Results
At study entry, 98 records were usable out of the 100 participants who
consented for the study. The Faces Scale was completed by all
participants (100%) and the posture measure had a completion rate of
68%. The child report PedsQL had more missing data (59 complete out of
98 participants = 60%) and was slightly more skewed than the parent
proxy PedsQL (95 participants completed = 97%). The data set was first
examined with descriptive analysis to confirm normality, which revealed
approximately normal distributions for most variables. Only the child
report PedsQL subscales of pain, appearance, worry, and treatment
anxiety had a skew / SE skew > 2.0,
but these results were still in the 2-3 range, and therefore
approximately normally distributed. The child report had a floor effect
but no outliers on the high end; therefore, we used parametric tests for
the primary analysis, with follow-up nonparametric tests as a
sensitivity analysis whenever slight levels of non-normality might have
affected the results. Missing data were < 10% of the total
and were handled by case-wise deletion. Although the child report PedsQL
had more missing data, only one item was not missing at random
(communication) and because of the low rate of missingness this likely
did not bias our results. Reliability in this study was acceptable with
Cronbach’s α > .90 on all subscales. Descriptive summary
data can be viewed in Table 1.