibly, because the entry reads as a record of the day.

The value was read from a device that stores its own readings, and those stored readings have not been retained. Originality fails.

One correct value, three failures. Every one of them invisible to a check that compares the case report form against the note, because both say 138/88.

What this looks like when it goes wrong

A site uses a single shared system account for its research nurses, established years ago so that cover arrangements would not require account changes.

Every entry and every correction across four studies is attributed to that account. The site knows who did what, because its own staffing roster records who worked which days.

An inspector asks who entered a specific value on a specific date. The site can produce a roster showing three nurses on duty. It cannot say which of them entered it, and the system's audit trail records only the shared account.

The finding is not that a value is wrong. No value is wrong. The finding is that the site cannot attribute any record in four studies, which means every record is weaker evidence than it appears, and there is no remedy that reaches backwards.

Key takeaways