Priority accuracy by dispatch centers and Emergency Medical Services professionals in trauma patients: a cohort study

This multisite, cohort study aimed to evaluate the accuracy of priority-setting by dispatch centers and EMS professionals. Five different EMS regions (Brabant Midden-West, Brabant-Noord, Gelderland-Zuid, Rotterdam-Rijnmond, Utrecht) and seven trauma regions in the Netherlands participated in the current study. These EMS regions transport nearly 380,000 patients annually, covering an area of approximately 6700 square kilometers with 4.7 million inhabitants. Dispatch centers professionals are nurses with additional training for dispatch [8]. The ambulances are staffed by a nurse that is licensed to provide advanced life support care and a driver that is qualified to provide medical assistance. In the Netherlands, three levels of priority exist: (i) A1 is the highest priority and implies an acute threat to the patients’ vital functions only to be excluded after an on-scene evaluation by the EMS professional, (ii) A2 priority indicates a request for care without a direct threat to life, but may involve serious damage to the patients’ health, and (iii) low-priority are scheduled transports (e.g., inter-facility transfers) [8, 9]. The participating trauma regions contain seven higher-level trauma centers and 60 lower-level trauma centers. All hospitals in these regions include a trauma-receiving emergency department and only level I trauma centers (i.e., higher-level trauma centers) are designated to treat patients in need of specialized care. Level II and level III trauma centers (i.e., lower-level trauma centers) are capable to provide care to mildly and moderately injured patients. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [10]. The Medical Ethical Committee of the University Medical Center Utrecht judged this study as not subject to the Medical Research Involving Human Subjects Act (reference number 20/500747).

All trauma patients assigned with A1 or A2 priority by the dispatch center that were transported by ground ambulance to a trauma center were included. Low-priority transports, inter-facility transfers, patients that were not transported to a participating trauma center (e.g., treatment by EMS professional on-site), duplicates, and non-trauma patients were excluded. The patient inclusion strategy is depicted in Fig. 1. A selection tool that was developed in prior research was used to automatically identify trauma patients in unfiltered EMS records [11]. This tool demonstrated an accuracy of 98.9% (95% CI, 98.3–99.2) through external validation.

Ambulance records from the participating EMS regions were prospectively collected from January 1, 2015 through December 31, 2017. These records consisted of dispatch and transport priority, patient demographics, injury locations, pre-hospital time intervals, vital signs, and the receiving hospital. The vital signs included systolic blood pressure, respiratory rate, and the Glasgow Coma Scale. Patient locations were converted to coordinates and the driving distance to destinations was calculated using Bing Maps, accounting for day of the week and hour of the day [12]. Response time was defined as the time between ambulance dispatch and arrival at the scene of injury, whereas transport time covers the time between departure from the scene of injury and arrival at the trauma center. The ambulance records were matched with data from regional trauma registries. These registries contain, among others, diagnosed injuries, ISSs, and mortality status of all admitted trauma patients. Patients that were discharged from the emergency department were assumed not to be in need of specialized care, nor to be severely injured, as verified in data from prior research [13]. EMS and hospital records were linked with unique EMS identifier. Patients with a missing identifier in either pre-hospital or in-hospital data were linked using a prediction model from prior research that demonstrated an accuracy of 100.0% [11]. Several variables, such as date of injury and transportation destination, are incorporated in this model.

The main outcome of this study was the sensitivity of dispatch and transport priority, defined as the proportion of patients requiring specialized trauma care assigned with the highest priority. Patients were classified as being in need of specialized care when they were (i) admitted to the intensive care unit after initial evaluation in the emergency department, (ii) underwent an emergency intervention within 24 h, (iii) died within 24 h after arrival at the trauma center, or (iv) when they were intubated pre-hospital (Table 1). This composite resource-based endpoint (i.e., early critical-resource use) is comparable to previously suggested definitions [11, 14‐16]. An Injury Severity Score (ISS) ≥ 16 served as a secondary reference standard for being in need of specialized care [17]. Secondary outcomes were the undertriage and overtriage rates in different priority subgroups. The proportion of patients requiring specialized care, according to both reference standards, initially transported to a lower-level trauma center was considered undertriage. Overtriage was defined as the proportion of patients not in need of specialized care that were transported to a higher-level trauma center.

Data were analyzed using R statistical software (R version 3.6.1) [18]. Multilevel multiple imputation was used to account for missing values across different sites using the package micemd [19]. Variables with missing values were dispatch priority (0.1%), transport priority (3.0%), systolic blood pressure (29.1%), respiratory rate (36.6%), Glasgow Coma Scale (13.6%), response time (1.4%), and transport time (4.1%). The predictor matrix that was used to impute these missing values included, among others, age, gender, vital parameters measured in the emergency department, and in-hospital outcomes. Forty-eight datasets were generated based on 20 iterations per set. Quantitative variables were described using median and interquartile ranges, or the mean and its corresponding standard deviation. Frequencies and proportions were computed for nominal variables. Groups were compared with the Mann–Whitney U or χ² test. p values below 0.05 were considered statistically significant. Analyses were applied to each of the 48 datasets. Point estimates of sensitivity, undertriage, and overtriage were calculated with the average results of all analyses and Agresti–Coull confidence intervals were computed, in accordance with Rubin’s rules [20, 21].

In total, 1,111,809 EMS records were collected between January 2015 and December 2017, of which 483,460 patients were transported from the scene of injury to a trauma center with A1 or A2 priority. After exclusion of non-trauma patients (n = 369,001), 114,459 patients were included (Fig. 1). The patients included in this cohort had a median age of 58.2 years, 56,798 (49.6%) were female and the dispatch center assigned 52,238 (45.6%) patients with A1 priority (Table 2). Ambulances dispatched with A1 and A2 priority had a median response time of 7.5 min and 11.0 min (p < 0.001), respectively. The 35,974 (31.4%) patients that were hospitalized had a mean ISS of 7.5 ± 6.4. In 20,285 (17.7%) patients, there were differences between dispatch and transport priorities (Table 3). Patients with a lowered priority (i.e., A1 dispatch priority lowered to A2 transport priority) more frequently had stable vital signs (p < 0.001), longer transport times (p < 0.001), fewer injuries to head or neck (p < 0.001), thorax (p < 0.001), and abdomen (p < 0.001), and had less critical-resource use (p < 0.001) than patients with a maintained A1 transport priority. Conversely, patients with a raised priority (i.e., A2 dispatch priority raised to A1 transport priority) more often had hypotension (p < 0.001), abnormal respiratory rates (p < 0.001), or an impaired Glasgow Coma Scale (p < 0.001) compared to patients with a maintained A2 transport priority. Their median transport time was comparable (0.1 min shorter, p = 0.725) and their median driving distance was 2.3 km longer (p < 0.001). Head or neck (p < 0.001), thoracic (p < 0.001), and abdominal injuries (p < 0.001) were more prevalent, while fewer injuries of their extremities (p < 0.001) were seen. They required five times more critical resources (p < 0.001).

The dispatch center has set A1 priority in 83.8% (82.5–85.0) of the patients with early critical-resource use (Table 4). The dispatch priority sensitivity in patients with an ISS ≥ 16 was 82.5% (81.0–83.8). The transport priority had a sensitivity of 74.5% (73.0–76.0) and 71.5% (69.8–73.2) according to both reference standards. Figure 2 illustrates the triage rates in the different priority subgroups according to the primary reference standard. The undertriage rate in patients with A1 dispatch and transport priority based on early critical-resource use was 22.7% (95% CI, 21.1–24.4). The 17,625 (15.4%) patients of whom the priority was lowered by the EMS professional had an undertriage rate of 34.5% (29.9–39.3). Evaluation of these priorities using the ISS resulted in undertriage rates ranging from 16.7 (15.1–18.4) to 34.8% (30.2–39.7). Otherwise, 65.5% (61.0–69.8) of the patients in need of specialized care with A2 dispatch and transport priority were transported to lower-level trauma centers. The undertriage rate of these patients decreased to 37.6% (28.3–47.8) when their transport priority was raised. According to the secondary reference standard (ISS ≥ 16), a raised priority by the EMS professional was associated with a decrease in undertriage from 55.7% (50.8–60.4) to 37.2% (27.4–48.3).