Posted by Sten Westgard, MS
Here's a question that came in via email the other day. At heart, it's a common question, so I thought we could share the answer so others benefit as well:
Question:
[W]hat happens when you calculate that the only rule you need to use is the 3s rule? I know you say that this might be all that is needed with modern hematology analyzers, but I also thought we are supposed to have one rule for precision and one for accuracy? Shouldn't I also use a rule for accuracy? Or, is the accuracy so great for the instrument that because the range is so tight, if something went wrong, it would show up as a 3s problem? In our lab, I have noticed times when there does appear to be trends and shifts without a 3s incident. I am not in charge of the QC so I don't know details about when they recalculate ranges etc depending on problems etc. but am wondering in general what should be used and why.
With the QC Design process, you are determining which are the best
control rules for run rejection. You may still want to add some additional rules
that are only warning rules. Violations of warning rules will give you
advance notice that the system might be deteriorating from stable
performance. Those warning rules violations will not be rejection rules
- you would not reject the run, repeat specimens or repeat a control
based on those signals.
Here's the crux of the issue: there are warning rules and there are "warning" rules. The classic "Westgard Rules" included a 2s warning rule. That "warning" was only to tell you that it was time to check if any of the other rules had been violated. This was back when all QC was done manually, so it was a way to ease the burden of checking all the rules all the time.
As anyone with experience of the 2s rule knows, there's a lot of false rejection involved. The problem with 2s violations is that laboratories get used to ignoring them. So that warning rule became a "warning" that was almost always ignored and people began to get used to the concept of just ignoring all QC flags completely. Now our modern "Westgard Rules" recommendations avoid the use of warning rules and stick only to rejection rules. Partly this is because the monitoring is usually automated now, so the computer software is checking all the rules all the time - no need for a 2s "heads up."
In short, the classic "warning" rule only told you to look and see if the other rules had been violated. A modern warning rule is completely independent of a 2s violation.
For some labs, it would be preferable to get advanced notice before an outright failure occurs. So they may add a warning rule to their QC procedure, or they may add patient data monitoring, etc. There's a big difference between the compliance lab and a lab that wants to do quality right.
To summarize, you can add additional rules to your QC procedure if you wish. A mean rule (6 measurements on one side of the mean, or 8 or 9 or 12, etc.) is usually one of those warning rules that may indicate a systematic drift or deterioration. Just bear in mind that those are not rejection rules - and make sure your techs don't get confused between the two kinds of rules. If they start thinking a 3s violation is just another warning rule, the whole point of customizing and optimizing the QC Design is lost.
[Note: we welcome your questions, even if we can't always immediately answer them. We hope to post more like this in the future]
Comments