Investigate a failed EQA result in this order: clerical error, the sample, internal QC on the day, lot and calibration changes, the method group, then the trend across surveys. The order runs from the cheapest check to the most expensive and from the most common cause to the least. Record what was checked and what was found at each step, including the steps that found nothing, because ISO 15189:2022 clause 7.3.7.3 expects the review and the action to be on record.
1. Clerical error
Check first that the result submitted is the result the analyser produced. Wrong units, a decimal place, two samples entered against each other, and a result entered under the wrong analyte or method code are all common, and none of them says anything about the assay. Compare the submission with the instrument printout. If two samples are both out by amounts that would swap neatly, suspect a swap.
2. The sample
Was the material stored, reconstituted and mixed as the programme instructs, and analysed within its stability? Was it run like a patient sample? An EQA sample given special treatment, such as repeat testing until a preferred number appears, no longer tests the routine process.
3. Internal QC on the day
Look at internal QC for the run that produced the result and the days either side. QC in control with EQA out suggests a problem QC cannot see: a matrix effect in the EQA material, a calibration bias the controls share, or a problem at a concentration the controls do not cover. QC drifting the same way as the EQA result suggests a real shift in the assay.
4. Lots and calibration
List what changed before the analysis date: reagent lot, calibrator lot, a recalibration, maintenance, a software or parameter change. This step depends entirely on having an accurate Date in Use for each lot. A deviation that begins at a lot change and persists on that lot is the strongest evidence a laboratory usually gets.
5. The method group
Compare the result with the group using the same method and instrument, not only the overall target. If the whole method group sits away from the target, the cause is the method and belongs with the manufacturer. If the laboratory is alone within its group, the cause is local. Check also how many laboratories are in the group: a target from a handful of results is itself uncertain.
6. The trend
One result is a point. The last several surveys are a picture. A steady bias on one side of the target, a bias that grows with concentration, or a drift over the cycle each point to different causes. A single outlying result on an otherwise flat history, with QC in control, is often a random or clerical event and may be closed as such if the evidence supports it.
Closing the review
The record should state the result and its deviation, each check made and its finding, the cause concluded or the statement that none was found, whether patient results could have been affected and what was done about that, the corrective action, and who reviewed it and when. Results in the warning zone deserve a shorter version of the same review: they are the cheapest early warning a laboratory has.
How EQAWarden follows this order
EQAWarden runs the same sequence as fixed rules. It scores each result against its allowable performance specification, runs clerical screens first, tests the series and the cycle, tests for a step at each lot change, puts internal QC beside the EQA result, and compares the method group. It then ranks 21 possible causes with the evidence for and against each and suggests one of four outcomes. The reviewer decides, writes the conclusion and signs. It uses no AI, and every rule and threshold is published.