Analyses and statistics
We planned to evaluate a minimal relevant difference
of two to three days of hospital stay. The
postoperative stay for routine procedures in the
department was eight days (including inpatient
rehabilitation), based on patient records over the
past three years, whereas the hospital stay for
patients in the intervention group was expected to
be five days. Calculating a 5%risk of a type 1 error
and a 20% risk of a type 2 error, we concluded that
about 30 patients should be included in each group.
The groups were compared using non-paired
two-sided tests: Mann–Whitney test and Fisher’s
exact test. The statistician was blinded. The results
were analysed according to intention to treat,
except for the postoperative complications and
stay, because of the high risk for underestimated
results due to the inclusion of patients who were
not operated on in those analyses, thereby presenting
a lower frequency of complications and a lower
stay in hospital. Area under curve was used for
analyses of measurements over time in order to
reduce the number of tests. The level of significance
was 0.05. The subgroup analysis was performed as
planned by using multivariate analyses (SPSS for
Windows; SPSS Inc., Chicago, IL, USA).