What Counts as Source, and Who Decides

Part of Clinical Data Management and Integrity: Source Data and the Case Report Form.

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

Why this matters

"Where is the source for this?" is the most common question in a monitoring visit and in an inspection, and a site that has to think about the answer has already revealed a problem.

Source is not a category things fall into naturally. It is a designation, made in advance, recorded in a document, and known to everyone who records data. Where it has not been designated, it gets determined retrospectively by whoever is asking, and their answer may not be the one the site would have chosen.

The definition that actually works

Source data is the information in original records, and certified copies of them, of clinical findings, observations and other activities necessary for the reconstruction and evaluation of the trial. Source documents are where that information first appears.

The operative phrase is reconstruction and evaluation. The test for whether something is source is whether you would need it to reconstruct what happened. That test resolves most arguments.

A nurse's note recording that a participant reported dizziness: needed to reconstruct the adverse event. Source.

A scheduling email confirming an appointment: not needed to reconstruct anything clinical, though it may matter for a protocol window question.

A scribbled calculation on a sticky note used to derive a dose actually administered: needed, and therefore source, whether or not anybody wants it to be.

That last case is the uncomfortable one and it is the one that produces findings.

The source data location document

Every study needs a document that says, per data point or per category, where the source lives. It goes by various names and its function is the same: it is the answer to the inspector's question, written down in advance.

It needs to cover the cases that are not obvious. Which fields are direct entry, where the electronic system is itself the source. Where device output is retained. What happens with data recorded by a participant at home. Which of two systems holds the source where both contain the same value.

It must be agreed with the site rather than issued to it, because the site is what makes it true. A document asserting that a value's source is the case report form, at a site that also keeps a paper worksheet, is wrong in a way that will be discovered at the least convenient moment.

Two originals of the same fact

The commonest and most avoidable source problem is a site keeping a parallel record for a direct-entry field.

It is done for good reasons: the site wants its own record, or its own clinical practice requires a note in the medical record, or a nurse prefers paper. The consequence is two durable records of the same fact, either of which could be the original, and which sometimes disagree.

When they disagree, there is no principled way to determine which is right. Both are contemporaneous, both are attributable, and the discrepancy is unresolvable in the way that matters: nobody can demonstrate which reflects what happened.

The fix is a decision made in advance, not a rule issued after. If the site's clinical practice requires a note in the medical record, the medical record is the source and the case report form is transcription. That is a fine answer. What is not fine is designating the form as source and letting the parallel record exist anyway.

In practice

A data manager preparing a study builds the source location document with each site rather than distributing a template.

At one site the conversation surfaces something no template would have. Blood pressure is measured by a device that prints; the nurse records the value in the hospital's electronic medical record because hospital policy requires it; and the value is then entered into the case report form.

Three records of one measurement. The document as drafted named the case report form as source, which was wrong twice over.

The agreed position: the printed output is retained as the original, the medical record entry is a clinical record made under hospital policy, and the case report form is transcription verified against the printout. Documented, agreed, and known to the nurse.

Two hours of conversation. The alternative is discovering at an inspection that three records exist, that the designated source was the last of the three, and that nobody had noticed.

What this looks like when it goes wrong

A coordinator records visit observations on a personal notepad during the visit, then writes the formal note afterwards from those jottings, then discards the notepad.

The notepad was the first durable record: the original. The formal note is a transcription. The original has been destroyed, and the note is now presented as though it were contemporaneous with the visit.

No value is wrong and no participant is harmed. The record nonetheless overstates its own status, and when the practice is described to an inspector, as it eventually is, the question becomes how many notes across how many studies were produced this way.

The practice was never hidden. Nobody had ever told the coordinator that the notepad was a source document, because nobody had thought about it.

Key takeaways