Thoracic point- of- care ultrasound is an accurate diagnostic modality for clinically significant traumatic pneumothorax

Objective: There are conflicting data regarding the accuracy of thoracic point- of- care ultrasound (POCUS) in detecting traumatic pneumothorax (PTX). The purpose of our study was to determine the accuracy of thoracic POCUS performed by emergency physicians for the detection of clinically significant PTX in blunt and penetrating trauma patients. Methods: We conducted a retrospective institutional review board– approved study of trauma patients 15 years or older presenting to our urban Level I academic trauma center from December 2021 to June 2022. All study patients were imaged with single-view chest radiography (CXR) and thoracic POCUS. The presence or absence of PTX was determined by multidetector computed tomography (CT) or CXR and ultrasound (US) with tube thoracostomy placement. Results: A total of 846 patients were included, with

circumstances, the test characteristics of US in detecting PTX in the trauma bay must be reexamined. The purpose of our study was to determine the accuracy of thoracic POCUS performed by emergency physicians for detecting clinically significant traumatic PTX, as defined as PTX > 35 mm of pleural separation on CT or PTX detected on CXR or US followed by immediate tube thoracostomy placement prior to CT in blunt and penetrating trauma patients.

Study design and setting
We conducted a retrospective review of trauma patients presenting to our urban academic Level I trauma center between December 4, 2021, and June 3, 2022. The study was approved by our institutional review board and followed the Standards for Reporting of Diagnostic Accuracy guidelines. 12 Our emergency department (ED) sees approximately 100,000 patients per year and is staffed by physicians who have US certificates in accordance with the American College of Emergency Physicians guidelines. 5 Of these faculty members, seven have a fo- crush injuries; greater than 65 years of age; use of an anticoagulant despite mechanism; near-drowning; hemodynamic instability (defined as heart rate greater than 120 beats/min, respiratory rate greater than 30 breaths/min, systolic blood pressure less than 90 mm Hg), flail chest; suspected spinal cord injury; or penetrating injuries proximal to elbow or knee. Patients  The trainees performing the eFAST examination are overseen by emergency medicine faculty.
Trauma patients in our study who had a documented thoracic POCUS study, CXR, and a CT scan performed in the ED were included. We also included those trauma patients who had a chest tube placed after both a thoracic POCUS and a CXR even if a subsequent CT was not obtained in the ED. Study exclusion criteria included any of the following: (1) known PTX either by outside hospital imaging or placement of a thoracostomy tube prior to any imaging at our institution, (2) less than 15 years of age, (3) failure to meet the designated trauma alert criteria, and (4) received an eFAST but did not include thoracic images. The criterion standard to identify the PTX was the presence of PTX on CT or alternatively deployment of a chest tube after CXR and POCUS in the absence of dedicated confirmatory CT imaging. 13

Data collection, processing, and outcome measures
Two of the study authors (KM and SF) abstracted data from the electronic medical record (EMR) using clearly defined variables and standardized coding methods. These authors underwent detailed training, practiced with example cases, and inputted data directly into a predefined data abstraction form. Regular meetings were held to assess progress and monitor performance. The determination of whether an US was deemed positive or negative for PTX was assessed by what was documented in real time by scribes who are documenting events in the trauma resuscitation as they are happening. Additional pertinent clinical information obtained from the EMR included age, sex (gender), body mass index, systolic blood pressure, heart rate, mechanism of injury, image reports, time of imaging, and procedural interventions. The level of training of the clinician obtaining the POCUS images was recorded from QPath.
An expert-blinded "overread" by two emergency physicians (JRB, JMJ) with FPD in advanced emergency medicine ultrasonography was obtained for each patient included in the study. The experts overreading the USs were blinded to everything other than the images including the clinical scenario and the initial "real-time" read.
To standardize the reporting of PTX size and location, all CT scans with the reported presence of PTX on the initial CT report were retrospectively reviewed by a senior attending radiologist To assess the reliability of the data collection, 43 patient records (5% of the included cohort) were randomly selected using a randomnumber generator for repeat assessment by a third abstractor (JRB) who was blinded to the data abstracted by KM and SF.

Primary data analysis
The sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios for thoracic US were calculated for all traumatic PTX. Test characteristics were then calculated with miniscule PTX excluded (miniscule PTX defined as ≤10 mm of pleural separation) and clinically significant PTX (clinically significant defined as PTX with >35 mm of pleural separation) or CXR or US positive with immediate tube thoracostomy placement prior to CT.

Means, proportions, and associated standard deviations (SD)
were calculated as appropriate. When the calculations of accuracy were made, each lung scanned was evaluated separately, i.e., two sides, right and left, per patient. Continuous variables were compared using a t-test or Mann-Whitney test as appropriate, while categorical variables were compared using a chi-square or Fisher's exact as appropriate. Cohen's kappa coefficient was used to measure interrater reliability. All statistical analyses were performed with Vassar stats and nQuery power analysis software. All significance tests were two-sided, and p < 0.05 was considered significant.
A post hoc power analysis demonstrates a sample size of 614, a 5% two-sided significance level will have 90.8% power to detect the difference between two diagnostic tests whose paired sensitivities are 0.9 and 0.99. This assumes that the prevalence of the disease is 0.25 and the proportion discordant is 0.12.

Characteristics of study subjects
A total of 11,101 trauma patient records were reviewed. Patients were excluded for the following reasons: (1) 35 patients were excluded for having a known PTX before arriving to the ED, (2) 59 patients were excluded for age < 15 years old, (3) 37 were excluded for failure to meet our trauma alert criteria, and (4) eight were excluded due to absence of dedicated thoracic windows on eFAST.
This left 846 patients for our final analysis. All included patients had both a CXR and a thoracic POCUS performed, while 837 also underwent CT imaging. The remaining nine patients had only POCUS and CXR performed before a chest tube was placed ( Figure 1).
Characteristics of the study population are described in Table 1.
For patients who had a CT scan, these studies were obtained in a median (interquartile range [IQR]) 16 (9-26) minutes after POCUS was started.

Clinically significant
When only clinically significant PTXs were included, defined as and specificity of 100%. Thoracic POCUS test characteristics in all three groups are presented in Table 3. A comparison of the clinical and practice variables was examined in the false-negative and true-positive PTXs with miniscule PTX excluded and is presented in Table 4.

Overread
All POCUS images were reexamined by expert-blinded reviewers.

DISCUSS ION
Guidelines to aid medical decision making of traumatic PTX are changing based on the size of the PTX. 3 In concordance with guidelines, we defined clinically significant PTX as one that best models clinical practice: >35 mm of maximal pleural separation on CT or bedside imaging with CXR and US followed by immediate tube thoracostomy placement. We found thoracic US identified the majority of clinically significant PTXs and was more sensitive than CXR.
Our study of all patients meeting our inclusion criteria allowed us to examine the accuracy of thoracic US to detect PTX in the hands of our ED clinicians. This is in contrast with results of the recently pub-  was an effective tool to decrease unnecessary tube thoracostomy placements with no statistical difference in observation failure, length of hospital stay, or overall mortality. 3,10 Our study found that thoracic US is very accurate in identifying PTX that will likely require chest tube placement (87% sensitive, 100% specific), with only one clinically significant PTX missed on thoracic US that was detected on CT and required a tube thoracostomy.
Our study intentionally included tube thoracostomies performed in the ED to accurately reflect the population for which this examination was designed. It is accepted that thoracic US may miss miniscule, clinically insignificant PTXs. 15,16 This is not the cohort of patients that thoracic US is intended to identify.
Rather, it was designed as a rapid triage tool to diagnose potentially life-threatening traumatic injuries. Our data demonstrate an appreciably higher false-negative rate when clinically insignificant PTXs were included, consistent with the current published literature.
Thoracic US is used clinically as a rapid triage tool to rule in potentially life-threatening PTXs requiring immediate decompression or tube thoracostomy placement. The eFAST examination is performed quickly, on average in 1-5 minutes, allowing for the timely identification of potential pathology and the ability to treat rapidly. 17,18 As such, eFAST was designed for its specificity. Our study findings demonstrate an outstanding specificity across all three groups (all, miniscule excluded, and clinically significant PTX).
The expert-blinded US overread process demonstrated an improved sensitivity in two of the three subgroups, owing to the expertise with FPD certification. In addition, the expert-blinded overread process brought to light frequent data acquisition errors that may have contributed to false-negative thoracic US interpretations. Common themes included too much depth and gain making it difficult to appreciate pleural movement, distracting hand movement, clips taken of only a single rib space, and clips taken adjacent to the heart, which may erroneously mimic lung sliding or mimic a lung point sign. 19,20 It is interesting to note that the expert-blinded overread process did not yield a higher sensitivity for clinically significant PTX (87% for both).
This is in part thought to be related to the fact these reviewers were blinded to all data from the patient encounter therefore they were unable to take into account clinical information. This is in contrast to real-time sonographers where, for example, a patient with a penetrating chest wall injury would result in a much higher pretest probability.
It is interesting and encouraging that the sensitivity and specificity of clinically significant PTX was consistent among real-time and expertblinded overread, suggesting that even early learners can have comparable performance in the detection of clinically significant PTX.
The most common reason for a missed PTX identified on expert-blinded overread was visualization of a lung point sign that was found on the overread process but missed during the real-time interpretation. This was an incidental finding that was noted during the overread process that was incorrectly interpreted as normal Curvilinear and phased array 1 (7) Note: Data are presented as mean (±SD) and number (%).
Abbreviations: BMI, body mass index; PGY, postgraduate year. lung slide in real time. Importantly both false-negative USs in the clinically significant group were due to a missed lung point sign. This interpretation error has been described previously and postulated to be due to inadequate understanding of the underlying pathophysiology of a lung point sign, despite a historically cited 100% specificity. 21 Another possible reason that this may be missed in particular by younger learners, as in our study, is missing the forest for the trees, i.e., that the fact lung sliding on the screen is noted, this is erroneously interpreted as negative for PTX, without taking into account the entire image whereby absent lung slide is also present.
This is not to say that additional clips should be sought out to search for a lung point, but rather to encourage educational initiatives to improve institutional sensitivity and specificity of point-of-care thoracic US. However, these two PTXs were also detected on blinded overread, decreasing our concern for bias in this very small proportion of PTXs included.

LI M ITATI O N S
Another limitation our retrospective design and reliance on accurate and complete EMRs. While the CXR is always performed prior to the CT, we cannot confirm the radiologist reading the CXR was blinded to the CT. Finally, our study only included a small percentage (5%) of penetrating trauma patients. While we did not find a difference between blunt or penetrating trauma patients, our results regarding penetrating trauma should be interpreted with caution due to our small sample size. Importantly, a select number of patients with a PTX ≤ 35 mm received a tube thoracostomy, highlighting the importance of patient hemodynamics and the clinical scenario. The decision to place a tube thoracostomy should not be adjudicated exclusively on the size of the PTX. Future research should focus on this cohort to answer clinically important questions related to accuracy of thoracic US.

CON CLUS IONS
Thoracic ultrasound identified the majority of clinically significant traumatic pneumothoraxes and outperformed single view portable chest X-ray.

AUTH O R CO NTR I B UTI O N S
Stephanie Balderston drafted the manuscript. All authors critically reviewed and approved the manuscript for publication.

CO N FLI C T O F I NTE R E S T
The authors declare no potential conflict of interest.