Measuring adherence to pulmonary rehabilitation 39 RAdMAT-NL The RAdMAT-NL is a 16-item questionnaire that uses a four-point rating scale (never = 1, occasionally = 2, often = 3, always = 4) to evaluate clinic-based adherence [16]. The original RAdMAT consists of three subscales: Attendance/participation ‘follows the prescribed rehabilitation plan’ (items 1-5, range 5-20 points), Communication ‘asks questions about his/her rehabilitation’ (items 68, range 3-12 points), and Attitude/effort ‘gives 100% effort in rehabilitation sessions’ (items 9-16, range 8-32 points). The total scale range is 16-64 points (maximal adherence). Like the SIRAS, the RAdMAT-NL was completed by the physiotherapist after three months of rehabilitation, independent of the patient and not in their presence. Statistical analysis Data were analyzed using R version 4.0.3 [21] using the {psych} and {eRm} package. Data were screened for outliers and tested for normal distribution. Only complete sets of data were included in the analysis. Descriptive statistics were used to evaluate the baseline variables of the patients and of their disease. Variables were expressed in percentages or as the median with interquartile range (IQR). To test the factorability of our relatively small data set (n = 193) the significance level of Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin measure of sampling adequacy (KMO) were calculated. The KMO was 0.90, and all values for individual items were > 0.76, exceeding the recommended minimum value of 0.7. Bartlett’s test of sphericity was, c2 (120) = 1759.6, p < 0.001, below the needed p < 0.05 [22], indicating the appropriateness of conducting a factor analysis. Explorative principal axis factor analyses were conducted to examine the dimensionality (structural validity) of the RAdMAT-NL. The absence of cross loading items was considered of evidence of structural validity of the RAdMAT-NL. A principal axis factoring (PAF) analysis was chosen over principal component analysis because the primary aim was to detect the underlying structure (latent variables) rather than to simply reduce the number of items [23]. PAF was performed followed by a parallel analysis and associated scree plot to explore how many factors the RAdMAT-NL consists of. After the factor solution as indicated by the parallel analysis was performed, a forced one factor solution was performed to determine whether this solution was better or worse than the proposed factor solution. Both solutions were evaluated by the number of cross-loadings (items loading higher, r > 0.40, on multiple factors [24]), the root mean square of the residuals (RMSR) (a value of 0.05 or lower is optimal), the Fit based upon off diagonal values (values > 0.95 are good), the Tucker Lewis Index (TLI) (a larger value is better; > 0.90 is good), the root mean square error of approximation (RMSEA) (smaller value is better; ) [25] and the Bayesian information criterion (BIC) (a smaller value indicates a better fit) [22].
RkJQdWJsaXNoZXIy MjY0ODMw