Excluding Hollow Viscus Injury for Abdominal Seat Belt Sign Using Computed Tomography

Key Points

Question  Do negative findings on computed tomography (CT) in patients with abdominal seat belt sign safely exclude operative hollow viscus injury?

Findings  This cohort study that included 754 patients found no patients with a completely negative CT scan who required operative intervention for hollow viscus injury.

Meaning  These findings suggest that patients with an abdominal seat belt sign can be safely discharged in the setting of a high-quality, negative CT scan when in agreement with good clinical judgment.

Importance  Abdominal seat belt sign (SBS) has historically entailed admission and observation because of the diagnostic limitations of computed tomography (CT) imaging and high rates of hollow viscus injury (HVI). Recent single-institution, observational studies have questioned the utility of this practice.

Objective  To evaluate whether a negative CT scan can safely predict the absence of HVI in the setting of an abdominal SBS.

Design, Setting, and Participants  This prospective, observational cohort study was conducted in 9 level I trauma centers between August 2020 and October 2021 and included adult trauma patients with abdominal SBS.

Exposures  Inclusion in the study required abdominal CT as part of the initial trauma evaluation and before any surgical intervention, if performed. Results of CT scans were considered positive if they revealed any of the following: abdominal wall soft tissue contusion, free fluid, bowel wall thickening, mesenteric stranding, mesenteric hematoma, bowel dilation, pneumatosis, or pneumoperitoneum.

Main Outcomes and Measures  Presence of HVI diagnosed at the time of operative intervention.

Results  A total of 754 patients with abdominal SBS had an HVI prevalence of 9.2% (n = 69), with only 1 patient with HVI (0.1%) having a negative CT (ie, none of the 8 a priori CT findings). On bivariate analysis comparing patients with and without HVI, there were significant associations between each of the individual CT scan findings and the presence of HVI. The strongest association was found with the presence of free fluid, with a more than 40-fold increase in the likelihood of HVI (odds ratio [OR], 42.68; 95% CI, 20.48-88.94; P < .001). The presence of free fluid also served as the most effective binary classifier for presence of HVI (area under the receiver operator characteristic curve [AUC], 0.87; 95% CI, 0.83-0.91). There was also an association between a negative CT scan and the absence of HVI (OR, 41.09; 95% CI, 9.01-727.69; P < .001; AUC, 0.68; 95% CI, 0.66-0.70).

Conclusions and Relevance  The prevalence of HVI among patients with an abdominal SBS and negative findings on CT is extremely low, if not zero. The practice of admitting and observing all patients with abdominal SBS should be reconsidered when a high-quality CT scan is negative, which may lead to significant resource and cost savings.

An abdominal seat belt sign (SBS) has historically entailed admission and observation. This approach is based on both previous estimates that abdominal SBS is associated with hollow viscus injury (HVI) in 15% to 64% of cases and concern about the diagnostic accuracy of computed tomography (CT) imaging.^1-8 The 2002 Eastern Association for the Surgery of Trauma Practice Management Guidelines also recommended that patients with SBS be admitted for observation and serial physical examination.⁹ Although these guidelines have since been retired, the most widely accepted approach to abdominal SBS remains admission and observation regardless of CT findings, because of the morbidity associated with delayed identification of an injury.^1,7,8,10,11

These concerns and recommendations predated the introduction of 64-slice CT scanners in 2004 and the widespread adoption of multislice CT scanning.¹² Furthermore, the high incidence of HVI with abdominal SBS reported in prior studies may have been related to differences in safety between 2-point and modern 3-point restraints. Regardless, multiple studies have challenged the notion of limitations of CT in blunt abdominal trauma and abdominal SBS, each showing an HVI detection rate of nearly 100%.^13-16 However, these findings have been limited to single-institution studies. Concerns about generalizability remain, and there has been no widespread adoption of discharging patients with an abdominal SBS and negative CT findings from the emergency department.

This study sought to prospectively collect data from patients across multiple trauma centers, hypothesizing that patients presenting with abdominal SBS and negative CT findings would not have HVI requiring operative intervention.

Institutional review board approval was obtained and waiver of consent granted to collect data for patients (age ≥18 years) with trauma and an abdominal SBS who underwent a CT scan of the abdomen and pelvis with intravenous contrast. Pregnant women and patients who underwent an operation without abdominal CT were excluded. Patients were enrolled at 9 trauma centers in August 2020 through October 2021. Rather than provide a specific definition of abdominal SBS to study centers, we left its identification to the discretion of the treating team.

Study data were collected using REDCap.¹⁷ Data included age, sex, Injury Severity Score, Glasgow Coma Scale score, and vital signs (initial heart rate, maximum heart rate in 24 hours, initial temperature, maximum temperature in 24 hours [considered febrile at >38° C], and initial systolic blood pressure). Also recorded were initial and repeated test results for white blood cell count and serum lactate. The abdominal examination was collected, including the presence and characterization of tenderness (localized vs diffuse) and guarding. If more than 1 physical examination was documented, the examination from the most senior member of the trauma team was used.

Imaging data included the results of a Focused Assessment With Sonography for Trauma (FAST) examination and CT scan of the abdomen and pelvis. A FAST scan was considered positive if free fluid was identified within the abdomen. The data for each CT scan included abdominal wall soft tissue contusion (consistent with abdominal SBS), free fluid and density in Hounsfield units, bowel wall thickening, mesenteric stranding, mesenteric hematoma, bowel dilatation, pneumatosis, and pneumoperitoneum. A negative CT was defined as the absence of any of these findings. If a repeat CT abdomen/pelvis was performed, the elapsed time, findings, and whether the repeat CT led to an operation were recorded.

Details of any abdominal operation were collected, including time between emergency department presentation and operation, whether an HVI was suspected preoperatively, type of operation (ie, laparoscopy vs laparotomy), location and severity of HVI (ie, full thickness vs partial thickness), and any interventions. Outcome data collected included mortality, length of stay, intensive care unit (ICU) length of stay, ventilator days, packed red blood cell transfusions, postdischarge presentation to the emergency department, and hospital readmission. Data were collected on index hospitalization and 30-day medical record review. Reviews of CT scans were performed by a senior member of the research team at each institution to ensure concordance with official radiology report and to identify any missed findings. If discrepancies were found, they were discussed and confirmed with the attending radiologist of record (n = 26, 3.4%).

The primary outcome (HVI) was the presence of an enteric injury (full or partial thickness) identified during operative exploration. If a patient did not undergo an operation, they were considered negative for HVI.

Descriptive statistics were used to compare patients with HVI and patients without and reported as median with IQR (continuous), mean with SD (continuous), or count with percentage (categorical). Comparisons were made using appropriate bivariate tests (t test, χ² test, and Wilcoxon rank sum test) and reported as P values with significance <.05. Each variable was reported and considered individually. Any variables with more than 10% missing data are noted in the Results section.

Simple, bivariate logistic regression was used to model the odds of HVI. These associations were reported as odds ratios (ORs) with 95% CIs. The area under the receiver operator characteristic curve (AUC) was reported for each variable associated with HVI to determine the performance of each variable as a binary classifier of HVI. Analyses were performed in Stata version 17 (StataCorp). The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

There were 754 patients with abdominal SBS with an HVI incidence of 9.2%, or 69 patients. There was no difference in median (IQR) age for patients with and without HVI (37 [26-56] vs 36 [24-46] years, respectively; P = .10). Patients with HVI had a higher median (IQR) Injury Severity Score (14 [10-24] vs 5 [2-12], respectively; P < .001) and increased mortality (4/69 patients [6%] vs 9/685 [1.3%]; P = .006) compared with those without HVI. The HVI group had a longer median (IQR) length of stay (9 [6-16] vs 2 [2-4] days; P < .001) and longer ICU length of stay (2 [0-6] vs 0 [0-0] days; P < .001), more commonly received transfusions (26/69 patients [38.2%] vs 31/685 [4.6%]; P < .001), and were readmitted after discharge more often (8/69 patients [11.6%] vs 21/685 [3.1%]; P = .001) (Table 1).

There was no difference on presentation between cohorts for heart rate or temperature, but the HVI group in the first 24 hours had a higher mean (SD) heart rate (118/min [22] vs 105/min [19]; P < .001) and were more likely to be febrile (18/69 [26.1%] vs 37/685 [5.5%]; P < .001). Most patients were normotensive, though the HVI group had a higher prevalence of hypotension (7/69 [10.1%] vs 10/685 [1.5%]; P < .001). Tenderness was more common in the HVI group (58/69 [84.1%] vs 329/685 [48.2%]). The presence of guarding was more frequent in the HVI cohort (17/69 [25%] vs 22/685 [3%]; all P < .001), as well as a higher initial median (IQR) white blood cell count (14 700/μL [11 400 to 17 900] vs 10 800/μL [8200 to 14 200]; P < .001) and lactate level (18 mg/dL [12.6-26.1] vs 24.3 mg/dL [18.9-34.2]; P < .001), although the latter was not routinely collected (missing values: 31%). The remaining characteristics are summarized in Table 1.

Most patients underwent a FAST scan (84.5%), and patients with HVI more commonly had a positive scan (17/60 [28.3%] vs 19/575 [3.3%]; P < .001). All patients underwent abdominal CT, and each of the 8 findings were more common in the HVI group (Table 2). The most common CT finding among patients with HVI was free fluid (60/69 patients [87%]), and it was more common compared with patients without HVI (92/685 [13.5%]; P < .001). There was no difference in the mean (SD) density of fluid (38.3 [17.3] HU vs 34.4 [12.6] HU; P = .20). However, measurement of free fluid density was not always reported (missing: 29.6%). There was only 1 patient (1.5%) in the HVI group with a negative CT based on the a priori definition, compared with 255 patients in the no-HVI group (37.2%, P < .001). The patient with a negative CT in the HVI group did have a pericolonic hematoma, which was identified but thought not to arise from the mesentery (ie, it did not meet criteria for mesenteric hematoma). However, it was recognized as concerning, and the patient underwent laparotomy, which demonstrated the hematoma was arising from the colonic mesentery.

A total of 93 patients (12%) underwent an abdominal operation (Table 3). The majority underwent laparotomy, either as the initial surgery (n = 61, 65.6%) or from conversion to laparotomy after laparoscopy (n = 14, 15.1%). In 10 of 14 cases, laparoscopy was converted to laparotomy on identification of an injury. For the remaining 4 cases, 3 were converted because of inadequate visualization, and 1 case was converted after inability to repair the injury laparoscopically. A total of 69 patients (74.2%) undergoing abdominal operation had 1 or more bowel injuries identified, and the majority were treated with resection and anastomosis (n = 42, 60%). For 24 patients (26%), a bowel injury was not identified: 9 patients (10%) were taken for a suspected bowel injury, but no injury was identified (ie, nontherapeutic operation), and 15 patients had a therapeutic intervention for a non-HVI injury (6 traumatic diaphragmatic hernias, 4 solid organ injuries, 3 abdominal wall injuries, and 1 bladder perforation).

Of the surgeries that remained laparoscopic (n = 13, 14%), 5 bowel injuries were identified, 2 of which underwent therapeutic intervention. The remaining 8 laparoscopic surgeries were negative.

The association between select variables and HVI was based on an unadjusted logistic regression for each of the covariates of interest. The predictive value of each of the covariates as a binary classifier of HVI was further assessed using the AUC. A summary of these findings can be found in Table 4.

Among the variables collected prior to CT, tenderness to palpation was the best binary classifier for HVI with an AUC of 0.70 (95% CI, 0.65-0.74) and a nearly 6-fold increased likelihood of having HVI (OR, 5.73; 95% CI, 3.19-10.29; P < .001). The highest association with HVI was seen among patients who presented with hypotension (OR, 7.61; 95% CI, 2.80-20.69; P < .001) and patients with a positive FAST scan (OR, 11.61; 95% CI, 5.63-23.95; P < .001), although each performed poorly as a binary classifier (AUC, 0.54; 95% CI, 0.51-0.58; and AUC, 0.63, 95% CI, 0.57-0.68, respectively).

Imaging findings had the highest association with HVI. The presence of free fluid had the highest effect size and performed best as a binary classifier (OR, 42.68; 95% CI, 20.48-88.94; P < .001; AUC, 0.87; 95% CI, 0.83-0.91). Selecting for only patients with free fluid in the absence of solid organ injury reduced both the association with HVI and its performance as a binary classifier (OR, 19.75; 95% CI, 11.24-34.69; P < .001; AUC, 0.79; 95% CI, 0.73-0.84).

A CT scan with completely negative findings had a strong association with the absence of HVI (OR, 41.09; 95% CI, 9.01-727.69; P < .001; AUC, 0.68; 95% CI, 0.66-0.70). Removing radiographic SBS from the definition of a negative CT increased the association with the absence of HVI, as well as increasing its utility as a binary classifier (OR, 250.49; 95% CI, 54.5-4423.58; P < .001; AUC, 0.89; 95% CI, 0.86-0.90).

The practice of admitting patients with abdominal SBS has persisted despite its origin dating to when CT technology was less sophisticated. Over the last 5 years, 3 single-institution studies that included 841 patients have examined abdominal SBS and found no instances of missed HVI.^14-16 The current study sought to improve on previous studies by providing a prospective evaluation of abdominal SBS across 9 trauma centers. Although 1 patient in the study who had HVI did not meet the a priori criteria for a positive CT, no patients with a normal CT required operative intervention for HVI. Choosing to discharge patients with a normal CT could result in a cost savings of approximately $2000 to $8000 for each avoided admission.¹⁸

The prospective design of this study was similar to the single-institution studies published by many of us and shares a similar finding: Operative HVI is exceedingly low, if not zero, in the presence of a normal CT.¹⁵ Free fluid remained the most common finding in patients with HVI, occurring in nearly 90% of patients. Our prior prospective study had 1 patient whose only finding on CT was an abdominal wall contusion,¹⁵ which was not replicated in this study.

Another key difference between our prior study and this multicenter study was the significance of clinical parameters (heart rate, systolic blood pressure, fever, higher initial lactate values and white blood cell count, diffuse abdominal tenderness/guarding on examination).¹⁵ This reinforces that a wide constellation of findings may suggest HVI, as noted previously.^19,20 However, the clinical utility of each of these findings in isolation is questionable. Rather, there should remain a high suspicion for HVI in patients with concerning clinical parameters. This holds true for the utility of FAST in detecting HVI. Although a positive FAST examination is more common with HVI, it misses 75% of injuries and is inadequate in isolation, a nearly identical finding to the work of Stassen et al.²¹

The association between imaging and HVI in the present study showed similar effect sizes but tighter confidence intervals when compared with the single institution study.¹⁵ The presence of free fluid on CT increased the likelihood of HVI more than 40-fold and performed well as a binary indicator for HVI (AUC, 0.87; 95% CI, 0.83-0.91). Surprisingly, removing patients with solid organ injury from the model reduced its performance as a binary classifier for HVI (AUC, 0.79; 95% CI, 0.73-0.84). Area under the receiver operator characteristic curve accounts for both the sensitivity and specificity of any given test, in this case, HVI. Though the specificity for HVI increased (ie, fewer false positives) when excluding patients with solid organ injury (90.7% from 86.4%), there was a significant reduction in sensitivity (the ability to correctly identify patients with a disease) for the detection of HVI (66.7% from 87%) due to an increase in false negatives. This suggests that free fluid in the presence of solid organ injury should not be reassuring for the absence of HVI. This finding is consistent with a previous report of the association between solid organ injury and HVI.²²

The association between a negative CT and the absence of HVI was also examined using a logistic regression model, and there was a strong association with the a priori definition of a negative CT (no radiographic SBS and no intra-abdominal findings). However, it performed poorly as a binary classifier (AUC, 0.68). On the surface, this seems to be contrary to the primary hypothesis of the study. However, this reduction in AUC is driven by the high number of false positives (ie, patients with a positive CT and no HVI). This association is improved by removing radiographic SBS from the definition of negative CT (AUC, 0.89; 95% CI, 0.86-0.90). Because there were no patients within the current study where radiographic SBS was the sole finding of HVI, a negative intra-abdominal CT has a higher sensitivity for the absence of HVI (78.6% vs 37.2%) without any reduction in specificity (98.6%). However, this finding needs to be put into the context of the previous single-institution studies, where there was a single patient with HVI whose sole finding on CT was the presence of a radiographic SBS.¹⁵ Perhaps the most conservative approach to what constitutes a negative CT is the inclusion of radiographic SBS, although the current study would suggest it is unnecessary. Regardless of the definition for a negative CT, clinical judgment remains crucial, considering the dire consequences for diagnostic delay.¹¹ While the findings of this study should reassure clinicians who want to discharge patients with negative scans who otherwise appear well, such an approach should be used with caution for patients with a concerning examination or other findings where observation would be prudent.

The lack of standardization surrounding what constitutes an abdominal SBS has previously been noted, and it is possible that providing a specific description or picture-based definition may have changed the findings of this study. However, this would not be consistent with current practice, which relies on physician judgment. Shreffler et al¹⁶ attempted to differentiate between seat belt abrasions and ecchymoses, discovering that both were significantly associated with intra-abdominal injury, and Jiang et al²³ showed that an abdominal SBS below the anterior superior iliac spine had the same risk for intra-abdominal injury as not having an abdominal SBS. While each of these studies provides insight on what constitutes true abdominal SBS, examination findings ultimately rely on physician experience and putting the finding into context.

This raises the question about whether to move to a strictly radiographic definition of abdominal SBS. Johnson and Eastridge²⁴ proposed CT-driven characterization for abdominal SBS and showed that the need for operation was higher in patients with radiographic SBS compared with patients with clinical examination findings only (19.7% vs 4.1%). If a similar methodology was applied to the current study (ie, comparison of radiographic SBS to all SBS), the incidence of HVI would increase from 9.2% to 12.7%. However, it is important to remember that approximately 25% of patients in the HVI group did not have an abdominal wall contusion on CT. Although radiographic SBS should further raise suspicion, it is inadequate to fully exclude HVI.

It is possible that limiting the definition of HVI to an operative finding underestimates the true incidence of HVI because serosal injuries, small mesenteric injuries, or patients who die before operation may go undiagnosed. However, we feel this does not detract from the primary hypothesis of the study that patients with negative findings on CT will not require an operation for HVI. It is also important to note that the radiology reports used for this study were final reads, meaning that an attending radiologist reviewed the scans. Final reads may not always be available when decisions surrounding care are being made, and it is possible that missed injury rates would increase if relying solely on preliminary reports. Another limitation of this study is its generalizability outside of trauma centers that use high-resolution, multislice CT scanners. That said, in most locations, abdominal SBS is a field indicator for trauma center triage, and most centers have high-quality CT scanners.

The findings of this study provide further evidence that CT can be used to safely exclude the presence of operative HVI in the setting of an abdominal SBS. While there has been evidence supporting the safe discharge of patients with blunt abdominal trauma and negative diagnostic studies for more than 20 years, current practice has excluded patients with abdominal SBS.²⁵ Over the course of the last 5 years, between multiple single-institution studies and now this prospective multicenter study, more than 1500 patients with abdominal SBS have been reported without a single missed operative HVI. This suggests that abdominal SBS alone, in the setting of negative findings on CT, should not mandate admission and observation, and likely these patients can be discharged from the emergency department safely, thereby avoiding unnecessary admission and associated health care expenditures.

Accepted for Publication: April 30, 2022.

Published Online: July 13, 2022. doi:10.1001/jamasurg.2022.2770

Corresponding Author: Patrick T. Delaplain, MD, Division of Trauma, Burns, & Surgical Critical Care, University of California, Irvine, 333 City Blvd West, Ste 1600, Orange, CA 92868 (pdelapla@hs.uci.edu).

Author Contributions: Drs Delaplain and Nahmias had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Delaplain, Biffl, Kim, Nahmias.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Delaplain, Biffl, Schaffer.

Critical revision of the manuscript for important intellectual content: Delaplain, Tay-Lasso, Biffl, Sundel, Behdin, Ghneim, Costantini, Santorelli, Switzer, Schellenberg, Keeley, Kim, Wang, Dhillon, Patel, Campion, Robinson, Kartiko, Quintana, Estroff, Kirby, Nahmias.

Statistical analysis: Delaplain, Behdin, Switzer, Kirby.

Administrative, technical, or material support: Delaplain, Tay-Lasso, Biffl, Kim, Dhillon, Patel, Kartiko, Quintana, Estroff.

Supervision: Schellenberg, Campion, Nahmias.

Conflict of Interest Disclosures: None reported.

Meeting Presentation: This paper was presented at the 93rd Annual Meeting of the Pacific Coast Surgical Association; February 11, 2022; Maui, Hawaii.