Posted by Sten Westgard, MS
We got the following question emailed to the website recently:
Greetings, I was [wondering] if the following case represents [violation] of 4 1s rule?
Control (A) point 1= EXCEEDS 2 SD /-2 SD lines >> warning
Control (A) point 2= WITHIN 2 SD/-2 SD lines >> accepted
Control (A) point 3= EXCEEDS 2 SD /-2 SD lines >> warning
Control (A) point 4= WITHIN 2 SD/-2 SD lines >> rejection
all the point[s] on the same side of the mean.
What do you think? The answer, after the jump
This is a great question about multirules and "Westgard Rules" - and the interpretation depends on which version of the "Westgard Rules"you are implementing.
There is the classic set of "Westgard Rules" which only triggers the 4:1s after a 1:2s warning. And in this case, while you have had two 1:2s warnings go off, there is no 2s violation at point 4, therefore you don't interpret the 4:1s and there is no rejection of the run.
Two things to say in follow-up:
1. The classic "Westgard Rules" with 2s warning rule was intended for labs that were only doing their QC with pen and paper and using the medical technician's brain to process the rule interpretation. Now in our modern era, we usually have software, either directly on the instrument, or in an LIS, or in middleware, or even offline spreadsheets, that can plot the data and interpret the rules. When computers or informatics are interpreting the rules, we no longer recommend waiting for a warning rule to be triggered. Instead, interpret all necessary rejection rules with every data point. We sometimes refer to this as the "modern" or "automated" version of the "Westgard Rules". And if this modern adaptation of "Westgard Rules" are applied, then your example IS OUT-OF-CONTROL.
2. A deeper question to ask is whether or not the 4:1s rule is NEEDED for the particular analyte. We have found that assays that perform at 5 or 6 Sigma quality do NOT need the 4:1s rule. A 6 Sigma assay can be controlled with limits set at 3s and just run 2 or 3 controls - none of the other "Westgard Rules" are necessary. So we strongly encourage labs to calculate the Sigma-metrics of their test methods and use those results to customize QC. When a method is 4 Sigma or lower in performance, then the 4:1s rule IS needed and should be used to monitor the method.
There are a number of tools to help select QC rules: "Westgard Sigma Rules" are the simplest, OPSpecs charts and Normalized OPSpecs charts are more complicated but more powerful. We have free versions of all these tools available for download on the website westgard.com
I realize this makes life more complicated in the short run. But in the long run, this optimization and customization of QC for each method will make the lab more efficient and safe. You won't over-QC and over-reject methods that are performing very well, but you will also still know which methods are poor in performance and you can concentrate your efforts on better monitoring and quality improvement.