Comorbidities, injury severity and complications predict mortality in thoracic trauma

A retrospective study was performed on patients admitted to Haukeland University Hospital (HUS), a level one trauma center, between January 2009 and December 2018. Until December 2014, the patients were registered in the local hospital trauma registry and after that, in the Norwegian National Trauma Registry (NTR) [17]. The project was approved by the Regional Medical Ethics Committee in Norway under registration number 2017/293-8 [18]. Inclusion criteria were thoracic trauma with an AIS chest score of ≥ 3 and an age of 18 years or older at the day of the trauma. Patients not admitted directly to HUS and patients dead on arrival were excluded.

Baseline demographic information and clinical data were extracted from the HUS trauma registry and the NTR. The HUS trauma registry was prospectively maintained by dedicated trauma registrars which screen emergency patient admissions fulfilling predefined criteria in the local trauma registry until 2015. Thereafter, registration continued in the web-based NTR. In both registries, patients were labeled according to the AIS score (2005 version, update 2008) [19]. Patients were identified in the databases by selecting for thoracic injury AIS codes. Patient data was collected using a standardized case report form. Data on comorbidities, medication use and complications were supplemented by additional retrospective data collection. Patients with no record of comorbidities or medication use were distinguished from the cases where the absence of comorbidities was explicitly recorded. ISS, Glasgow Coma Scale (GCS), revised trauma score (RTS) and the trauma and injury severity Score (TRISS) were calculated at patient admission [16, 20, 21]. ISS ≥ 16 was applied as cutoff for polytrauma patients. Comorbidity data were scored according to the American Society of Anesthesiologists (ASA) score, Charlson comorbidity index (CCI) and comorbidity polypharmacy score (CPS) [22‐24]. Complications were scored with the trauma adapted Clavien–Dindo scale (CDS) which classifies complications upon the most intensive therapy required complication treatment [25, 26]. To give an overview of all complications to a patient, the comprehensive complication index (CI) was calculated and analyzed [27].

Primary end-point was 30-days mortality. The following factors and variables were analyzed for their contribution as a risk factor for in-hospital mortality. Age was categorized in four groups: 0–39, 40–59, 60–74 and 75 years and older. Comorbidities scored with CCI were categorized as 0, 1 and ≥ 2, according to common practice [28]. CPS was categorized as 0–9 and > 9 as appropriate [29] and also analyzed continuously. ISS was divided into three categories: minor 0–9, moderate 10–15, and severe with 16 and higher. RTS was stratified by odds of survival: low risk of death < 5% RTS ≥ 7.2, an intermediate risk of death ≥ 5% < 50% RTS ≥ 3.4 < 7.2, and a high risk of death ≥ 50% RTS < 3.4 [30]. TRISS was analyzed as a categorical variable in three categories with probability of survival (PS) 0–50, 51–75 and 76–100. ASA, Intensive care unit (ICU) and use of anticoagulants were regarded dichotomously. CDS was categorized in two groups, no or minor complications and severe CDS ≥ 3. Complications scored with CI were categorized as no or minor complications CI < 30, and severe CI ≥ 30 as previously reported [31]. The cut-off value for the CI was set at 30 to differentiate between major and minor complications based on the calculation of all possible comprehensive complication indices considering the worst complication as CDS grade two.

Categorical variables were presented using frequencies and percentages and tested using Pearson’s chi square test or Fischer exact test depending on the expected count > 5: Pearson, < 5: Fischer. Normally distributed variables were presented using means and standard deviations and tested using t tests and repeated measures ANOVAs. Non-Gaussian variables were presented as median and inter quartile range (IQR) and tested with the Kruskal–Wallis test.

Separate analyses for specific patient groups were conducted: (1) isolated thoracic injury with ISS < 16, (2) polytrauma without TBI which was defined as no AIS ≥ 3 head injury, and (3) polytrauma with AIS ≥ 3 head injury (TBI). Logistic regression analyses were executed with stepwise backward selection to identify risk factors for mortality following thoracic trauma. The p-value for inclusion of a variable in the logistic models was set at 0.157. As there were several variables that were used as a measure of the same factor, several different models were made to test each variable. To investigate a potential confounding influence of penetrating trauma, models with and without penetrating thoracic injuries were analyzed. The number of events (deaths, n = 4) in the isolated thoracic trauma patient group was too small to conduct a meaningful risk factor for mortality analysis with multivariate logistic regression. For each model, Hosmer–Lemeshow goodness of fit tests and areas under the Receiver Operating Characteristic (ROC) curve were calculated [32]. The best model was selected after calculating all the models, based on the Hosmer–Lemeshow test and the area under the ROC curve of the respective models. All statistical analyses were performed using StataCorp 2017 Stata Statistical Software: Release 15.

In total, 4042 patients were admitted during the inclusion period, of which 514 patients with thoracic injuries were included for analyses (Fig. 1). The mean annual incidence rate of severe thoracic trauma per 100,000 inhabitants remained the same during the study period, with a ten-year average of 13.3 cases (Fig. 2). Mean (SD) age was 51 [18] years and 78% were male (Table 1). Severe comorbidities according to CCI were present in 190 (40%) patients. Median (IQR) CPS was 0 (0–4). Most patients were in normal health status according to ASA classification (55%). Anticoagulants were used by 85 (17%) patients.

Most frequent mechanism of injury was road traffic accident (RTA) (Table 2). Overall, RTAs made up for 49% of mechanisms of injury with car crashes being the most frequent cause among RTAs (26%) followed by motorcycle crashes (9%), cycling accidents (7%) and injured pedestrians (3%). Falls made up for most of the remaining injuries (35%).

Median ISS was 17 (13–27) and median RTS was 7.8 (7.1–7.8) (Table 1). Most commonly co-occurring injuries were head and abdominal injuries. Rib fractures were the most prevalent thoracic injury, occurring in 426 (83%) patients (Table 3). In 19 (4%) patients an emergency thoracotomy was performed. In 126 (25%) patients a chest tube was placed. Complications occurred in 189 (37%) patients (Table 4, Fig. 3). Most common severe complication (CDS ≥ 3) was respiratory failure in 70 (14%) patients. Within the first 24 h after the accident 24 (5%) patients died (Fig. 4). Sixty patients (12%) died within 30 days after the accident. Three hundred thirteen (61%) patients were polytrauma casualties.

The analyses showed that female sex, more than nine comorbidities and medication use combined, a TRISS with a PS of lower than 50% and all the complications of a patient with a CI ≥ 30 were associated with an adverse outcome after trauma (Table 5). The final model (Hosmer–Lemeshow goodness of fit test (p = 0.4), area under ROC curve (0.93) consisted of the predictors mentioned above. Females had a 2.7 times higher risk of dying compared to male victims (OR 2.7, p = 0.04). Having more than nine comorbidities and medication use combined lead to nearly five times higher risk of dying (OR 4.8, p = 0.01). TRISS was also a strong risk factor for mortality where patients with a PS of 0 to 50%, having a 44 times higher risk of dying (OR 44, p < 0.001). Finally, patients with severe complications measured by CI, resembling a minimal complication of CDS grade ≥ 3, were 12.5 times more likely to die compared to those having a mild or no complication (OR 12.5, p < 0.001). Anticoagulants were no risk factors for mortality in this study. No confounding effect of penetrating thoracic injuries was observed. The analyses with penetrating thoracic trauma excluded showed the same significant results (online resource supplementary Table 1). Characteristics of the patients included and excluded in the risk factor for mortality analyses are given in Table 6.

Subgroup analyses were conducted for three specific groups: 201 isolated thoracic trauma patients, 215 polytrauma without TBI and 98 polytrauma with TBI patients (Table 2). Mean age (SD) of polytrauma patients with TBI was significantly lower (47, 19) than polytrauma without TBI patients (52, 19) and isolated thoracic trauma patients (53, 18), (p < 0.001). There were no differences in sex distribution or comorbidities between subgroups. Low-energy falls < 3 m were relatively frequent (15%) in isolated thoracic trauma compared to the other patient groups. High energy falls > 3 m were more often the causative mechanism in both polytrauma groups. Median (IQR) ISS was significantly higher in polytrauma without TBI (22, 17–29) and polytrauma with TBI (29, 22–41) than in isolated thoracic injury (11, 9–13), (p < 0.001). In the polytrauma without TBI group the highest number of AIS ≥ 3 thoracic injuries were observed with 102 (48%) patients with two or more AIS ≥ 3 thoracic injuries. There were 44 (45%) patients with severe complications in the polytrauma with TBI group. This was significantly higher than the 60 (28%) patients in the polytrauma without TBI group and the 19 (10%) patients in the isolated thoracic trauma group (p < 0.001). The highest mortality was observed in the polytrauma with TBI group with 25 (25%) deaths. In polytrauma without TBI the highest number of fatalities occurred within 24 h after trauma with 18 (58%) deaths, while the majority of deaths in polytrauma with TBI occurred on the second day with 11 (45%) deaths (Fig. 4). Main cause of death in the polytrauma without TBI group was exsanguination. In the polytrauma with TBI group main cause of death was brain injury. Risk factor for mortality analyses for specific subgroups did not reveal other risk factors for mortality than in the total thoracic trauma cohort in this study. Female sex was no risk factor for mortality in polytrauma without TBI (online resource supplementary Tables 2, 3).

Data was missing in the following variables: injury mechanism (5.3%), nature of the trauma mechanism (2.0%), GCS (3.3%), RTS (2.3%), TRISS (4.3%) ASA (11.5%) CCI (4.7%) CPS (5.8%) anticoagulants (6.4%), first emergency procedure (1.0%), chest tube (4.1%) complications (2.3%). Missing data were likely not to be random; mortality was significantly higher in the group of patients with missing data on any of the variables.

The average of 51 treated patients with severe thoracic injury at HUS per year is roughly comparable to other hospitals that have studied thoracic trauma patients [6, 9]. The incidence rate of polytraumatized patients with severe thoracic trauma in Western Norway (8/100,000 inhabitants) was comparable to that of the Netherlands (9/100,000 inhabitants). The percentage of car accidents as cause of severe thoracic trauma with polytrauma in HUS (50%), was relatively low compared to the university hospital of Helsinki (58%) and German hospitals (64%) [6]. It is known that Norway is one of the safest countries with regards to traffic and has one of the lowest traffic accident mortality rates in the world [33]. Mortality after severe thoracic trauma in polytrauma varies within the current literature, ranging from 10 to 18%, with no cohorts exactly matching this study [4‐6, 9, 34]. In this study, overall mortality in the polytrauma group, a cohort similar with the aforementioned literature, was relatively high with 18%. There were no general trends in different age, ISS or comorbidities if reported in those studies compared to HUS. Future studies are required to determine the cause of this difference in mortality.

Mortality in this study was not according to the classical trimodal pattern of trauma deaths by Trunkey et al. [35]. The classical model describes deaths on scene, these were not included in this study, early deaths within 6 h after trauma, and a third peak of late deaths three to 4 weeks after trauma. This late peak was not observed in this study. As shown before, it is hypothesized that the late peak of deaths due to sepsis and respiratory insufficiency does not occur due to improvements in health care systems [36]. In this study mortality in polytrauma without TBI was the highest in the first 24-h after trauma due to exsanguination. In polytrauma with TBI the highest mortality was later than 24-h after trauma, and the main reason for death in this subgroup was brain injury. Deaths due to exsanguination in this study show the second peak of early deaths of the classical trimodal pattern of trauma deaths, and primary brain injury deaths do not, with most of brain injury deaths occurring later than 24-h after trauma. This is in line with current literature [36‐38].

Comorbidities that have been shown to predict mortality after trauma are: hepatic and renal disease, congestive heart failure, hypertension, pulmonary disease, diabetes, malignancy, obesity, peripheral artery disease and neurologic disorders [5, 39‐43]. This study demonstrates patients having more than nine comorbidities and medication use combined measured by CPS, to have a considerably higher mortality risk compared to patients with relatively few or no comorbidities (OR 4.8, p = 0.011). This was demonstrated using a relatively novel, standardized instrument for quantifying comorbidities. A (geriatric) trauma study by Nossaman et al. in 2018 reported CPS to be a risk factor for mortality, although patients with a higher CPS demonstrated lower mortality risk [44]. The authors hypothesized that patients with known comorbidities and the respective medication use may have returned to a better physiologic conditions than their peers with unknown comorbidities and medication (a low measured CPS). These patients might have suffered from the same comorbidities but were not treated and therefore had a higher mortality risk. We hypothesize that the number of care avoiders in Norway is low due to a free health care system while as much as 45% of US citizens delays or avoids medical care due to costs [45]. Hence, Norwegian patients with a high CPS may be less healthy than Norwegian patients with a low CPS, they may thus have a higher mortality risk.

It has been debated whether higher mortality in older patients is caused by ageing itself or rather by more comorbidities in elderly patients [41, 46]. The present study demonstrates that comorbid conditions independently predict mortality. This is supported by Milzman et al. who showed that the effect of comorbidities on mortality in trauma patients proved consistent over all age and injury severity groups [40]. The main reason for adverse outcomes in patients with severe comorbidities is the reduced physiological reserve that is caused by comorbidities [11].

In the present study, CPS as a measure of comorbidities, appeared to be the strongest predictor of mortality, even stronger than the CCI. The literature on this topic is contradicting, since both CPS and CCI have previously shown to predict mortality in trauma [47, 48]. However, to the best of our knowledge, CPS and CCI have not been compared. In our study, the number of comorbidities combined with medication use is a more accurate predictor of mortality than the comorbidities included in the CCI. This suggests that there are other relevant comorbidities influencing survival that are not included in the CCI. With this result, we would like to emphasize, that the number of comorbid conditions and medication use is a helpful tool to estimate mortality risk in thoracic trauma patients.

Several trauma outcomes and injury severity scores are related to injury severity. In the present study injury severity in thoracic trauma is best measured by TRISS. The TRISS value is based upon registered ISS and RTS as well as age and type of trauma, hence giving an estimation of probability of mortality following trauma. It is well known that the anatomical severity of the injury combined with the physiological burden to the body is the best measure of injury severity and therefore the strongest predictor for mortality after trauma [49, 50].

This is, to our knowledge, the first study that provides an overview of all complications that occur after severe thoracic trauma with a standardized method. We find the overall complication rate of thoracic trauma to be comparable with complication rates mentioned in recent studies on rib fractures in polytrauma patients, both in polytrauma patients in general as well as in geriatric [5, 8, 51]. This study emphasizes the high risk of pulmonary complications occurring after thoracic trauma, and the higher mortality in this patient group. Pneumonia and respiratory failure were frequently observed in our material. Also, patients with a CI ≥ 30, representing severe complications of CDS grade 3 and higher were more likely to die (OR 12.5, p < 0.001). Of CDS grade 3 and higher complications, 65% were pulmonary complications. It has been observed that respiratory complications are a frequent cause of death in older patients [8, 14]. The multivariate models with CI were marginally better than the models with CDS, both were therefore risk factors for mortality in this study. The number of severe complications was significantly higher in the polytrauma with TBI group than in isolated thoracic trauma and polytrauma without TBI subgroups. This result can be explained by polytrauma with TBI being the most severely injured patient group with the highest ISS, and deaths due to brain injury occur relatively late compared to exsanguination, therefore there is more time to develop complications.

An unexpected finding in the present study was that age and anticoagulants were no independent risk factors for mortality. In none of the analyzed models, age was an independent risk factor for mortality. Advancing age leads to a higher mortality and has been found to be a risk factor for mortality in a large number of previous studies [4, 9, 12, 43, 52, 53]. The exact relationship between mortality, age and comorbidities remains debated. The present and other studies suggest that comorbidities accompanying advanced age contribute to mortality, as opposed to age alone [54].

Anticoagulant use predicts mortality after head trauma [55]. However, anticoagulant use did not emerge as predictor of mortality among our patients. The sample size may have been too small to demonstrate an effect of higher bleeding tendency on mortality. In addition, anticoagulant use also means that the patient has comorbidities. The models with CPS included, appear to be a stronger predictor for mortality than anticoagulant use alone. These effects could have caused anticoagulants not to become a significant predictor in our material.

Potential limitations of the present study arise mostly from its retrospective character. Underestimation of effects could be caused due to missing data on comorbidities, medication use and complications in patients that die within the first 24 h after the accident. Patients who died on scene are excluded. In Hordaland county, of which HUS is both the local and level one trauma center, 72% of patients who die after trauma, die at the scene or before arrival at a hospital [56]. This may cause an underestimation of the contribution of the investigated factors on mortality. Subgroups were relatively small and this may be the reason that no other risk factors were found in the subgroup analyses. Complications were defined as all adverse events graded by CDS occurring on the ward or in the ICU in patients after the initial stabilization and treatment. This could be a limitation in the risk factor for mortality analyses in patients who died in the first 48 h after trauma. There is a partial similarity between injury severity and complications. The significant risk factor complications might have become significant partially by the mortality in the first 48 h after trauma due to the severity of the injury.

In this study severe complications CD ≥ 3 were a significant risk factor for mortality and these complications occurred in 23% of all patients. It is important to be aware of patients at risk of developing a severe complication. These patients should be monitored intensively and treated aggressively as soon as complications occur to prevent further health deterioration. To the best of our knowledge, only one study on risk factors for complications in thoracic trauma exists, focusing solely on pneumonia [13]. Future studies should focus on identifying risk factors for complications following thoracic trauma, with subgroup analyses of patients who survive the first 48 h, to gain insight in patients at risk and in need of active trauma level one treatment.