 Figure 2. The prototype dose calculator running on the iPhone (it also works on desktop Web browsers). The opening screen is red and shows that a dose and drug concentration have not yet been provided. The tabs at the bottom of the screen allow the user to choose which numbers to enter; they allow users to enter numbers in any order, unlike an ordinary calculator, where changing order would create errors.

 

 Figure 3. Entering the drug concentration, using the keypad. The screen scrolls up, and the numeric keyboard appears when a number field is tapped. The “Rate” tab is red, indicating outstanding errors; at this point one of the errors is that the user has not finished entering the concentration.

indeed, the cognitive load of compiling a complex calculation would have reduced their error-detecting vigilance in general.

The ISMP report commissioned a small human factors study of the Abbott pump: It identified numerous problems [ 2]. Why aren’t devices made consistent with best clinical practice, so that operator training becomes simpler, rather than the other way around? Why does the report say in its recommendation 10A that nurses should be trained that “mL” on an infusion pump means “mL per hour”[ 2]? Why does recommendation 10B ask purchasers (hospitals) to do human factors studies of pumps? The same answer to both questions is that for the time being manufacturers—and national regulatory processes—can’t be relied on, and hospitals therefore have to train nurses to cope with bad design. That also means that when things go wrong, as they do, that the nurses or the training has failed: It’s then a very short step to blame the nurses or their management for the consequences.

September + October 2008

 Figure 4. Once all numbers are entered correctly, the main screen goes green and summarizes the dose details. It also confirms how long standard sizes of drug will last and what the daily dose is.

interactions

ward, it could have checked that dose rates were appropriate for standard drugs; more than 30 grams per day for fluorouracil should have raised warnings ( 1 gram a day is a high adult dose for fluorouracil). Unfortunately the pump provided no such checking.

The infusion pump was an Abbott AIM Plus. In the mode where the nurse should enter mL per hour, the display option is “mL” without the “per hour,” which is incorrect (see Figure 1). Moreover, the HELP button provides information on only two of the three options and does not give help for the incorrectly labeled option!

The pharmacy computer printed the label on the fluorouracil bag, which the nurse used to get the numbers for the calculation. The label confusingly included many numbers, 1.2mL/hr, 28.8mL/24h, 1312.5mg/24h … 15 numbers in all, not counting the date and patient-identification details. The numbers on the label break many recommendations: 1.2mL/hr rather than the correct 1. 2 mL per hr (the space before mL is required to help avoid the m being misread as 00), and showing a number pointlessly to five significant figures, and so on. Worse, in my opinion, the numbers were not organized in any way that related to the pump’s requirements. The bag label appears not to have been designed to help the nurse who has to use it.

Both nurses incorrectly calculated 28. 8, yet this incorrect number had also been printed on the label, which would have provided confirmation bias for the nurses and distracted their attention from relevant detail;

Alternatives Are Possible

I spent a day programming an Apple iPhone to explore ways of improving things (see Figure 2). With my prototype you can hold in your hands a working system that avoids some of the problems described above. It can be downloaded from harold.thimbleby. net/health.

The iPhone stimulates many ideas, For example, it has a camera and could photograph the drug barcode and check that it was what was expected; it could require a second nurse to check the calculation; and so on. On