Page 30 - Journal of Special Operations Medicine - Fall 2017

the absolute difference between the mean of each experimen- Table 1 Glove Group Results for All Users, Using Dunnett’s Method
tal group and the mean of a control group. If positive, the Absolute Difference
absolute difference minus the least significant difference (LSD) Minus Least Significant
determined the experimental group mean was more apart Group Difference (seconds) p Value
than the LSD from the control group mean and, therefore, Bare hands −10.4 1.000
was significantly different. For pairwise comparisons of group Examination gloves −9.11 .9998
means, a nonparametric method was used. The number of Flight gloves −6.61 .8667
pairwise comparisons for nine groups was 36 ([{9 – 1} × 9]/2). Glove liners −1.11 .0955
Both LSD and Bonferroni corrections were used for pairwise Leather gloves −4.61 .5028
comparisons. Pairwise comparisons of group means were then Glove liners & leather gloves −1.86 .1440
put into levels based on statistical significance. Means not
sharing a level were significantly different, but a mean may Mittens −3.36 .3014
have existed in more than one level if it was not significantly Cold gloves −5.36 .6455
different from the other means in the level. A connecting let- Cold gloves & mittens −0.61 .0717
ters report was used for comparisons for each pair using Stu-
dent’s t-test. Table 2 Time Required to Unwrap by Glove Group for All Users
Mean Time to Unwrap (seconds)
A mixed-model ANOVA was also used, which included the 95% Confidence
user as a random effect in the model. User effects were pre- Group Mean SD SEM Limits
sented as a percentage of the overall variance component Bare hands 12 2.6 1.3 7.6, 15.9
based on the restricted maximal likelihood variance method. Examination gloves 13 0.8 0.4 11.7, 14.3
Significance for results was established when p values were Flight gloves 16 4.8 2.4 7.9, 23.1
< .05. All statistical analysis was conducted by using SAS soft- Glove liners 21 8.7 4.4 7.1, 34.9
ware (JMP version 12.0; SAS Institute; http://www.sas.com) Leather gloves 18 5.2 2.6 9.2, 25.8
and MS Excel 2003 (Microsoft; www.microsoft.com). Glove liners & leather
gloves 20 8.1 4.1 7.3, 33.2
Results Mittens 19 3.8 1.9 12.7, 24.6
Cold gloves 17 1.7 0.9 14.0, 19.5
Success in Unwrapping Cold gloves & mittens 22 4.8 2.4 13.9, 29.1
In unwrapping the tourniquets, each user was successful in each SD, standard deviation; SEM, standard error of the mean.
of their nine tests. Unwrapping as a “yes-no” parameter re-
sulted in all results as “yes” (100%; 36/36). With no change in comparisons, no result remained significant, likely due to
performance over time, unwrapping results showed a flat line, overpenalization. With 36 different pairwise comparisons, the
because no user showed any learning for this binary metric. unadjusted p value would have to be less than .00138 to re-
main significant after such an adjustment. Each of the three
Time to Unwrap levels included either six or seven means. Three means were
The mean time to unwrap for all 36 tests was 17 ± 5.6 seconds in only one level: the fastest mean for bare hands was in the
(median, 16 seconds; minimum, 9 seconds; maximum, 34 sec- fastest level with five other means, and the slowest two means
onds; range, 25 seconds). for glove liners and cold gloves layered under mittens were in
the slowest level with five other means.
In a mixed-model ANOVA with user as a random effect in the
model, the mean times to unwrap by glove group analyzed for By user, mean times to unwrap for all nine tests ranged from
fixed effects were not statistically significant in their differ- 15 seconds to 19 seconds (Table 5). Variance by user as indi-
ences (p = .0961). Among the eight experimental group means cated by standard deviation was least for the cadet (3.5 sec-
compared with the one control group mean for bare hands us- onds) and most for the research associate (6.5 seconds). The
ing Dunnett’s method, results were not statistically significant analysis showed that 9% of the variance of time to unwrap
(p > .07, all eight pairs; Table 1). The mean times by glove results could be attributed to the users.
group ranged from 12 seconds for bare hands to 22 seconds
for cold gloves layered under mittens (Table 2). By glove group, when the time to unwrap as a mean for all
users was plotted, the results ranged from a minimum of 12
Among pairwise comparisons of difference between glove seconds with bare hands to a maximum of 22 seconds with
group means, eight of the 36 comparisons were statistically cold gloves layered under mittens in a twofold (1.8 [22/12])
significant (p < .0295, all eight pairs; Table 3), but after ad- spread of the data (Figure 3). The radial polar plot depicted
justment for multiple comparisons, no comparison remained the results in an outline shaped like a nautilus; for some glove
significant. Among the eight comparisons, four involved bare groups, however, there were peaks and valleys along the out-
hands, four involved examination gloves, four involved cold line, indicating longer and shorter times, respectively. In gen-
gloves, four involved mittens, and two involved glove liners; eral, bare hands were fastest, and cold gloves layered under
these last three were either by themselves or as layered with mittens were slowest. These findings were similar to those
another type of glove (Figure 2). found in the pairwise comparisons of means reported.

When pairwise comparisons of group means using the LSD When the time to unwrap as a mean for all four users was
method were put into levels based on statistical significance, indexed to the control (i.e., mean time by glove group divided
the connecting letters report showed three levels (Table by mean time with bare hands), the results varied from 1 to
4); however, after the Bonferroni adjustments for multiple 1.9 (Figure 4).

28 | JSOM Volume 17, Edition 3/Fall 2017