Page 159 - Demo

Implementing MDET: Evaluating adaptations and effectiveness1576Quantitative data analysisThe data were analysed using R version 4.3.2 (R Core Team, 2023) in R Studio version 2023.12.1.402 (Posit team, 2024). R packages imputeTS (Moritz & Bartz-Beielstin, 2017P), readxl (Wickham & Bryan, 2023), tidyverse (Wickham et al., 2019), and tsibble (Wang et al., 2020) were used for data cleaning and preparation, and psych (Revelle, 2024) was used for the descriptive statistics. Alpha levels were set at .05 for all analyses. First, the data were cleaned and prepared for analysis. For each care home, missing weekly data were imputed using linear interpolation. Then, the outcome variable for research question 2 (weekly recorded involuntary care) was calculated by the sum of weekly recorded involuntary care and unforeseen care, averaged per 10 clients to aid interpretability of the results. For research question 3, a variable was created of the number of weekly recorded incident reports, also averaged per 10 clients. The time series data of care homes in the intervention and CAU conditions were then aligned. This was done by randomly matching pairs of care homes in both conditions (one in the CAU condition and one in the intervention condition), finding the minimum and maximum overlapping weeks and years for each pair, and trimming both datasets to keep the rows that fall within the overlapping weeks and years. This way, the time series of each pair of care homes started and ended in the same week and year. Then, the variable indicating the start and end date of MDET of the care home in the intervention condition was copied to the care home in the CAU condition, which made it possible to compare the impact of MDET between care homes in both conditions. All pairs of care homes were then merged into the final dataset. Lastly, for each care home, a time variable was created, ranging from 1 to the total number of rows (weeks) in the data. Descriptive statistics (means, averages, frequency and range) were then calculated for all study variables, separately for care homes in the CAU and intervention condition.For research question 2, linear mixed effects models (package nlme; Pinheiro et al., 2023) were developed, with a random intercept for care home and an autoregressive correlation structure. Weekly recorded involuntary care was the outcome variable. The first model (model 1) was an unconditional means