Home   About Kurtosis   Courses   Course Calendar   Booking Information   Ideas   Technique   Clients   Contact

Two types of management information

Do we need a bit less here-and-now and a bit more there-and-then?

Last year, my favourite sentence with the word “data” in it was this one:

Some boards use data simply for reassurance, rather than the forensic, sometimes uncomfortable, pursuit of improvement.

If you don’t recognise it, it comes from Professor Sir Bruce Keogh’s Review into the quality of care and treatment provided by 14 hospital trusts in England, which was published on 16th July 2013.

I’m someone who believes passionately that data should be used for “the forensic, sometimes uncomfortable, pursuit of improvement”. But I often feel like I’m in the minority. It seems to me that people mostly use data for reassurance. Puzzled by this, I decided to try and work out why. And I decided to go right back to the beginning and think about how people actually think about data.

So let’s see if the following train of thought makes any sense. And I'm kind of thinking out loud here, so it may well not!

Once upon a time (in about 1986, to be a bit more exact), back in the days when the NHS was only just beginning to get its head around words like “computer” and “system” and “data”, there was a sort of prevailing conventional wisdom amongst data people that said there were two types of data.

First, there was operational data. This was the data that helped people do their jobs on an immediate, day-to-day level. For example, a patient administration system (PAS) would record a patient’s name, their address, their date of birth, what ward they were in, what consultant was responsible for their care. And data items such as this would be useful in helping clinicians and others do their jobs on an operational level. This was the stuff that used to get written down on paper but now it was starting to be captured digitally.

Secondly there was management data. I say “data”. In fact, you’d often find this expressed as management information, as if to emphasise the fact that it had somehow evolved from being just, well, data. Management data was the data you used for making management decisions. Decisions about resources. And it’s the data that—in an ideal world—you’d obtain as a by-product of the operational data. One of the textbook principles was that you should never really build computer systems to gather just management information. No, you should get your management information from a system that’s also doing something useful, by which was meant something operational.

I’m making this distinction because too many people who use the word “data” these days are blissfully unaware of it. To a lot of people, data is just data. Stuff you see on computers. Numbers on a spreadsheet. Reports containing tables and charts. Data. No distinctions made.

So let’s just recap on that distinction. Operational data is data we need in order to help get the job done. It’s the digital means by which we know that Mr Armstrong has got diabetes, that he’s taking drug X, that, that he’s currently an inpatient on Ward 6, and he was admitted on 29th December. All this we need to know in order to do just basic stuff for Mr Armstrong and so that we don’t keep having to ask him the same questions over and over again.

Management information, by contrast, is the data that tells us how full Ward 6 was on the day Mr Armstrong was admitted, and how many patients with diabetes get admitted to this hospital each year. It’ll help us evaluate how effective drug X is at treating whatever it is that Mr Smith has got, and it tells us whether this year was busier or quieter than last year.

So management information tends to be “bigger” than operational data. And by “bigger”, I mean that it tends to be summarized data as opposed to individual patient data.

So that’s the basic distinction, but to get closer to an explanation of why Professor Sir Bruce Keogh found that too many trusts “use data simply for reassurance, rather than the forensic, sometimes uncomfortable, pursuit of improvement”, we have to dig deeper into our understanding of what we mean by management information.

And when we dig deeper, we find—I think—that there are two types of management information. And if the first distinction—between operational data and management information—is not often made, then this second distinction is made hardly ever at all.

The first type of management information is here-and-now information. This is summarised data but it’s recent data (sometimes real-time data, in fact), and it’s telling you stuff that you wouldn’t ordinarily be able to just see (so it’s valuable) and it enables you to take evasive action or pre-emptive action or, of course, you just look at it to reassure yourself that everything is ticking along just nicely.

This first type of management information is the information you might expect to find on a dashboard.

The second type of information is there-and-then information. You use this data when you are looking back, when it’s too late to take evasive action because anything that could’ve gone wrong has already gone wrong.

Dashboard devotees often ask: “What’s the point of there-and-then data? Surely looking at the road ahead is more important than looking through the rear-view mirror?”

Well, the point of there-and-then data is that it allows you to reflect on what’s gone wrong. It’s the post-mortem. It’s the data you look at after you’ve lost control, after you’ve crashed the car, and after you’ve killed three pedestrians. Or, to put it another way, it’s the data you look at the end of a month when you’ve had 142 patients who breached the four-hour target, of whom 29 spent longer than eight hours in A&E, and of whom six died within a week of being admitted.

You use there-and-then data to find out what went wrong. And why it went wrong. You use it for “the forensic, sometimes uncomfortable, pursuit of improvement.”

And to use there-and-then data properly means that organisations have to do at least two things first.

Firstly, organisations have to be prepared to admit that things are not going as well as they ought to be. So if the organisation has a culture of just wanting to report good things happening, if nobody wants to admit that things are actually a bit (or a lot) rubbish and could do with being improved, then there-and-then data will never reach the boardroom table. Someone has to make the decision that a post-mortem is necessary. If nobody thinks the deaths are suspicious, then no post-mortems will take place, you’ll just carry on without them, and nobody will be any the wiser.

Secondly, the organisation has to be prepared to reflect, and learn, and dig until it finds the cause-and-effect relationships. The point is to find out why the crash happened so that corrective action can be taken in future so that they don't happen again.

So the conclusion I’m tentatively reaching is that although there-and-then management information can help provide us with answers to the difficult questions the NHS faces, it’ll only ever be in a position to provide these answers if the culture of the organisation enables it.

Using data for “the forensic, sometimes uncomfortable, pursuit of improvement” means a change in organisational behaviour, as well as a change in data analysis behaviour.

[3 January 2014]


Make a comment on this blogpost:


Job Title:


Email address:


© Kurtosis 2014. All Rights Reserved.