Designing for the task: what numbers are really used in hospitals?
Sarah Wiseman (PhD student, UCLIC)
The reason I wanted to know this was not because I was designing a new number-based Scrabble-type game, but because I wanted to know more about error rates in number entry. We know that there are different error rates associated with the frequency of a letter, but we don’t know if this applies to numbers. I wanted to find out. The first step was to look at a big set of numbers and see if there were any differences in the frequency of digits.
I’m interested in number entry in the medical domain so the big set of numbers I chose to look at were the log files from 32 drug-infusion pumps, used in different wards in a hospital.
If you’ve ever played Scrabble or Words With Friends you’ll know that each letter on the board has a different scoring value. This is decided by the frequency of that letter in the English language; the more frequent a letter, the easier it is to place on the board and thus fewer points are awarded for using it. A while ago I started to wonder how you might assign points to numbers, instead of letters. Do some digits occur more frequently than others or are they all equally likely?
After analysing all the pumps I got the following results, you can see that there was a lot of variation in the number of times each digit was used.
These logs recorded lots of information about the pump: when alarms had gone off, when the pump had been started and stopped, and also what the current fluid volume and rate of infusion were. It was these two pieces of information that I was interested in; these were the numbers that the medical worker would have had to type in to set the infusion going.
The 0 was by far the most common digit; it was used three times more than any other digit. The digits 1, 2 and 5 were also used more often than you’d expect on average.Ignoring the decimal point, there were more interesting findings when I looked at the frequency of digits on each of the wards. You can see the graph of digits for each of the wards below.
You can see how each ward generally follows the same sort of pattern overall, however, in the surgical ward you can see that suddenly the digit 9 becomes the most frequent. What might be causing that?We don’t know the circumstances under which these pumps were programmed, but we can start to guess at why we see this abundance of 9s in the surgical ward. If you were programming an infusion and wanted to administer the medication to the patient as quickly as possible what would you do? You might set the pump to work as fast as it can, by typing in the highest number you can, which in this case would mean filling the numeric display with the number 9. With this sort of information, we might think about how we can better design the number entry interface to match the tasks that are being completed on it. This happens in other areas of design, often on smart phones. For example, when entering an email address, the keyboard on some smart phones changes to provide you with the @ symbol, which is normally hidden in a punctuation menu and harder to get to. Equally when entering a web address (URL) in the browser address bar, the keyboard might change to give you a ‘.com’ button, because in this situation, it’s very likely you’re going to need to type that. Both of these design alterations involve changing the interface, to meet the needs of the task. Applying this idea to infusion pumps we might want to add a button to infusion pumps that automatically sets the pump to the maximum to stop people needing to keep pressing the 9 key. Or, we could make the popular keys easier to access on the interface or add buttons for common digit combinations, like they have in this optician’s interface (glasses prescriptions end in .00, .25, .50 or .75).
This is just the beginning of this line of research. We now need to know which interfaces suit these types of numbers. And we can also begin to investigate whether users make the same types of frequency related errors with digits and numbers as they do with letters and words. It will also be interesting to see if any other wards have unique patterns of digits, or indeed if these patterns are seen in different hospitals.
Sarah Wiseman on digit distribution: analysis of numbers entered into infusion pumps (CHI+MED blog, 7 December 2011).