Auto-Injectors, Nurses, and Mental Models

by Tim Reeves, PhD CHFP

Over the years we have run numerous usability studies with auto-injectors. And while auto-injectors provide significant usability advantages over more traditional syringes, from pre-set dosing to built-in sharps protection, we’ve certainly seen our share of users struggle to use these devices both safely and effectively. Curiously, those who sometimes have the most difficulty are not always who you’d think.

We’ve witnessed nurses have all manner of problems using auto-injectors – failing to remove an auto-injector’s safety release, holding the device at the wrong orientation (exposing themselves to needle sticks), pressing imaginary buttons to trigger activation, and failing to hold auto injectors in place long enough after activation to deliver all of the medication. It’s not unusual to see use error rates among nurses exceed those of untrained lay users.  And unfortunately, it’s not uncommon to have nurses who report training patients on the use of auto-injectors, have difficulties in our studies.

Why would nursing experience make it harder to learn and use an auto-injector correctly?

The answer lies with mental models. Cognitive psychologists use the concept of mental models to explain how we learn about and manage our interactions with objects in the world – from simple devices like auto-injectors, to complex “objects” like people (e.g., we have a mental model of our smartphone and a mental model of cousin Bill). When dealing with something new, we quickly develop a mental model of how it works, and for expediency, we sometimes borrow from existing models that seem reasonably analogous (my new Apple phone seems a lot like my old Blackberry, and my new neighbor seems a lot like cousin Bill).

We use mental models to make sense of the things around us, to understand how things work, and to guide our interactions. But models are by definition abstractions. To the extent that our mental models effectively represent how something works, we do well in our interactions. But if our models are impoverished, or worse, inaccurate, we do poorly.

Why do some nurses have difficulties with auto-injectors? They have poor mental models of how auto-injectors work. It seems likely that these nurses model auto-injectors as analogous to pre-filled syringes, and expect auto-injectors to work much the same way as pre-filled syringes do.

To use a pre-filled syringe, you first remove the needle cover, insert the needle through the patient’s skin, and fully press the plunger to deliver the medication, controlling the flow rate as you go. When the plunger is fully depressed, the medication has been given and the needle is safe to remove.

An EpiPen®, the most ubiquitous example of an auto-injector, doesn’t work this way. And the differences create predictable errors. If you mentally model an EpiPen® as analogous to a pre-filled syringe, you will have difficulties.

Pre-Filled Syringe       EpiPen®
Remove the needle cover. To the extent that nurses treat a needle cover and a safety lock as analogous, they expect the auto-injector needle to come out of the safety end.  Many auto-injectors are not designed that way, including the EpiPen®.
Push the plunger to deliver the medication. We have seen many nurses push imaginary buttons on simple auto-injectors in an effort to initiate dose delivery.
End of push means the medication is delivered. Perhaps the most common use error we’ve seen nurses commit is not holding an auto-injector in place, after activation. These nurses treat resistance after pushing as signifying the end of the injection, much like resistance from a plunger means that all of the medication has been expelled. These nurses don’t wait for the auto-injector to complete delivery.

We’ve also seen nurses commit these same errors, even after reading instructions that are otherwise effective in guiding lay users. We’ve also watched nurses finish failed injections (e.g., not removing the safety and as a result, delivering no drug at all) feeling quite confident they successfully delivered the medication. Unfortunately, with some auto-injector designs, the visual, auditory, and tactile cues are not strong enough to aid these nurses in detecting use problems. The culprit is in part, a confirmation bias that leads us to give inappropriate weighting to feedback that confirms our expectations while ignoring feedback that runs counter to them.  But that’s a topic for another post!