Page 51 - Demo

Impact of the CCA on recorded involuntary care492the tidyverse package (Wickham et al., 2019), checked for irregularities and outliers, and transformed into time series. Because the number of clients was not constant across the study period, we created a variable of the number of recordings of restrictive measures and involuntary care per 1,000 clients.Secondly, the data were decomposed into seasonal and trend components, using loess smoothing (stl). The data were detrended and examined for seasonality using the Webel-Ollech test (seastests package; Ollech, 2019), which combines results from the QS test (p < .01) and the Kruskall-Wallis test (p < .002) to investigate whether a time series contains seasonal variation. If seasonality was present, this was subtracted from the raw data. Thirdly, Poisson regression models were developed to answer the research questions. Weekly counts of restrictive measures and recordings of involuntary care were the outcome variables in all models. To test the effects of the impact moments, we developed Poisson regression models and examined changes in the intercepts and slopes (as recommended by Bernal et al., 2016; and previously followed by Schuengel et al., 2020, using data from %u2018s Heeren Loo). To understand the direction of the effects, plots were created showing both the observed and predicted counts of restrictive measures and recordings of involuntary care. In addition, standardised mean differences were calculated as effect sizes for significant changes in the intercept and slope, using the RCountD Shiny app by Coxe (2022).As a sensitivity analysis, the Theil-Sen method was used to examine changes in slope throughout the study period. This method calculates the slopes of the regression lines of all x and y values in a dataset, and takes the median of these slopes to estimate the slope of a regression line (Wilcox, 1998).