I recently read a paper in Trauma Surgery and Acute Care Open questioning the relationship between trauma audit filters and opportunities for improvement (OFI). Having concentrated on performance improvement for decades, I was taken aback by their conclusion that no individual or combination of filters performed well in predicting OFIs in their trauma patients.
Since their conclusion is in stark disagreement with my impressions of the value of audit filters, I, of course, had to read the paper in its entirety to see where my thinking had gone wrong. Or theirs.
Researchers in the Emergency & Internal Medicine and Public Health departments at Karolinska University Hospital conducted a retrospective review of 10 years of registry data on audit filter usage and subsequent opportunities for improvement identified at their morbidity and mortality conferences. These are the equivalent of the multidisciplinary trauma PI committees found at US trauma centers.
They found that the filters they studied had poor sensitivity and positive predictive value. The filters were:
- SBP<90
- Death within 30 days
- ISS>15 and no trauma activation
- Massive transfusion
- GCS<9 without intubation
- ISS>15 and not admitted to ICU
- >60 minutes to first major intervention
- >30 minutes to first CT
- Liver or spleen injury
- No anticoagulants within 72 hours after TBI
Without deeper reading, the authors imply that using static filters like these is of little value and that new tools should be developed.
Of course, I have to disagree. There are several issues here.
- The trauma morbidity & mortality conference (or multidisciplinary trauma PI committee in the US) should not be the only level of PI analysis. It was not clear if there were corresponding primary review or secondary review processes at this hospital. If every case with a filter violation is sent to the committee, there will undoubtedly be many that are uninteresting and would have been filtered out at our lower levels of review.
- I worry that this M&M conference is focusing only on the patient in question. The goal of any PI review is to help protect the next similar patient. They may not tolerate the relatively minor care issues that the current patient did.
- It’s important not to focus only on patient harm. Many audit filters flag items associated with potential process problems. There is no guarantee of finding an OFI. The “greater than 60 minutes to first intervention” filter is a good example. Most of the time, deeper analysis will identify legitimate reasons for the delay. But the entire process must be dissected to ensure there weren’t any opportunities for improvement that need to be addressed. Most patients will suffer actual harm from the issue. But some could.
- Trauma PI focuses more on high acuity, low occurrence events. Pulmonary embolism is rare but potentially deadly, so we devote a lot of time trying to streamline our systems for providing adequate and timely VTE prophylaxis. Thus, there will be a lot of PI “overtriage,” leading to low sensitivity and positive predictive value. It’s the price we pay to try to eliminate these “never” events.
Bottom line: In my opinion, audit filters offer a checklist of shortcuts that may identify potential opportunities for improvement. A filter violation does not guarantee that we will find anything. It doesn’t matter if the patient did well or did poorly. We certainly learn important lessons when the eventual outcome is suboptimal. But we can learn just as much by finding irregularities in care that caused no harm in this patient. It might for the next one.
A key filter that I use is something I call the grandmother test. Would I be happy if my grandmother (or my child or spouse) received this care? If not, it needs to be scrutinized even if this patient went home happy and well.
Reference: Performance of individual audit filters in predicting opportunities for improvement in adult trauma patients. Trauma Surgery & Acute Care Open. 2025;10:e001808. https://doi.org/10.1136/tsaco-2025-001808