Introduction

In many countries, physicians and trauma surgeons are occasionally dispatched to the scene of injury in cases of severe trauma. Although the effect of physician-staffed transport on patient outcomes has not been fully evaluated, previous studies have associated it with reduced mortality rates of patients with trauma due to better proficiency in procedures and diagnostic techniques compared to transports staffed by paramedics only1,2,3,4,5,6.

The primary patient transport methods used are ground and air ambulances. Previous studies comparing transport methods found better outcomes when patients were transported with helicopters7,8,9, whereas others found no differences between the two modes10,11. However, no study has clarified which transport method is better when a physician is on board. Although transportation time may be longer with physician-staffed ground ambulances (GA) than a physician-staffed air ambulance (AA), repeated assessments and therapeutic procedures can be performed during transport, whereas confined space and communication challenges in AAs limit such interventions5,6,9.

Physician-staffed GA are commonly used in Japan. Physician-staffed AA are also available; however, their operation is depends on the weather and daylight hours. Owing to the high operating costs and limited operating time of AA, the comparative effectiveness of a physician-staffed AA over a physician-staffed GA should be evaluated from the perspective of establishing an efficient prehospital emergency system. This study aimed to compare the effects of physician-staffed AA and GA on patient outcomes using a nationwide trauma registry.

Methods

Study design and context

This retrospective descriptive study used data from the Japan Trauma Data Bank (JTDB) spanning from April 2004 to December 2021. Since established in 2003, the JTDB serves as a nationwide trauma registry mandating the registration of all trauma patients with an abbreviated injury scale (AIS) score of ≥ 3, regardless of the anatomical site of injury. There are no exclusion criteria. Throughout the study duration, the JTDB collected records from 303 hospitals wherein 242 are tertiary emergency centers. Sixty-two hospitals operate physician-staffed AA services, while GA with a physician on board is operated in 279 hospitals. The database encompasses details on injury mechanisms, modes of transportation, prehospital timelines, and baseline patient characteristics, including vital signs at the injury scene and upon emergency department (ED) arrival, procedures performed, and survival status of patients at hospital discharge. Additionally, procedures performed in the emergency and operating rooms post-admission are documented.

In Japan, the operation of prehospital physician teams, including dispatch criteria and operating hours, varies depending on the medical control area, with coverage areas differing significantly between urban and rural regions. Physicians are transported either by car or helicopter, depending on the respective medical control area’s protocol. While not always trauma surgeons, these physicians typically work in EDs and are trained to provide fundamental prehospital trauma management, such as sonography assessment, tracheal intubation, chest drainage, intraosseous infusion, and temporary hemostasis using a tourniquet.

This study adhered to the principles outlined in the 1964 Declaration of Helsinki and its subsequent amendments and received approval from the Ethics Committee of Tsuchiura Kyodo General Hospital (approval number: 2022FY10). Due to its retrospective nature, the requirement for informed consent was waived by the Ethics Committee of Tsuchiura Kyodo General Hospital, with an opt-out method implemented, allowing individuals to decline participation via online information disclosure at the hospital.

Patient selection

Inclusion criteria included patients aged ≥ 15 years who sustained injuries with an Injury Severity Score (ISS) of > 15 and patients who were directly transferred from the injury scene with the availability of specific information regarding injury times, physician contact, and hospital arrival. Exclusion criteria were patients who experienced cardiac arrest during contact with paramedic, those with unsalvageable injuries (AIS = 6), individuals with missing essential data for analyses, and cases with unrealistic or outlier values in prehospital timelines, which were identified statistically and subsequently removed. To examine common emergency transport cases, cases with outlier values where the time from injury to physician contact exceeded 90 min were also excluded based on clinical perspective.

Data collection

Data extracted from the JTDB included patient demographics (age, sex, injury year), prehospital vital signs, transportation details, prehospital treatment, vital signs upon ED arrival, Glasgow Coma Scale and Japan Coma Scale (JCS) scores12, AIS, ISS, surgical interventions, and survival status at hospital discharge. The JCS has four main categories that can be subdivided according to eye response test13. It evaluates consciousness levels using a single-, double- or triple-digit numeric code, categorized into three levels of responsiveness (Supplemental Table 1). Eligible patients were categorized into two groups: those transported by physician-staffed GA (GA group) and those transported by physician-staffed AA (AA group). Patients were differentiated based on the timing of physician contact and hospital arrival, with injury times divided into four 6-h zones starting at 00:00. The primary study outcome was the in-hospital mortality rate.

Definition and outcomes

Consciousness levels at the injury scene were assessed using the JCS12. Injury time aligned with the time of the emergency medical service call. Day and night shifts were defined as 08:00 a.m. to 17:59 p.m. and 18:00 p.m. to 07:59 a.m., respectively. Primary outcome measures included in-hospital mortality, while secondary outcomes comprised the time from injury to ED arrival, time from injury to physician contact, prehospital treatment, and changes in vital signs at physician contact and ED arrival.

Statistical analysis

Given the uneven distribution of patient characteristics between the GA and AA groups, propensity score matching analysis was employed for outcome comparison. A logistic regression model estimated the propensity score for each patient based on various factors, including age, sex, injury mechanism, vital signs, AIS, and ISS, which were selected based on a chronological perspective and subject matter knowledge. Propensity score matching yielded 1:1 matched pairs from the GA and AA groups, with match balance assessed using the absolute standardized mean difference. Subgroup analysis within the propensity score-matched cohort explored potential beneficiaries of physician-staffed GA transport management.

Several sensitivity analyses were performed: (i) a multivariate logistic regression model using the same variables used in the primary analysis, (ii) a propensity score-matching analysis, in which age, pre-hospital vital signs, and vital signs upon ED arrival were treated as categorical variables, and (iii) a propensity score-matching analysis in the population among whom the time from injury to physician contact was less than 120 min. Statistical analyses were conducted using R 4.3.1. A significance level set at two-sided p < 0.05 for all analyses.

Results

A diagram of the patient selection process is shown in Fig. 1. Data from 16,181 patients were used for the analyses, with a study period from April 2004 to December 2021. The number of patients in the GA group was 4,170, while that in the AA group was 7,838. After propensity matching, 3,508 patients in each group were matched. The backgrounds of all patients and matched patients are presented in Table 1. All the absolute mean differences of the variables used for propensity score estimation was < 10%, indicating well-matched balance. The median ages of the matched patients in the GA and AA groups were 61 and 62 years, respectively. The GA and AA groups comprised 2509 (75.4%) and 2511 (71.5%) male patients, respectively. The most common type of trauma was blunt trauma, which accounted for 3397 (96.8%) and 3403 (97.0%) cases in the two groups, respectively. Regarding the AIS codes for trauma, both groups had a median score of 3 for the head and chest, and a median score of 25 for ISS.

Fig. 1
figure 1

Diagram of the process of patient selection. AIS abbreviated injury scale, ISS injury severity score.

Full size image
Table 1 Characteristics of the patients before and after propensity score matching.
Full size table

Regarding the primary outcome, in-hospital mortality was observed in 810 patients (23.0%) in the AA group and in 894 patients (25.4%) in the GA group; the AA group had a significantly lower mortality rate (the odds ratio, 0.86; 95% confidence interval, 0.77–0.96, p < 0.001). A comparison of prehospital treatment, time from injury to physician contact, time from injury to ED, vital signs in the ED, hospital treatment, and mortality for both groups is shown in Table 2. In prehospital treatment, chest compression, intubation, mechanical ventilation, pelvic binder use, and tourniquet use were performed more frequently in the GA group. Both the time from injury to physician contact and the time from injury to ED arrival were shorter in the GA group than in the AA group. The vital signs in the ED were similar between the two groups. Surgical intervention was performed more frequently in the AA than in GA group (48.0% vs. 52.0%). Subgroup analysis (Fig. 2) showed a trend towards lower mortality in the AA group for most items.

Table 2 Post-transportation procedures and information in the ED.
Full size table
Fig. 2
figure 2

Subgroup analysis for the effect of air ambulance transport on in-hospital management. CI confidence interval.

Full size image

The sensitivity analyses using a logistic regression model, a propensity score matching model using categorized variables, and a propensity score-matching model in a population among whom the time from injury to physician contact was less than 120 min showed similar results to the main analysis (Supplemental Tables 2, 3, 4, and 5).

Discussion

This study compared the mortality rates of patients with severe trauma transported by physician-staffed GA with those transported by physician-staffed AA. More frequent treatments during transportation and shorter transport times to a hospital were observed in the GA group, while the AA group had a lower mortality rate. This study represents the first attempt to directly compare the efficacy of physician-staffed GA transport with that of physician-staffed AA transport. Regarding in-hospital mortality, no specific significant interaction was observed in the subgroup analysis, indicating the consistency of the main results. The sensitivity analysis, using categorical variables for age and vital signs based on clinically meaningful cut-off values, demonstrated similar results from the main analysis, which corroborated the robustness and clinical significance of the main analysis.

Considering that the survival rate was lower in the GA group, which received more treatments during transport and had a shorter total transport time, suggests that improving patient survival rates simply by prehospital treatment or transport time is challenging. In many cases in Japan, patients with cardiac arrest are not transported by AA but by GA with a physician1,3,6,14,15. Although this study excluded patients who experienced cardiac arrest at the time of contact with paramedics, those experiencing cardiac arrest at the time of contact with physicians could not be excluded because of a lack of information, potentially leading to worse outcomes in the GA group. This can be considered from the fact that the PR and HR0-39 bpm values are higher in the GA group for the categorized pre-hospital and emergency department vital signs. In addition, patients in the AA group were likely to be transported to facilities providing high-level trauma care, such as favorable accessibility for appropriate emergency surgery, since AA can transport a patient to the appropriate hospital in the distant place in a shorter time than GA. Notably, the AA group had significantly more cases of surgical intervention than the GA group (52.0% vs. 48.0%).

In this study, the AA group had a longer transport time than did the GA group. Similarly, the time from injury to physician contact was longer in the AA group, with ED arrival delayed by an additional 10 min. One possible reason for this is the nature of AA, where any therapeutic intervention during flight is challenging owing to unstable circumstances, such as noise and rolling in confined spaces9,10,14,15,16. Flight physicians are generally required to complete the necessary treatments before take-off, potentially extending total prehospital time17,18. Another reason could be the time required to transfer a patient from a primary-dispatched GA, which is an ambulance dispatched from a fire department to a physician-staffed AA. The rendezvous of AA and primary-dispatched GA is permitted at pre-specified limited locations, which sometimes requires additional time for ground transportation to contact AA. Because accessibility to large hospitals is favorable in most of Japan, the advantage of high helicopter speed may be canceled out1,14,15. While it takes longer transport time in the AA group, it might be beneficial for patient outcomes that a flight physician achieves temporal fixation of physiological status before take-off.

Limitations

This study had several limitations. As this was a retrospective study using a large database, residual confounding factors existed, including changes in vital signs after paramedic contact, proficiency of the dispatched physician, accessibility for emergency surgery, and the level of the hospital to which the patient was transported. The information on the level of the hospital, which could have had a significant impact on survival rates, was not available in the JTDB. These potential residual confounding might have caused selection bias. In addition, regional factors were not considered. As prehospital trauma care systems in Japan are developed in each medical control area, they may differ. The direction and extent of the effects may have differed depending on the distribution density of tertiary emergency medical centers and differences in regional prehospital systems such as the criteria to select a hospital to transport. Although data from a nationwide database were analyzed, registration errors may have occurred at different institutions. Finally, because prehospital activities by physicians were not uniformly protocolized, the quality of the activities might have differed.

Conclusions

In emergency transport with physicians for critical trauma patients, air ambulance transport was significantly associated with lower mortality rates than ground ambulance transport. However, considering potential biases including the lack in information on regional and hospital disparities, further research will be needed.