"Westgard Rules"
What we get rong about Westgard Rules
As we celebrate the 50th Laboratory Professional Week, Westgard Rules are also 44 years old. How have the rules changed, and how have we changed using them?
What we get Rong about the Westgard Rules
April 2025
Sten Westgard, MS
The Westgard Rules celebrated their 44th birthday in March of 2025. For many, these rules seem older than that, hearkening back to an ancient time of flame photometers, the SMAC (IYKYK), and pipetting by mouth. Over the decades, misunderstandings and misinterpretations have accumulated, so this lab week, we thought we’d take a moment to address some of the most common ones.
Not all rules are Westgard Rules.
There are statistical QC rules. There are Westgard Rules. These two things are not identical. Westgard Rules are a multirule combination of statistical QC rules, so all Westgard Rules are statistical QC rules. But not all statistical QC rules are Westgard Rules. For example, there’s a 1:2.5s rule, a 7:T rule, and we’ve recently seen some bizarre, bespoke rules like R:3s and 10:T, which we’ve never encountered before. (If we haven’t seen a rule, that’s saying something.) More importantly, if your lab has a strange rule, there may not be any knowledge about the performance characteristics such as error detection and false rejection, which would be demonstrated in a power function curve. If no one knows the capabilities of a bespoke rule, if that work has not been completed, that is just mystery. It might help, it might hurt, your ability to detect medically important errors.
See the Best Practices for Westgard Rules to see a list of all the acceptable, known combinations of the multirule procedure.
https://westgard.com/lessons/westgard-rules/westgard-rules/lesson74.html
2 SD is no longer part of Westgard Rules. Not as a rejection. Not even as a warning.
The 1981 Westgard Rules included 1:2s as a warning rule. Even back then, the 1:2s habit was ingrained in most laboratories. The original Westgard Rules can be seen as a way to wean labs off of 2 SD limits for rejection, transitioning them to using 2 SD limits as a warning instead. 2 SD limits produce 9-14% false rejection, and even the classic, complete Westgard Rules had less than half that rate of false rejection. But even this frequency of false warnings is too much today, given the size of the menu and volume of modern laboratories. In the US, too many labs are out of control every day, if not multiple times per day, with the 1:2s as the main source of this pain.
See Weaning yourself off Westgard Rules
https://westgard.com/essays/basic-qc-practices/weaning-off-warning-rules.html
A Warning Rule is not a Repeat Rule
Too often, we see that the 2 SD outlier is taken as a trigger for repeating the control. Got an outlier? Repeat the control. Repeat didn’t work? Try a new control. New control didn’t work? Repeat the new control. And on and on. This makes great profits for the control vendors, but a lot of unnecessary waste and frustration for the laboratory, especially when a great proportion of these so-called outliers are simply false rejections, a result of using limits that are too narrow.
In the classic Westgard Rules, the 2 SD outlier was used as a trigger not to repeat, but to check other rules. Over the decades, a lot of labs have unlearned this lesson, and fallen back to ancient habits. You may enjoy repeating controls, you may even have convinced yourself that doubling your effort to get just one thing done is an acceptable expense, but there's a better way. Set alarms that are meaningful, so when the alarm goes off, you know it's something you have to troubleshoot, not something where you roll the dice again.
See Repeated, Repeated, Got Lucky
https://westgard.com/essays/basic-qc-practices/essay39.html
Westgard Rules are not one-size-fits-all.
In 1981, when the original Westgard Rules were introduced, instrumentation was far more imprecise and inaccurate. Using all of the Westgard Rules on every test was a better choice than applying the 1:2s rule on all tests. That’s why adoption was so rapid and widespread – of course it helped that the rules were completely free.
44 years later, do we see improvements in instrument performance? Absolutely. So much has improved in method performance, to the point where we no longer need all rules on all tests. It's better to optimize how we apply QC, using all the rules on the highest risk (lowest performing) tests, while we lean out on QC for the best tests.
The most recent evolution of Westgard Rules are the Westgard Sigma Rules, where we use the analytical Sigma-metric so assess performance and reduce the number of rejection rules to the barest minimum required.
See Westgard Sigma Rules.
https://westgard.com/lessons/westgard-rules/westgard-rules/westgard-sigma-rules.html
You can’t implement Westgard Rules with manufacturer ranges. Or peer group ranges. Or adjusted, widened, tweaked ranges.
The temptation to widen your limits is very strong, often encouraged by the instrument manufacturer, particularly if you use their controls. Wider limits mean fewer outliers, fewer false rejections, less troubleshooting, and numbers go out the door faster. There’s just one huge problem: if your limits are too wide, you miss errors. The errors go out the door and hit clinicians and patients, and you may not even know it until the error gets very large.
When you use a manufacturer’s range, your 1:3s rule might actually be a 1:6s rule. Your 2:2s rule might be a 2:4s rule. Your R:4s rule might be an R:8s rule. Your 4:1s rule might be a 4:2s rule. You see the issue? You’re not actually using the Westgard Rules. You’re using a set of rules that are far less sensitive and may not provide any useful error detection.
See Double-Blind QC
https://westgard.com/lessons/basic-qc-practices-l/double-blind-qc-manufacturer-ranges.html
Better QC doesn't mean eliminating QC
Often we see the frustration with the way QC is currently practiced boil over into a "get rid of it all" mentality. Labs aren't doing QC correctly? Get rid of all QC. Replace it with an exotic technique, perhaps built on other materials like patient specimens. PBRTQC has a place, but the demands of implementation often require considerable customization and informatics - and the false rejection problem can get even worse than traditional QC if the implementation is poor. Another proposal is to ramp up the use of EQA/PT to the point it can replace QC entirely. That certainly would make more money for the EQA/PT providers, but swapping one set of controls for another isn't a cure if the root cause lies in the statistical implementation.