We performed a retrospective observational cross-sectional study based on anonymised public health data collected prospectively from prehospital electronic records in the canton of Vaud.

We included patients ≥ 18 years old for whom an ambulance or helicopter was dispatched between 1 January 2018 and 31 April 2023 consecutive to a trauma. We excluded patients < 18 years old and without significant injury, defined by the National Advisory Committee for Aeronautics (NACA) scale of 0 and 1. We excluded patients with a NACA of 7 who died at the scene because they were not transported to hospital. As we expected undertriage of older patients, we included all patients for whom an ambulance was dispatched and not only those suspected of major trauma based on initial assessment or retrospectively by the Injury Severity Score (ISS).

The Emergency Medical Service (EMS) of the canton of Vaud is a two-tier system in Switzerland covering more than 800,000 inhabitants. In the event of a trauma, the dispatch centre alerts the nearest ambulance. If the dispatch centre or the ambulance in the field identifies severity criteria, a mobile intensive care unit (MICU) staffed by a prehospital emergency physician is dispatched by ground ambulance or by helicopter. Severity criteria include compromised airway or breathing, haemodynamic instability, unconsciousness, spinal cord injury resulting in paralysis, penetrating injuries, limb amputation, incidents involving vehicle ejection, the need for extrication or scenarios involving more than three injured people.

There are 14 ambulance bases in the canton (between 23 and 29 ambulances can be dispatched simultaneously according to time of day), 6 MICUs with ambulances staffed by emergency physicians, a Helicopter Emergency Medical Service (HEMS). The canton of Vaud has seven general hospital and one university hospital which is both the regional trauma centre and a local hospital.

The primary outcome is admission to the resuscitation room (RR) of the regional trauma centre with trauma team activation (TTA).

We expected that older patients would have less care and less access to trauma centres than younger people. Undertriage could also be observed with less dispatched MICU and Helicopter Emergency Medical Service (HEMS). Secondary outcomes are admission to trauma centre, MICU or HEMS dispatched, prehospital interventions such as intubation, analgesia, tranexamic acid and vasoactives drugs.

We assessed outcomes by age group (< 55 years, 55–74; ≥ 75 years) and by age as a continuous variable.

As we expected that admission to RR with TTA would vary with the presence of major trauma, we estimated the risk of severity using validated prognostic models. Prehospital triage refers to the clinical assessment of the probability of life-threatening injury based on vital signs, level of consciousness, anatomical criteria and mechanism of injury. To estimate which patient should have been transported to the regional trauma centre, we estimated the risk of major trauma using validated prognostic model including these criteria available at the site of the accident. The BATT score estimates the risk of haemorrhage by predicting traumatic death from bleeding and early death [7]. The MGAP score estimates the risk of major trauma by predicting traumatic death at 28 days [8]. The BATT score has been externally validated in the UK and Switzerland and predicts the risk of death from bleeding and early death [7, 9]. We performed adjusted analysis using the BATT score and the MGAP. We did not include the Injury Severity Score (ISS) as a potential confounder because it was not estimated in the prehospital database. Moreover, the ISS is not useful for prehospital triage because it cannot be estimated in the prehospital setting and because less severe injuries can be life-threatening for older patient. The ISS is a criterion for retrospective evaluation of the triage protocol but not for deciding which patient should be transported to the trauma centre.

We first described the characteristics of the injured patients by age and compared the proportions using the Chi-2 test. Continuous variables were expressed as mean and SD or median and interquartile range, depending on the distribution. Student t-test was used to compare the means of continuous variables.

We examined the difference in prehospital interventions by age categories. We used interaction tests (Chi-squared test of homogeneity) to see whether the prehospital interventions differed by age.

As most older injured patients have a low risk of major trauma and don’t meet the criteria for full trauma team activation, we expected an interaction with the BATT and the MGAP scores, which estimate the risk of major trauma at baseline. We will examine the effect of undertriage in the different categories of baseline risk. We defined different risk categories for the BATT score: risk unlikely: BATT score < 3 (mean death from bleeding < 0.5%); Low risk: BATT score 3 to 4 (< 1%); Intermediate risk: BATT score 5 to 7 (1 to 2%); High risk: BATT score ≥ 8 (10%) and for the MGAP score: low risk: MGAP 23 to 29; intermediate risk: MGAP 18 to 22; high risk: MGAP 3 to 17.

We estimated crude odds ratio by category and test the homogeneity of the association with age across categories using CHI-2 test.

We explored the relationship between age and outcome by estimating odds ratios and 95% confidence intervals adjusted with potential confounders: sex, BATT score, MGAP score, mechanism of injury (MOI), systolic blood pressure, Glasgow Coma Scale (GCS) and distance from the accident to the trauma centre. We selected covariates in a parsimonious manner, using a Direct Acyclic Graph (supplement files, Figure 1). We then developed different models to test plausible interaction: age and sex, age and baseline risk, age and MOI, age and NACA, age and BATT score, age and High energy trauma, age and penetrating trauma, age and distance to the trauma centre. Continuous variables were included in the model with their linear and polynomial terms. We fitted a multivariate model including confounders and interaction variable. As we also expected an interaction between sex and undertriage, we explored if the undertriage varied by sex.

Finally, we examined the characteristics of undertriaged older patients. As the criteria for RR admission and TTA were not collected in the database, we considered a patient with a high risk of major trauma who was not admitted in the RR to be undertriaged. We defined a high risk of major trauma as a BATT score ≥ 8 and/or a MGAP score < 18. We compared these patients with patients with the same criteria who were admitted to the RR with TTA.

All analyses were performed using STATA software (version 18.0; Stata Corp., College Station, TX, USA).

This study followed the Strobe guidelines.

Age was missing for 6 patients (0.1%). We did not report missing value for the primary outcome. The detail of missing value for each variable are summarized in the Table 1. We performed multiple imputation with chained equation for physiological parameters. We imputed 20 datasets.