In addition to analyzing the group differences of overall costs for the entire inpatient period
we tested, if the expected decrement of treatment costs over time was actually present in our
data. For all partial categories of costs, i.e. medication, personnel and infrastructure we tested
group x time effects and the main effect of time by a repeated measures ANOVA between means
of biweekly intervals for weeks 1 to 8 (i.e. averages of week 1–2, 3–4, 5–6, 7–8) where the number
of inpatients in both groups was still sufficient to perform such a test. The national costs
accounting system actually predicted a decrement reaching a constant level within this interval.