Page 33 - JSOM Winter 2025

DETAILED METHODS AND RESULTS

Methods Results
A retrospective chart review was conducted with the ap- Of all 65 trauma patients, (23, 35%) had a fibrinogen level
proval of the University of Alberta, Research Ethics Office below 1.6 g/L. Thirty-five (53.8%) required MHP activation,
(Pro00088974). STROBE reporting guidelines were followed. including 19 (29.2%) who were part of the low fibrinogen
22
The study used Shock Trauma Air Rescue Society (STARS) group. In total, 34 of the 65 (52.3%) patients met code red
Electronic Patient Care Reports (ePCR), charts from the Uni- criteria (SBP <90) after infusion of one unit of pRBC. Of these
versity of Alberta Hospital (UAH) (level 1 trauma center) and 34 patients, 19 (29.2%) had a hospital-measured fibrinogen
the Royal Alexandra Hospital (RAH) (level 2 trauma center) level below 1.6 g/L.
in Edmonton, Alberta, and included data from the Alberta
Trauma Registry. The study period was January 2015 to Au- The first measured fibrinogen level across all patients was 1.91
gust 2019. We enrolled consecutive patients of any age, over (SD 0.96), median 1.7 g/L. The average fibrinogen level for the
the weight of 50kg, with blunt or penetrating trauma trans- patients in the high group was 2.42 (SD 0.8), median 2.25g/L,
ported by STARS who received at least one unit of packed red in comparison, the fibrinogen level for the patients in the low
blood cells (pRBC) by STARS HEMS. In total, we collected group was 0.98 (SD 0.3), median 1 g/L. Patients meeting any
data from 65 trauma patients. Included charts were abstracted code red criteria had a mean fibrinogen level of 1.5 (SD 0.8),
by JT, XC, and PC to an EXCEL spreadsheet with 10% of while patients that did not meet CRC had a mean fibrinogen
charts double extracted by SP and DO in a blinded fashion to level of 2.4 (SD 0.94) g/L.
assess inter-rater reliability. Data abstractors were trained and
experienced at abstraction. The first measured mean INR in the ED across all patients was
1.3137 (SD 0.5), median 1.2. Fourteen patients were in the
The primary outcome variable was the first measured fibrino- INR ≥1.5 group, the average INR level for patients in the ≥1.5
gen level in the ED. For the statistical analysis, the first mea- group was 2.06 (SD 0.69), median 1.8. Fifty-one patients were
sured fibrinogen level in the ED was transformed into a in the INR <1.5 group, the average INR level for patients in
dichotomous variable, where all values below 1.6g/L were the <1.5 group was 1.17 (SD 0.15), median 1.2.
classified as the low fibrinogen group, and those at or above
1.6g/L were classified as the high fibrinogen group. Secondary Primary Analysis
outcome variables and predictor variables used in the analysis For first fibrinogen level and the five primary predictors, a
are specified in Table 1. These variables were selected for anal- significant model was returned χ² =31.0, P<.001. McFadden
9
ysis based on their clinical relevance. R =0.42, indicating excellent model fit. Of the variables in-
2
cluded in the model only CRCpost emerged as a significant
Three sets of analyses were conducted: 1) The first analysis predictor of low fibrinogen levels, P=.04, OR 6.6 (95% CI
assessed the effects of the primary predictor variables on first 1.1–40.2), those with a Code Red remaining after a pRBC
fibrinogen levels in the ED, and the impact of first fibrinogen infusion had 6.6 times greater odds of having fibrinogen
levels in the ED on Discharge and 14-day mortality; 2) The <1.6g/L. The mean fibrinogen level for the CRCpost NO code
second analysis assessed the effects of the primary predictor red remains was 2.4 (SD 0.94), median 2.1 g/L, while for YES
variables on the secondary outcomes; 3) The third set of anal- code red remains was 1.5 (SD 0.8), median 1.3g/L.
ysis explored the relationship between the primary outcome
variable and secondary predictor variables. For fibrinogen level and discharge status, a significant model
was returned χ² =10.0, P=.002. McFadden R =0.14, indicat-
2
1
Direct multivariate logistic regression was used for all anal- ing adequate model fit. The odds of being discharged from
yses, except for 14-day mortality which had three outcome hospital alive were 6.8 times higher for those with a fibrinogen
levels and so multinomial logistic regression was used. All level ≥1.6g/L, P=.003, OR 6.78 (95% CI 1.96–23.5)
missing data was determined to be missing completely at
random (MCAR). As the total missing data was minimal and For fibrinogen level and 14-day mortality, a significant model
MCAR despite the small sample size, the sample was adequate was returned χ² =12.8, P<.002. McFadden R =0.14, indicating
2
2
for the intended number of predictor variables to be included adequate model fit. Fibrinogen level lower than 1.6g/L was
in the primary analyses (Tabachnik and Fidell, 2013). Variance significantly associated with 21.6 times greater odds of mor-
Inflation Factor (VIF) was used to examine multicollinearity tality from hemorrhage or multiorgan failure P=.006, OR 21.6
between predictor variables, a VIF >10 indicates high col- (95% CI 2.4–193.9). There was no association with death
linearity. For the analysis of the secondary outcome variable from other causes P=.15, OR 3.08 (95% CI 0.66–14.3).
Surgery/Angiography, SBP and SI (prehospital) had a VIF >10,
SI was dropped from the analysis as it is partly derived from There was no effect for injury type on fibrinogen level,
SBP. For First INR measured in the ED all variables except for χ² =0.55, P=.46, McFadden R =0.008, Blunt versus penetrat-
2
1
CRC had a VIF >10; CRCpre and CRCpost were determined ing P=.45, OR 1.6 (95% CI 0.46–5.7).
to be collinear along with SBP and SI (prehospital); CRCpre
and SI (prehospital) were dropped and the analysis was car- Secondary Analysis
ried forward. No VIF >10 was identified in any other instance, Prediction of discharge status based on the five primary pre-
with most VIF being around 1, indicating no collinearity. All dictor variables returned a significant model χ² =25.7, P=.002.
9
statistical analyses were conducted using jamovi, a graphical McFadden R =0.41, indicating excellent model fit. Of the
2
user interface for R. variables included in the model, CRC emerged as a significant

HEMS Predictors of Low Fibrinogen | 31