Sam Riley's Challenge
How do we make SPC the default option?
At a Measurement for Improvement event in London on Monday, Sam Riley issued a challenge to the delegates…
…and as someone who spends a significant part of their life driving up and down the motorways of the UK teaching teams of NHS analysts how to “do” SPC, it’s safe to say that control charts are things I have views on! So here are three practical actions I think analysts can take, all of which are—as it happens—joined together by a common theme, which I’ll expand on a bit at the end.
1 / Is it a process?
SPC charts only have resonance if everyone is in agreement that the things being measured by them are processes. As I point out at every available opportunity during my Health Service SPC training courses, the P in SPC stands for “process”, so if the thing you’re measuring isn’t a process, then don’t use a run chart or a control chart to describe it.
This means that analysts have to have the curiosity, the motivation and the confidence to ask the question of managers and clinicians: “This thing we’re measuring here: is it a process?” Do your own research to find a user-friendly definition of “process” that you’re comfortable with. But I usually begin with something along the lines of: “Are we expecting (roughly) the same measurement each time we measure it?” If the answer is yes, it’s a process.
2 / What do those dotted lines mean? And how have you calculated them? And why?
Run charts and control charts aren’t just dots joined together by lines. There are superimposed lines, too. The average line on a control chart (the process average) is easy enough to grasp but the upper and lower control limits usually need words of explanation. Analysts need to have a clear sense in their own heads of what these control limits mean, and why they’re important. We need to have “off-the-shelf” ways of explaining them easily and quickly to a layperson audience. It also helps—particularly when talking to medics who are well-versed in the language of P-values and confidence intervals—if you can explain control limits in that context, too.
And it’s not just the control limits on control charts that need explaining. The median lines on run charts often need to be accompanied by a few words about probability, so we need to have a script for this, too. Without these lay-person explanations, our credibility will be undermined and our attempts to mainstream SPC will falter.
3 / Is the middle line in the right place?
I think it helps if the analyst drawing the control chart holds a view on whether the middle horizontal line in it – the process average – is in the right place or does the process need to change so that the line is lower or higher? One of the dangers of SPC – and it’s got a lot to do with its language – in particular the word “control” – is that by saying a process is “in control” we can unwittingly help foster a sense of complacency about the measurements. We need to remember that it’s possible for a sub-optimal process to be “in control”. And if there is already a large amount of unwarranted variation in the process, then the control limits will be wide (as in: too wide) and the process average line will be lower or higher than it should be. If your desired state is a process average in a different place with narrower control limits then analysts need to know this if we are to explain things clearly.
The common theme running through these three ideas is the need for three-way conversations between analysts, managers and clinicians so that we can find out the story that lies behind the data. Analysts need to know the twists and turns of the plot that will help explain the upwards and downwards movements of the dotted lines.
But the beauty of SPC (and its insistence on the time-series chart as the basis for all of its visualizations) is that the charts themselves will help you uncover the story. Run charts and control charts serve as great props to take with you when you venture to the coalface to converse with managers and clinicians. And of course each run chart and control chart is a story—the graph itself is an unfolding narrative crying out to be annotated, a bit like what I think Kate Cheema is referring to in this Tweet—which I think is why they have such a visceral appeal to so many healthcare professionals.
[22 November 2017]