Strengths and limitations of this study

Risk stratification of the need for artificial airway procedures was determined by X-tile software.

Two scores were developed based on different feature compositions to better fit the clinical scenarios.

The large study population ensured sufficient statistical power.

Whether patients establish artificial airways was mainly decided by the emergency doctor, and some of these decisions are jointly decided by the emergency doctor and the surgeon.

The performance of this model is limited by a single piece of data on admission.

Introduction

Trauma is one of the major causes of death and disability worldwide. Trauma accounts for 16% of the global disease burden; 16 000 people die from trauma, and millions of people temporarily or permanently become disabled owing to trauma.1 In low and middle-income countries, injury-related mortality and disability account for approximately 90% of the global burden.2

As the classic trauma management procedure, the initial assessment of Advanced Trauma Life Support3 starts with airway management, which aims to identify obstructed or potentially obstructed airways and relieve the obstruction. Adequate airway management has been identified as one of the means to reduce preventable trauma-related death.4

The Eastern Association for the Surgery of Trauma proposed the indications of endotracheal intubation for trauma patients in 2012. The level 1 recommendations included airway obstruction, hypoventilation, persistent hypoxaemia (SaO2<90%) despite supplemental oxygen, severe cognitive impairment (Glasgow Coma Scale (GCS) ≤8), severe haemorrhagic shock and cardiac arrest. The level 3 recommendations included facial or neck injury with the potential for airway obstruction, moderate cognitive impairment (GCS>9), persistent combativeness refractory to pharmacological agents, respiratory distress, preoperative management and cervical spinal cord injury with any evidence of respiratory insufficiency.5 The indications in level 1 recommendations are easy to identify, while the indications in level 3 cannot be quantified and lack a uniform definition, which creates a challenge for medical staff to recognise injured patients in need of artificial airways (including endotracheal intubation and tracheotomy).

Therefore, the objective of this study was to develop and validate a model for predicting the need for artificial airway procedures for acute trauma patients in the emergency department (ED).

MethodsStudy design and sample selection

This study was conducted from September 2020 to August 2022. The requirement for written informed consent was waived by the Biomedical Ethics Committee of West China Hospital, Sichuan University, because this study was retrospective and there was no intervention implemented. This work has been reported in compliance with the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis reporting guideline.6

This was a retrospective case–control study. All trauma patients admitted to the ED of West China Hospital, Sichuan University, within 24 hours of injury and who were admitted from 1 August 2012 to 31 July 2020 were included. The exclusion criteria were as follows: (1) pregnant patients; (2) patients aged <16; (3) patients lacking any one of the following information: current medical history, physical examination or imaging examination results; (4) patients who had established artificial airways, such as endotracheal intubation, tracheotomy and cricothyroid membrane puncture, when transferred to the ED; (5) patients with a Japan Coma Scale (JCS) score of 3 (indicating inability for being aroused by any forceful stimuli); and (6) patients with out-of-hospital cardiac arrest.

The study outcome was the establishment of an artificial airway within 24 hours of admission to the ED. The included patients were divided into an artificial airway group and a control group according to the study outcome. Logistic regression was applied to develop the prediction model, a minimum of 10 events per variable are recommended.7 Eight variables were evaluated in the logistic regression model, so the sample size in the derivation stage was at least 80 events. Considering the proportion of patients in the artificial airway group was 3.03% in this cohort, thus we planned to collect at least 2641 cases for the prediction cohorts.

Patient and public involvement

Patients or the public were not involved in the design, or conduct or reporting of our research.

Variables

The variables, including age, gender, vital signs when admitted to the ED, medical history, physical examination and imaging results, were collected retrospectively and integrated into two feature categories based on clinical meanings and acquisition time. One feature category included basic variables, including age, gender, trauma mechanism, the time interval between admission to the ED and establishment of an artificial airway, JCS, vital signs on arrival at the ED (pulse rate (PR), respiratory rate (RR), systolic blood pressure (SBP), diastolic blood pressure, pulse oxygen saturation) and eye response. The other category is the trauma score consisting of the Abbreviated Injury Scale (AIS) and Injury Severity Score (ISS). Based on the patients’ variables on admission to the ED, the basic variables can be evaluated rapidly, while the trauma score may require more time to accumulate, which requires the information of the patient’s physical examination and imaging results.

The trauma mechanisms consisted of penetrating injury and blunt injury (traffic accident injury, high falling injury, flat ground falling injury, heavy pound injury (namely injury caused by heavy objects), scald burn and combined injury), which were determined by the patient’s medical history. The JCS is an extensively adopted scale for assessing the patients’ consciousness levels and was described in 1974. The scale is composed of three main categories: JCS-1 indicates awake without stimuli, JCS-2 means arousable with some stimuli (but reverts to the previous status if the stimulus stops) and JCS-3 indicates not able to be aroused by any forceful stimuli.8 The JCS has a good correlation with the GCS9 and is more concise than the GCS.

Whether the pupils were of equal diameter was recorded according to the physical examination, and the pupils were classified as equal, unequal and unable to be checked. Any limited examination caused by eye trauma was recorded as unable to be checked. The eye response was also recorded in accordance with the physical examination, which was classified as sensitive, slow, absent and unable to be checked. The JCS and pupillary-related examinations were recorded according to the physical examination.

The vital signs on arrival at the ED included PR (<60 beats per minute (bpm) or unmeasurable or ≥110 bpm, 60–109 bpm) and RR (<12 bpm or unmeasurable or ≥22 bpm, 12–21 bpm), SBP (<90 mmHg or unmeasurable, ≥90 mmHg), diastolic blood pressure (<60 mmHg or unmeasurable, ≥60 mmHg) and pulse oxygen saturation (<90% or unmeasurable, ≥90%).

AIS is an internationally recognised trauma severity scoring system based on anatomy. The AIS score was determined retrospectively according to the physical examination record and imaging results. The AIS scores were categorised into two categories: <3 and ≥3.10 According to the different body regions, AIS is divided into six parts: head and neck, face, chest, abdomen, limbs as well as skin. ISS is scored on the basis of AIS, which equals the sum of squares of AIS of the three most severely injured body parts. The score ranges from 0 to 75; the higher the score is, the greater the severity of trauma. The ISS scores were grouped into three categories: <16, 16–24 and ≥25.11 12

Statistical analysis

SPSS (V.23.0; IBM SPSS) was applied to analyse the data and construct a predictive model. MedCalc (V.18.2.1) was used to compare the area under the curve (AUC) value between the two scores. X-tile software (V.3.6.1) was applied to perform risk stratification. The measurement data obeying a normal distribution are represented as the mean and SD and were compared using an independent samples t-test. The measurement data obeying a skewed distribution are represented as the median and IQR and were compared using the Mann-Whitney U test. Numeration data are represented as counts and percentages and were compared using the χ2 test. To preserve the study sample and reduce bias, the median value was used to handle the missing measurement data, while the missing values of numeration data were replaced with the mode values.

The included patients were randomly divided into a development group and a validation group. The former accounted for 70% of the total patients, and the latter accounted for 30% of the total patients. To increase the clinical application, the continuous variables were converted into classified variables based on the Youden index or clinical meanings. In the development cohort, univariable logistic regression analysis was applied to select predictors, and multivariable logistic analysis was used to identify the coefficient β and OR. The selected predictors were weighted on the basis of the respective coefficient β. The Hosmer-Lemeshow test was applied to assess the goodness of fit for the developed system, and an adequate fit was assumed if p>0.05. The data in the validation group were used to verify the model. The performance of this model was evaluated with AUC. There was a significant difference if the two-sided p value <0.05.

To better fit the clinical scenarios, two scores were developed: one was based on the basic information to achieve a quick evaluation, and the other was based on both the basic information and trauma score to identify the function of the trauma score to artificial airway demand.

Results

A total of 13 685 patients were included, and 5397 patients were excluded based on the exclusion criteria. Finally, 8288 patients were analysed in this study, including 251 in the artificial airway group and 8037 in the control group. There were 5801 patients in the development group and 2487 in the validation group (figure 1). Artificial airways were established in 170 (2.93%) patients in the development group and 81 (3.26%) patients in the validation group.

Figure 1

Patient selection. JCS, Japan Coma Scale.

Among the 251 patients requiring artificial airways, 227 underwent endotracheal intubation and 24 underwent a tracheotomy. The median time interval between admission to the ED and establishment of an artificial airway was 1.67 hours, with an IQR of 0.58–3.47.

Baseline feature comparison

The study population is characterised in online supplemental table 1. Compared with the control group, the age in the artificial airway group was significantly older (49.39±18.72 vs 46.69±16.51, p=0.03). There was no difference in gender composition between the two groups. Additionally, the SBP, diastolic blood pressure and oxygen saturation in the artificial airway group were significantly lower than those in the control group, while the PR and RR were significantly higher. The head AIS, face AIS, chest AIS, abdomen AIS and ISS in the artificial airway group were significantly higher than those in the control group. There was no significant difference in the AIS of limbs and AIS of skin between the two groups.

Predictive score based on basic information

Continuous variables were converted into classified variables according to the largest Youden index or clinical meaning for convenient clinical application. The variables with p<0.05 in the univariate logistic regression (online supplemental table 2) were further evaluated with multivariate logistic regression.

For the variables in the basic information category, a total of seven variates, including age >60 years, SBP<90 mmHg, PR, RR, pulse oxygen saturation <90%, eye response and JCS, were found to be related to artificial airway demand for injured patients. JCS-2 achieved the largest OR of 5.94. According to coefficient β, these variates were weighted, and the respective score is displayed in table 1. The scoring system was summarised by the mnemonic ‘O-SPACER’ (Oxygen saturation, Systolic blood pressure, Pulse rate, Age, Coma Scale, Eye response, Respiratory rate) (table 2). The Hosmer-Lemeshow test was applied to examine the goodness of fit for the ‘O-SPACER’ system with a p value of 0.31.

Table 1

Multivariate logistic regression for identifying the injured patients needing artificial airway in development group based on two feature compositions

Table 2

Two scores developed based on the different feature compositions

The patients were then divided into three risk groups according to their O-SPACER score: low risk, score 0; medium risk, score 1–2; and high risk, score 3–9. In both development and validation groups, the OR values of medium-risk and high-risk groups were significantly higher than the low-risk group (table 3).

Table 3

Performance of the O-SPACER score and IO-SPACER score

Predictive score based on both the basic information and trauma score

There were eight variables independently associated with artificial airway demand in multivariate logistic regression (table 1). Among the eight variables, ISS was the newly added on the basis of the variables in the O-SPACER score. Likewise, the mnemonic ‘IO-SPACER’ (Injury Severity Score, Oxygen saturation, Systolic blood pressure, Pulse rate, Age, Coma Scale, Eye response, Respiratory rate) was developed, which is presented in table 2.

According to the IO-SPACER score, the patients were stratified into low-risk (0–1), medium-risk (2–3) and high-risk (4–11) groups (table 3). Using the low-risk group as a reference, IO-SPACER scores of 2–3 and >4 were associated with ORs of 6.32 (95% CI 3.14 to 12.72, p<0.001) and 60.06 (95% CI 31.34 to 115.08, p<0.001), respectively, for artificial airway demand. In the validation cohort, a medium-risk and high-risk O-SPACER score was associated with increased demand for artificial airways (OR=4.40, 95% CI 1.82 to 10.65, p=0.001; and OR=40.16, 95% CI 18.13 to 88.97, p<0.001, respectively).

Receiver operating characteristic curves were drawn, and AUCs were calculated to evaluate the performance of the two scores developed in this study (figure 2, online supplemental table 3). In the validation group, the AUC of the IO-SPACER was significantly larger than the AUC of the O-SPACER (0.88 with 95% CI 0.84 to 0.92 vs 0.85 with 95% CI 0.80 to 0.89, p=0.002, in the DeLong test).

Figure 2

Receiver operating characteristic curves of the IO-SPACER (Injury Severity Score, Oxygen saturation, Systolic blood pressure, Pulse rate, Age, Coma Scale, Eye response, Respiratory rate) and O-SPACER (Oxygen saturation, Systolic blood pressure, Pulse rate, Age, Coma Scale, Eye response, Respiratory rate) scores.

Discussion

The present study developed the O-SPACER score and IO-SPACER score for predicting the need for artificial airway procedures in acute trauma patients based on a large sample, and these scores provide new tools for risk assessment and airway management in acute trauma.

Several studies13–20 have explored the predictive system of artificial airways or surgical airways for injured patients, and the risk factors found in those studies included age,13 gender,13 18 American Spinal Injury Association (ASIA),13 16 ASIA Motor Score,13 16 ISS,13 state of consciousness,17 haemodynamic instability,14 history of smoking,15 history of lung diseases,15 trauma mechanism,17 trauma site,15–18 pulmonary complications14 and so on. The sample sizes of these studies ranged from 146 to 788 patients.13–16 19 20 Limited sample sizes make it difficult to identify the risk factors for artificial airway demand for patients with acute trauma. Furthermore, the time of the establishment of the artificial airway was not clearly defined. The major reasons for establishing an artificial airway within 1 day after trauma are different from the reasons for establishing an artificial airway within 5 days after trauma. The former is closely related to trauma, while the latter may be more correlated with trauma-related complications.

The two scoring systems derived in this study are consistent with the standard trauma assessment procedures, and they start with a primary assessment based on the patient’s basic physical examination, which is followed by a secondary assessment based on anatomical information provided by imaging examinations. In addition, considering the time dependence of calculating AIS and ISS, it may not be practical to guide clinical practice according to the IO-SPACER score for patients requiring urgent airway intervention during initial care in the ED, although the performance of the IO-SPACER was greater than that of the O-SPACER score in terms of AUC. Therefore, the IO-SPACER score was a better fit for patients needing planned intubation or tracheotomy early after trauma.

This risk stratification, if validated in prospective studies, is a potentially important tool for the initial clinician, with the potential to identify those patients at high risk on admission. As an adjunct to the existing risk assessment of airway management, this scoring system may aid in the preparation and prediction. In consideration of the serious adverse effect of unplanned intubation, we can pay more attention to high-risk patients, and this can optimise the utilisation of medical resources, especially in the context of busy emergency medical work, which includes a seriously unbalanced proportion of medical staff and patients.

There are several unavoidable limitations. First, the proportion of patients in the artificial airway group was only 3.03%, which was similar to the research by Okada et al17 and Hayashida et al,18 so this model may increase the risk of overfitting. Second, in this study, whether patients establish artificial airways and what kind of artificial airway to establish are mainly decided by the emergency doctor, and some of these decisions are jointly decided by the emergency doctor and the surgeon. Therefore, there may be some limitations on the generalisation of the model. Multicentre research should be carried out in the future to fix this deficiency. Third, a single piece of data certainly cannot cover as much information as dynamic data, so further studies based on dynamic data can be carried out to dynamically assess the risk. Fourth, to rule out endotracheal intubation due to late complications such as lung infection, the study outcome was set as the establishment of an artificial airway within 24 hours of admission to the ED. While the cut-off point of time establishing an artificial airway needs to be further explored in future research.

To conclude, the O-SPACER score may permit risk stratification of injured patients requiring urgent airway intervention in the ED and may be useful in guiding initial management. The IO-SPACER score may assist in further determining whether the patient needs planned intubation or tracheotomy early after trauma.

Data availability statement

Data are available upon reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study was approved by the Biomedical Ethics Committee of West China Hospital, Sichuan University (approval ID: 2020(1030)).