Direct Entry and the Disappearing Source

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

Direct entry, where the electronic form is completed with no prior record, is genuinely better for data quality. It removes transcription, which removes an entire class of error, and it removes the delay between observation and record, which improves contemporaneity.

It also moves the source document out of the site's filing cabinet and into a sponsor-contracted system, and that move has consequences that teams consistently fail to think through until the system is being decommissioned.

What changes when the form is the source

The system must have an audit trail from the first entry. Not from the first correction. A first-record field's initial value is source data, and the record of who entered it and when is part of the record's integrity.

The site must be able to access its own source data. During the study this is usually fine. After the study, when the sponsor's license ends, it frequently is not, and the site is left unable to produce source for its own participants.

Retention must outlive the system. Study data is retained for years, commonly far longer than a sponsor holds a license for a particular platform. If the retention plan is "the data is in the system," there is no retention plan.

The site cannot keep a parallel record. Which means it must be told, clearly, which fields are direct entry, or it will keep one for perfectly sensible reasons.

The question to settle before enrollment

For each direct-entry field, four things must be decided and written down:

Who may enter it, with the delegation to match.

What the site receives, so it holds a record of its own participants' source data. This is usually a periodic certified copy, and it is the step most often omitted.

What happens at study end, specifically how the site retains access for the full retention period.

What happens if the system is unavailable, because it will be, and a site that cannot record an assessment needs a defined fallback that does not become an undocumented parallel record.

That fourth one deserves attention. Systems go down. A site mid-visit with a participant in front of them will record the assessment somewhere, and if no fallback is defined they will use paper and the paper will become an accidental original.

A defined fallback, with a documented process for entering the data afterwards and retaining the paper as source, turns an incident into a procedure.

The certified copy back to the site

The usual arrangement is that at study end, and periodically during it, the site receives a copy of its own participants' data, verified as complete and accurate, with the verification documented.

This is not a courtesy. It is what allows the site to meet its own retention obligations and to respond to an inspection years later, when the sponsor's system no longer exists and the sponsor may not either.

Teams skip it because it is administrative, it costs money, and nothing breaks when it is skipped. Nothing breaks for several years.

In practice

A study uses direct entry for symptom assessments recorded during visits.

The data manager works through the four questions with each site. Three produce nothing surprising. The fourth site raises something none of the others did: its network is unreliable, and it has already had two visits where the system was inaccessible.

The team defines a fallback: a controlled paper worksheet, numbered and accounted for, completed when the system is unavailable, entered within 24 hours with the entry annotated, and the worksheet retained as source for those specific entries.

This is more work than saying "the system will be available." It is also the difference between a documented exception process and four uncontrolled sheets of paper in a drawer that nobody will be able to explain.

What this looks like when it goes wrong

A sponsor runs a study with extensive direct entry, completes it, archives the data, and ends its contract with the platform vendor three years later.

Six years after study completion, an authority inspects one of the sites in connection with a marketing application.

The inspector asks for source data for four participants. The site has none: it was correctly instructed not to keep parallel records, and it never received a certified copy because the periodic-copy step was not implemented.

The sponsor has the data in its archive, in another country, as exported datasets. It is complete and it is not the site's source document, and the site cannot produce its own participants' source data.

The finding is against both. The sponsor designed a study in which sites could not meet their retention obligations, and it is not fixable six years later.

Key takeaways