Page 75 - JSOM Winter 2017

8
Selection of Participants approximately 50%. With 50 subjects, each making a deter-
We recruited a convenience sample of emergency medicine mination as to the presence of a foreign body of a given size,
physician volunteers to participate in the study. The inclusion our study had power to detect a 20% difference in sensitivity
criteria were that participants be residents, attending physi- across the range of foreign body sizes.
cians, and physician assistants (PAs) with self-reported prior
ultrasound training. The extent of prior training could vary Investigators aggregated all hardcopy data into an Excel da-
from a 2-day introductory course to completion of a formal tabase (version 14; Microsoft,www.microsoft.com). We used
ultrasound fellowship. Of note, all the emergency medicine SPSS version 22 (IBM, www.ibm.com) for statistical analysis.
residents and PAs recruited to participate had been previously We calculated test characteristics for the use of ultrasound
required to complete a dedicated 4-week ultrasound rotation for foreign body detection including sensitivity, specificity,
during which trainees must perform 175 proctored ultrasound and positive and negative likelihood ratios. We generated
examinations to include 25 soft-tissue and musculoskeletal 95% confidence intervals (CIs) for these estimates, using
studies. The exclusion criterion was lack of any prior experi- bootstrap techniques. We further stratified these calculations
ence using ultrasound for foreign body detection. by subject education level and foreign body size. The primary
outcome analysis used a χ test to compare subject sensitivity
2
Study Protocol in foreign body detection between foreign bodies of alterna-
We used a previously described tissue simulator of food-grade tive sizes.
chicken thigh to model human tissue. 10,13,14,21 Common wooden
toothpicks were the wooden foreign bodies and were cut to stan- Results
dard similar sizes. Each participant performed ultrasound exami-
nations on 10 separate models and were blinded to the presence Characteristics of Study Subjects
or absence of foreign bodies. Five models were free of any foreign Investigators enrolled 50 subjects for study participation via
body. The remaining five models each contained one wooden convenience sampling, all of whom were eligible and agreed
toothpick of a different known length (1mm, 2.5mm, 5mm, to participate. Of these participants, there were 37 residents
7.5mm, and 10mm). Investigators inserted the toothpicks by (74%), eight attending physicians (16%) without ultrasound
pulling back the tissue model skin and inserting directly into the fellowship training, three PAs (6%), one ultrasound fellow
subcutaneous tissue at a 30° angle. Investigators then performed (2%), and one attending physician (2%) with ultrasound fel-
ultrasound to confirm a depth of insertion between 5mm and lowship training. All subjects completed the study approxi-
10mm. After insertion, investigators replaced the anatomic situa- mately within the 30-second time limit.
tion of the model skin so there were no marks visible to the study
subjects to suggest the underlying placement of a foreign body. Main Results
The pooled sensitivity for all 250 examinations of tissue mod-
The order in which participants examined each of the 10 tissue els with foreign bodies was 48.4% (95% CI, 42.1%–54.8%).
models was randomized using a random number generator to The overall specificity for all 250 examinations of tissue
produce tissue model sequences. Each participant examined models without foreign bodies was 67.6% (95% CI, 61.3%–
each tissue model once. Participants performed all ultrasound 73.3%). These values corresponded to a positive likelihood
examinations using standard coupling gel and a 7.5mHz lin- ratio of 1.5 (95% CI, 1.2–1.9) and negative likelihood ratio of
ear probe with a Sonosite M-Turbo machine (https://www 0.8 (95% CI, 0.7–0.9).
.sonosite.com). Investigators did not permit any direct contact
between subjects’ hands and the individual tissue models. In- There was no statistically significant difference in sensitiv-
vestigators instructed each participant to examine each tissue ity based on foreign body size (p = .709; Table 1). Moreover,
model for the presence of a foreign body for no more than 30 sensitivity and specificity were broadly comparable across all
seconds per model. The time started when the subject placed education levels (Table 2). Similarly, sensitivity stratified by
the ultrasound probe on the tissue model and ended by the education level remained largely stable across all foreign body
subject reporting their determination of the presence or ab- sizes studied (Figure 1).
sence of a foreign body. Subjects were not privy at any point
in the study as to the presence or absence of a foreign body in Table 1 Sensitivity of Ultrasound in Detecting Wooden Foreign
each individual tissue model. Similarly, subjects were not privy Body Stratified by Object Size (N = 50)
to the proportion of models with or without foreign bodies. Scans Identifying
Object Size, mm Object, No. Sensitivity, % 95% CI
Investigators verbally asked each subject about their level of 1 22 44.0 31.2–57.7
education (i.e., resident, PA, attending physician, ultrasound 2.5 21 42.0 29.4–55.8
fellow, or attending physician with formal ultrasound fellow- 5 25 50.0 36.6–63.4
ship training). Investigators also asked each subject whether 7.5 26 52.0 38.5–65.2
he or she identified a foreign body in each tissue model. Inves- 10 27 54.0 40.4–70.6
tigators recorded all data on a hardcopy data collection form. CI, confidence interval.

Outcomes
The primary outcome measure was subject determination as to Discussion
the presence or absence of a foreign body for each tissue model. Overview of Findings
This study found generally poor diagnostic accuracy for ultra-
Analysis sound to detect wooden foreign bodies in a chicken-tissue
We based our sample size estimate on α = .05 and β = .20 model. We did not identify any association between ultra-
with two-sided testing. We expected overall sensitivity to be sound sensitivity and foreign body size ranging from 1mm to

Emergency US to Detect Wooden Foreign Bodies | 73