Multicenter observational prehospital resuscitation on helicopter study

Abstract

Background: Earlier use of in-hospital plasma, platelets, and red blood cells (RBCs) has improved survival in trauma patients with severe hemorrhage. Retrospective studies have associated improved early survival with prehospital blood product transfusion (PHT). We hypothesized that PHT of plasma and/or RBCs would result in improved survival after injury in patients transported by helicopter.

Methods: Adult trauma patients transported by helicopter from the scene to nine Level 1 trauma centers were prospectively observed from January to November 2015. Five helicopter systems had plasma and/or RBCs, whereas the other four helicopter systems used only crystalloid resuscitation. All patients meeting predetermined high-risk criteria were analyzed. Patients receiving PHT were compared with patients not receiving PHT. Our primary analysis compared mortality at 3 hours, 24 hours, and 30 days, using logistic regression to adjust for confounders and site heterogeneity to model patients who were matched on propensity scores.

Results: Twenty-five thousand one hundred eighteen trauma patients were admitted, 2,341 (9%) were transported by helicopter, of which 1,058 (45%) met the highest-risk criteria. Five hundred eighty-five of 1,058 patients were flown on helicopters carrying blood products. In the systems with blood available, prehospital median systolic blood pressure (125 vs 128) and Glasgow Coma Scale (7 vs 14) was significantly lower, whereas median Injury Severity Score was significantly higher (21 vs 14). Unadjusted mortality was significantly higher in the systems with blood products available, at 3 hours (8.4% vs 3.6%), 24 hours (12.6% vs 8.9%), and 30 days (19.3% vs 13.3%). Twenty-four percent of eligible patients received a PHT. A median of 1 unit of RBCs and plasma were transfused prehospital. Of patients receiving PHT, 24% received only plasma, 7% received only RBCs, and 69% received both. In the propensity score matching analysis (n = 109), PHT was not significantly associated with mortality at any time point, although only 10% of the high-risk sample were able to be matched.

Conclusion: Because of the unexpected imbalance in systolic blood pressure, Glasgow Coma Scale, and Injury Severity Score between systems with and without blood products on helicopters, matching was limited, and the results of this study are inconclusive. With few units transfused to each patient and small outcome differences between groups, it is likely large, multicenter, randomized studies will be required to detect survival differences in this important population.

Level of evidence: Level II.

Figure 1. Distribution of ISS for study sites

Figure 1 presents boxplots of the distribution of ISS for each center of the study. Note that the medians of the sites with blood available on the helicopter all have distributions that are shifted higher than those of the sites without blood available. This shift shows that the ISS of patients arriving to sites with blood available prehospital are generally higher than the ISS of patients arriving to sites without blood available prehospital.

Figure 2. Distributions of propensity scores among matched and unmatched patients

Panel A shows the distribution of propensity scores for all patients in the highest risk group who did not receive prehospital transfusion (PHT) (n=916, blue bars) relative to the distribution of propensity scores for all patients in the highest risk group who did receive PHT (n=142, red bars). Panel B shows the distribution of propensity scores in the matched patients who did not receive PHT (n=66, blue bars), relative to the distribution of propensity scores in the matched patients who received PHT (n=43, red bars). Among those patients who did not receive a PHT, very few had a high propensity score. Similarly, among those patients who did receive a PHT, very few had a low propensity score. This striking difference produces a small area of overlap between the two distributions (denoted by the purple area), representing the limited pool of similar patients across treatment groups from which to form suitable matched pairs. As expected, the distribution of the matched sample shows better overlap in purple, but still represents a small number of patients compared to the total population. Also note that the scale on the y-axis is different for panels A and B.

Figure 2. Distributions of propensity scores among matched and unmatched patients

Panel A shows the distribution of propensity scores for all patients in the highest risk group who did not receive prehospital transfusion (PHT) (n=916, blue bars) relative to the distribution of propensity scores for all patients in the highest risk group who did receive PHT (n=142, red bars). Panel B shows the distribution of propensity scores in the matched patients who did not receive PHT (n=66, blue bars), relative to the distribution of propensity scores in the matched patients who received PHT (n=43, red bars). Among those patients who did not receive a PHT, very few had a high propensity score. Similarly, among those patients who did receive a PHT, very few had a low propensity score. This striking difference produces a small area of overlap between the two distributions (denoted by the purple area), representing the limited pool of similar patients across treatment groups from which to form suitable matched pairs. As expected, the distribution of the matched sample shows better overlap in purple, but still represents a small number of patients compared to the total population. Also note that the scale on the y-axis is different for panels A and B.