Radiology

Claude B. Sirlin, MD
Miche`le A. Brown, MD
Reena Deutsch, PhD
Olga A. Andrade-Barreto,
MD
Dale A. Fortlage
David B. Hoyt, MD
Giovanna Casola, MD

Index terms
Abdomen, injuries, 66.761, 70.41,
81.4134
Abdomen, US, 70.1298
Emergency radiology
Trauma, **.412
Published online
10.1148/radiol.2293030285
Radiology 2003; 229:766 774
Abbreviation:
MTOS Major Trauma Outcome
Study
1

From the Departments of Radiology

(C.B.S., M.A.B., O.A.A.B., G.C.), Family and Preventive Medicine (R.D.),
and Surgery (D.A.F., D.B.H.), University of California, San Diego, UCSD
Medical Center, 200 W Arbor Dr, MC
8756, San Diego, CA 92103. Presented at the 2002 RSNA scientific assembly. Received February 20, 2003;
revision requested May 14; revision received June 4; accepted July 23. R.D.
supported in part by National Institutes of Health grant M01 RR00827.
Address correspondence to C.B.S.
(e-mail: csirlin@ucsd.edu).

**. Multiple body systems

Screening US for Blunt

Abdominal Trauma: Objective
Predictors of False-Negative
Findings and Missed Injuries1
PURPOSE: To determine the risk for missed injury in patients with blunt abdominal
trauma and negative findings at screening ultrasonography (US) and with coexistent
hematuria or fracture of the sixth through 12th ribs, lumbar spine, or pelvis.
MATERIALS AND METHODS: From a database of 4,000 patients screened with
US for blunt abdominal trauma at a level 1 trauma center, the 3,679 patients with
negative US findings were retrospectively classified by consensus of two authors into
high-risk (n 494) and low-risk (n 3,185) groups based on the presence of
hypothetical predictors of missed injury: hematuria (n 96) or fracture of the sixth
through 12th ribs (n 216), lumbar spine (n 105), or pelvis (n 174). Outcome
in each patient was determined by the same two authors consensually after retrospective review of the trauma registry and all radiologic, surgical, and autopsy
reports. The risk for missed abdominal injury was determined for each patient risk
group and for each hypothetical predictor. Risks were statistically compared by
using the Pearson 2, Fisher exact, or Fisher-Freeman-Halton exact test, depending
on expected frequencies.
RESULTS: High-risk patients were 24 times more likely to have abdominal injuries
after negative US findings (30 [6.1%] of 494) than were low-risk patients (eight
[0.25%] of 3,185) (P .001). Among high-risk patients, the absolute risks for missed
abdominal injury associated with specific predictors were 15.6% (15 of 96 patients)
for hematuria, 6.0% (13 of 216) for lower rib fractures, 7.6% (eight of 105) for
lumbar spine fractures, and 5.2% (nine of 174) for pelvic fractures. Each of these
risks was significantly higher for patients in the high-risk group than for those in the
low-risk group (P .001).
CONCLUSION: Hematuria and fracture of the lower ribs, lumbar spine, or pelvis are
objective predictors of missed abdominal injury in patients with blunt abdominal
trauma and negative US findings, and such patients may benefit from additional
screening with computed tomography.

Author contributions:
Guarantor of integrity of entire study,
C.B.S.; study concepts, C.B.S., M.A.B.,
O.A.A.B., D.B.H., G.C.; study design,
C.B.S., M.A.B., G.C.; literature research, C.B.S., M.A.B., G.C.; clinical
studies, C.B.S., M.A.B., G.C., D.B.H.;
data acquisition, C.B.S., M.A.B., G.C.,
O.A.A.B., D.A.F.; data analysis/interpretation, C.B.S., D.A.F., R.D.; statistical analysis, C.B.S., R.D.; manuscript
preparation and manuscript editing,
C.B.S., R.D., G.C.; manuscript definition of intellectual content and manuscript revision/review and final version
approval, all authors

RSNA, 2003

766

RSNA, 2003

Ultrasonography (US) is the primary method used to screen for blunt abdominal trauma at
several trauma centers in Europe and Asia, as well as at selected centers in the United States
(110). Although false-negative results with screening US are rare (1%) (10 14), they may
cause a delay in diagnosis and management of abdominal injuries. Most abdominal
injuries missed at screening US are detected at subsequent computed tomographic (CT)
examinations ordered because of clinical suspicion of missed injuries (15). Reliance on
clinical symptoms as a prompt for further screening is problematic, however, because
many symptoms (eg, abdominal pain) are subjective, and others (eg, decreasing hematocrit levels) may take time to develop. Therefore, we sought to define objective factors that
could help identify patients with blunt abdominal trauma who are at high risk for
false-negative findings at screening US. The delineation of such factors, we believed, would
not only expedite diagnosis and management of abdominal injury but also could help
decrease morbidity and mortality associated with missed injury.

Radiology

We hypothesized that hematuria and

fractures of the sixth through the 12th
ribs, lumbar spine, and pelvis might be
predictors of missed abdominal injury in
patients with negative US findings. Thus,
the primary purpose of our study was to
determine the risk for missed abdominal
injury in patients with blunt abdominal
trauma and negative findings at screening US and with coexistent hematuria or
fracture of the sixth through the 12th
ribs, lumbar spine, or pelvis.

MATERIALS AND METHODS

Patients
From a database of 4,000 patients who
underwent US for blunt abdominal
trauma at a level 1 trauma center between April 1994 and July 2001 and who
qualified for the Major Trauma Outcome
Study (MTOS) (16), we retrospectively
identified 3,679 patients in whom the
results of screening US were prospectively interpreted as negative for abdominal injury. The average age of the 3,679
patients was 40 years (range, 199 years):
2,597 patients were male (mean age, 38;
range, 195 years) and 1,082 were female
(mean age, 42; range, 4 99 years). The
difference in mean age between the sexes
was statistically significant (P .001, unpaired two-tailed Student t test). This
study received full approval from the institutional review board. Informed consent was not routinely obtained for the
use of patient data, because the institutional review board waives this requirement for records of patients in whom
abdominal US is performed in the trauma
setting.
After screening US, patients were admitted to the trauma surgery service for
observation and management of other
injuries (median hospital stay, 1.8 days;
range, 6 hours to 430 days). Patients who
were discharged after screening US without being admitted did not meet MTOS
criteria and were not included in the
study. Anteroposterior chest radiography
and urinalysis were performed in all patients at admission. CT and other tests
including repeat US, cystography, diagnostic peritoneal lavage, exploratory
laparotomy, and radiography of the
spine and pelviswere performed at the
discretion of the trauma service. The
number of patients in whom radiographs
of the spine or pelvis were obtained was
not recorded. Because all study patients
met MTOS entry criteria, all were considered to have nontrivial mechanisms of
trauma. Survivors were examined at folVolume 229

Number 3

low-up appointments in the trauma

clinic 1 week after discharge. Autopsies
were performed in all patients who died.
The results of autopsy and any additional
follow-up radiologic or surgical examinations served as the standard of reference
for this study. If no autopsy or additional
examination was performed, clinical outcome was used as the standard, as previously described (4,10 13,15). The date
and time of imaging were not recorded.
Urinalysis results were typically available
within 15 minutes of urine specimen collection, but the exact time was not recorded.
The trauma service entered patients
prospectively into the trauma registry.
The registry contains information about
all documented injuries and is updated
regularly by the trauma service as new
injuries are discovered. To identify any
missed injuries after discharge from the
hospital, the trauma service performed
systematic monthly audits of all trauma
centers in the county. In these monthly
audits, all errors and delays in diagnosis
of traumatic injuries in all county trauma
centers were discussed, tabulated, and
entered into the trauma registry of each
institution. To our knowledge, this is the
only systematic county-level trauma auditing system in the United States.

with an alternator. After April 2001, images

were archived digitally with a picture
archiving and communication system
(IMPAX; Agfa Healthcare, Ridgfield Park,
NJ) and reviewed on a 19-inch color monitor (MWD 421; Barco Display Systems,
Kortrijk, Belgium) with resolution of
1,000 1,000 pixels.
US images were interpreted prospectively by the resident and staff radiologist
on the US service at the time of the examination. Images were considered positive for abdominal injury only if they
depicted free fluid, extraparenchymal hematoma, or parenchymal abnormality
indicative of injury; all other images were
considered negative. In the trauma setting, any parenchymal abnormalities except fatty infiltration of the liver or
clearly visualized cysts, gallstones, renal
calculi, or calcified granulomas are considered to indicate possible injury. On
the basis of a previously reported study
(12) and in the absence of other suspicious findings, we considered small accumulations of anechoic fluid measuring
less than 3 cm in the anteroposterior dimension and isolated in the cul-de-sac in
women of reproductive age to be physiologically normal and not indicative of
traumatic injury.

Data Collection and Statistical

Analysis

US Technique
US examinations were performed during trauma resuscitation by one of 15
certified sonographers (with 120 years
of experience) using one of two scanners
(ATL HDI 3000, Advanced Technologies
Laboratories, Bothell, Wash; or 128-XP,
Acuson, Mountain View, Calif) and a
2.25-, 3.5-, or 5.0-MHz sector transducer
or a 5.0-MHz curved-array transducer
with Doppler capabilities. Prior to scanning, the patients bladders, if empty,
were distended with 200 300 mL of sterile saline solution administered via Foley
catheter. Seven abdominal regions were
examined for free fluid, including the upper quadrants, the pelvis, the paracolic
gutters, and the renal fossae. Visceral organs were evaluated for parenchymal abnormalities indicative of injury. Total
scanning time typically was 35 minutes,
but the precise duration of each US examination was not recorded.

US Image Interpretation
Until April 2001, US examinations were
recorded with an imaging system (Kodak
Image Link; Eastman Kodak Company,
Rochester, NY), and images were reviewed

The trauma registry and all radiologic,

surgical, and autopsy reports for each patient in the study were reviewed by two
authors (C.B.S., M.A.B.) in consensus to
determine the presence of fracture of the
lower ribs, lumbar spine, or pelvis, or of
hematuria at admission urinalysis. Lower
rib fractures were defined as those involving one or more of the lower seven ribs
(the sixth through 12th ribs). The exact
portion of a rib that was fractured was
not documented. Fractures involving the
first through the fifth ribs were not considered because they did not satisfy our a
priori hypothesis. Any documented fractures of the lumbar spine or pelvis qualified. For lumbar spine fractures, we recorded whether fractures were limited to
the posterior elements (transverse or spinous processes, pedicles, and laminae) or
involved at least one vertebral body. We
did not record the exact site or type of
pelvic fracture. We did not grade the severity or score the number of individual
fractures involving the ribs, lumbar
spine, or pelvis. Hematuria was defined
as more than 50 red blood cells per highpower field, because this was the definition used in the trauma registry. The

Screening US for Blunt Abdominal Trauma

767

Radiology

gross appearance of the urine specimen

was not recorded in the trauma registry
and therefore was not included in our
analysis.
On the basis of our hypothesis, patients were retrospectively divided into
two risk groups by the same two authors
in consensus. The high-risk group (n
494) consisted of patients with hypothetical predictors of missed injury: hematuria (n 96) or one or more fractures in
the lower ribs (n 216), lumbar spine
(n 105), or pelvis (n 174). The lowrisk group (n 3,185) had no hematuria
or fractures.
The sum of the numbers of patients
with each individual hypothetical predictor exceeded the number of patients
in the high-risk group because some
high-risk patients had more than one
predictor. Whereas 408 patients had a
single predictor, 76 patients simultaneously had two predictors, nine simultaneously had three predictors, and one
simultaneously had all four predictors.
Outcomes in high- and low-risk groups
were then compared by calculating the
absolute and relative risks for missed abdominal injury (any, surgical, and nonsurgical). Absolute risk was defined as the
percentage of patients with negative
findings and missed injury at screening
US. Relative risk was calculated by dividing the absolute risk in a given group by
the absolute risk in the low-risk group.
Comparisons were repeated with data
stratification according to patient sex.
Abdominal injuries were defined as contusion, laceration, or hematoma of intraabdominal structures, or as unexplained
intraabdominal fluid if no discrete injury
was identified. Surgical injuries were defined as those for which laparotomy was
performed. Nonsurgical injuries were defined as those treated successfully without surgical intervention or considered
minor at autopsy. The general location of
injuries (intraperitoneal, extraperitoneal,
or both) and the specific organs involved
were documented. The delay in hours between negative findings at screening US
and eventual diagnosis of missed injury
was recorded.
Each of the four hypothetical predictors (hematuria, rib fracture, lumbar
spine fracture, and pelvic fracture) was
analyzed as an independent risk factor
for missed injury. This analysis was performed in two separate stages. In the first
stage, absolute risk and relative risk were
calculated for missed injury associated
with each of the parameters. Patients
were included in the risk calculation for
each hypothetical predictor that they
768

Radiology

December 2003

had; thus, patients who had more than

one predictor were counted more than
once in the analysis. The specific sites
and general locations (intraperitoneal,
extraperitoneal, or both) of missed abdominal injuries associated with each
predictor were recorded.
In the second stage of analysis, the individual hypothetical predictors were statistically compared. To achieve accurate
statistical comparison, we had to ensure
that patients with more than one predictor were not counted more than once. In
contrast to the first stage, in which patients were counted as many times as the
number of predictors they had, in the
second stage, patients with more than
one predictor were excluded from analysis. For example, to compare the three
categories of fracture, we restricted our
analysis to the 378 patients who had fractures involving only one part of the skeleton (ribs [n 182], lumbar spine [n
71], and pelvis [n 125]), regardless of
whether they had hematuria. To compare hematuria with the various fractures, we restricted our analysis to 408
patients: the 60 patients with hematuria
and no fracture, and the 348 patients
with only one type of fracture and no
hematuria (169 with fracture of the ribs,
70 with fracture of the lumbar spine, and
109 with pelvic fracture). To further evaluate lumbar fracture as a predictor of
missed abdominal injury, we restricted
our analysis to the 105 patients with lumbar fracture (independent of hematuria
or other fractures): 39 with one or more
fractures limited to the posterior elements of the spine, and 66 with one or
more fractures involving at least one vertebral body.
Finally, each patient was assigned a numeric risk score from 0 to 3, according
to the number of hypothetical predictors
present in that patient. The risk for
missed injury was then calculated for
each risk score.
For relative risk, 95% CIs were calculated. Proportions were statistically compared by using the Pearson 2 test, Fisher
exact test, or Fisher-Freeman-Halton exact test, depending on the expected frequencies. A Wilcoxon rank sum test was
used to determine differences in number
of injuries per patient between risk
groups. The Somer D test for association
was applied to assess for a trend of increasing risk for missed injury as the risk
score increased from 0 to 3. P values less
than .05 were assumed to indicate statistically significant difference for each
analysis, with no adjustments being
made for multiple comparisons. Power to

detect reported differences was calculated

post hoc by using two-tailed tests. Statistical software was used for these analyses
(JMP version 5.0, SAS Institute, Cary, NC;
SPSS version 10.1.0 for Windows, SPSS,
Chicago, Ill; and StatXact-5 version 5.0.3,
Cytel Software, Cambridge, Mass).

RESULTS
Comparison between Risk Groups
Of 3,679 patients with negative findings at screening US, 494 (13.4%) retrospectively were assigned to the high-risk
group because of hematuria (50 red
blood cells per high-power field) at admission urinalysis or because of fracture
of the sixth through the 12th ribs, the
lumbar spine, or the pelvis. A total of 60
patients had hematuria only, 398 had
fractures only, and 36 had both fractures
and hematuria. Of the 494 high-risk patients, 30 (6.1%) had missed abdominal
injuries, including 10 (2.0%) with surgical and 20 (4.0%) with nonsurgical injuries (Fig 1). The 30 injured patients in the
high-risk group included 16 with isolated
intraperitoneal injuries, 10 with isolated
extraperitoneal injuries, and four with
combined intra- and extraperitoneal injuries. There were 55 cumulative injuries
in this group (1.8 injuries per injured patient), including 10 retroperitoneal hematomas and eight renal, seven adrenal,
10 bowel or mesenteric, nine hepatic,
nine splenic, and two other injuries (Table 1).
For comparison, 3,185 patients with
negative findings at screening US retrospectively were assigned to the low-risk
group because they had no fractures and
no hematuria. Of the 3,185 low-risk patients, eight (0.25%) had missed abdominal injuries, including four (0.1%) with
surgical injuries and four (0.1%) with
nonsurgical injuries (Fig 1). The eight
injured patients in the low-risk group
included three with isolated intraperitoneal injuries, four with isolated extraperitoneal injuries, and one with combined intra- and extraperitoneal injuries.
There were 10 cumulative injuries in this
group (1.3 injuries per injured patient),
including three retroperitoneal hematomas and two bowel or mesenteric injuries
(Table 1). The difference between the
numbers of injuries per injured patient in
the two patient risk groups (ie, 1.8 vs 1.3)
was not statistically significant (P .06).
Patients in the high-risk group were 24
times more likely than patients in the
low-risk group to have a missed abdominal injury (relative risk, 24; 95% CI: 11,
Sirlin et al

Radiology

52) (Fig 2). The relative risk was 16 (95%

CI: 5, 51) for surgical injuries and 32
(95% CI: 11, 94) for nonsurgical injuries
(Fig 2). The differences in risk for missed
injury (any, surgical, or nonsurgical) between high-risk and low-risk patient
groups were statistically significant (P
.001 for each comparison). Our study had
higher than 99% power to detect these
reported differences.
The risks for missed abdominal injury
were virtually identical for male patients
and female patients. Of the 494 high-risk
patients, 337 were male and 157 were
female. Of the 337 male patients, 21 (6%)
had missed abdominal injuries detected
after negative screening US, including 10
(3%) with surgical injuries. Similarly, of
the 157 female patients, nine (6%) had
missed injuries, including four (3%) with
surgical injuries.
The delay between negative findings at
screening US and eventual diagnosis of
missed abdominal injuries in patients of
each risk group is shown graphically in
Figure 3. All missed injuries in the eight
low-risk patients were diagnosed within
24 hours of initial US. By contrast, missed
injuries in 26 (87%) of the 30 high-risk
patients were diagnosed within 24 hours;
missed injuries in the other four high-risk
patients were diagnosed at 31, 63, 72,
and 144 hours.

Figure 1. Graph shows absolute risk for missed abdominal injuries

according to risk group and for the study population. Bars show the
percentages of low-risk and high-risk patients with surgical or nonsurgical abdominal injuries detected after negative screening US and
the percentage of all patients with such injuries. Patients in the
high-risk group were significantly more likely than those in the lowrisk group to have a missed abdominal injury after negative findings
at screening US (P .001).

TABLE 1
Number and Location of Missed
Abdominal Injuries by Patient
Risk Group

Assessment of Individual Predictors

A total of 216 patients had fractures in
the lower ribs, 105 had fractures in the
lumbar spine, and 174 had pelvic fractures. All fractures were visible on radiographs and documented in dictated reports except for three transverse process
fractures in lumbar vertebrae, which were
detected at CT in patients who had not
undergone radiography of the spine. Because these three fractures involved gross
displacements that likely would have
been visible on radiographs if radiography had been performed, they were included in the analysis. Ninety-six patients had hematuria; the number with
gross hematuria was not recorded. Because some patients had more than one
hypothetical predictor and therefore
were counted more than once, the total
number of patients with predictors exceeds the number of patients in the highrisk group.
The risk for missed abdominal injury
was significantly higher in patients with
any of the hypothetical predictors than
in patients with no predictors (P .001
for each comparison). As shown in Table
2 and Figure 4, 5.2%7.6% of patients
Volume 229

Number 3

Risk Group
Location of Injury

Figure 2. Graph shows relative risk for

missed abdominal injuries in the high-risk patient group; the low-risk group is the reference.
Lines indicate 95% CIs. High-risk patients were
16 32 times more likely than low-risk patients
to have a missed abdominal injury, depending
on the type of injury.

with fractures had missed abdominal injuries (ie, the percentage varied according
to the type of fracture), including 1.2%
2.9% with surgical injuries and 4.0%
4.8% with nonsurgical injuries. Moreover, 15.6% of patients with hematuria
had missed injuries, including 5.2% with
surgical injuries and 10.4% with nonsurgical injuries (Table 2, Fig 4). Thus, compared with the risks for patients with no
predictors, the relative risks of missed in-

Intraperitoneal
Liver
Spleen
Bowel or mesentery
Other
Extraperitoneal
Kidney
Adrenal gland
Retroperitoneum
Other
Total

Low
Risk

High
Risk

0
1
2
1

9
9
10
1

0
1
3
2

8
7
10
1

10

55

Note.Eight patients (0.25%) of 3,185 in the

low-risk group and 30 (6.1%) of 494 in the
high-risk group had one or more missed abdominal injuries. One patient in the low-risk
group and four in the high-risk group had
combined intra- and extraperitoneal injuries.

jury were 24 for patients with rib fracture

(95% CI: 10, 57), 30 for patients with
lumbar spine fracture (95% CI: 12, 79),
21 for patients with pelvic fracture (95%
CI: 8, 53), and 62 for patients with hematuria (95% CI: 27, 143) (Fig 5). Our study

Screening US for Blunt Abdominal Trauma

769

TABLE 2
Risk of Missed Abdominal Injury by Predictor

Radiology

Patients with Missed Abdominal Injury

Predictor
None
Fracture
Lower rib
Lumbar spine
Pelvis
Hematuria

Patients with
Predictor

Total

Patients with
Surgical Injury

Patients with
Nonsurgical Injury

3,185

8 (0.25)

4 (0.1)

4 (0.1)

216
105
174
96

13 (6.0)
8 (7.6)
9 (5.2)
15 (15.6)

3 (1.4)
3 (2.9)
2 (1.2)
5 (5.2)

10 (4.6)
5 (4.8)
7 (4.0)
10 (10.4)

Note.Data are numbers of patients, with percentages in parentheses. Patients with more than
one predictor were counted more than once.

(three [8%] of 39 patients) had the same

risk for missed abdominal injury as did
those with lumbar fractures involving
the vertebral body (five [8%] of 66 patients).
In the calculations of risk and relative
risk and in the tabulation of specific injuries associated with each hypothetical
predictor, patients with more than one
predictor were counted more than once.
In contrast, in the statistical comparison
of individual predictors, patients with
more than one predictor were excluded
so that they would not be compared with
themselves.

Assessment of Risk Score

Figure 3. Comparison of the delay between negative findings at US

and identification of missed injury according to patient risk group.
Bars indicate the number of abdominal injuries detected in low-risk
versus high-risk patient groups at various time intervals after initial
screening US. All eight (100%) low-risk patients versus 26 (87%) of 30
high-risk patients with missed abdominal injuries had their missed
injuries detected within 24 hours of screening US.

had a power of 98%99% for detection of

these reported differences.
Cumulatively, the numbers of missed
injuries in high-risk patients were as follows: There were 20 missed injuries in
patients with lower rib fracture, 17 in
those with lumbar fracture, 18 in those
with pelvic fracture, and 28 in those with
hematuria (Table 3). The majority of injuries were intraperitoneal for patients
with rib or lumbar fracture and extraperitoneal for patients with pelvic fracture or
hematuria. The most common specific
injuries associated with fracture of the
lower ribs were injuries to the liver (n
5) and spleen (n 4); with lumbar fractures, retroperitoneal hematoma (n 5)
and injuries to the spleen (n 4) and the
bowel or mesentery (n 4); with pelvic
fractures, retroperitoneal hematoma (n
5) and injuries to the kidney (n 4); and
with hematuria, kidney (n 7) and adrenal (n 5) injuries.
A total of 378 patients had a single
770

Radiology

December 2003

fracture type. In these patients, the small

differences in risk for missed injury 4%
(eight of 182 patients) for lower rib fracture, 3% (two of 71 patients) for lumbar
spine fracture, and 2% (three of 125 patients) for pelvic fracturewere not statistically significant (P .7). By contrast,
among patients with only one hypothetical predictor, the risk associated with hematuria (nine [15%] of 60 patients) was
significantly higher than the risk associated with any of the fracture types (P
.014 for lower rib, P .01 for lumbar
spine, and P .001 for pelvic fractures).
Post hoc, the power of our study to
achieve statistical significance with these
reported differences between hematuria
and fracture patients depended on the
fracture type: Power was 65% for rib, 60%
for lumbar spine, and 94% for pelvic fractures. In patients with fractures of the
lumbar spine, the type of lumbar fracture
was not significant; patients with fractures isolated to the posterior elements

A total of 3,185 patients had no hypothetical predictors (risk score of 0), 408
had one predictor (risk score of 1), 76 had
two predictors (risk score of 2), and 10
had three or more predictors (risk score of
3). The percentage of patients with
missed injuries increased in a stepwise
fashion as the risk score increased (Fig 6):
Eight (0.25%) of 3,185 patients with
missed injuries had a risk score of 0, 19
(4.7%) of 408 had a risk score of 1, eight
(10.5%) of 76 had a risk score of 2, and
three (30.0%) of 10 had a risk score of 3.
A similar trend was observed in the group
of injured patients with regard to the
need for surgery: four (0.1%) of 3,185
patients with a score of 0, seven (1.7%) of
408 with a score of 1, andwith the two
upper risk score categories combined to
compensate for the small sample size
three (3.5%) of 86 with a score of 23
had injuries that required surgery. These
trends were statistically significant (P
.001 and P .01, respectively). Our study
had a power higher than 99% for detection of these reported trends.

DISCUSSION
False-negative findings at screening US
are rare in patients with suspected blunt
abdominal trauma; in our experience,
only 1% of patients with negative screening US have missed injuries (10,15). To
detect missed injuries, trauma surgeons
at our institution have usually relied on
subjective or slowly developing clinical
parameterssuch as abdominal pain,
tenderness, decreasing hematocrit levels,
hemodynamic instability, and signs of
sepsis (15). With the aid of these clinical
parameters, 50% of missed injuries are
recognized within 12 hours and 89% are
found within 24 hours of screening US
(15). Although this diagnostic delay is
relatively short, it would be better to
Sirlin et al

Radiology

identify patients at high risk for falsenegative findings at US even sooner; these
patients could then undergo screening
CT, a more definitive test, even before
missed injuries manifest clinically. This
would be beneficial because the rapid diagnosis and treatment of abdominal injury is a key factor in decreasing morbidity and mortality associated with blunt
trauma.
The results of this study confirm our
hypothesis that hematuria and fractures
of the lower ribs, lumbar spine, and pelvis are predictors of missed injury in patients with negative findings at screening
US for blunt abdominal trauma. Patients
with at least one of these predictors of
injury were significantly more likely to
have a missed injury than were patients
without any predictors (6.1% vs 0.25%,
relative risk of 24). The differences remained statistically significant when injuries were stratified as surgical (2.0% vs
0.1%, relative risk of 16) or nonsurgical
(4.0% vs 0.1%, relative risk of 32). Thus,
although high-risk patients constituted
only 13% (494 of 3,679) of all patients in
the study, they accounted for 79% (30 of
38) of those with missed injuries, including 71% (10 of 14) of those with surgical
injuries and 83% (20 of 24) of those with
nonsurgical injuries. In addition to their
prognostic value, a distinct advantage of
these predictors is that they are objective
criteria that can be assessed at initial
work-up.
As opposed to patients with coexistent
predictors of injury, patients with negative findings at screening US and no predictors had abdominal injuries extremely
rarely. In our study, only one in 398 lowrisk patients with negative US findings
had a missed abdominal injury and only
one in 796 had a missed injury requiring
surgery. It should be emphasized, however, that screening US examinations at
our institution are performed by experienced certified sonographers using highend US scanners and include a thorough
evaluation of peritoneal and retroperitoneal compartments for free fluid and
for parenchymal abnormalities. Because
equipment, sonographer experience, and
examination protocols may vary from institution to institution, our results may
not be generalizable to all trauma centers.
Our choice of specific predictors was
made a priori and was based on several
observations and assumptions. We chose
hematuria as a potential predictor, for
example, because we assumed that in the
setting of trauma, hematuria should be
considered a surrogate marker of urinary
Volume 229

Number 3

Figure 4. Graph shows absolute risk for missed injury for patients
with each hypothetical predictor. Bars show the percentages of patients with no predictors; patients with fractures of the lower ribs,
lumbar spine, or pelvis; and patients with hematuria in whom surgical or nonsurgical abdominal injuries were detected after negative
findings at screening US. Patients with at least one hypothetical
predictor were significantly more likely to have a missed abdominal
injury than patients with no predictors, and patients with hematuria
were significantly more likely to have a missed injury than patients
with any other predictor. Fx fracture, LS lumbar spine, Rib
lower rib.

Figure 5. Graph shows relative risk for missed abdominal

injury in high-risk patients by predictor of missed injury.
The low-risk patient group is the reference. Vertical lines
indicate 95% CIs. Fx fracture, LS lumbar spine, Rib
lower rib. Depending on the specific predictor, high-risk
patients had a risk of missed injury 21 62 times higher
than that of low-risk patients.

tract injury. This assumption is supported by our data: The most commonly
injured organs in patients with hematu-

ria were the kidneys and the neighboring

adrenal glands. We chose the various
fractures as predictors because fractured

Screening US for Blunt Abdominal Trauma

771

TABLE 3
Number and Location of Missed Abdominal Injuries by Predictor

Radiology

Predictor
Location of Injury

Lower Rib
Fracture

Lumbar
Fracture

Pelvic
Fracture

Hematuria

5
4
3
0

3
4
4
0

2
1
3
0

3
3
4
1

3
2
3
0

0
1
5
0

4
3
5
0

7
5
4
1

20

17

18

28

Intraperitoneal
Liver
Spleen
Bowel or mesentery
Other
Extraperitoneal
Kidney
Adrenal gland
Retroperitoneum
Other
Total

Figure 6. Graph shows risk for missed injury

as a function of risk score (score of 0, 1, 2, or
3). Bars indicate the percentage of patients
with no, one, two, or three or more individual
predictors of missed injury in whom abdominal injury was detected after negative findings
at screening US. Because of small sample sizes,
surgical and nonsurgical injuries were pooled
for each risk score. The risk for missed injury
increased significantly as the risk score increased.

bones could lacerate tissues and cause direct injury to organs. We chose the sixth
through the 12th ribs because these ribs
are in contact with portions of the diaphragmatic pleura and would be most
likely, when fractured, to puncture the
abdomen. Similarly, we chose lumbar
spine and pelvic fractures because of the
proximity of these bones to abdominal
structures. In our study, the most commonly injured organs in patients with
lower rib fractures were the liver and
spleen, and the most common injury in
patients with lumbar spine or pelvic fractures was retroperitoneal hematoma.
However, not all abdominal injuries could
be directly explained by a specific predictor. Intraperitoneal solid organ and bowel
or mesenteric injuries, for example, occurred in patients with hematuria and
772

Radiology

December 2003

lumbar or pelvic fractures, although not

as commonly as did extraperitoneal injuries. Nonetheless, these particular fractures were important surrogate markers
of important injury, regardless of direct
causality.
Among the individual predictors, hematuria had a higher association with
missed abdominal injury than did any of
the fractures, whereas the various fractures conveyed similar risks. These statistical results should be interpreted cautiously. One might conclude that
hematuria is a better marker of missed
injury than the other predictors; however, the higher risk associated with hematuria could be related to the somewhat arbitrary definition of hematuria in
the trauma registry (50 red blood cells
per high-power field), which served as
the source of urinalysis data in our study.
It is likely that hematuria, if defined by a
lower threshold number of red blood
cells, would have had a lower association
with missed injury. This hypothesis was
not analyzed in our study because the
trauma registry did not contain data on
such urinalysis specimens. Moreover, for
the sake of simplicity, we made no attempt to grade the severity of individual
fractures. It is likely that displaced, comminuted, or multiple fractures may have
increased the risk for abdominal injury,
but this hypothesis also was not evaluated. Although the sample sizes are small,
our data indicate that patients with isolated posterior element fractures had a
risk for missed abdominal injury similar
to that of patients with vertebral body
fractures; thus, patients with even relatively minor lumbar fractures should undergo additional testing to exclude occult
abdominal injury. This finding supports
previous observations regarding the im-

portance of transverse process fractures

in the assessment of trauma patients (17).
As we expected, the risk for missed abdominal injury increased as the number
of predictors increased. Although patients with multiple predictors were relatively uncommon (only 2.0% [76 of
3,679] had a risk score of 2 and only 0.3%
[10 of 3,679] had a risk score of 3), they
were at particularly high risk for missed
injury (10.5% and 30.0%, respectively).
These findings lead us to conclude that
such patients should receive a more thorough initial assessment. It also should be
emphasized, however, that nearly 5% of
patients with only one predictor (risk
score of 1) had missed abdominal injuries. Thus, even patients with a single
predictor should undergo additional testing.
On the basis of these results, we propose a new procedure for triaging blunt
abdominal trauma patients (Fig 7): If abdominal injury predictors are known or
suspected to be present in a patient before screening US is performed, we suggest that the patient be screened instead
with CT. If no predictors are known or
suspected, the patient should be screened
with US. However, because screening US
is not a definitive test for blunt abdominal trauma injuries, patients with negative findings at US and with no known or
suspected predictors should be admitted
for observation. The required admittance
of trauma patients with negative US findings would not impose an additional burden on most level 1 trauma centers, because trauma patients even those with
negative findings at CTare routinely
admitted. If predictors are discovered
during observation, patients with negative US findings should undergo further
evaluation with CT.
It is important to note that the presence of these various predictors may not
be evident prior to US. If it is not evident,
then the patient should undergo initial
screening with US. Any uniform attempt
to assess for these predictors prior to the
performance of US may be time consuming and could incur unnecessary costs
and radiation exposure to patients.
If we had followed the proposed triage
procedure from the inception of screening US at our institution, only eight
(0.2%) of the 3,679 patients with negative US findings who were included in
this study would have been inappropriately triaged (ie, admitted for observation
without undergoing CT). The missed injuries in all eight of these patients were
discovered during the first 24 hours of
observation, so any negative effect of apSirlin et al

Radiology

plying this procedure would presumably

have been small.
Two limitations of our study should be
emphasized. Because of the retrospective
nature of the study, some data could not
be reliably collected. For example, the exact times at which urinalysis and radiographic examinations of the chest, lumbar spine, and pelvis were performed and
interpreted were not recorded prospectively. Typically, patients underwent
chest radiography initially, followed by
abdominal US. Additional radiographic
examinations were performed in some
patients after screening US, on the basis
of clinical suspicion of spine, pelvic, or
other fractures. Urine specimens were
usually obtained prior to the performance of US.
Another limitation of our study is that
definitive findings of missed abdominal
injury (eg, findings at CT, laparotomy, or
autopsy) were available for only a minority of the patients included in our study.
As a result, clinical suspicion and findings at follow-up diagnostic testing were
used as the reference standard for the
majority of our patients (3,343 [90.9%] of
3,679). To optimize the validity of a clinical reference standard, we limited our
study to patients who met MTOS entry
criteria. In other words, all surviving patients were admitted either for a minimum of 6 hours to the intensive care unit
or the intermediate care unit, where continuous monitoring is possible, or for a
minimum of 72 hours to the surgical
ward. Patients who were discharged after
screening US without admission did not
meet MTOS entry criteria and were not
included in the study. Finally, all county
trauma centers were audited monthly to
detect potential missed injuries after discharge. Although some discharged patients with missed injuries might have
received subsequent treatment at institutions outside the county, such patients
would likely have been few.
Similarly, although urinalysis and
chest radiography were performed in all
patients, radiographs of the lumbar spine
and pelvis were obtained on a case-bycase basis at the discretion of the trauma
service, because our institutions trauma
assessment procedure includes selective
radiographic screening of the lumbar
spine and pelvis on the basis of clinical
suspicion (18). Thus, an imaging reference standard for lumbar and pelvic fractures was not available in all study subjects, and it is conceivable that some
patients in the low-risk group may have
had undiagnosed fractures. Moreover,
because radiographs were not obtained
Volume 229

Number 3

Figure 7. Flow chart of proposed procedure for triaging patients

who are believed to have blunt abdominal trauma. Patients who have
or are suspected of having predictors of missed injury (ie, patients at
high risk) should be screened immediately with CT. Patients who do
not have predictors of missed injury (ie, patients at low risk) should be
screened with US. Patients with negative findings at screening US
should be admitted for observation. If predictors of missed injury are
discovered after findings at screening US were interpreted as negative,
patients should undergo confirmatory CT to exclude missed abdominal injury. BAT blunt abdominal trauma, Lap laparotomy.

in all patients, lumbar transverse process

fractures were detected at CT in three
patients who did not undergo spinal radiography. In our opinion, inclusion of
these fractures in our analysis was warranted because the fractures were grossly
displaced and likely would have been detected on radiographs. Radiography was
not performed in these patients, however, because it would have provided no
benefit with regard to patient care. Moreover, our study was designed to retrospectively assess the predictive value of
various fractures and included no prospective assessment of radiographically
detectable fractures. Finally, because the
criteria for obtaining ancillary radiographs in trauma victims may vary
among institutions, our results may not
be generalizable to all trauma centers.
It is noteworthy that the male patients
in our study population, on average,
were much younger than the female patients. One possible explanation for this
divergence in age is that boys and young
men are more likely than girls and young
women to be trauma victims. In addition, because of the large sample size, a
small absolute difference (4 years) in
mean age achieved high statistical significance. Nevertheless, the age difference
between the male and female patient
subgroups did not bias our results, because the risk for missed injury was unaffected by patient sex.

In conclusion, the results of our study

show that hematuria and fractures of the
lower ribs, lumbar spine, and pelvis are
objective predictors of missed abdominal
injury in blunt abdominal trauma patients with negative findings at initial
screening with US. It is important to note
the presence of these predictors in
trauma patients, because the rapid diagnosis and treatment of abdominal injury
is a key factor in decreasing morbidity
and mortality associated with blunt
trauma. These objective predictors can be
assessed immediately after the patients
arrival in the resuscitation suite. We believe that patients with these predictors
for missed injury should be screened immediately with CT rather than US,
whereas appropriate care for low-risk patients can be determined on the basis of
US findings alone. Patients who have no
predictors for missed abdominal injury
need not undergo CT or diagnostic peritoneal lavage, both of which involve a
risk of associated complications. Lastly,
the presence of one or more of these predictors in patients with positive US findings could indicate an increased likelihood that their injuries will require
surgery, but additional study is needed to
test this hypothesis.
References
1. Hoffman R, Nerlich M, Muggia-Sullam M,
et al. Blunt abdominal trauma in cases of

Screening US for Blunt Abdominal Trauma

773

Radiology

2.

3.

4.

5.

6.

7.

774

multiple trauma evaluated by ultrasonography: a prospective analysis of 291 patients. J Trauma 1992; 32:452 458.
Rothlin MA, Naf R, Amgwerd M, Candinas D, Frick T, Trentz O. Ultrasound in
blunt abdominal and thoracic trauma.
J Trauma 1993; 34:488 495.
Yoshii H, Sato M, Yamamoto S, et al. Usefulness and limitations of ultrasonography in the initial evaluation of blunt abdominal trauma. J Trauma 1998; 45:45
51.
Healy MA, Simons RK, Winchell RJ, et al.
A prospective evaluation of abdominal ultrasound in blunt trauma: is it useful?
J Trauma 1996; 40:875 885.
McGahan JP, Rose J, Coates TL, Wisner
DH, Newberry MS. Use of ultrasonography in the patient with acute abdominal
trauma. J Ultrasound Med 1997; 16:653
662.
McKenney MG, Martin L, Lentz K, et al.
1,000 consecutive ultrasounds for blunt
abdominal trauma. J Trauma 1996; 40:
607 612.
McKenney KL, Nunez DB, McKenney
MG, Asher J, Zelnick K, Shipshak D.
Sonography as the primary screening
technique for blunt abdominal trauma:

Radiology

December 2003

8.

9.

10.

11.

12.

13.

experience with 899 patients. AJR Am J

Roentgenol 1998; 170:979 985.
Rozycki GS, Ochsner MG, Schmidt JA, et
al. A prospective study of surgeon-performed ultrasound as the primary adjuvant modality for injured patient assessment. J Trauma 1995; 39:492500.
Rozycki GS, Ballard RB, Felliciano DV,
Schmidt JA, Pennington SD. Surgeon-performed ultrasound for the assessment of
truncal injuries. Ann Surg 1998; 228:557
567.
Brown MA, Casola G, Sirlin CB, Budorick
N, Patel N, Hoyt DB. Blunt abdominal
trauma: screening US in 2,693 patients.
Radiology 2001; 218:352358.
Brown MA, Casola G, Sirlin CB, Patel N,
Hoyt DB. Importance of evaluating organ
parenchyma during screening abdominal
ultrasonography after blunt trauma. J Ultrasound Med 2001; 20:577583.
Sirlin CB, Casola G, Brown MA, et al. US
of blunt abdominal trauma: importance
of free pelvic fluid in women of reproductive age. Radiology 2001; 219:229 235.
Sirlin CB, Casola G, Brown MA, Patel N,
Bendavid EJ, Hoyt DB. Quantification of
fluid on screening ultrasonography for
blunt abdominal trauma: a simple scor-

14.

15.

16.

17.

18.

ing system to predict severity of injury. J

Ultrasound Med 2001; 20:359 364.
Pinto F, Bignardi E, Pinto A, Rizzo A, Scaglione M, Romano L. Ultrasound in the
triage of patients after blunt abdominal
trauma: our experience in 3,500 consecutive patients (abstr). Radiology 2002;
225(P):358.
Sirlin CB, Casola C, Brown MA, Hoyt DB.
Clinical value of the negative ultrasound
screening examination in patients with
blunt abdominal trauma (abstr). Radiology 2000; 217(P):481.
Champion HR, Copes WS, Sacco WJ, et
al. The major trauma outcome study: establishing national norms for trauma
care. J Trauma 1990; 30:1356 1365.
Patten RM, Gunberg SR, Brandenburger
DK. Frequency and importance of transverse process fractures in the lumbar vertebrae at helical abdominal CT in patients with trauma. Radiology 2000; 215:
831 834.
Niedens BA, Gross EA. Validation of a
decision instrument to limit pelvic radiography in blunt trauma. Acad Emerg
Med 2003; 10:475 476.

Sirlin et al