Source Records in a Clinical Setting

Part of Principal Investigator (PI) Training: The Data You Sign For.

About 4 minutes. Free, in your language, no account needed.

Why this matters

Physicians already keep clinical records to a professional standard. Research records are held to a similar standard for a different purpose: not only to support the care of one patient, but to allow a third party, years later, to reconstruct exactly what happened and verify that the reported data is true.

The most common source finding at a site is not fabrication. It is a medical record that was perfectly adequate for clinical care and does not support the data that was reported from it.

The requirement

Section 2.12.1 makes the investigator responsible for ensuring data integrity when data are generated, recorded and reported, whatever the media used. Section 2.12.2 asks for adequate source records — records that are attributable, legible, contemporaneous, original, accurate and complete, with changes traceable, not obscuring the original entry, and explained where necessary.

The guideline's Glossary sets out data-integrity attributes directly. Note that the familiar acronym ALCOA and its variants appear nowhere in E6(R3). They are useful industry shorthand for the same idea, and the attributes in the Glossary are the requirement.

Deciding what source is, before the trial

Section 2.12.2 also asks that the source records, how they are captured and where they are located be defined before the trial starts, and updated as needed — and that unnecessary transcription be avoided.

That defining exercise is a source data location document, and it is worth thirty minutes of your time at start-up rather than the coordinator's alone. For each data point the protocol requires, three questions:

The transcription question is the one that pays. Every copy is an opportunity for an error and a thing to verify. Where data can be captured once, capture it once.

The clinical-record collision

Research at a hospital or clinic runs into a structural problem: the medical record belongs to the institution and serves clinical care, while the trial needs specific things recorded in specific ways. Four recurring conflicts, with the practical resolution:

ConflictResolution
Trial visits not distinguishable from routine care in the notesRecord the protocol visit identifier in the clinical note
Consent taken but the medical record does not show itA consent note in the medical record, dated, referencing the version used
Adverse events described clinically, without the elements the protocol needsA structured event note: onset date, term, severity, action taken, outcome
Worksheets used as source, with the medical record silentDecide which is source and say so in the source data location document — do not leave it ambiguous

Deciding these once, at start-up, prevents the situation where a monitor finds the trial and the clinic telling different stories about the same afternoon.

Corrections, and the habit that survives inspection

Section 2.12.6 governs corrections: dated, attributable, not obscuring the original, and explained where the reason is not obvious.

On paper: a single line through the error, the correction beside it, initials and date, and a short reason where needed. Never overwrite, never use correction fluid, never erase. In electronic systems the audit trail does this automatically — which is fine, provided nobody works around it by having one person enter data under another's login. That is a §2.12.10 breach and it is treated as serious.

Two habits carry most of the weight. Contemporaneous means the same day; a note written a week later is not wrong but it is worth much less, and it should say when it was written. Attributable means an inspector can tell who made every entry — which is why the delegation log from Module 2 and the source records have to agree.

Key takeaways