Comparison of Injury Severity Score (ISS) and New Injury Severity Score (NISS) in the Evaluation of Thoracic Trauma Patients: A Retrospective Cohort Study

Objective To explore the value of the injury severity score (ISS) and the new injury severity score (NISS) for evaluating injuries and predicting complications (pneumonia and respiratory failure) and poor prognoses (in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death) in patients with thoracic trauma. Methods The data of consecutive patients with thoracic trauma who were admitted to the department of cardiothoracic surgery of a tertiary hospital between January 2018 and December 2021 were retrospectively collected. ISS and NISS were calculated for each patient. The study outcomes were complications and poor prognoses. The differences in ISS and NISS between patients with complications and poor prognoses and patients without the abovementioned conditions were compared using the Mann‒Whitney U test. Discrimination and calibration of ISS and NISS in predicting outcomes were compared using the area under the receiver operating characteristic (ROC) curve (AUC) and Hosmer‒Lemeshow (H-L) statistic. Results A total of 310 patients were included. ISS and NISS of patients with complications and poor prognoses were greater than those of patients without complications and poor prognoses, respectively. The discrimination of ISS in predicting pneumonia, respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death (AUCs: 0.609, 0.721, 0.848, 0.784, 0.763, 0.716, and 0.804, respectively) was not statistically significantly different from that of NISS in predicting the corresponding outcomes (AUCs: 0.628, 0.712, 0.795, 0.767, 0.750, 0.750, and 0.818, respectively). ISS showed better calibration than NISS for predicting pneumonia, respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, and ICU admission but worse calibration for predicting prolonged ICU stay and death. Conclusion ISS and NISS are both suitable for injury evaluation. There was no statistically significant difference in discrimination between ISS and NISS, but they had different calibrations when predicting different outcomes.


Introduction
Trauma is the leading cause of death and a major public health problem worldwide that results in 5.8 million deaths each year and substantial healthcare costs [1][2][3][4].Toracic trauma accounts for 10%-15% of all trauma cases and is responsible for approximately 25% of trauma-related deaths [5][6][7][8].Injuries due to thoracic trauma afect the chest wall and internal organs of the thorax to diferent extents and may lead to complications such as pneumonia and respiratory failure, and patients with thoracic trauma frequently require admission to the intensive care unit (ICU) and mechanical ventilator support [9,10].Accurate evaluation of thoracic trauma severity in the early posttrauma phase is highly important for predicting complications and the prognosis, assessing the need for intensive care, and making the optimal clinical decision [11,12]; therefore, accurate evaluation of thoracic trauma severity is the basis for efective treatment.
Trauma scoring systems have been developed to quantitatively evaluate trauma severity and have been used for predicting mortality and morbidity, triaging patients, stratifying the risks for research purposes, and benchmarking trauma outcomes [13][14][15][16].Among the various trauma scoring systems, the injury severity score (ISS) is the most commonly used [17][18][19].Te ISS was proposed by Baker et al. [20] in 1974 based on the abbreviated injury scale (AIS).To calculate the ISS, the whole body is divided into six regions and the sum of the squares of the highest AIS values in each of the three most severely injured body regions is taken [20,21].Although the ISS has been widely used, an important limitation of the ISS is that the ISS does not account for more than one injury within the same body region [18,19,22].To overcome this limitation, Osler et al. [23] introduced a simple modifcation of the ISS called the new injury severity score (NISS).Te NISS is defned as the sum of the squares of the AIS values of the three most severe injuries regardless of the body region injured [23].Numerous studies have compared the ability of the ISS and NISS to predict trauma outcomes in patients in a specifc age group such as pediatric patients, patients with injuries in a specifc body region such as head injuries, patients with injuries caused by a specifc type of trauma such as penetrating trauma, and patients with injuries caused by a specifc mechanism such as a frearm [13,18,24,25].To date, however, studies comparing the capacities of these two scoring systems in evaluating patients with thoracic trauma are rare, and the capabilities of the two scoring systems to predict diferent prognostic parameters in patients with thoracic trauma have not been fully elucidated.
Te aim of this study was to validate the value of the ISS and NISS for injury evaluation and to compare the performance of these two scoring systems for the prediction of complications and poor prognoses, including pneumonia, respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay and death, in patients with thoracic trauma.

Ethics Approval.
Te study was approved by the Ethics Committee of 926th Hospital of Joint Logistics Support Force of PLA (no.2022-126).Due to the anonymization of the data, the need for informed consent was waived.

Study Design.
Tis study was conducted in a tertiary hospital in Yunnan Province, China.Consecutive patients with thoracic trauma who were admitted to the department of cardiothoracic surgery between 1 January 2018 and 31 December 2021 were retrospectively recruited in this research cohort.Te inclusion criteria for this study were admission to the hospital within 24 hours of trauma and a hospital stay of 2 or more days.Patients younger than 16 years were excluded from the analyses.

Data Collection.
Demographic data such as sex and age and injury characteristics including injury mechanism and the AIS value for each of the body regions injured were extracted from the registry.AIS values were coded using the AIS-2005.Te ISS and NISS for each patient were then calculated as primary exposures in this study.Complications (pneumonia and respiratory failure) and poor prognoses (in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death) were examined from the registry as the study outcomes.
In this study, pneumonia is referred to as hospitalacquired pneumonia (HAP) and ventilator-associated pneumonia (VAP) [26].Te defnitions for HAP and VAP from Chinese guidelines for the diagnosis and treatment of hospital-acquired pneumonia and ventilatorassociated pneumonia in adults (2018 Edition) [27] were used as follows: HAP was defned as pneumonia not incubating at the time of hospital admission and occurring 48 hours or more after admission in patients not receiving invasive mechanical ventilation during hospitalization.VAP was defned as pneumonia occurring >48 hours after endotracheal intubation or tracheotomy to receive mechanical ventilation.Pneumonia occurring within 48 hours after weaning from mechanical ventilation and extubation was also considered VAP.A clinical diagnosis of pneumonia was established when the patient met the following criteria: a new or progressive infltrate, consolidation, or ground glass opacity was revealed on chest radiograph or CT scan, plus two or more of the following three criteria: fever >38 °C, purulent airway secretions, and peripheral white blood cell count of >10 × 10 9 /L or <4 × 10 9 /L [27].Respiratory failure was defned as a decrease in arterial oxygen tension (PaO 2 ) to less than 60 mmHg with or without an increase in arterial carbon dioxide tension (PaCO 2 ) on room air [28].In-hospital tracheal intubation was defned as a procedure involving the placement of a tube into the trachea during a hospital stay.Extended length of hospital stay was defned as a length of hospital stay of ≥10 days [19,22].ICU admission was defned as admission to the ICU during 2 Emergency Medicine International hospitalization.Prolonged ICU stay was defned as an ICU stay of >14 days [29].Death was defned as all-cause death during a hospital stay.

Statistical Analyses.
Te primary objective of this study was to estimate the discrimination index, namely, the area under the receiver operating characteristic (ROC) curve (AUC).Assuming that the incidence of the study outcome was 20%, a sample size of 310 would produce a 95% confdence interval with a width of 0.149 (upper to lower bounds) and a targeted AUC of 0.75.If the incidence of the study outcome was 10%, the width would extend to 0.205.Te frequency and percentage of the qualitative variables were computed.Te Kolmogorov-Smirnov test or the Shapiro-Wilk test was used to assess the normality of the continuous data.Te mean and standard deviation of the data were calculated when the continuous data followed a normal distribution; otherwise, the median and interquartile range of the data were calculated.Te Mann-Whitney U test was employed for statistical comparison of the data that were not normally distributed.
Measures of discrimination and calibration were used to compare the performance of the ISS and NISS in predicting pneumonia, respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death.Discrimination indicates how well the model diferentiates those at higher risk of having an event from those at lower risk, and calibration describes the accuracy of absolute risk estimates [30].Discrimination was measured by the AUC and calibration was measured by the Hosmer-Lemeshow (H-L) statistic [15].An AUC of 1 indicates perfect discrimination, whereas an AUC of 0.5 indicates that the predictive ability of the model is no better than chance.Te following scale is usually used to illustrate the predictive power represented by the AUC: 0.5 < AUC < 0.6 indicates failed prediction, 0.6 < AUC < 0.7 poor prediction, 0.7 < AUC < 0.8 fair prediction, 0.8 < AUC < 0.9 good prediction, and 0.9 < AUC < 1 excellent prediction [31].A lower H-L statistic value indicates better calibration.Te confdence intervals of the AUC were binomial exact confdence intervals; the confdence intervals of the cutof and Youden index were estimated using the bias-corrected and accelerated (BCa) bootstrap method; the confdence intervals of the sensitivity and specifcity were "exact" Clopper-Pearson confdence intervals; the confdence intervals of the likelihood ratios were calculated using the log method; and the confdence intervals of the predictive values were the standard logit confdence intervals.
A p value <0.05 was considered to be statistically signifcant.Statistical analysis was performed using SPSS version 20 and MedCalc version 20.

Results
A total of 310 thoracic trauma patients who met the inclusion criteria were enrolled in this study over a 4-year period (Figure 1).Te descriptions of the patients included are shown in Table 1.Tere were 231 males (74.5%) and 79 females (25.5%).Te mean age was 49.47 years (standard deviation � 12.15), ranging from 20 to 83 years.Te median and interquartile range of the ISS and NISS were 13 (9, 17) and 17.5 (11,22), respectively.Te most frequent mechanism of injury was falls (56.1%), followed by trafc accidents (27.1%).Te median and interquartile range of the length of hospital stay was 13 (8,19) days.In the whole sample, 198 patients (63.9%) were admitted to the ICU.Te median and interquartile range of the length of ICU stay of these 198 patients who were admitted to the ICU was 4 (2, 7) days.Overall, there were 4 deaths for a mortality rate of 1.3%.
Te ISS and NISS of patients with complications (pneumonia and respiratory failure) and poor prognoses (in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death) were statistically greater than those of patients without complications and poor prognoses (Table 2).
Te results of the ROC analysis of the two scoring systems, the ISS and NISS, for predicting the outcomes are shown in Table 3.

Comparison of the Performance of the ISS and NISS for the Prediction of the Outcomes in Patients with Toracic Trauma
3.1.1.Pneumonia.We did not fnd a statistically signifcant diference in discrimination between the ISS and NISS for predicting pneumonia (AUC: 0.609 versus 0.628; p � 0.5385).Te calibration of the ISS was superior to that of the NISS (H-L: 2.411 versus 17.204).In addition, the calibration of the NISS was statistically insufcient (p � 0.016) (Figure 2(a) and Table 4).

Respiratory Failure.
Tere was no statistically significant diference between the ISS and NISS in their discrimination for predicting respiratory failure (AUC: 0.721 versus 0.712; p � 0.7907).Te ISS showed a better calibration than the NISS (H-L: 1.982 versus 6.162) (Figure 2(b) and Table 4).
All thoracic trauma patients within study period: n= 325 Total included: n= 310 Excluded: (1) length of hospital stay less than 2 days: n=9 (2) younger than 16 years: n=6 Emergency Medicine International

In-Hospital Tracheal Intubation.
Tere did not exist a statistically signifcant diference in discrimination between the ISS and NISS when predicting in-hospital tracheal intubation (AUC: 0.848 versus 0.795; p � 0.1722).Te calibration of the ISS was better than that of the NISS (H-L: 6.885 versus 11.294) (Figure 2(c) and Table 4).

Extended Length of Hospital Stay.
A statistically signifcant diference in discrimination between the ISS and NISS for predicting the extended length of hospital stay was not found (AUC: 0.784 versus 0.767; p � 0.4563).Te ISS was with a better calibration than the NISS (H-L: 1.699 versus 9.612) (Figure 2(d) and Table 4).

ICU Admission.
No statistically signifcant diference was observed in discrimination between the ISS and NISS for the prediction of ICU admission (AUC: 0.763 versus 0.750; p � 0.5841).Te calibration of the ISS was better than that of the NISS (H-L: 2.836 versus 10.781) (Figure 2(e) and Table 4).

Prolonged ICU Stay.
Tere was no statistically signifcant diference between the ISS and NISS in their discrimination to predict prolonged ICU stay (AUC: 0.716 versus 0.750; p � 0.6089).Te NISS outperformed the ISS in terms of calibration (H-L: 3.958 versus 8.067) (Figure 2(f ) and Table 4).

Discussion
Toracic trauma can cause respiratory complications and poor prognoses.Accurately evaluating trauma severity and predicting the occurrence of complications and poor prognoses are crucial for developing efective treatment plans for patients while also optimizing the allocation of 4 Emergency Medicine International medical resources.In the present study, we found that the ISS and NISS were both well suited for injury evaluation in patients with thoracic trauma.For predicting the outcomes including complications (pneumonia and respiratory failure) and poor prognoses (in-hospital tracheal intubation, extended length of hospital stay, ICU admission, prolonged ICU stay, and death) of thoracic trauma patients, the ISS and NISS were not statistically signifcantly diferent in discrimination.Te calibration of the ISS and NISS in predicting diferent outcomes varied.Te ISS demonstrated better calibration than the NISS for predicting pneumonia, respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, and ICU admission, whereas the NISS showed better calibration for predicting prolonged ICU stay and death.Falls are the second leading cause of death from unintentional injuries, and most fall-related injuries afect the head, spine, and chest [32].Traumatic injuries, particularly to the chest, have increased steadily as a result of the increase in trafc accidents [3].Assessment of the mechanisms of injury in the present study showed that the two most frequent mechanisms of injury were falls and trafc accidents, which was in accordance with the fndings from another study including patients with multiple traumas [33].
Te more severe the patient's injuries are, the greater the risks of complications and poor prognoses.In this study, the ISS and NISS were signifcantly greater in patients with complications and poor prognoses than in those without complications or poor prognoses.Tese fndings indicated that the ISS and NISS could well refect the severity of the trauma, so the two scoring systems were both suitable for injury evaluation in patients with thoracic trauma.
Although the ISS was developed to predict death, it has also been used to model many other outcomes [34].Since the NISS was proposed, many studies have compared the ability of the ISS and NISS to predict the outcomes of trauma patients.However, the results have been controversial.Some studies have shown that the NISS is superior to the ISS, while others have suggested that the ISS and NISS have similar predictive abilities [17][18][19][35][36][37].To comprehensively explore the scope of optimal application of the ISS and NISS, it is important to conduct studies that compare the performance of the ISS with that of the NISS in accordance with certain parameters related to trauma outcomes, such as age, injury mechanism, and body region injured [34].
Several studies exploring the ability of the ISS and NISS to predict outcomes according to specifc body regions have been performed.Lavoie et al. [18] reported that the NISS outperformed the ISS in terms of both discrimination and calibration for predicting ICU admission in patients with moderate to severe head injuries.Lavoie et al. [34] also found that the NISS showed advantages over the ISS both in discrimination and calibration for predicting in-hospital death in patients with head/neck injuries, facial injuries, thoracic injuries, and abdominal injuries.Jang et al. [38] reported that the AUCs of the NISS were larger than those of the ISS when predicting laparotomy and in-hospital mortality in patients with abdominal trauma, while the AUC of the ISS was larger than that of the NISS when predicting long hospital stay.
When predicting in-hospital mortality in patients with thoracic injuries, Lavoie et al. [34] reported the superiority of the NISS over the ISS in terms of both discrimination and calibration, but the NISS was not much better than the ISS in terms of discrimination (for the AUC, NISS: 0.809, ISS:    Emergency Medicine International 0.802, and p � 0.04; for the H-L statistic, NISS: 20.9 and ISS: 50.1).Another study showed that the AUC of the NISS was larger than that of the ISS for predicting mortality in patients with thoracic trauma, but there was no statistical comparison of the discrimination between the two scoring systems (for the AUC, NISS: 0.876 and ISS: 0.867) [39].Tese results were similar to our results, which showed that the calibration of the NISS was better than that of the ISS and that the NISS had a larger AUC without a statistical diference in discrimination compared with the ISS (for the AUC, NISS: 0.818, ISS: 0.804, p � 0.9069; for the H-L statistic, NISS: 4.180 and ISS: 7.919).In these three studies, both the ISS and NISS showed good capability according to the AUC (0.8 < AUC < 0.9) for the prediction of mortality.Hu et al. [40] reported that the AUC of the NISS was signifcantly larger than that of the ISS for predicting respiratory failure in patients with chest trauma (for the AUC, NISS: 0.91, ISS: 0.84, and p � 0.045), but the calibration analysis was not reported.In our study, the results did not support this fnding, as the discrimination was not statistically signifcantly diferent between the ISS and NISS; moreover, the AUC of the ISS was larger than that of the NISS and the calibration of the ISS was better than that of the NISS (for the AUC, ISS: 0.721, NISS: 0.712, and p � 0.7907; for the H-L statistic, ISS: 1.982 and NISS: 6.162).

Emergency Medicine International
In addition to death and respiratory failure, the other fve outcomes, namely, pneumonia, in-hospital tracheal intubation, extended length of hospital stay, ICU admission, and prolonged ICU stay, are also common in patients with thoracic trauma.However, to date, few studies have been conducted to compare the performance of the ISS and NISS for predicting these fve outcomes in patients with thoracic trauma.To close the gap in the literature, our study compared the performance of the ISS and NISS in predicting these outcomes.We found that for predicting these fve outcomes in the present study, the ISS and NISS did not statistically signifcantly difer in terms of discrimination.
To our knowledge, this is the frst study to compare the performance of the ISS and NISS for predicting pneumonia in trauma patients.However, for predicting pneumonia in patients with thoracic trauma, both the ISS and NISS had poor predictive ability (0.6 < AUC < 0.7) (for the AUC, ISS: 0.609 and NISS: 0.628).Tese fndings indicate that when predicting pneumonia, in addition to the ISS and NISS, which refect the anatomical injury severity, the importance of exploring the infuence of other risk factors, such as age, sex, physiologic conditions, and comorbidities, should be considered.
For predicting in-hospital tracheal intubation in patients with thoracic trauma, this study showed that although the ISS and NISS were not statistically signifcantly diferent in terms of discrimination, the AUC of the ISS represented good predictive power, whereas the AUC of the NISS represented fair predictive power, and in addition, the calibration of the ISS was better than that of the NISS (for the AUC, ISS: 0.848, NISS: 0.795, and p � 0.1722; for the H-L statistic, ISS: 6.885 and NISS: 11.294).Studies comparing the ability of these two scoring systems to predict in-hospital tracheal intubation in trauma patients are scarce.A study on trauma patients who were admitted to the ICU reported that the NISS outperformed the ISS for the prediction of the need for intubation (for the AUC, NISS: 0.863, ISS: 0.788, and p < 0.05; for the H-L statistic, NISS: 9.1 and ISS not reported) [41].
For predicting extended length of hospital stay in thoracic trauma patients, the results of the present study showed that the ISS was not statistically signifcantly diferent from the NISS in terms of discrimination but had a larger AUC than that of the NISS, and the ISS had a better calibration.Tese fndings were similar to those reported by Tamim et al. [19].Tey reported that the ISS was statistically better than the NISS in terms of discrimination and had a better calibration for predicting the length of hospital stay of ≥10 days in trauma patients (for the AUC, ISS: 0.72, NISS: 0.70, and p � 0.03; for the H-L statistic, ISS: 11.0 and NISS: 15.78).In contrast, Balogh et al. [22] demonstrated that the NISS was a better predictor of the extended (≥10 days) length of hospital stay than the ISS in patients with multiple orthopaedic injuries.
In this study, although no statistically signifcant difference was found between the ISS and NISS in discrimination for the prediction of ICU admission in thoracic trauma patients, the ISS had a larger AUC and better calibration.A study conducted by Tamim et al. [19] also showed that the AUC of the ISS was larger than that of the NISS and that the ISS had a better calibration for predicting ICU admission in trauma patients (for the AUC, ISS: 0.81, NISS: 0.70, and p � 0.0001; for the H-L statistic, ISS: 7.83 and NISS: 39.07).Lavoie et al. [18] reported that there was no statistically signifcant diference in discrimination between the ISS and NISS for predicting ICU admission in trauma patients; the AUC of the ISS was larger than that of the NISS, but the NISS showed an improved calibration compared to the ISS (for the AUC, ISS: 0.843, NISS: 0.839, and p � 0.08; for the H-L statistic, ISS: 611 and NISS: 309).However, two other studies involving severe blunt trauma patients and patients with multiple orthopaedic injuries found that the NISS was superior to the ISS in predicting ICU admission [17,22].
Our results indicated that the NISS was not statistically signifcantly diferent from the ISS in discrimination for predicting prolonged ICU stay among patients with thoracic trauma, but the NISS had a larger AUC and superior calibration (for the AUC, ISS: 0.716, NISS: 0.750, and p � 0.6089; for the H-L statistic, ISS: 8.067 and NISS: 3.958).Te fndings of this study were similar to those of two other studies.Li et al. [17] reported that the NISS had a larger AUC and better calibration than the ISS for predicting prolonged ICU stay (>14 days) in severe blunt trauma patients (for the AUC, NISS: 0.772, ISS: 0.760, and p � 0.0460; for the H-L statistic, NISS: 38.82 and ISS: 43.58).Harwood et al. [35] demonstrated that there was no statistically signifcant diference between the ISS and NISS in discrimination for predicting the length of ICU stay of >7 days in trauma patients, but the ISS had a larger AUC than the NISS (for the AUC, ISS: 0.763 and NISS: 0.762).Our study, along with the other two studies, demonstrated that both the ISS and NISS had fair predictive ability (0.7 < AUC < 0.8) for prolonged ICU stay, although one of the studies considered the length of ICU stay of >7 days as prolonged ICU stay.
At present, the ISS is considered an essential measurement in trauma evaluation globally and has long been employed for trauma research and benchmarking [17].As a modifcation of the ISS, the NISS has two main advantages over the ISS: from a clinical perspective, the NISS is more logical than the ISS because it assigns the same importance to all injuries, even if they exist within the same body region; from a practical perspective, the NISS is easier to calculate and more efective compared to the ISS, as the NISS does necessitate the division of AIS codes into body regions [18,42].Meanwhile, two scenarios may cause the overestimation of injury severity for NISS when a patient has two injuries in the same body region: adjacent injuries potentially due to the same impact and a secondary injury resulting from a primary injury.Te advantages and disadvantages of the NISS may lead to uncertain results when comparing the performance of the ISS and NISS for predicting the outcomes of trauma patients in diferent subsets.Te fndings of our study and previous studies support this viewpoint, as the predictive performance of the ISS and NISS varies among diferent subgroups.Tus, it has recently been suggested that the ISS should not be replaced by the NISS [36].Terefore, in clinical practice, the ISS and NISS should be selected based on their optimal scope of application.Further research is needed to explore the best application scope of ISS and NISS in more detail.

Strengths and Limitations.
Compared to studies previously conducted to explore the predictive performance of the ISS and NISS not only in thoracic trauma patients but also in other subgroups or in the general trauma population, this study included a broad range of complications and prognostic indices.Tis is the frst strength of this study.Second, as far as we know, this study is the frst to elucidate and compare the predictive performance of the two scoring systems for pneumonia, which is one of the most common complications of thoracic trauma, thereby bridging the gap in the literature.Tird, when assessing the predictive performance of the two scoring systems, calibration was calculated in addition to discrimination in this study, whereas many other studies only calculated discrimination.Te results of this research provided a more thorough depiction of the predictive efcacy of the two scoring systems.
Our study has several limitations.First, this was a retrospective study, which meant that there were potential biases and limitations inherent to this type of study design.Second, the sample size included in the current study was not overly large, and the patients were from a single center, which may limit the generalizability of our fndings.

Conclusion
Tis study revealed that the ISS and NISS of thoracic trauma patients with complications and poor prognoses were signifcantly greater than those of thoracic trauma patients without complications and poor prognoses.Terefore, both the ISS and NISS are suitable for injury evaluation in patients with thoracic trauma.In terms of predicting complications and poor prognoses in thoracic trauma patients, there was no statistically signifcant diference in discrimination between the ISS and NISS.Except for their ability to predict pneumonia, both scoring systems had a fair or good ability to predict other complications and poor prognoses.For predicting pneumonia, both the ISS and the NISS showed poor predictive ability, and the ISS had a smaller AUC than the NISS but better calibration.Te ISS had a larger AUC and better calibration than the NISS in predicting respiratory failure, in-hospital tracheal intubation, extended length of hospital stay, and ICU admission, respectively.Te NISS had a larger AUC and better calibration than the ISS in predicting prolonged ICU stay and death, respectively.Tese results can provide a reference for the clinical application of these two scoring systems for evaluating injury and predicting complications and poor prognoses in patients with thoracic trauma.

Figure 1 :
Figure 1: Overview of the patient inclusion process for the analysis.

Figure 2 :
Figure 2: Comparison of the receiver operating characteristic (ROC) curves of the injury severity score (ISS) and new injury severity score (NISS) for the prediction of (a) pneumonia, (b) respiratory failure, (c) in-hospital tracheal intubation, (d) extended length of hospital stay, (e) intensive care unit (ICU) admission, (f ) prolonged ICU stay, and (g) death.

Table 1 :
Description of the patients.

Table 2 :
Comparison of the ISS and NISS between diferent subgroups.

Table 3 :
ROC analysis of the ISS and NISS for predicting the outcomes.

Table 4 :
Comparison of the ISS and NISS for predicting the outcomes.
ISS, injury severity score; NISS, new injury severity score; AUC, area under the ROC curve; H-L, Hosmer-Lemeshow statistic; ICU, intensive care unit; ROC, receiver operating characteristic.* Comparison of AUC between the ISS and NISS.