Data Queries and the Irreconcilable Value
Part of Clinical Trial Assistant (CTA) Training: Vendor and Sponsor Coordination.
About 4 minutes. Free, in your language, no account needed.
Why this matters
Queries are how a study asks a site to explain or correct something. They are a normal part of data cleaning, and they become a problem when they accumulate, when they are answered without being resolved, or when one of them is not really a query at all.
Your role in the query cycle
You typically do not raise clinical queries or decide whether an answer is adequate. You keep the cycle moving: queries reach the right person, ageing is visible, and closure is real rather than nominal.
Two metrics tell you most of what you need. How many queries are open, and how old is the oldest. A study with a stable count and a growing oldest is answering the easy ones and avoiding the hard one, and the hard one is usually the interesting one.
Ageing is a signal, not an annoyance
A query open for sixty days is not slow; it is telling you something. Usually one of three things: the person who can answer has left, the answer requires source data nobody can find, or the query is unanswerable as written.
All three need escalation rather than another reminder. A fourth chase on a query nobody can answer wastes the site's goodwill.
The irreconcilable value
Sooner or later a value will not reconcile. A laboratory result the site says is wrong and the laboratory says is right. A visit date in the source that contradicts the system. A dose that does not match the dispensing record.
The correct handling is precise and it is largely about what not to do.
Do not decide which source is correct. Do not ask the site to change the data to match. Do not close the query because the discrepancy is small.
What you do is preserve both values with their provenance, establish what each party's record actually says and where it came from, and escalate to the people who can adjudicate, which typically means data management and the investigator, and the medical monitor where a participant's safety could turn on it.
The reason for that discipline is simple. A discrepancy that is investigated and documented is a normal finding. A discrepancy that was quietly harmonized is a data integrity problem, and it will look like one in the audit trail.
Closing honestly
A query is closed when the discrepancy is explained or corrected with an audit trail, not when somebody has replied. "Confirmed as recorded" with no explanation is a reply, not a resolution.
A worked irreconcilable value
A potassium result in the database reads 5.8. The site's source document reads 3.8. The laboratory report, retrieved, reads 5.8.
The site believes the participant's result was normal. The laboratory believes it issued 5.8. Both are confident.
What you do not do. Ask the site to correct the source. Ask data management to correct the database to match the source. Close the query because the participant was asymptomatic.
What you do. Record both values with their provenance: the database entry with its audit trail, the source document with its date and author, and the laboratory report with its accession number. Note that they diverge and by how much. Escalate to data management, the investigator, and the medical monitor, because 5.8 is a value that could require clinical action and 3.8 is not.
What it usually turns out to be. A transcription error at the site when copying the result into the source, a sample mix-up, or a result filed against the wrong participant. Each has different consequences, and two of them affect more than one participant.
Why the discipline is worth it
If somebody had simply changed the database to 3.8, the study would have a participant with a potentially clinically significant value that nobody reviewed, an audit trail showing an unexplained change, and no way to find out which of the three explanations was true.
The discrepancy was the signal. Preserving it is the whole contribution.
Ageing as diagnosis
When a query has been open sixty days, stop chasing and ask why. The answer is almost always that the person who could answer has left, the source cannot be found, or the query is unanswerable as written.
A fifth reminder solves none of those and costs the site's willingness to answer the next one.
Key takeaways
- Keep the cycle moving; open count plus age of the oldest tells you most of it.
- A stable count with a growing oldest means the hard query is being avoided.
- Never decide which source is right, never ask for data to be changed to match, never close on size.
- Preserve both values with provenance and escalate to the people who can adjudicate.
- A reply is not a resolution.