Page 62 - The efficacy and effectiveness of psychological treatments for eating disorders - Elske van den Berg

62 Chapter 3
load records were used and all patient evaluations monitored. Irregularly, to monitor adherence, audiotaped sessions were reviewed.
Statistical analysis
Statistical analyses were performed with SPSS version 25 and R statistical programming environment. Differences between groups were expressed as differ- ence in proportion for categorical data and as mean difference for continuous data. Chi-squared (χ2) or Fisher’s exact tests (as appropriate) were used to compare categorical measures between two groups, and t-tests or Mann-Whitney U tests (as appropriate for the distribution of data) to compare continuous data. Paired t tests for continuous data were used to compare differences within groups. Linear mixed model analyses, for the analysis of the repeated continuous outcomes, were conduct- ed according to a two-level structure (patient and repeated measures). To control for possible confounding, we added the baseline score of the dependent outcome varia- ble plus the variables from the baseline analysis with a p <.05. Effect sizes were calcu- lated using Cohen’s d. Treatment outcome analyses were performed on the respond- ers (complete measures available at baseline and EOT) and on an imputed dataset. Multiple imputation (with 50 imputations for each missing observation) was used for missing outcome data, under a missing-at-random (MAR) assumption; compar- ing both patient groups with and without EOT EDE-Q global score, analyses showed that with regard to gender, age, eating disorder diagnosis, DASS global score, EDE-Q global score and baseline BMI, no significant differences were found. Analyses were performed using SPSS, first on the imputed datasets separately. Next, the outcomes of the 50 imputations were combined using Rubin’s rules (Rubin, 1987).With regard to cost-effectiveness calculations, all analyses were performed on intention-to-treat basis (ITT), using the imputed dataset. Unless indicated otherwise, the cost-effective- ness results are based on the multiple imputed data.
Cost-effectiveness calculations
Treatment costs in euros were established for each patient by multiplying stand- ard Dutch cost prices, index year 2016 (Zorginstituut Nederland, 2016), with the number of outpatient contacts, number of days in day care and/or inpatient stay; both contacts with patients as well as contacts concerning patients are included. As the time horizon, mean time frame between start and EOT, was little under a year, no discounting for future costs/effects was applied.
Differences in costs and effects between 2012-2014 and 2015-2017 samples were calculated as the difference in cumulative direct costs. A total of 2,500 nonparametric