Test Characteristics of Focused Assessment of Sonography for Trauma for Clinically Significant Abdominal Free Fluid in Pediatric Blunt Abdominal Trauma

Abstract

ACADEMIC EMERGENCY MEDICINE 2011; 18:477–482 © 2011 by the Society for Academic Emergency Medicine

Abstract

Objectives:  Focused assessment of sonography in trauma (FAST) has been shown useful to detect clinically significant hemoperitoneum in adults, but not in children. The objectives were to determine test characteristics for clinically important intraperitoneal free fluid (FF) in pediatric blunt abdominal trauma (BAT) using computed tomography (CT) or surgery as criterion reference and, second, to determine the test characteristics of FAST to detect any amount of intraperitoneal FF as detected by CT.

Methods:  This was a prospective observational study of consecutive children (0–17 years) who required trauma team activation for BAT and received either CT or laparotomy between 2004 and 2007. Experienced physicians performed and interpreted FAST. Clinically important FF was defined as moderate or greater amount of intraperitoneal FF per the radiologist CT report or surgery.

Results:  The study enrolled 431 patients, excluded 74, and analyzed data on 357. For the first objective, 23 patients had significant hemoperitoneum (22 on CT and one at surgery). Twelve of the 23 had true-positive FAST (sensitivity = 52%; 95% confidence interval [CI] = 31% to 73%). FAST was true negative in 321 of 334 (specificity = 96%; 95% CI = 93% to 98%). Twelve of 25 patients with positive FAST had significant FF on CT (positive predictive value [PPV] = 48%; 95% CI = 28% to 69%). Of 332 patients with negative FAST, 321 had no significant fluid on CT (negative predictive value [NPV] = 97%; 95% CI = 94% to 98%). Positive likelihood ratio (LR) for FF was 13.4 (95% CI = 6.9 to 26.0) while the negative LR was 0.50 (95% CI = 0.32 to 0.76). Accuracy was 93% (333 of 357, 95% CI = 90% to 96%). For the second objective, test characteristics were as follows: sensitivity = 20% (95% CI = 13% to 30%), specificity = 98% (95% CI = 95% to 99%), PPV = 76% (95% CI = 54% to 90%), NPV = 78% (95% CI = 73% to 82%), positive LR = 9.0 (95% CI = 3.7 to 21.8), negative LR = 0.81 (95% CI = 0.7 to 0.9), and accuracy = 78% (277 of 357, 95% CI = 73% to 82%).

Conclusion:  In this population of children with BAT, FAST has a low sensitivity for clinically important FF but has high specificity. A positive FAST suggests hemoperitoneum and abdominal injury, while a negative FAST aids little in decision-making.

Children less than 15 years old comprise 21.7% of emergency department (ED) visits.¹ Unintentional injuries are the leading cause of mortality for children ages 1–19 years and account for over 43% of deaths.² Among unintentional injuries, traffic collisions (automobile vs. automobile, pedestrian, and bicycle) and falls predominate. These mechanisms frequently cause blunt abdominal trauma (BAT), which is a leading cause of morbidity and mortality.³

The focused assessment of sonography in trauma (FAST) evaluates for free fluid (FF) around the heart and three areas of the abdominal–pelvic cavity.^4,5 The majority of traumatic deaths occur early and are due to injuries in the airway, brain, or thoracoabdominal cavity.^5–7 FAST has become a standard adjunct to the secondary survey in trauma resuscitation and, in our institution, has replaced diagnostic peritoneal lavage. FAST can identify FF in the peritoneal cavity, which immediately directs operative intervention in the unstable patient. In the trauma setting, FF is presumed to be blood, but FAST does not localize the source.⁵

The test characteristics of FAST have been determined in adults with positive and negative likelihood ratios (LRs) of 98 and 0.02, respectively.⁸ However, its performance has been inadequately assessed in children. Previous studies had few patients,^9,10 were retrospective,^11,12 used unrealistic definitions of FF, or had overly stringent criterion references (need for operation).^13–15 Consequently, reports of test characteristics vary widely.^{9–11,13–17} Pediatric trauma management rarely involves surgery, but important decisions may depend on FAST, such as the need for intensive care, transfer to higher level care, prioritization of multiple victims, and type and volume of resuscitation.

The first objective was to determine test characteristics for clinically important FF in pediatric blunt abdominal trauma using computed tomography (CT) or surgery as criterion references. The second objective was to determine test characteristics of FAST to detect any intraperitoneal FF. To our knowledge, we present the largest prospective observational study of clinician-performed pediatric FAST in consecutive patients.

Methods

Study Design

This was a prospective, descriptive observational study comparing the results of FAST performed in the ED compared to CT or laparotomy in the setting of pediatric BAT. The local institutional review board approved the study design. Informed consent was required for enrolled subjects. Informed consent was obtained from parent or legal guardian, and patient assent was obtained from subjects older than 7 years.

Study Setting and Population

We studied consecutive patients aged 0–17 years who suffered BAT requiring trauma team activation at a tertiary care Level I trauma center university hospital from 2004 to 2007. We considered blunt mechanism as falls, motor vehicle crashes, automobile versus pedestrian collisions, nonaccidental blunt trauma, and battery. Annual ED patient census was 36,000, with trauma activation volume 2,000 to 2,200 per year. The proportion of all trauma patients with Injury Severity Score (ISS) greater than 15 was 24.5%. Seventy-nine percent of patients were victims of blunt injury. Pediatric trauma volume was 283 patients per year, or 850 patients during the study period (including total blunt and penetrating trauma). We assured 100% capture of pediatric BAT by reviewing the trauma center patient log each morning.

Study Protocol

The FAST exams were performed by emergency medicine (EM) residents (62% of total, 51% of total by third-year EM residents [EM3s]), attending emergency physicians (EPs; 21%), and ultrasound (US) fellows and surgeons (8% each), using a B+K Hawk 2102 (Copenhagen, Denmark), Sonosite Titan, or Sonosite Micromaxx (Bothell, WA) US machine with 3.5- to 5.0-MHz curved array (Hawk), 2- to 4-MHz convex array (Titan), or 2- to 4-MHz phased array transducers. All EM3s and EPs had substantial training in FAST in the ED, each with at least 300 total US exams, well exceeding criteria for US competency outlined by the American College of Emergency Physicians.¹⁸

By protocol, all eligible patients had FAST at arrival, followed by CT of the abdomen and pelvis within 30 minutes, or underwent laparotomy (n = 1). All patients had CT with intravenous contrast, while we used oral contrast routinely for the first 18 months of the study, but not after. Only an initial FAST was part of the protocol. Trauma patient management was at the discretion of the treating physicians. The decision to proceed to CT or operating room (OR) was made by the surgeon in attendance, who had knowledge of the FAST interpretation. However, all study patients were evaluated with CT by preexisting protocol, unless they were unstable and required an immediate operation.

Research personnel recorded demographic information on a standardized data form. The physician performing the US entered FAST results on the data form. Research personnel accessed radiology or OR records for results of the CT or laparotomy, respectively.

Measures

We used a composite criterion reference of CT findings and operative report. We defined clinically important FF on CT as moderate or more, reported by the attending radiologist or if the patient went to the OR immediately for intraabdominal injury. Conversely, the terms “trivial,”“trace,” and “small” were considered not clinically important. For FAST, any amount of FF in any of the three windows (hepatorenal, splenorenal, suprapubic) was considered positive. We did not routinely interrogate the chest for pleural fluid. The CT scans were interpreted by radiologists blinded to the FAST results and clinical findings. Any FF at surgery was considered clinically important. We defined “organ injury” on CT as a report by the radiologist of solid or hollow intra- or retroperitoneal organ injury. Conversely, reports of “occult injury” based on findings of FF alone and bony fractures were not considered injuries. ISS were abstracted from the trauma database for 334 patients (94%).

Data Analysis

Descriptive statistics were calculated using Stata (version 10.1, StataCorp, College Station, TX). Exact binomial confidence intervals (CIs) are reported for proportions. Because of small cell sizes, p-values for contingency tables were calculated using Fisher’s exact test. FAST results are compared to the criterion reference in contingency tables. Test characteristics, calculated using the conventional definitions, are presented with exact CIs. Accuracy is the percentage of FAST results that are consistent with the criterion reference.

Results

We approached the families of 475 potentially eligible pediatric trauma victims for consent. A total of four hundred thirty-one (91%) consented. We excluded 74 (21 with no signatures, 46 no confirmatory studies, and seven with wrong mechanism of injury) and analyzed data on 357 (75% of all potentially eligible subjects). Demographics and mechanism of injury of analyzed patients are show in Table 1.

Table 1.    Patient Characteristics (n = 357)

Characteristic	n	% of Patients (95% CI)
Age, years
0–2	34	9 (7–13)
2–6	88	25 (20–29)
7–12	79	22 (18–27)
13–17	156	44 (38–49)
Sex
Male	230	64 (59–69)
Female	127	36 (31–41)
Mechanism of injury
Automobile vs. pedestrian/bicycle	144	40 (35–46)
Traffic collision	125	35 (30–40)
Fall	52	15 (11–19)
Blunt other (sports, animal, object)	24	7 (4–10)
Battery	7	2 (1–4)
Unknown	5	1 (0.5–3)

Our patients were moderately injured. The interquartile range for ISS was 4–12, with 15% above 15. Sixteen of 357 (4.5%) patients had surgery: six abdominal, five orthopedic, three neurosurgical, and two thoracic injury. Patients were admitted to the pediatric intensive care unit (29%), surgical intensive care unit (3%), or ward (68%).

Table 2 shows the training level of sonographers for all 357 patients. There was no statistically significant difference in the proportion of positive FAST by level of training (p = 0.10, Fisher’s exact test). All FASTs were interpreted at the bedside by the sonographer. There were 14 board-certified radiologists who interpreted the CT scans; only one was a pediatric specialist, who read 24% of the CTs.

Table 2.    Level of Sonographer Performing FAST

Level of Sonographer	n	% of Patients	True-positive FAST*
EM = emergency medicine year of training; FAST = focused assessment with sonography in trauma; FF = free fluid; US = ultrasound. *For clinically important FF.
Attending physician	69	19	8
Surgery resident (all levels)	35	10	0
US fellow	28	8	1
EM3 resident	173	48	13
EM2 resident	27	8	0
Unknown (including EM1 residents)	25	7	0

Results for Objective 1

Table 3 shows the FAST results with clinically significant hemoperitoneum. Twenty-three patients had significant hemoperitoneum (22 on CT and one at surgery). Twelve of the 23 had true positive FAST (sensitivity = 52%; 95% CI = 31% to 73%). FAST was true negative in 321 of 334 (specificity = 96%; 95% CI = 93% to 98%). Table 4 shows the FAST results for any amount of hemoperitoneum. Twelve of 25 patients with positive FAST had FF on CT (positive predictive value [PPV] = 48%; 95% CI = 28% to 69%). Of 332 patients with negative FAST, 321 had no significant fluid on CT (negative predictive value [NPV] = 97%; 95% CI = 94% to 98%). The positive LR for FF was 13.4 (95% CI = 6.9 to 26.0), while the negative LR was 0.50 (95% CI = 0.32 to 0.76). Accuracy was 93% (333 of 357, 95% CI = 90% to 96%).

Table 3.    FAST results by Clinically Significant Hemoperitoneum

FAST	Clinically Significant Hemoperitoneum
	Present	Absent	Total
FAST = focused assessment with sonography in trauma.
Positive	12	13	25
Negative	11	321	332
Total	23	334	357

Table 4.    FAST Results for Any Level of Hemoperitoneum (Large, Moderate, Small, Trace)

FAST	CT Positive or Odds Ratio	CT Negative	Total
FAST = focused abdominal sonographic exam for trauma.
Positive	19	6	25
Negative	74	258	332
Total	93	264	357

Four patients with clinically significant hemoperitoneum went to the OR for intraabdominal injuries. One of these patients was a 10-year-old female who fell off of her bicycle; she had a negative FAST, but was taken immediately to the OR from the ED. She had an exploratory laparotomy, small bowel resection, large bowel resection, and mesenteric tear repair. The remaining three patients had positive FAST.

Results for Objective 2

Ninety-three patients had some level of FF on CT or at laparotomy (large, moderate, small, or trace). FAST was positive in 19 of these (20% sensitive, 95% CI = 13% to 30%) and negative in 74. A total of 264 patients did not have any FF on CT. FAST was negative in 258 of these (98% specific, 95% CI = 95% to 99%) and positive in six. The PPV was 76% (95% CI = 54% to 90%), and NPV was 78% (95% CI = 73% to 82%). Accuracy was 78% (277 of 357, 95% CI = 73% to 82%). Positive LR was 9.0 (95% CI = 3.7 to 21.8) and negative LR was 0.81 (95% CI = 0.7 to 0.9). Two additional patients from this group went to the OR for a total of six patients whose intraabdominal injuries were managed operatively. Both of these patients had trace, small FF on CT, and neither had positive FAST. One patient had severe hydronephrosis/suspected ureter injury and went to the OR for a ureteral stent placement. The other patient had a jejunal and cecal injury and went to the OR for a bowel resection. Of the 264 patients with no FF at all by CT, 12 had solid organ injury, but none went to the OR.

FAST was more likely to be positive with increasing FF on CT (Figure 1, p < 0.0005, Fisher’s exact test). Of the 52 patients with abdominal injury identified on CT, 19 (37%) had clinically important FF, 21 (40%) had trace to small amounts of FF, and 12 (23%) had no FF (Table 5). Two of these 12 with no FF had false-positive FAST. Ten of these 12 patients had a solid organ injury, and 2 of 12 had hollow organ injury. No patient without FF on CT went to the OR.

Figure 1.  Percentage* of patients with positive FAST by FF on CT. *Excludes one patient who went to surgery with unquantified FF (n = 356; p < 0.0005). CT = computed tomography; FAST =  focused abdominal sonographic exam for trauma; FF = free fluid

Table 5.    Statistics* for FF on CT and Results of FAST and Presence of Organ Injury, Regardless of Presence of FF on CT

FF by CT	Positive FAST		Negative FAST		Total
	CT Showing Organ Injury	No Organ Injury	CT Showing Organ Injury	No Organ Injury
CT = computed tomography; FAST = focused abdominal sonographic exam for trauma; FF = free fluid. *Excludes one patient with a negative FAST who went to surgery with unquantified FF.
Large	6	0	0	0	6
Moderate	6	0	7	3	16
Small	5	0	11	27	23
Trace	0	2	5	20	27
None	2	4	10	248	264
Total	19	6	33	298	356

Discussion

Computed tomography is the standard for imaging pediatric patients with BAT.¹⁹ It evaluates stable trauma victims, locates injury, and identifies free air and FF. Any amount of FF on CT is suspicious for injury. Further evaluation is via serial abdominal exams, repeat CT, or exploratory laparotomy. CT is sensitive and accurate but inappropriate for unstable patients.^19,20 Furthermore, there may be increased risk of cancer from diagnostic radiation. A recent review estimated one fatal cancer per 1000 CT scans in children.²¹

For the EP or trauma surgeon, FAST is convenient and portable,¹⁶ provides rapid data, and can be used even during resuscitation.²² FAST detects intraperitoneal FF, and thus many intraabdominal injuries, but does not detect retroperitoneal or bowel injury without FF.^23–25

Pediatric FAST is very specific for hemoperitoneum (83% to 98.3%), but sensitivity varies widely (33% to 100%).^9,11,15–17 These studies suggest that FAST is less sensitive than CT, but overall useful to screen for pediatric blunt trauma. However, they compare FAST to variable standards such as need for operation, any organ injury on CT including solid alone, and any FF found on CT.

We found specificity consistent with previous studies, with very low false-positive rates.²⁶ Conversely, FF detected by FAST is likely a true positive. For example, in our population, the pretest probability of clinically important intraabdominal FF is 6.4% (23 of 357). Given the positive LR of 13.4, using the Fagan nomogram the posttest probability rises to 48%. Thus, when positive, FAST significantly alters the suspicion of bleeding. Conversely, the negative LR in our study (0.50) scarcely lowers the pretest probability to 3%.²⁷ Hence when negative, FAST does not inform clinical decision making.

We studied a more realistic criterion reference: FF more than physiologic, but not necessarily requiring operation. Our rationale was that some BAT victims would be menstruating teenagers with physiologic FF (10.6% of our patients fell into this category).²⁸ A previous study showed that CT scans interpreted as “physiologic” and “questionable occult injury” had no complications or return visits.²⁹ Therefore, we felt comfortable assigning these patients as true-negatives for our clinical outcome. Finally, management of pediatric BAT, even with considerable hemorrhage, is largely nonoperative. Despite this, our definition of clinically significant FF did not catch all patients who had intraabdominal injuries, nor did it catch all those patients who were managed operatively. Two patients with trace to small amounts of FF on CT went to the OR for intraabdominal injuries. However, these patients had a bowel injury and a ureteral injury, respectively. Twelve patients without any FF on CT had intraabdominal injuries, including 10 with solid organ (liver, spleen, or kidney) injuries.

Because of the temporal order of FAST versus CT, new or further hemoperitoneum can develop during the interim between FAST and CT. FAST cannot predict future bleeding. Therefore, FAST suffers from an inherent diagnostic disadvantage. In our study, 30 minutes (mean; range = 20 to 95 minutes) elapsed from arrival to CT, with FAST typically performed in the first 5 minutes.

Our enrollment of patients was planned to be consecutive, but parents of critically ill children more often declined, limiting the true-positive FASTs that we can report and decreasing our reported sensitivity and predictive value.³⁰ Had we enrolled all critically ill patients increasing our prevalence of disease (FF), our strong positive LR may have generated higher posttest probabilities. Because of this, and the temporal delay of FAST as compared to CT, we report what is likely a minimum sensitivity of pediatric FAST.

The Trendelenberg position places the patients at a downward angle and encourages FF to settle in Morrison’s pouch, a dependent area between the liver and right kidney. We recommended this position for our patients as well as a darkened room, but compliance was variable (30 and 47%, respectively). These subgroups were too small to allow valid comparisons.

Limitations

Perhaps the largest limitation of our study was the definition of “clinically significant.” Our definition failed to predict all patients with solid organ injury or who went to the OR. When all definitions of FF are included in the data analysis, the sensitivity of FAST is low (20%). Twelve patients without any FF still had intraabdominal injuries, but none of these patients went to the OR.

Additionally, there is inconsistency between the interval from arrival to FAST scan and from FAST to CT. This may explain the greater number of positive CTs for FF than was found with FAST. There is a customary delay of 30 minutes between FAST and CT in our institution. We did not re-FAST our patients to determine if this time interval affected FAST test characteristics.

Although this appears to be is the largest clinician-performed prospective series to date, we found a relatively small proportion of positive FAST results (6%), abdominal injuries (15%), and patients whose intraabdominal injuries were managed operatively (2%). Definitions of amounts of FF used by different radiologists were not standard. We did not calculate interrater reliability among radiologists. We did not study utility of FAST to identify cardiac or thoracic injuries.

The institutional review board requirement of obtaining informed consent in pediatric blunt trauma victims resulted in 21 eligible patients ultimately not included as subjects. Although we cannot publish the characteristics of these subjects, they tended to be sicker patients. Understandably, parents of these severely injured children were difficult to approach and consent, secondary to their emotional distress.^30,31 This may have biased our research to reflect less injured patients.

Conclusions

We found that focused assessment of sonography in trauma does not have a high enough sensitivity to prelude the use of CT in a pediatric blunt abdominal injury population. The focused assessment of sonography in trauma exam has a high specificity. A positive focused assessment of sonography in trauma suggests hemoperitoneum and abdominal injury, while a negative focused assessment of sonography in trauma aids little in decision-making.