Posted by Sten Westgard, MS
In the Levey-Jennings chart below, just focus on the two points the +2.0 sd point and the -3.0 sd point.
Your QC is actually on the line, so what do you do?
If we're using the 2 sd warning rule, is it violated?
If we're using the 3 sd rejection rule, is it violated?
What do you think?
(after the jump, the answers)
It's admittedly an unusual occurrence (then again, maybe your lab sees it all the time): a point that is actually on the line of your rule.
The traditional way that control rules are defined is that a 1:2s rule is violated when you EXCEED +2 sd or -2 sd. So if you're on the line at 2.0, you have not actually exceeded the rule, and therefore you are IN CONTROL!
This is also how most control rules are programmed in software. You don't write the code to violate the 1:2s rule with the value is greater than 1.99999. You write the code so that if "x>2.0" then "1:2s rule is violated". So, in most practical implementations, you have to exceed the control limit of the stated rule.
There's also a key factor that impacts these interpretations: data rounding. If you only round to one significant digit, say 1.6, and the control limit is set at 1.6, you might see more values "on the line" than if you would expose another significant digit. If you expanded to 1.59, 1,58, 1.61, 1.62, 1.63, etc. you have more values that will fall above and below the limit line. With one significant digit, all those previous values might be rounded into "1.6" and not exceed the limit, even though the additional decimal place shows much more variation. It's often useful to include an extra decimal place for some assays that report out in a simpler format - you don't need to report 1.61, but you can certainly use it within the laboratory to get better QC handling.
Still, even if we add some extra decimal places, we may see that we've got one point at +2.00, or -3.000, do we still consider those controls "in"? It may seem like we've reached some absurd cul-de-sac of QC, where a tiny 0.01 or 0.001 change could be the difference between "report the patients" and "reject the run."
At some point in the decimal places, we reach the same place as Monroe Stahr in The Last Tycoon, by F. Scott Fitzgerald.
As Monroe talks with a pilot, looking down on the mountains below the plane, he explains the difficulty of making some decisions:
"'Suppose you were a railroad man,' he said. 'You have to send a train through there somewhere. Well, you get your surveyors' reports, and you find there's three or four or half a dozen gaps, and not one is better than the other. You've got to decide - on what basis? You can't test the best way - except by doing it. So you just do it....You choose some one way for no reason at all - because that mountain's pink or the blueprint is a better blue.'"
Ultimately, we choose 2.0 and 3.0 sd as our control limits because we're human. We could have chosen 1:1.99s and 1:2.99s as our standard control rules, but the simplicity of whole numbers appeals to us. So that's what we chose, and that's what we should implement as we look at our charts. It may be that there's no real difference between a +1.99 sd control point and a +2.01 sd control point, but once we build the rules, we have to enforce them.
Hope that's not too philosophical of an answer.
It is a potential non-conformance. The lab has to be pro-active and take corrective actions after performing RCA. We should not wait for the failure to occur and then take CA.
Posted by: Dr CVAnand | December 06, 2019 at 12:11 AM