The Chaos of Triage: A Model for Early Exclusion of Cardiac Injury in Chest Gunshot Wound Patients


 IntroductionHeart injury caused by thoracic gunshot wounds (GSW) is especially life-threatening and require prompt diagnosis and treatment. Heart injury is especially life-threatening and requires prompt diagnosis and treatment. During the pre-hospital phase and initial triage in the emergency department (ED), early recognition of a patient with heart GSW is difficult but important. The purpose of this study was to evaluate the predictability of heart injury in patients with chest GSWs. MethodsThe National Trauma Data Bank was queried for patients with chest GSW treated at all US trauma centres from July 1, 2009, to June 30, 2016. Patients with and without heart injuries (ICD-9: 861.00-03, 861.10-13) were compared and analyzed. Multivariate logistic regression was performed to evaluate independent factors of heart injury which could be obtained during the pre-hospital or triage phase only. Step-backward selection was used to establish a model for such patients. We used the receiver operating characteristic (ROC) curve to test the accuracy of this model and Youden’s J statistic to find the cutoff value of sensitivity/specificity. Level 1 trauma registry data from Stroger Hospital of Cook County (July 1, 2016, to June 30, 2017) was used for external validation of this prediction model.Results47,044 patients with chest GSW were evaluated in the ED and 8.6% of them had heart injuries. The mortality rates of patients without cardiac injury versus those with cardiac injury were 9.0% (3864/42997) and 21.7% (879/4047) respectively. Patients with heart injuries were significantly younger (28.4 vs. 29.3, p<0.001), had lower SBP (34.7 vs. 103.8 mmHg, p<0.001), had lower GCS (5.1 vs. 11.2, p<0.001) and a higher probability of apnea (58.3% vs. 14.7%, p<0.001), higher rate of pulselessness (59.9% vs. 12.0%, p<0.001), and more self-inflicted injuries (9.7% vs. 8.5%, p<0.001) than patients without heart injuries. The cutoff values of SBP and GCS for prediction of heart injury were 61mmHg (AUC: 0.783) and 5.5 (AUC: 0.768) respectively. Integration of six independent factors (age, SBP, GCS, apnea, lack of pulse and suicide intent) with multivariate logistic regression showed an AUC: 0.823 and specificity of 88.8% in the heart injury prediction model. External validation with the local database showed 95.6% specificity.ConclusionEarly diagnosis of heart injury is important in the management of patients with chest GSWs. Our model has high specificity and can be beneficial for early triage of cardiac injury in patients with GSW to the chest.


Introduction
Cardiac injury can be lethal. Historically, clinical experience with cardiac injury due to gunshot wounds (GSWs) was uncommon. Recently with an increased access (1)(2) and use of rearms throughout the U.S., the incidence of penetrating injuries to the chest and heart has risen (3). Some studies suggest that the mortality rate in cardiac injury patients is as high as 60-74% (4). Early recognition of patients with cardiac injuries is necessary for prompt diagnosis and treatment to reduce the high mortality associated with cardiac injuries.
Several studies suggest that mortality can be reduced when major trauma patients are transported directly to a trauma center (5)(6)(7)(8). However, the triage and decision to transfer can be prolonged or delayed due to an unpredictable pattern of injury seen in GSW patients. Early identi cation of cardiac injury in these patients could be lifesaving. Currently, prehospital personnel are unable to predict cardiac injury in patients with GSW to the chest.
The current study examines the incidence of cardiac injuries and identi es independent factors that predict such injuries. We used the National Trauma Data Bank (NTDB) to predict heart injuries in patients with GSW to the chest. We developed and validated a model that can be used both in the pre-hospital environment and Emergency Department (ED) to identify patients with cardiac injury.

Method
The National Trauma Data Bank (NTDB) was analyzed between July 1, 2009 to June 30, 2016. The NTDB is the largest multi-institutional information repository prospectively gathered from Trauma Centers and maintained by the American College of Surgeons (ACS). In the current study, patients with intra-thoracic injuries (ICD-9:860.0-862.9) due to GSW were included. Non-GSW penetrating injuries, non-chest injuries and records with missing key values were excluded. Patients with and without heart injuries (ICD-9: 861.00-03, 861.10-13) were compared and analyzed. Data extracted included demographics, vital signs, Injury Severity Score (ISS) and Glasgow Comma Scale (GCS). Pre-hospital data such as intention of suicide with the use of a rearm and patients' condition on arrival in the ED such as pulselessness and apnea were also included.
A univariate analysis was created to study the characteristics of the patient population. A bivariate analysis comparing mortality in patients with and without heart injury was created. Variables were compared between patients with and without heart injuries using chi square testing and independent ttest. Multivariate logistic regression (MLR) analysis was performed to evaluate independent variables predictive of cardiac injury. Variables gathered during the pre-hospital phase included age, race, Systolic Blood Pressure (SBP), GCS, pulselessness, apnea and intention of suicide. A step-backward selection method was used to create an accurate model that predicted heart injury in chest GSW. Odds ratio with 95% con dence intervals were calculated and statistical signi cance was set at a p-value < 0.05.
Receiver operating characteristic (ROC) curves were created for SBP and GCS values, along with the MLR model to predict a cut off score for cardiac injury in this patient population via the Youden's J statistic (sensitivity/1-speci city). Level 1 trauma registry data from Stroger Hospital of Cook County (July 1, 2016 to June 30, 2017) was used for external validation of this prediction model. All original les from the NTBD with required data were merged and analyzed using R (V3.3.1) (9). Microsoft Excel (V15.32) was used for data entry and generating the associated gures.
Results 47,044 patients with GSW to the chest were evaluated. Patients with heart injuries comprised 8.6% (4,047) of this population. The mean age was 29 years and 89.7% (42,192) were male. The mean ISS was 20.9 and mortality rate was 10.1% (4,743) in all patients (Table 1). Table 2 compares survivors and nonsurvivors among all GSW patients.
Multivariate regression analysis (Table 4) showed that age, SBP, GCS, apnea and pulselessness were independent risk factors for heart injury in patients with GSW to the chest. For every unit increase in SBP, the odds of a heart injury decreased by 0.9% and for every unit increase in GCS, the odds of heart injury decreased by 7.3%. Apnea and pulselessness increased the odds of a heart injury by 13.8% and 2.22-fold respectively. GCS and SBP were selected to create the ROCs. The cutoff values of SBP and GCS for prediction of heart injury were 61mmHg (Area Under Curve, AUC: 0.783) and 5.5 (AUC: 0.768) respectively. Both values of AUC for the stated variables were considered fair values.
We performed a step-backward selection for the multivariate logistic regression model to include only independent factors for heart injury in the study population ( Table 4). The following formula was derived as a heart injury prediction model: From this formula, another ROC was created with the following independent risk factors for heart injury: age, SBP, GCS, apnea, pulselessness, and suicide intent. The heart injury prediction model speci city was 88.8% with AUC = 0.823. The cut off score was determined to be 29 per Youden's J index. Patients with GSWs to the chest with a score below 29 were determined to have an 88.8% chance of not having a heart injury (Fig. 1).
The level 1 trauma registry data from Stroger Hospital of Cook County in Chicago, Illinois from July 1, 2016 to June 30, 2017 was used for external validation of the suggested heart injury prediction model. Patients with the same inclusion and exclusion criteria showed a 95.6% speci city in our model.

Discussion
Cardiac injuries due to GSWs are uncommon but highly lethal. Early detection and triage are vital for survival. There are several scoring systems to help predict patient mortality such as Trauma Injury Severity Score (TRISS) and Thorax Trauma Severity Score (TTSS). However, there is no reliable score for predicting cardiac injury in patients with GSWs to the chest. We found that cardiac injuries can be identi ed through a combination of easily measurable variables in patients with GSW to the chest.
The cardiac box is an anatomical region de ned as the area on the anterior thorax limited by the clavicles superiorly, midclavicular lines laterally and the line connecting midclavicular lines at the costal margin inferiorly. Degiannis et al. (10) found that mortality from injuries outside the cardiac box were higher compared to those in the box. Literature suggesting the use of cardiac box as a predictor of cardiac injury was limited by small size (11,12). Simply relying on the anatomic borders of the cardiac box to predict cardiac injury is inadequate (13).
In 1996, Focused Assessment with Sonography in Trauma (FAST) was introduced as a diagnostic tool to detect pericardial effusions in trauma patients (14). While, FAST has high sensitivity and speci city in traumatic pericardial effusion, a negative FAST examination does not exclude penetrating cardiac injury (15). Ultrasound use is not feasible for early identi cation of cardiac injury during prehospital triage due to lack of availability and lack of trained staff (16). MGAP (Mechanism, GCS, Age and arterial Pressure) utilizes parameters that are simple to measure. However, the prediction for mortality is inferior to that of TRISS (17). TTSS is not the best model to easily and quickly detect cardiac injuries in patients with GSW to the chest.
Asensio et al (23) created a predictive model for penetrating cardiac injury outcomes using the NTDB database and included 2016 patients with penetrating cardiac injuries and 1264 patients with GSW as the mechanism of injury. The study suggested that in patients with cardiac GSW's eld CPR, the absence of spontaneous ventilation, the presence of an associated abdominal GSW, need for ED intubation and aortic cross-clamping were independent predictors of mortality. This study provides insightful information on penetrating cardiac injury; however, it's more useful when evaluating in-hospital patients.
In our study, we have evaluated a larger (47,044) but speci c population (GSW patients to the chest) with 4,047 patients with cardiac injury. We primarily focused on triage and early identi cation of patients with cardiac injury. This model strati es a patient's risk of having cardiac injury due to GSW by incorporating easily measured variables including age, presence of self-in icted injury, pulselessness, apnea, SBP and GCS. As expected, the predicted model demonstrated an excellent result with 95.6% speci city during external validation with the local database from Stroger Hospital of Cook County.
We found a close relationship between suicide intent and cardiac injury in GSW to the chest. GSWs are most commonly the result of assaults (24), however, self-in icted injury with rearms has higher mortality rate (25). Our study shows that cardiac injury after GSW to the chest is more common in younger patients. According to Center for Disease Control (CDC), suicide by rearms was the second leading cause of death among 15 to 35 years old in 2017 (26). Prevalence of GSW in the younger age group has been studied extensively and is explained by a greater access to rearms (1, 2).
Our study shows that patients with a cardiac injury had signi cantly lower SBP (34.7 mmHg) compared to patients without cardiac injury (103.8 mmHg), and that most patients were found to be pulseless [59.9% (2423) vs. 12% (5166), p < 0.001]. Our results showed that one more unit of SBP (mmHg) in the ED decreased the probability of heart injury by 0.9%, and patients arriving with immeasurable pulse had a 2.2 times higher mortality rate. Despite proper triage, patients with cardiac injury require immediate resuscitation to prevent irreversible cardiac damage resulting from increased stress and anaerobic loading onto myocardial tissues and cells. Teixeira et al. found similar ndings in that mortality was higher among patients with cardiac injury that presented with SBP < 90 mmHg compared to patients with SBP > 90 mmHg (85% vs 65%, p = 0.01) (27). This is consistent with other research ndings, which stated that systolic BP is the best univariate predictor in pre-hospital trauma patients.
A GCS < 8 is a common nding in cardiac injury associated with increased mortality (28). As in previous studies, GCS was found to be an independent variable in our predictive model. GCS and SBP both had a 71.5% speci city and AUC of 0.768 and 0.783 respectively. Combination of variables in our suggested model including age, SBP, GCS, apnea, pulselessness and suicide intent show a speci city of 88.8% with AUC of 0.823.
The primary limitation of this study was its retrospective nature both during development and validation of the suggested model. The study is also limited by the nature of NTDB. External validation of the suggested model at a single institution might also be considered a limitation, therefore, a multiinstitutional validation is required. However, our model is e cient, robust and relies on easily obtainable factors such as pulselessness and suicidal intent. We developed a model speci c for patients with GSW to chest that is quick and easy to use during triage both by paramedics and ED personnel.

Conclusion
Early diagnosis of heart injury is important in the management of patients with chest GSWs. Our established model has acceptable sensitivity and speci city which is proven to be bene cial for primary screening of GSWs to the heart. Our model has high speci city suggesting that the paramedics and ED personnel can con dently assess and identify whether patients with GSW have cardiac injury, compared to those who do not.  Table 3. Comparisons of the characteristics between patients with and without heart injury in chest GSW patients (N=47044)

Declarations
This study only included deidenti ed, Health Insurance Portability and Accountability Act-compliant data from a national database; hence, it was exempt from institutional review board approval. All authors have no con icts of interest to disclose. This research did not receive any speci c grant from funding agencies in the public, commercial, or not-for-pro t sectors.