Potential utility of prehospital blood for trauma patients in a dense, urban setting: simulations of prehospital travel times

Abstract

Background

There is increasing interest in allowing emergency medical services (EMS) to transfuse blood for patients with traumatic hemorrhage. National estimates suggest that prehospital blood could save thousands of lives annually, but the usefulness of prehospital blood in dense, urban settings is unclear.

Methods

This retrospective cohort study examined all 294 patients arriving via EMS to our urban trauma center in 2016 to 2019 who received blood within 4 hours. The feasibility of prehospital blood transfusion was investigated by simulating EMS travel times. Generalized additive modeling examined associations between EMS travel time and mortality.

Results

The median [Q1, Q3] EMS travel time was 9 [6, 13] minutes. Simulations indicated that 11% (95% CI = 9%–14%) of travel times are long enough to transfuse a full unit of blood (>15 minutes). Travel time was positively associated with mortality after penetrating injury (odds ratio = 1.15) but not blunt injury (odds ratio = 1.01).

Conclusions

This study suggests a potential mortality benefit from earlier blood product resuscitation following penetrating injury. However, simulations suggest that <15% of eligible patients could benefit from prehospital blood due to short EMS travel times. Urban trauma systems should likely not expect mortality reductions from prehospital blood that are proportionate to national estimates.

Despite multiple advances in trauma care, hemorrhage and its consequences remain a leading cause of preventable death among trauma patients.¹ Experience from the military has demonstrated that early transfusion of blood products improves outcomes.² In the civilian population, similar findings were reported in the PAMPer trial, which demonstrated a significant survival benefit with prehospital plasma administration.³ However, no mortality benefit was observed for prehospital transfusion in more urban settings with shorter transportation time by emergency medical services (EMS).⁴ Transfusion of blood products during transport by EMS requires sufficient time to evaluate the patient, establish intravenous access, assess the need for transfusion, and administer an emergency blood product.

Cold stored low titer O-positive whole blood transfusion has become an important part of the resuscitation of patients in hemorrhagic shock at our comprehensive trauma center. Prehospital whole blood transfusion programs are under development and growing quickly at centers with wide catchment areas and long transport times to definitive care.⁵ The THOR-AABB group continues to advocate for the development of prehospital blood product transfusion programs.⁶ The city of Dallas, Texas, also recently began a program for prehospital packed red blood cell transfusion through Dallas Fire-Rescue.⁷ The primary aim of this retrospective study was to evaluate whether trauma patients who received a blood transfusion upon arrival to our comprehensive trauma center within Dallas would have experienced transportation times from the scene or from referring hospitals of sufficient length to administer blood products if they were available to EMS personnel under the constraints of the current protocol, which requires crews to request blood products be delivered to the scene after confirming transfusion criteria are met.⁷ We also sought to evaluate the impact of transportation time on mortality in this population.

METHODS

The institutional trauma registry of an American College of Surgeons verified trauma center in Dallas, Texas was queried to identify patients who received blood product transfusion within 4 hours of hospital arrival between January 1, 2016, and December 31, 2019, and were transported to the hospital by either ground ambulance or helicopter. This period predates the development of a low titer O-positive whole blood program at our institution as well as potential COVID disruptions. Our trauma center’s catchment area includes 54 ZIP codes covering 623 square miles, ranging from dense and urban to semirural. Focused chart review was utilized to address missingness in the registry data. Data collected included patient vital signs, time of EMS arrival to the scene (when the ambulance/helicopter stops moving, not when EMS is with the patient), EMS starting ZIP code (location of injury or referring hospital), time of EMS departure from scene (wheels up or moving time), and time of EMS arrival at the hospital. This retrospective cohort study was approved by the hospital’s institutional review board with a waiver of informed consent.

To examine the primary aim of feasibility of prehospital blood transfusion, we fit a log Poisson generalized additive model to predict EMS travel times using thin plate smooths to estimate nonlinear effects of age and injury severity score (ISS) as well as injury mechanism, EMS transportation mode (ground ambulance vs. helicopter), transfer-in status, and the EMS scene ZIP code. After evaluating model fit, the model was used to derive 10,000 samples of the posterior predictive distribution via a Metropolis-Hastings Markov Chain Monte Carlo simulation; these simulated data were used to determine the probability that future EMS travel times would be sufficiently long for a unit of blood to transfuse (>15 minutes).⁸ Although patients who receive less than a full unit of blood before hospital arrival may benefit from the transfusion, this analysis provided a conservative estimate of the share of patients who could benefit from a full transfusion in the field. The 15-minute threshold was also selected since transfusion delays >15 minutes are associated with increased risk of morbidity and mortality.⁹ We also examined the share of patients with transport <10 minutes to know if waiting for blood delivery would exceed expected transport times. This was conducted because current prehospital protocols require EMS crews to evaluate patients for transfusion eligibility, call for blood delivery to the scene, and wait up to 10 minutes for blood delivery.⁷

For the secondary aim of examining associations between prehospital transportation time and mortality, we fit a logit binomial generalized additive model. The model estimated the association between mortality and EMS travel time stratified by injury type (blunt vs. penetrating) to examine potential differential effects of travel time by injury type. The model also controlled for the nonlinear effects of age, ISS, EMS scene time, and initial scene systolic pressure via penalized thin plate smooths. Finally, as only 8 patients (2.8%) had EMS transport times >30 minutes, transport time was censored at 30 minutes for this analysis.

Continuous variables were described using median [25th percentile, 75th percentile], and categorical variables were described using frequency (percent). After chart review, data were 97% complete for variables critical for the primary aim (e.g., EMS times, ZIP codes, initial emergency department vitals, outcomes), with the largest degree of missingness occurring for scene vital signs (69%–78% complete). As scene vitals were considered likely to be missing not at random, data missingness was addressed with pairwise deletion; however, results were largely comparable when missing data were multiply imputed. All analyses were conducted in R (version 4.4.2, R Core Team, Vienna, Austria), with the additional critical libraries of tidyverse, mgcv, and gratia.^10–12

RESULTS

The initial sample included 300 patients. Six patients were excluded based on chart review findings (two with no blood product administration; two with no trauma service involvement; one with a delayed presentation >24 hours after injury; and one with an incomplete health record). The final sample included 294 patient encounters. The sample was predominantly male (76.2%) and White (69.8%), with a median age of 36 [26, 54] years. A narrow majority had blunt injury mechanisms (54.4%), with a median ISS of 22 [13, 34]. Slightly more than 11% of patients did not survive beyond the emergency department, and overall mortality was 27.2%. Additional clinical variables are summarized in Table 1.

Table 1.

Summary of clinical variables

Variable	Overall	EMS transport
		≥15 min	<15 min
N	294	50	244
Age	36 [26, 54]	32 [25, 59]	36 [26, 52]
Male sex	224 (76.2%)	35 (70.0%)	189 (77.5%)
Blunt mechanism	160 (54.4%)	32 (64.0%)	128 (52.5%)
Injury severity score	22 [13, 34]	26 [14, 34]	22 [13, 33]
Scene SBP	120 [90, 140]	118 [86, 140]	120 [90, 140]
Scene heart rate	90 [70, 110]	80 [57, 97]	92 [71, 111]
Scene GCS	15 [6, 15]	15 [3, 15]	15 [6, 15]
EMS scene time, min	11 [7, 16]	13 [8, 20]	11 [7, 16]
EMS travel time, min	9 [6, 13]	18 [16, 25]	8 [6, 11]
ED SBP	113 [87, 136]	119 [78, 143]	113 [89, 136]
ED heart rate	98 [77, 119]	92 [74, 112]	99 [77, 120]
ED GCS	14 [3, 15]	14 [3, 15]	14 [4, 15]
Full trauma team activation	241 (82.0%)	34 (68.0%)	207 (84.8%)
Transfusions in 4 h
Red blood cells, mL	825 [550, 1925]	688 [550, 1925]	1100 [550, 1925]
Plasma, mL	525 [350, 1225]	525 [350, 1225]	525 [350, 1225]
Platelets, mL	100 [100, 200]	100 [100, 200]	100 [75, 200]
Cryoprecipitate, mL	0 [0, 100]	0 [0, 200]	0 [0, 100]
MTP activated	16 (5.4%)	5 (10.0%)	11 (4.5%)
ED LOS, min	66 [28, 134]	100 [32, 170]	58 [28, 130]
Any ICU admission	212 (72.1%)	34 (68.0%)	178 (73.0%)
ICU LOS, days	4 [3, 9]	4 [2, 11]	4 [3, 8]
ED mortality	33 (11.2%)	10 (20.0%)	23 (9.4%)
Overall mortality	80 (27.2%)	17 (34.0%)	63 (25.8%)
Hospital LOS, days	6 [1, 14]	5 [1, 12]	6 [1, 15]

Data summarized as n (%) or median [25th percentile, 75th percentile].

ED indicates emergency department; EMS, emergency medical services; GCS, Glasgow coma scale; ICU, intensive care unit; LOS, length of stay; MTP, massive transfusion protocol; SBP, systolic blood pressure.

Prehospital providers were on the scene for a median of 11 [7, 17] minutes and en route to the hospital for 9 [6, 13] minutes; 244 patients (84.7%) had transport times <15 minutes and 95% had transport times <23 minutes (Figure 1). Only 21 (7.1%) patients had transport times >20 minutes, and six of these patients were transfers from outside facilities. When combining time spent on scene by EMS providers and in transit to the hospital (median: 20 [16, 38] minutes), 76.3% of patients had contact with EMS >15 minutes. However, not all this time reflected time under treatment by EMS; it included time spent searching for the patient, making the scene safe for EMS personnel, or performing extrications.

Figure 1.

Emergency medical services (EMS) travel times. The dashed line represents the empirical cumulative density function, which provides the percentile for each value.

After estimating EMS travel times, the Poisson model was deemed an adequately good fit to the data to be used to simulate travel times via the posterior predictive distribution (R² = 0.85, root mean square error = 4.97). As shown in Figure 2, the simulation indicated that the median probability of future EMS travel times being sufficiently long for a unit of blood products to transfuse fully (>15 minutes) was 0.11, with 95% of simulations estimating the probability between 0.09 and 0.14. The median probability of patients having transport times <10 minutes was 0.61, with 95% of simulations estimating the probability between 0.55 and 0.66.

Figure 2.

Random draws (n = 100) from the posterior predictive distribution of emergency medical services (EMS) travel time. Each translucent gray line represents a randomly sampled draw from the posterior predictive distribution that was created with the log Poisson model. Values with higher densities are more likely.

Descriptively, there was no appreciable difference in unadjusted censored EMS travel times between those who survived to discharge versus those who did not (alive: 8.5 [6, 12] minutes; dead: 9 [6, 12] minutes). As shown in Table 2 and Figure 3, the logit binomial model indicated that for patients with blunt mechanisms, the association between EMS travel time and mortality was essentially null (adjusted odds ratio = 1.01, 95% confidence interval [CI] = 0.94–1.09, P = 0.79). However, for patients with penetrating mechanisms, longer EMS travel times were associated with increased mortality (adjusted odds ratio = 1.15, 95% CI = 1.02–1.30, P = 0.02).

Table 2.

Associations with all-cause in-hospital mortality and EMS travel time by mechanism of injury*

Mechanism of injury	Adjusted odds ratio	95% CI	P value
Blunt	1.01	0.94–1.09	0.79
Penetrating	1.15	1.02–1.30	0.02

*Adjusted for the nonlinear effects of age, injury severity score, time spent on scene by emergency medical services (EMS), and initial scene systolic pressure. Excluded patients transferred in. Due to missing scene systolic blood pressures, the model included 202 patients (68.7% of full sample). Adjusted odds ratios for EMS travel times represent the increase in the odds of mortality per additional minute of travel within each injury mechanism category.

Figure 3.

Estimated mortality by emergency medical services (EMS) travel time and injury mechanism. Mortality estimates derived from a logit binomial model controlling for nonlinear effects of age, injury severity score, EMS time on scene, and initial scene systolic pressure. Shaded areas represent 95% confidence intervals.

DISCUSSION

Despite significant and noteworthy advances in trauma care over the last two decades, hemorrhage remains a leading cause of preventable death.^13,14 Efforts to reduce this have included devices, changes to transfusion practices, and the deployment of blood products closer to the point of injury. Recent efforts across the nation, the state of Texas, and locally within Dallas County have focused on blood product resuscitation in the field by EMS personnel.^15,16 Early experience with such programs has demonstrated that paramedic administration of blood products can improve outcomes for trauma patients.¹⁷ However, this investigation suggests that the full benefits of earlier transfusion are not likely to be universally applicable. Specifically, this 4-year retrospective cohort study of hemorrhaging trauma patients presenting to our urban trauma center indicates that as few as 11% of patients would have EMS travel times long enough to fully transfuse a unit of blood, and up to 60% of patients could face treatment delays if they had to wait for blood delivery to the scene. The study also suggests that prolonging EMS travel times for patients with penetrating injuries may increase mortality.

This study’s findings are in notable contrast to recent estimates suggesting that prehospital blood products could reduce national trauma mortality by approximately 5300 deaths annually.¹⁸ These two sets of results are not necessarily in conflict; rather, they highlight the importance of considering local factors when evaluating national estimates. For example, the estimate of 5300 fewer deaths per year is based on modeling that used a sample with a mean of 38 minutes from EMS scene arrival to hospital arrival and 42% of patients having EMS travel times >20 minutes. This is a substantially longer prehospital period than the present investigation, which had a mean scene to hospital arrival time of 24 minutes, with only 7% of patients having EMS travel times >20 minutes. Thus, it is unlikely that our urban trauma center’s patient population would see mortality benefits proportionate to national estimates.

This should not be taken to mean that no patients in our catchment would benefit from prehospital blood. Ceteris paribus, earlier initiation of transfusions will likely improve survival for some patients. However, as the proportion of each transfusion that occurs in the field decreases, survival benefits will diminish, and for patients with very short transportation, this may approach a benefit that is statistically indistinguishable from zero. As such, the average treatment effect of prehospital transfusion in the authors’ catchment area is likely to be substantially smaller than national estimates because only 11% of patients are likely to receive the full benefits of transfusion in the field, and it is unclear if earlier partial transfusion in the field is superior to expediting definitive care.

It is also worth noting that the estimate of 5300 fewer deaths annually assumes that 100% of patients in prehospital hemorrhagic shock had access to prehospital blood products.¹⁸ For these assumptions to be tenable, EMS personnel would have to have access to approximately 370,000 additional units of O-negative blood per year—roughly 4% of the nation’s total blood product utilization.^19,20 The increased demand may be manageable if dedicated donors can be recruited and maintained, as some regional trauma systems have accomplished;¹⁵ however, the risks of further stressing regional blood supplies should not be ignored.

While early programs within the US civilian trauma system have been promising, some previous studies have failed to demonstrate a benefit to prehospital blood product transfusion. This includes the COMBAT trial, which was stopped for futility after failing to demonstrate a mortality benefit from prehospital plasma administration in an urban setting.⁴ Notably, the mean transport time in the COMBAT trial was 19 minutes. The authors noted that the financial burden of a prehospital transfusion program would not be justified in an urban setting with short distances to comprehensive trauma centers. This same sentiment has been noted more recently with the additional concern that prehospital transfusions may delay definitive care when transport times are already short.²¹ Interestingly, the logistics of Dallas’s prehospital transfusions partially match those of the COMBAT trial, as patients in the trial’s plasma arm experienced transfusion delays of 6 to 7 minutes waiting for the plasma to thaw and Dallas EMS could wait up to 10 minutes for blood deliveries.^7,22

However, there is still a potentially valuable role for prehospital transfusion. Post hoc analysis of data pooled from the PAMPer and COMBAT trials suggests a benefit of prehospital plasma transfusion for transport times in excess of 20 minutes.²³ This finding is mirrored by the prehospital whole blood experiences of the Southwest Texas Regional Advisory Council, which covers 26,000 square miles and frequently has long prehospital transport times.¹⁵ In this setting, prehospital transfusions have reduced mortality for both trauma and nontrauma patients.^5,24 Practically speaking, this is due to the logistics of initiating transfusion during transport, which requires sufficient time for EMS personnel to evaluate the patient, establish intravenous access, assess the need for transfusion, and administer an emergency blood product. Additionally, it is noteworthy that the PAMPer trial exclusively and the Southwest Texas Regional Advisory Council very commonly used helicopter EMS with blood products immediately available to EMS crews, whereas 100% of the COMBAT trial and 98% of this study’s patients relied on ground EMS that faced transfusion delays. Thus, it appears that for short transport distances commonly serviced by ground EMS or when there are transfusion delays in the field, patients may be able to arrive to definitive care and have a larger team of providers available to assess and treat them before emergency transfusions can be completed by EMS providers.

Due to its retrospective design, this study was subject to multiple limitations, including inability to assess causality and susceptibility to confounding. We attempted to address the latter by controlling for nonlinear effects of probable confounding variables in our analyses, but the possibility of residual confounding remains. While there was some missing data, especially for scene vital signs, results were nearly identical when missing data were imputed. Additionally, inclusion criteria allowed for normotensive patients and those needing minimal resuscitation to be included, although sensitivity analysis (not reported) indicated that primary results were robust to this matter. Also, analysis of transport times did not include all time EMS spent in contact with patients; however, data limitations were such that this was the only period we knew with certainty that EMS crews were in contact with patients. Finally, while this study was not intended to generalize to other settings but rather to highlight potential effects of local conditions, it did intend to generalize to future patients from this locale; it is possible that these pre-COVID data will not be applicable to future patients.

In conclusion, this study sought to evaluate the feasibility of a local prehospital transfusion program within our catchment area based upon review of transport times to the hospital among patients who received blood products upon arrival. Local experience suggested that the majority of patients requiring immediate resuscitation had transport times under 15 minutes, with very few outliers beyond 20 minutes. Simulations confirmed this institutional suspicion and demonstrated empirically that our trauma center should expect <15% of patients in this cohort to be able to fully benefit from prehospital blood transfusions due to their short EMS travel times, and it could be faster to transport up to 60% of patients to our hospital than to wait for blood to be delivered to the scene. Additionally, there was no difference in mortality as a function of transport time after blunt injury. This reflects the reality of our trauma center as part of a broader trauma system with four comprehensive trauma centers concentrated within an urban environment. This analysis should not be taken as an advocation against prehospital transfusion programs, which do have great potential to reduce mortality. Rather, we are advocating for thoughtful consideration of local factors that may influence the effectiveness of such programs. Trauma care has long embraced the motto of “the right patient in the right place at the right time.” This project highlights the importance of the right intervention for the right patient in the right time and place.

Disclosure statement/Funding

The authors report no funding or conflicts of interest.