A A A
CHI+MED logo banner
Loading

Writing incident reports

Key points

  • Working with NHS England, we have reviewed over 8,000 incident reports about the use of medical devices. We have demonstrated the limitations of the reports for learning about design, training and practice about infusion pumps.
  • In particular the free text fields of incident reports are open to many interpretations.
  • This limits their value for deep learning beyond the local organisation, though they may have value for spotting general trends.
  • Reports also typically focus on local factors without regard for wider policy implications.
  • This has led to our making recommendations to NHS England guidelines for training people in writing incident reports

Background
Reports on untoward incidents are collected by healthcare organisations and submitted to the National Reporting and Learning Service (NRLS). NRLS reports comprise both form and free-text fields. Most analyses of incident reports have focused attention on the form fields, or on a small number of serious incidents. In principle, the free text fields of reports should contain the most information, and thus provide the best support for learning. Given our focus on the design and use of infusion devices, we were keen to investigate how useful incident reports are for informing future design and use. We were given access to over 8000 anonymised incident reports relating to infusion devices.

It quickly became clear that there was very little information in the reports that would inform design, because it was very hard to link the details contained in the narrative with a single make and model of infusion device or what user interactions with the devices had taken place. Our attention therefore turned to the question of what could be learned from the free text fields of reports, covering both minor (no harm / low harm) incidents as well as more serious ones.

Writing better incident reports
In the first study we focussed on how people detect errors in infusions. We found that many were detected during routine safety checks – e.g., at the start of a shift. This indicates that increasing the number and quality of such checks, and making it easy for staff to compare the programming of pumps with the patient’s prescription, could increase the number of errors detected, and thereby improve patient safety. We followed this up with an analysis on the use of language in the reports. As a result we have recommended to NHS England that a style guide is introduced and staff are trained in writing incident reports.

Policy issues
In a parallel study, we analysed a different subset of reports looking for misalignments between the patient, product, practice, and policy: the “4 Ps” framework devised by Pat Baird’s of Baxter Healthcare. Policy issues rarely featured: reports focused almost entirely on the “sharp end” of what went wrong, and very rarely on the “blunt end” of underlying organizational factors. We had also hoped to develop and test a protocol for analyzing free text fields. However, independent analysts could not reliably agree on the factors that contributed to incidents. This shows how difficult it is to have confidence in the interpretation of the free text without access to other contextual information about the situation in which the incident occurred.

Improving Learning
Overall. the free text of incident reports were found to be a poor resource for deep learning. We have highlighted some of their limitations for learning about design, training and practice: the free-form text entry fields are used inconsistently, often giving insufficient information about who was involved in an incident, what actually happened, what devices were involved, or what the circumstances were. Consequently we have made recommendations to NHS England about training in writing incident reports.

See also

Publications
Amos, S. (2015) The detection of errors in infusion rates on infusion devices: an analysis of incident reports from the National Reporting and Learning System (NRLS). MSc dissertation.