Trauma performance improvement has always been around in one form or another. The processes were relatively informal and very institution dependent until the designation and verification of trauma centers became more commonplace in the 1980s. And although the nomenclature and structural elements of PI have changed over the years, the basic framework is much the same.

Harborview Medical Center in Seattle is one of the grand-daddies of US trauma centers, as well as one of the busiest in the country with over 6,000 admissions per year. They have a very mature trauma PI program, and many of their alumni are movers and shakers in both trauma care and performance improvement. They published a paper fifteen years ago that detailed an analysis of their own preventable deaths from 1996 to 2004.

This group has now duplicated their study using a more modern time-frame, from 2005 to 2014. They sought to identify the similarities and differences in their experience between the two decades, and how their continuous performance improvement activities impacted patient mortality. The level of detail maintained within their PI program is impressive, allowing them to finely dissect the timing and causation of adverse events.

As in most PI programs, there are multiple levels of analysis that take place for each case. At the lowest level, there are meetings of the trauma PI program personnel. There is also a weekly M&M conference. Significant quality concerns can then be escalated to a departmental quality officer, or pushed up to the multidisciplinary trauma PI committee or hospital-wide quality committee. Harborview scrutinizes not only the usual adverse events, but they also dissect unexpected mortalities defined by TRISS methodology (mortality with calculated probability of survival > 50%).

For the purpose of this study, all deaths were included that were the result of an error. “Error” was broadly defined as “the failure of a planned action to achieve its desired goal.” There were two general classes of error: errors of commission (something that was done) and errors of omission (something that was not done). Several error types were also defined, including an error in diagnosis, treatment, prevention. or other. The phase of care where the error occurred was also analyzed, and included initial assessment and resuscitation, secondary survey and tests, interhospital transfer, initial interventions, ICU care, ward care, and rehabilitation. Even the cause of error was analyzed, and was divided into input errors where information was incorrectly perceived, intention error where an incorrect plan was formulated, and an execution error where the wrong action was performed.

Some Things Stayed The Same

The number of deaths associated with an error remained constant across the two ten-year study periods. During the early period, the authors identified 2,594 deaths from over 44,000 admissions, and only 64 were associated with an error (2.5%). In the later period, there were 2,659 deaths out of nearly 61,000 admissions, with 77 associated with an error (2.9%). This was remarkably constant across the two decades of study.

In both study phases, errors in treatment were most common at about 60%, followed by prevention errors (20%), and errors in diagnosis (13-21%). Errors of intention (incorrect treatment plan) were the most common cause (50%), with errors of input or execution evenly splitting the rest.

There has been speculation over the years that seasonal volume swings and the start of new residents in July might increase errors and deaths. The authors were able to analyze their data by season and by month and found no correlation with either of these. Similarly, the change to the 80-hour work week has often been bemoaned, but no differences in error-related deaths were attributable to that, either.

The Significant Differences

But some things did change. Death from failure to control hemorrhage declined significantly from 28% to 13%. This was most likely due to fewer delays in control of thoracic, abdominal, or pelvic hemorrhage. The authors suggested that this may have been a result of their implementation of a number of practice guidelines to avoid this problem.

Offsetting this decline in bleeding deaths was an increase in airway and respiratory deaths from 16% to 32%. This was driven by a number of esophageal intubations in the field and aspiration events in elderly patients on the ward.

There was also an obvious shift in the time of occurrence of death. During the early period, this tended to occur earlier in the hospitalization, frequently in the ICU. In the later study period, death on the hospital ward was more common.

Finally, the average age and injury severity increased over the duration of the study. Patients dying during the early period had an average age of 46, versus 57 in the later period. Average ISS increased somewhat as well, from 26 to 29. And the days in hospital until death increased from 1 to 2.

Bottom Line

As you can see, the PI processes at Harborview are extremely robust. They have adopted a detailed (and useful) taxonomy to slice and dice their mortality statistics. PI personnel analyzed their data at a variety of levels, and have implemented numerous practice guidelines based on their findings as part of their loop closure process. This scrutiny and the interventions developed certainly led to the improvements seen between the two study periods.

Here are some takeaways from this interesting paper:

• A good trauma PI program is generally self-sustaining and can improve care for decades. You cannot cut corners and maintain a high-quality trauma program.
• Thorough analysis requires an understanding of the elements of the item being studied. In this case, mortality was divided into classes, types, causes, locations, times, and more. These were very helpful in teasing out meaningful trends. Develop a meaningful taxonomy for the problem at hand. Don’t rely on overly complex, one size fits all taxonomies like TOPIC. They urge you to collect more detail than needed for your problem.
• Practice guidelines can play a major part in reducing preventable deaths from errors. As major issues were identified by the PI program, guidelines were rolled out to address them. The authors listed no less than ten that were implemented during the total study duration. Look for opportunities to adopt your own practice guidelines to combat the problems you find. And don’t be embarrassed to borrow (<steal>) from others to avoid reinventing the wheel!
• The Heisenberg principle applies to trauma PI. Studying the problem changes it. And with well thought out action plans, these changes should be in the direction of reducing errors.
• The shift toward more elderly trauma patients will continue to have a significant impact for years to come. Our population will only get older, as will your trauma patients. You must incorporate guidelines for care of the frail and elderly in your trauma program processes. This includes guidelines for initial management, proper placement in the hospital (ICU vs stepdown vs ward), goals of therapy discussions, therapies, sleep hygiene, medication selection, and much more.

Reference: Changes in error patterns in unanticipated trauma deaths over 20 years: in pursuit of zero preventable deaths. J Trauma publish ahead of print, August 6, 2020.