Handling Queries and Corrections

Part of Clinical Research Coordinator (CRC) Training: Source Documentation and Data.

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

Why this matters

A query is a question about your data. Queries are not accusations, and a study with no queries is usually a study nobody is checking.

What matters is how you answer them. A rushed answer creates two problems: the original error stays, and the record now also contains a wrong explanation.

Where queries come from

• Automated checks in the EDC system. Values outside a range, dates in an impossible order, required fields left blank, or entries that contradict each other.
• The data management team. People reviewing the accumulated data across all sites.
• Your monitor, comparing what is in the CRF with what is in source.
• Central or statistical review, looking for patterns across sites.
• Safety reconciliation, where the adverse event data in EDC and the safety database disagree.

What queries are usually about

• Missing data. A required field is empty.
• Inconsistency. Two records disagree — a visit date in EDC that does not match the source, a medication start date after its stop date.
• Out-of-range values. A laboratory result or vital sign outside expected limits.
• Unclear entries. A free-text field that a reader cannot interpret.
• Protocol issues visible in the data. A visit outside its window, a missing assessment, an adverse event with no end date and no explanation.

A method for answering

Step 1: Read the query properly. Identify the participant, the visit, the field, and exactly what is being asked. Many bad answers are answers to a question that was not asked.

Step 2: Go to source. Open the actual record: the clinic note, the laboratory report, the printout, the diary. Do not answer from memory and do not answer from the CRF you are being asked about.

Step 3: Decide what is true. There are only four outcomes.

• The source and the CRF agree, and the CRF is right. Confirm the value and explain why it is correct.
• The CRF is wrong and the source is right. Correct the CRF.
• The source itself is wrong or incomplete. Correct or complete the source properly — single line-through, initials, date, reason — and then correct the CRF.
• The data are genuinely missing. Say so, say why, and use the system's missing-data provision.

Step 4: Answer specifically. "Confirmed correct" is a weak answer. "Value confirmed against laboratory report dated 14 March; participant fasted as required" is a useful one. Someone will read this in three years.

Step 5: Answer promptly. Queries age badly. The people who remember what happened move on.

Step 6: Escalate what is not yours. A query that asks for a medical judgment — is this event related, is this result clinically significant — goes to the investigator. Do not answer it yourself.

What you must never do

• Never change source data to match the CRF. The direction is always source first.
• Never close a query by entering a value you have not verified.
• Never let someone else answer a query under your account.
• Never answer "as per site records" without saying which record.

Preventing queries

Most repeat queries have a system cause. Look for the pattern.

• The same field is queried every visit → the source worksheet has no place to record it, or staff do not know it is required.
• The same participant generates most queries → check whether a particular staff member or a particular visit type is involved.
• Date errors recur → the source worksheet and the CRF probably ask for dates in different orders.
• Laboratory queries recur → check whether the local reference ranges on file are current.

Fix the cause, not the instance. Then record what you changed, so the fix survives you.

Data lock

At the end of a study, or at a defined interim point, the database is locked: no further changes without a formal process. Before lock, all queries must be closed, all data entered, and the investigator sign-off complete.

Sites feel this as an intense few weeks. It is far less painful at sites that kept their queries under control throughout, which is the real argument for answering them promptly.

In practice

Nadia receives a query on a VT-201 participant: the visit 5 date in EDC is 3 August, but the adverse event start date recorded at that visit is 28 July, before the previous visit.

She goes to source. The clinic note confirms visit 5 was on 3 August. The participant had reported a headache that began on 28 July, between visits. The CRF is correct on both dates; they only look contradictory.

She answers with that explanation, naming the clinic note and its date, rather than writing "dates confirmed".

She then notices that this is the fourth query of its kind across three participants. The site's source worksheet asks for "date reported", while the CRF asks for "date of onset". Staff have been recording one and entering the other.

She changes the worksheet to ask for both dates, retrains the team, and documents the change. The queries stop.

One query is a question. Four of the same query is a process problem.

Key takeaways

• A query is a question about the data, not an accusation, and a study with no queries is usually one nobody is checking.
• Always answer from source, never from memory and never from the record being queried.
• Corrections flow from source to CRF, never the reverse, and source corrections use a single line-through with initials, date and reason.
• Queries requiring medical judgment, such as causality or clinical significance, are answered by the investigator.
• Repeated queries of the same type indicate a process cause — usually a mismatch between the source worksheet and the CRF — and fixing the cause is what stops them.