Chapter 8 156 will be used to create and analyze five multiple imputed datasets. Incomplete variables will be imputed under fully conditional specification, using the default settings of the mice 3.0 package [32]. The parameters of substantive interest will be estimated in each imputed dataset separately, and combined using Rubin’s rules. Data will be screened for outliers and tested for normal distribution. Descriptive statistics will be used to summarize the baseline characteristics of the physiotherapists, and baseline demographic and clinical characteristics of the patients. Variables will be expressed in percentages, in the mean ± standard deviation (SD) or as the median with interquartile range (IQR), depending on which is appropriate. The outcomes are repeatedly measured in clusters. So, to evaluate changes in health outcomes, clinical supervision, and adherence over time, a multilevel linear mixed-model analysis will be performed [33]. The model will include health related quality of life per patient from baseline to T4 as the dependent variable. The mixed model will include a random intercept per physiotherapy practice. A correlation structure will be chosen for the repeated measurements on the level of patients by selecting the best fitting variance-covariance matrix. A similar approach will be used for the outcome adherence, number of scheduled appointments, exercise capacity, MRC-score, disease burden, and lung attacks. The discriminatory ability of the PATCH tool (the validity) will be determined by comparing the proportion of patients correctly classified as adherent or nonadherent with the actual classification measured by the RAdMAT-NL, quantified as the area under the receiver operating characteristic curve (AUROC). Using the AUROC, the threshold of the PATCH tool will be determined. In a prospective study in COPD patients in PR, the current PATCH tool had a sensitivity of 67.9% and specificity of 75.9% [17]. For all tests, p-values < 0.05 will be considered statistically significant. Analysis qualitative data The semi-structured interviews will be analyzed using N-vivo14. As a strategy for analysis systematic text condensation (STC) will be used [43]. STC is a modification of Giorgi’s phenomenological analysis and encompasses thematic analysis of meaning and content of data across cases [43]. Finally, the results of the two-level mixed model will be integrated with the results of the thematic analysis using joint displays [38]. Joint display brings qualitative and quantitative data together through a visual means to “draw out new insights beyond the information gained from the separate quantitative and qualitative results” [38].
RkJQdWJsaXNoZXIy MjY0ODMw