Page 40 - JSOM Summer 2018

analysis was done for combinations of data by model to assess User symptoms occurred in four SXT tests but in no C-A-T
how robust the difference in performance was. test. Such symptoms were feeling hot and sweaty, having a
rod pinching a finger against a clip, having hands get tired
Descriptive statistics were used to portray results. Categorical and a little beat up by turning a rod, and having hand slightly
data were analyzed by contingency tables, and likelihood ratios bruised by turning a rod.
were calculated. Continuous data were summarized by mean,
which was analyzed using analysis of variance (ANOVA) to The mean glove-donning time (± standard deviation [SD])
see differences. Fixed-effect tests were made by tourniquet was 23 seconds (SXT, 22 ± 5 seconds; C-A-T, 23 ± 6 seconds;
model. For pairwise comparisons of means, a nonparametric p = .2733). The analysis showed that 43% of the variance of
Wilcoxon method was used. Pairwise comparisons were then results in time could be attributed to the users. Users were in
put into levels on the basis of statistical significance. A mixed- two levels of significance by ANOVA, with users 1 and 3 being
model ANOVA was used for analysis of the user as a random slow and users 2 and 4 being fast (p < .0001, four pairs).
effect in that model. User effects were presented as a percent-
age of the overall variance component based on the restricted The mean time to unwrap was 16 seconds (mean ± SD: SXT,
maximal likelihood variance method. R is reported as the per- 16 ± 4 seconds; C-A-T, n = 17 ± 4 seconds; p = .0115). Two
2
centage of the response variable’s variance that is explained by reasons for a difference were recorded: (1) Approximately 20%
a linear model. Significance for results was established when of SXTs had no paper instructions for use (IFU), because the
values were p < .05. Statistical analyses were conducted with shipment was delivered so early in production, and C-A-T IFU
SAS software (JMP, version 12.0; SAS Institute; http://www were wedged between the clips to keep the IFU and tourni-
.sas.com) and MS Excel 2003 (Microsoft; www.microsoft.com). quet together. (2) If present, the SXT IFU often fell away when
unwrapped, whereas the always-present C-A-T IFU, on occa-
sion, were manually extracted from the clips. Users noted both
Results
reasons caused an intermodel difference in time. The analysis
Patient status ended in bleeding for three tests and as stable showed that 3% of the variance of results in unwrap time could
for 157 (effectiveness: SXT, n = 78; C-A-T, n = 79). The inter- be attributed to the users who were in one level of significance.
model and interuser differences were not statistically signifi- Interestingly, one user recalled an SXT IFU falling away right
cant (p = .6, both). after a test in which the assessor specifically observed and
noted that it had been manually extracted from its wrapper.
Most tests (n = 154; 96%) had satisfactory results in a “go”
status (SXT, n = 78; C-A-T, n = 76). Only user 3 had each trial The mean pretime was 39 seconds (mean ± SD: SXT, 38 ± 6
receive “go” status, but differences among models and users seconds; C-A-T, 40 ± 8 seconds; p = .0166). The analysis
were not significant (p = .4 and .9, respectively). showed that 31% of the variance of results in pretime could
be attributed to the users. Users were in two levels of signifi-
Tourniquet placement was always correct, with neither model cance, with users 1 and 3 being slow and users 2 and 4 being
nor user differences (p = 1, both). Bleeding control was as- fast (p < .0001, four pairs). On average, 41% of pretime (16
sessed in 159 tests (SXT, n = 79; C-A-T, n = 80). Model and of 39) was due to unwrapping and 59% (23 of 39) was due to
user differences were not significant (p = .2 and .4, respec- glove donning.
tively). Pulse stoppage was assessed in 157 tests (SXT, n = 79;
C-A-T, n = 78). The two cadets forgot that assessment three The mean time to determination of bleeding control was 66
times; however, model and user differences were not signifi- seconds (mean ± SD: SXT, 70 ± 30 seconds; C-A-T, 62 ± 18
cant (p = .6 and .2, respectively). seconds; p = .0075). The analysis showed that 44% of the
variance of results in time could be attributed to the users.
Tourniquets were damaged in two tests of the SXT by user 4. Users were in two levels, with users 1 and 2 being slow and
Model and user differences were not significant (p = .1, both). users 3 and 4 being fast (p < .0049, six pairs).
Manikin damage occurred in 18 tests (SXT, n = 15; C-A-T,
n = 3; p = .002). The scientist, user 1, detected 12 damages; The mean total trial time was 116 seconds (mean ± SD: SXT,
user differences were significant (p < .0001). Users noted glove 121 ± 43 seconds; C-A-T, 111 ± 26 seconds; p = .0038). The
damage in 14 tests (SXT: n = 10, tears were attributed to its analysis showed that 67% of the variance of results in time
rod; C-A-T: n = 4, tears were attributed to user errors). User could be attributed to the users. Users were in three levels,
4 noted gloves catching on the SXT rod in seven other tests; with users 1 and 2 being slowest and next to slowest, respec-
no such instances were reported when using the C-A-T. The tively, and users 3 and 4 being fast (p < .0012, six pairs).
assessor even abraded the table with the SXT rod in another
test. Users noted that the SXT rod, which was metal, rigid, and The mean post-time was 51 seconds (mean ± SD: SXT, 51 ± 25
textured coarsely like a wood file, often abraded the wound, seconds; C-A-T, 50 ± 15 seconds; p = .4883). The analysis showed
a hard plastic portion of the manikin. However, users noted that 40% of the variance of results in time could be attributed
the C-A-T rod (a composite of fiberglass and plastic, semi- to the users. Users 1 and 2 were slowest and next to slowest,
rigid, and not coarsely textured) contacted and moved over respectively, and users 3 and 4 were fast (p < .0028, five pairs).
the wound surface without causing damage. Furthermore, its SXT would have had a higher mean and lower variance and user
rod could be lifted slightly to avoid contact altogether, and 4 would have been slower if he had not forgotten to check bleed-
lifting was most needed at higher turn numbers. However, the ing control; as a result, that one test had no post-time.
SXT rod could not lift at higher turn numbers, because it see-
sawed rigidly through the central fulcrum to drive the opposite The mean blood loss volume was 508mL (mean ± SD: SXT,
tip of the rod down into the manikin skin, which blocked both 505mL ± 135mL; C-A-T 511mL ± 120mL; p = .6219). The
tipping and lifting. analysis showed that 62% of the variance of results in blood

38 | JSOM Volume 18, Edition 2/Summer 2018