CHI+MED logo banner

Understanding the language used when reporting user error

Key points

  • Working with NHS England, we have reviewed over 6,000 incident reports about the use of medical devices with respect to the language used. We particularly focused on how incident reports classified as user-error within this sample differred to those that were not.
  • When reporting incidents that are classified as involving user-error the reports are significantly shorter than those classified in other ways. They are also significantly less likely to be written in the first person, from the reporter's point of view. They are also far more likely to write in a passive style, and avoid naming anyone, even patients.
  • Given user-error classified incidents are classified as such because the error is linked in some way to a person, or user, we would expect to see the opposite. This suggests that those writing the reports are actively removing themselves and others from the report in the language they use, and this may link to a blame culture.
  • This work suggests that techniques from linguistics may help understand aspects of blame culture and even track a changing blame culture in healthcare over time.

Healthcare professionals in the United Kingdom are encouraged to report any incidents that have caused, or have the potential to cause, patient harm. However, there is widespread under-reporting of patient safety incidents in both the USA and UK. It has been widely suggested that there is a blame culture within healthcare and that this affects the way incidents are reported and the potential for learning from adverse events .

In order to better understand the relationship between a blame culture and reporting practices, we analysed the language used in incident reports that had been classified as involving 'user error', ie they involved a person making a mistake of some kind when using a medical device. Intuitively, incidents classified in this way are likely to involve an individual with whom responsibility lies, and, so 'is to blame'. It is rarely that simple with many other factors contributing to mistakes that are out of the control of the person concerned.

Exploring the language of incident reports
We used a combination of techniques from linguistics to investigate the text filled in to summarise the incidents. As this is a free text box, the person entering the report can decide what information to include, and the length and detail of the summaries varies greatly. By comparing these summaries for incidents classified as involving user error, with those that are not, we were able to explore how incident reporting differs depending on the kind of incident.

We found that user error linked incidents are significantly less likely to be reported in the first person: they do not talk about "I did this", "I did that". They are also much more likely to have an absent narrator, where the person writing the report is not a part of the story as told in any way. Things just happened. User-error incident reports are also significantly shorter than other types of report.

This shows that medical staff use a different approach to reporting incidents that involve people making mistakes. They are more likely to avoid naming people who might be held to blame and if they do refer to individuals it is likely to be other staff members or patients, not themselves, while the role of the person making the report remains ambiguous. This brevity and ambiguity can reduce the quality of the reports. It may make it harder for those reading them to understand exactly what happened, and so make it harder for improvements to be made to prevent such mistakes happening again.

If the way reports are written does cast a spotlight on blame culture, it may be possible to measure improvements towards a learning culture and how it changes over time. Alternatively, the work supports other work analysing these reports using different methods, that guidance and training in how to write such reports would be helpful, and in particular the need to identify those involved and always be clear about the roles played including the person writing the report. Encouraging people to write in the active voice will help to garner more precise incident details, but also involves locating responsibility and identifying individuals at least by role if not by name. However, this can only be effective in the absence of a blame culture where those involved do not need to fear the consequences of honest and open reporting. In addition this work shows that techniques from linguistics can be useful in giving a deeper understanding of incident reports and their links to blame culture.