MV observed the following lack of distinction in scene time for penetrating trauma mortality, which I did not give the proper attention in EMS Time and Survival from Blunt and Penetrating Trauma. I will try to correct my mistake here.
I find it interesting (good) that they actually published those graphs. I’m not sure what you can really say based on the graphs, as the error bars are way too wide to make good conclusions. For instance, there is no real difference in death rate due to scene time on penetrating injuries that can be determined by this study based on Figure 5.
Here is Figure 5.
On multivariate regression of patients with penetrating trauma, we observed that a scene time greater than or equal to 20 minutes was associated with higher odds of mortality than scene time less than 10 minutes, with an odds ratio (OR) of 2.90 (95% confidence interval [CI] 1.09 to 7.74). Scene time of 10 to 19 minutes was not significantly associated with mortality (OR 1.19; 95% CI 0.66 to 2.16).
I should have paid more attention to this, rather than just looking at the brief description of the 0-9 minute vs. the 10-19 minute scene time intervals, which are not graphed vs. mortality.
Clearly, this graph is not a graph of 0-9 minute, 10-19 minute, and 20-29 minute time periods, but there is no explanation of the numbers represented by the graph.
If we eliminate the blunt trauma part of the graph, the wide error bars on the penetrating trauma percent dying show a lot of overlap. The error bars cover as much as a 50% difference in survival in one time period, while the narrowest error bars cover over at least a 15% difference in survival.
Overlap on error bars is usually an indication that the results are not statistically significant, but it depends on the type of error bar. The text describes CIs (Confidence Intervals) when comparing the 0-9 minute vs. 10-19 minute scene time intervals. However, the graphs do not specify which type of error bar is used, so we do not know how the error bars should be interpreted.
Error bars are a graphical representation of the variability of data and are used on graphs to indicate the error, or uncertainty in a reported measurement. They give a general idea of how accurate a measurement is, or conversely, how far from the reported value the true (error free) value might be. Error bars often represent one standard deviation of uncertainty, one standard error, or a certain confidence interval (e.g., a 95% interval). These quantities are not the same and so the measure selected should be stated explicitly in the graph or supporting text.
There should not be so much overlap if there is any significant difference. It appears that the only way there is a statistically significant difference is by covering a ten minute time period.
Click on images to make them larger.
The last time period (≈ 26 minute scene time) has error bars that overlap with all of the other time periods, including the ≈2 minute scene time. The ≈2 minute scene time is the only time period where the median is not within the error bars for the ≈ 26 minute scene time.
Even the shortest scene time (≈2 minute scene time) has wide enough error bars to include the medians from two other time periods (≈6 minute scene time and ≈10 minute scene time). These would all fall within the 0-9 minute scene time group and the larger group can have enough patients to narrow the error bars. Wide error bars are often an indication of small numbers. For an example of this, look at the graph that contains the blunt trauma patients – those error bars are much narrower than the error bars for the penetrating trauma patients. The study had 16,170 blunt trauma patients and only 2,997 penetrating trauma patients. That means 5.4 times as many blunt trauma patients.
Paradoxically, the much larger group of blunt trauma patients did not produce any statistically significant difference in outcomes based on scene times or transport times. with more penetrating trauma patients, would the statistical significance disappear?
We do not know.
The authors state that the choice of time periods was made before analysis of the data, so this does not appear to be an example of trying to find the right groups to make the data fit the hypothesis.
We categorized out-of-hospital times into 10-minute intervals a priori with the intent of choosing an interval that is operationally practical, clinically feasible, and politically acceptable.
I would like to see what the graph of 0-9 minute, 10-19 minute, and 20-29 minutes would look like.
None of this suggests that the concept of the Golden Hour should survive.
It is crucial for medical researchers to critically examine concepts such as the golden hour that are widely accepted but are in fact not scientiﬁcally supported. We frequently strive to push ever higher the ceiling of medical knowledge, but we must also ensure that the knowledge base upon which we stand is solid.
Ignorance may be bliss for some people, but we will not improve outcomes for our patients by promoting ignorance.
 Emergency Medical Services Out-of-Hospital Scene and Transport Times and Their Association With Mortality in Trauma Patients Presenting to an Urban Level I Trauma Center.
McCoy CE, Menchine M, Sampson S, Anderson C, Kahn C.
Ann Emerg Med. 2013 Feb;61(2):167-74. doi: 10.1016/j.annemergmed.2012.08.026. Epub 2012 Nov 9.
PMID: 23142007 [PubMed – in process]
Lerner EB, & Moscati RM (2001). The golden hour: scientific fact or medical “urban legend”? Academic emergency medicine : official journal of the Society for Academic Emergency Medicine, 8 (7), 758-60 PMID: 11435197
McCoy CE, Menchine M, Sampson S, Anderson C, & Kahn C (2013). Emergency Medical Services Out-of-Hospital Scene and Transport Times and Their Association With Mortality in Trauma Patients Presenting to an Urban Level I Trauma Center. Annals of emergency medicine, 61 (2), 167-74 PMID: 23142007