Describing pain with numbers
Sometimes at Flowopoly workshops we have whiteboards on easels positioned around the room. Press-ganged participants have the job of placing red and green magnets on these whiteboards in a way that represents the spacing and clustering of arrivals and departures during a day in the life of a hospital.
Testing out the magnetic whiteboards at St Andrew's House, Edinburgh: 27 October 2014
The easels, the whiteboards, the magnets: they're not just gimmicks. There's a point to them, which is that one of the first things we have to do with data is describe the status quo. It's a bit like the sentence that is famously not part of the Hippocratic Oath: "First do no harm." Well, for number-crunchers the equivalent precept should be: "First describe things as they are." That's the platform for anything else we might want to do that might be more sophisticated.
But we often fall at this first hurdle. Yes, we try to describe the status quo, but we fail to connect with our audience. And one reason why this happens is that we revert to type. We're data analysts and when you give us data, we summarise it, we count it, we sum it, we average it, we calculate percentages from it. We do everything apart from the thing that NHS managers and clinicians want us to do, which is: show some patients. Show some dysfunction. Show some pain.
This is a tricky problem for data analysts to resolve. It goes against our nature and our training. We have this urge to summarise, and yet there are these other voices saying "No! Show us the detail first!"
Here are three ideas for how we might ease this tension.
First, we could think ourselves into a new way of working by turning to the academic who towers colossus-like above the discipline of information design. Edward Tufte wrote on page 92 of The Visual Display of Quantitative Information that the fundamental principle of good statistical graphics is: "Above all else show the data." And my reading of this is: "Show the raw data—as well as the summarised data—wherever possible and appropriate." We should lean towards tables and charts that let the viewer see individual days, particular patients, specific interventions. By doing this we can align the analyst view with the clinician view.
Secondly, arrange to have conversations with clinicians and managers to talk about how they see the world. Here—by way of example—is an extract from an email exchange with the clinical director of an Acute Medical Unit, who is being critical of me presenting an average percentage bed occupancy figure to him:
I'm not sure that a daily-averaged single figure for occupancy makes much sense to me, because it is the instantaneous bed state that matters. As an extreme example, if AMU was completely empty at midnight and didn't get any admissions until midday, then suddenly they were inundated to beyond capacity for the next 12 hours, you would get 50% mean occupancy but still have a terrible day overall, including A&E receiving the subsequent medical patients. Although this example is far-fetched, it makes the point that it is the peaks of activity that make the system grind to a halt. I think basing any decisions on the daily average occupancy means making the same kind of mistake as the statistician who drowned after wading confidently into a river because it was four feet deep on average.
And here's a third way. Mindful of that famous quote at the beginning of Alice in Wonderland ("What is the use of a book", thought Alice, "without pictures or conversations."), I thought it as well to say something about pictures as well as conversations here.
I've recently been working with two collaborators on a project called Anaximaps (and if you click on the link you'll see that this project parades its own debt to that opening page of Alice in Wonderland) that involves getting healthcare staff to draw pictures or diagrams or sketches or maps of how their workplaces look to them. It's based on the notion that if an information analyst can see—by looking at the drawn map and by listening to the recorded conversation that took place whilst the map was being drawn—what's in a clinician's mind's eye, then the analyst will be better equipped to describe that workplace with numbers.
And that's where the idea of the coloured magnets came from. Time and time again, when we talked to managers and clinicians about how a day of unscheduled care looks and feels when they live inside it, they mentioned how the spacing and clustering of arrivals and departures has a significant effect on patient flow.
Clustering and spacing of ARU arrivals and departures at Edinburgh's Western General Hospital: 17 October 2014
So we created a data exhibit. On a whiteboard. On an easel. Using coloured magnets.
[27 October 2016]