Exceptions to the rule

Sometimes we have to modify the fullness hypothesis

One of my more vivid Flowopoly memories is of a workshop on a wintry day at Wishaw General Hospital two-and-a-half years ago. Back in those days the method we used for choosing which bad day and which good day to replay wasn't as sophisticated as it is now (we used to just pick days when A&E four-hour compliance was either "bad" or "good"), and I remember that the thing that made the good day good wasn't the usual thing (that there were empty beds available downstream of A&E for the patients who needed to be admitted to flow into). Instead it was that none of the A&E attendances that day needed to be admitted until mid-afternoon, by which time the Assessment wards had had time to recover from their perilous 8:00am fullness situation and release beds thus enabling a relatively breach-free day. The hospital's unscheduled care system was effectively "rescued" by the late arrival times of its majors in A&E.

Patient flow up close and personal in a Flowopoly replay.

Hold that Flowopoly anecdote in your mind while I introduce a seemingly unconnected thought. There's a video of a Brian Cox TED talk I sometimes use in training workshops to illustrate the Ernest Rutherford quotation about how all science is either Physics or stamp-collecting. But just before he quotes Rutherford at 5'30", you see a colourful diagram on the screen (which I now know to be the Standard Model for Particle Physics) and I couldn't help but notice that this diagram bore a striking visual resemblance to Flow_ology grids.

The diagrammatic representation of the Standard Model of Particle Physics that appears on the screen at 3'34" in this TED video.

The idea behind Flow_ology grids is twofold. Firstly, they help reinforce the idea that the three basic 'metrics' of patient flow—how many, how long and how full—are held together vertically by arithmetic (otherwise known as Little's Law). And—secondly—they encourage people to see the cause-and-effect relationships in the unscheduled care system that are made visible by horizontal and diagonal lines in the grid. For example, fullness in the Assessment area is related to length of stay in A&E. And, further downstream, the fullness of the specialty beds is related to the fullness of the Assessment beds.

An example of a Flow_ology grid for a hospital that achieved 88% compliance with the four-hour target in calendar year 2016.

These horizontal and diagonal cause-and-effect relationships make up what I pompously call "The Fullness Hypothesis". (And these relationships are a lot easier to see when you place two Flow_ology grids next to each other so that you can see the bad days alongside the good days and then draw your own conclusions about the impact of fullness on patient flow.) But before you go off thinking this is some top-down theory that I'm trying to impose on an unwilling NHS empirical reality, it's worth remembering that I didn't just dream up the fullness hypothesis. It's based on what countless managers and clinicians have told me about how unscheduled care systems actually work—or don't work—in practice. Fullness downstream causes flow problems upstream. But it's important to remember that the fullness hypothesis is not inviolable. It's not always fullness that causes problems, and—similarly—emptiness doesn't always solve patient flow problems. There's more to it than just fullness.

And for reasons that I can't quite explain, I was seduced by the idea that I could somehow adapt the pink column in the Standard Model diagram for Flow_ology purposes. Now I should state here that I have no idea whatsoever what those things—gluons, bosons, whatever—in the pink column actually are, but I got it into my head that the pink column could represent factors or variables that would modify the other elements in the grid. It's all a bit hazy in my fevered imagaination but what the heck, I want to take this as far as I can.

This idea of having numbers or variables that help to explain exceptions is important. One of the things about the scatterplots you can draw to show these horizontal and diagonal relationships is that you need to be able to explain the dots that don't fit the general pattern as well as the dots that do. So I've long thought that it'd nice to be able to somehow build the exceptions to the rule into the general theory, as it were.

A scatterplot showing the relationship between downstream fullness and Assessment fullness.

I've done a bit of thinking about what these exceptions—I want to call them modifiers—are. And I'm going to propose tentatively and provisionally (this is the Kurtosis blog, so it's "thinking out loud", remember!) that the modifiers to Little's Law might be: (a) casemix as measured—crudely—by the percentage of patients who get transferred onwards into the system (majors as a percentage of all ED attendances, for example); (b) the spacing and clustering of majors (this was the thing that rescued the hospital I mentioned at the beginning of this piece), which means we need to find some way of measuring this (the standard deviation of arrival times, maybe..?) and (c) some measure of staffing that helps us understand the extent to which that might cause us a problem.

But at this stage it's just ideas. Please do feel free to email me (neil.pettinger@kurtosis.co.uk) if you want to make suggestions.


[21 August 2017]