Ires Ghielen

48 Chapter 3 variance. Figure 1 shows the distribution of the BAI items over the five subscales, with the corresponding factor loadings. Associations between anxiety, depression, cognitive and motor dys- function The residuals of both the total BAI score and subscales of the BAI had a positively skewed distribution, as shown by histograms. Normality was not improved after transformation of the data. Therefore, we used the original data for the analyses to maintain interpretability of results. Homoscedasticity and non-multicollinearity of the data was confirmed. The VIF ranged from 1.02 to 1.44 and the correlation matrix is displayed in supplementary table S1. Table 3 shows the results of the multiple regression analyses with the BAI total score and scores on the subscales of the BAI as dependent variable. The final, adjusted models are displayed, including all independent variables in the associations with the dependent variables. The unadjusted and adjusted models can be found in Supplementary table S2. As displayed in table 3, the BAI total score and the affective subscale of the BAI were both significantly associated with the BDI score. The BAI total and affective scores were negatively associated with the MoCA score. In the final adjusted model, the associations between the MoCA score and BAI total and affective scores were no longer statistically significant due to confounding by the BDI score. No other significant associations were found. Discussion In this study, we replicated the findings of our previous study [18] in an independent PD patient sample with neuropsychiatric symptoms necessitating referral to an expert neuropsychiatric outpatient clinic. From the data collected from the current CNP patient sample, PCA uncovered four clinically interpretable factors, representing subscales of the BAI ( affective, thermoregulation, cardiopulmonary, and unsteadiness ). As in the results published by Rutten et al. [18], there was a partition of psychological and somatic symptom dimensions and the highest percentage of the variance was explained by the affective subscale. This demonstrates the generalizability to an independent PD sample and confirms the robustness of our results. While the affective and thermoregulation