This question recently came in:

"We have two specific questions regarding the practical application of the 7T rule:

  1. "When applying this rule, is it mandatory to use the 1:2s rule as a starting trigger, or can the laboratory choose to use 7T independently without the 1:2s initiation? (Personally, I believe 1:2s initiation is necessary….)
  1. "When a 7T violation is flagged, is it required that the difference between the first and last data points exceeds a certain range (e.g., 3s)? In some cases, even if 7 consecutive trending points are observed (without 1:2s initiation), the difference between the first and last point is within 2s, which may represent normal random fluctuation rather than a clinically meaningful shift.

"What principles do you recommend for the proper use of this rule?"

Thanks for your question. They show some deep understanding of the deep details of QC.

Let's start by pointing out that the 7:T rule, while certainly it is described on the Westgard website, is NOT part of any official Westgard Rules combination. Not the version in 1981, which had the 1:2s warning rule, not the more recent versions that no longer have a 1:2s warning rule, and certainly not any of the Westgard Sigma Rules.

Thus, from that Westgard Rules perspective, the use of a 1:2s rule is not necessary in order to trigger the 7:T rule. So we would answer "no" to your question 1.

From another perspective, however, the fact that the 7:T rule is on its own, in a sense, then the laboratory is free to use it in any combination. That means a "yes" to your question 1. Do you want to use a 1:2sW/7:T combination? Good for you. Go for it. We can't help you characterize how good that choice is, because we have never done the simulation studies to determine error detection (P:ed) and false rejection (P:fr). But perhaps someone else in the world has done it, and you can check out their power function graphs.

As for the necessity of some distance between the 1st point and the 8th point in a 7:T rule, the rule description does not specify a distance. It only requires that each successive point is further away in one direction from the last point. So you could potentially have a 7:T rule violation with all the points falling within 1 SD, or even less. They could all fall within 1 SD of the mean. That's one of the drawbacks of the rule. It could be catching an extremely tiny trend.

Again, since the 7:T is a rule "in the wild", you are free to impose additional constraints on its implementation. If you want to require that the distance between the 1st point and the 8th point should exceed 2 SD, again, more power to you. It would be interesting to see the power function charts of different requirements like that.

Our last advice, you may not be surprised to hear, is not to use the 7:T rule at all. There are other rules that can carry the same weight.

P.S. Did you notice that the 7:T rule requires 8 points? The first point is the base where you start, then you need 7 more that are moving in one direction. 1+ 7 = 8. Alternatively, you could implement it with 7 points, but then it would only be catching a 6:T rule.

Send in more questions!