Introduction
Injuries are an important cause of morbidity and mortality, both in the developed world and the developing world [
1]. Although the global burden of injuries has declined over the past years morbidity and mortality caused by injuries are still substantial [
1]. To treat trauma patients, many countries have adopted a regionalized trauma network. Trauma care within the Netherlands is set up to contain designated trauma centers (TCs) spanning eleven trauma regions. Each region has a designated level I TC for the treatment of major trauma (MT) patients (Injury Severity Score (ISS) > 15) and level II and III centers for the stable patients. With such an exclusive trauma system survival rates of trauma patients have increased the past 15 years [
2‐
5]. Within this system a trauma classification scheme (level I to III) has been established, with optimal 24/7 resources in level I TCs.
In the Netherlands, level I TCs are frequently academic teaching hospitals which have to meet minimum volume standards regarding the number of MT patients. Level II TCs offer less specialized trauma care and are not equipped for acute neuro-surgical procedures for head trauma and less prepared for severely injured patients [
6,
7]. This description highlights the differences between level I and level II TCs, however, currently there is no consensus concerning whether there is a difference in outcomes between level I and level II TCs within an established and mature trauma system [
8‐
11].
Prehospital triage guidelines are important to ensure trauma patients are admitted as quick as possible to their respective TCs. Although this concept seems simple, the decision to which hospital, and thus the right level of care a patient needs to be transported is made by emergency medical service (EMS) providers. Compliance rates of triage protocols and experience of EMS providers are decisive to make the right judgement and dependent of a large number of variables, such as assessment of injury severity and local health care context.
Whether injured trauma patients have worse outcomes when secondarily admitted via transfer to a level I TC instead of being primarily admitted is unclear [
12,
13]. Some studies suggest a difference, reporting patients with traumatic brain injury and severe injuries [
14,
15] have a survival benefit when primarily admitted to a level I TC. In contrast, other studies found no difference in outcomes between level I and level II TCs (8), examined transfers without comparison to primary admission [
16] or analyzed level I and level II TCs combined [
15,
16].
The aim of this study is to compare in-hospital mortality between trauma patients primarily admitted to a level I TC, trauma patients primarily admitted to a level II TC and patients secondarily transferred to a level I TC. The secondary aim was to identify predictors for primary admission to a level I or level II TC, and for secondary admission from level II to level I TCs via transfer.
Methods
Study design
Data for this study were obtained from trauma region Southwest of the Dutch National Trauma Registry (DNTR), which is a database that is maintained by 11 administrative TCs nationwide. The DNTR handle general inclusion criteria; all patients admitted to the emergency department (ED) within 48 h after trauma, followed by either hospitalization, transfer to other hospitals or death are included, excluding patients that are dead on arrival. Information on patient demographics, prehospital care and injuries are coded using the Abbreviated Injury Scale (AIS), and outcomes are registered [
17]. In total, the region Southwest Netherlands consists of one Level I TC and five Level II TCs.
A retrospective cohort study was performed including all trauma patients ≥ 18 years who were admitted to the level I or level II TCs between January 1 2015 and December 31 2018. Transfer is defined as primarily presented at a level II TC and transferred to a level I TC within 48 h. Patients with isolated burn injuries were excluded because these patients are treated at one of the three nationally coordinated burn centers, of which one is located within the trauma region Southwest. Patients treated in level III TCs or originating from level III TCs were excluded due to the large difference in case mix of patients presented and/or admitted to level III TCs compared to patients admitted to level I and II TCs.
The primary outcome measure was in-hospital mortality.
Type of Hospital Admission: primary and secondary
Patients were divided into three groups: trauma patients presented and admitted at the level I TC (PA level I), trauma patients primarily presented at a level II TC followed by a secondary transfer to the level I TC (ST level I), and trauma patients presented and admitted at a level II TC (PA level II).
Statistical analysis
First, a descriptive analysis was executed for the three patient groups regarding patient characteristics, injury characteristics and outcome characteristics. For continuous and ordinal variables, medians with 25th and 75th percentiles were reported. For nominal variables, frequencies with percentages were reported. The transfer group was further divided into ISS > 15 and ISS < 15 for a subgroup analysis.
Second, missing values were imputed with multilevel multiple imputation dependent on mechanism of missingness [
18‐
20]. Outcome measures were not imputed since these variables had no missing values.
Third, a random effects logistic regression model was made to evaluate the association between type of hospital admission and in-hospital mortality. This provided an unadjusted estimate. After that, we added the following confounders to the model: age, gender, mechanism of injury, ISS, and prehospital vital parameters (systolic blood pressure (SBP), Glasgow Coma Scale (GCS), respiratory rate (RR)). Additionally, an ISS > 15 (MT) subgroup analysis was done. Continuous variables were tested for non-linearity using restricted cubic splines. When non-linearity was assumed, variables were split into two continuous linear variables with the optimal cut-off achieved from the restricted cubic spline [
21].
To identify predictors determining whether trauma patients are primarily admitted to a level I or level II TC and to identify what predictors determine secondary admission to the level I TC univariate analyses were performed using two logistic regression models, one for the outcome primary admission to level I versus secondary transfer to level I and one for the outcome primary admission to level II versus secondary transfer from level II to level I. The following variables were analyzed for an association with hospital admission: age (continuous), sex, mechanism of injury, AIS scores for head, chest, abdomen and extremities (continuous), and prehospital vital parameters (SBP, GCS, RR, continuous).
After that, two multivariable logistic regression models were created, one with outcome primary admission to level I versus secondary transfer to level I and one for the outcome primary admission to level II versus secondary transfer from level II tot level I, in which the following variables were analyzed for an association with hospital admission: age (continuous), sex, mechanism of injury, AIS scores for head, chest, abdomen and extremities (continuous), and prehospital vital parameters (SBP, GCS, RR, continuous). Continuous predictors were tested for non-linearity using restricted cubic splines and non-linear variables were cut into categories based on the restricted cubic spline.
Data were analyzed using the R Software Environment (version 3.5.1, the R Foundation for Statistical Computing, Vienna Austria). This study was done in accordance with the STROBE Statement.
Discussion
The present study demonstrated no difference for in-hospital mortality between patients that are primarily admitted at level I TCs, patients that are transferred from level II to level I TCs, and patients that are primarily admitted at level II TCs. Furthermore, the present study found several possible predictors regarding type of admission of trauma patients. Compared to patients primarily presented at level I TCs, transferred patients were found more likely to have a higher GCS, a lower SBP, more severe head and spine injuries, and less severe face, thorax, lower extremities and skin injuries. In addition, transferred patients compared to patients primarily presented at level II TCs, were more likely to be younger, have a lower GCS, more severe head, neck, spine and abdominal injuries.
Differences in clinical outcomes measures between level I and level II TCs have been studied in the past [
22,
23]. The present study demonstrated similar in-hospital mortality for level I and level II TCs, and transfer patients. These findings were previously reported for general trauma populations [
22,
23]. Other studies did find a difference in mortality between level I and level II TCs [
6,
11,
24]. A possible explanation for such contradicting results is that trauma systems in high-income countries have matured. Additionally, it is possible that geographical factors play a more prominent role in larger countries in patient allocation. Another possible explanation is that analyses were limited by case-mix differences, therefore not adjusting sufficiently for confounders. The present study found no difference for in-hospital mortality in MT patients. This is not in accordance with findings by other studies [
23]. A possible explanation is that while the trauma system in the Netherlands has been maturing, and along with public health in general, mortality rates have steadily been decreasing, also among MT patients. Also Dutch Level II TCs do not have neurosurgery available which explains the high amount of secondary referred neurotrauma patients. Studies often exclude transfer patients, the studies that do include transfer patients are in agreement with the findings of the current study and find no difference in outcomes in patients primarily and secondarily admitted to level I and level II TCs [
12,
13,
25,
26].
This study found several possible predictors regarding admission of trauma patients. Compared to patients primarily presented at level I TCs, patients transferred from a level II TC to a level I TC are more likely to have a higher GCS, a lower SBP, more severe head and spine injuries, and less severe facial, thorax, lower extremities and skin injuries. If compliance with prehospital triage is assumed, this demonstrates that patients with unnoticed or not yet apparent signs of TBI are more likely to be transferred when first admitted to a level II TC. This assumption is consistent with previous studies describing that most deaths caused by undertriage are secondary to severe TBI [
27].
Compared to patients primarily presented and admitted to level II TCs, transferred patients were more likely to be younger, have a lower GCS and more severe head, neck, spine and abdominal injuries, indicating injuries of the central nervous system. Other studies found similar results, showing that transferred patients presented with more severe head and abdominal injuries and with lower GCS [
16,
26]. The findings concerning age could be explained by a study which found nearly half of older trauma patients to be undertriaged [
28]. However, another study found transferred patients to be older [
26]. These different findings might be due to undertriage of older trauma patients resulting in these patients staying in non-level I TCs whereas discovery of these undertriaged patients could result in an increase of transfers.
Prehospital identification of MT patients and identification of MT patients at the ED could better focus on specific subgroups that are often undertriaged. The results show that triage leading to admission to a level 1 TC does not always coincide with an ISS > 15. This raises the question whether the ISS > 15 is a reliable tool regarding the identification of patients that are in need of an admission to level I TCs. A possible alternative suggestion could be a more multidimensional approach by not just taking into account the three body regions with the highest AIS scores as done when calculating the ISS but to also asses which body regions have AIS scores above zero, paying particular attention to head, neck and spine. An example of this is the New Injury Severity Scale (NISS), which takes into account the three highest AIS scores regardless of body region whereas the ISS takes into account the three highest AIS scores from the three different most severely injured body regions. The NISS has been proven to identify more MT patients than the ISS [
29]. The trouble with identifying patients that need level I trauma care and patients that need level II trauma care or lower level care is a phenomenon that encompasses many countries and triage protocols, as findings show nearly all triage protocols are unable to properly identify severely injured patients [
30].
Limitations
The present study included validated data of a large number of patient records. However, the retrospective design is traditionally limited with several biases. The biggest limitation of our study is the fact that there could be a negative selection bias because transferred patients tend to have worser outcomes than level II TC patients since that is the reason for a transfer to a Level I TC. In non-randomized trials, the observed study results observed may be because of unmeasured factors or variables. There is a chance of confounding bias due to case-mix differences. Additionally, the sample size of the transferred patient group could be too small and therefore lacking in power. Results of the transfer group should be interpreted with caution, especially regarding in-hospital mortality. It is possible there is a statistically significant difference in mortality that was not detected due to the small sample size. Another limitation is focusing on in-hospital mortality only Further research regarding outcomes such as quality of life and functional outcomes could also provide more information about which level of trauma care is appropriate for certain trauma populations.
Acknowledgements
Author collaborator group Dutch Trauma Registry Southwest: J.M. van Buijtenen: Department of Surgery, Sint Franciscus Gasthuis, Kleiweg 500, 3045 PM, Rotterdam, The Netherlands. P.T. den Hoed: Department of Surgery, Ikazia Ziekenhuis, Montessoriweg 1, 3083 AN, Rotterdam, The Netherlands. T.S.C. Jakma: Department of Surgery, Albert Schweitzer Ziekenhuis, Albert Schweitzerplaats 25, 3318 AT, Dordrecht, The Netherlands. G. de Klerk: Department of Surgery, Admiraal de Ruijter Ziekenhuis, ‘s-Gravenpolderseweg 114, 4462 RA, Goes, The Netherlands. G.R. Roukema: Department of Surgery, Maasstad Ziekenhuis, Maasstadweg 21, 3079 DZ, Rotterdam, The Netherlands.