**3. Results**

**Hypothesis 1.** *We tested this hypothesis by using three illness representations variables, that is cyclical timeline, personal control and treatment control*.

First, cyclical timeline representations of PD were stable between pre- and post-DBS assessments as showed by a rANOVA, *F* (2.025, 32.397) = 0.867, *p* = 0.431, η2 = 0.020, ω<sup>2</sup> = 0.000. Because of sphericity violation (Greenhouse-Geisser ε = 0.675), we used Greenhouse-Geisser correction to interpret the analyses.

In contrast with the rANOVA above, SEM analyses showed that patients perceived PD as less cyclical after surgery than before, as demonstrated with significant global mean slope changes in both the step (*b* = −1.482, SE = 0.536, *p* = 0.006) and the free (*b* = −0.556, SE = 0.235, *p* = 0.018) models. The latter model showed a significant change between T0 and T1 (*b* = 3.122, SE = 0.523, *p* < 0.001) with an e ffect size (*d* = 0.540) estimated as medium to large according to the criteria of Cohen [47]. In FIML estimation, the free model obtained the best statistical fit to the data as depicted in Table 1; ye<sup>t</sup> using Bayesian estimation the step model was better adjusted. Considering that the magnitude of the T1-T2 change was, although significant, very small (*b* = 2.804, SE = 0.528, *p* < 0.001, *d* = 0.053), our data sugges<sup>t</sup> that the change toward less cyclical representations of PD after DBS does not reinforce but stabilizes in the later post-surgical assessment sessions.

Second, patients did not report change in their representation of personal control over PD between pre- and post-DBS assessments as showed by a rANOVA, *F* (3, 48) = 0.044, *p* = 0.987, η2 = 0.001, ω<sup>2</sup> = 0.000. Because of sphericity violation (Greenhouse-Geisser ε = 0.867), we used Huynh-Feldt correction to interpret the analyses.

In the SEM analyses, although the three tested models showed good statistical fit (see Table 2), the step model was significantly better than the level, suggesting that representations of personal control over PD decreased after surgery (*b* = −1.393, SE = 0.627, *p* = 0.026). The free model specified

that this diminution had a small to medium e ffect size in the DBS pre-post transition (βT1 = 1.776, SE = 0.687, *p* = 0.010, *d* = 0.267) and continued after surgery, albeit with a much smaller amplitude, βT2 = 2.026, SE = 0.774, *p* = 0.009, *d* = 0.039. Bayesian analyses were adequate for all models, except for the slightly unstable autocorrelation trace plots of the free model.

Third, patients did not report change in their representations of treatment control on PD between pre- and post-DBS assessments as showed by a rANOVA, *F* (2.197, 35.146) = 0.1.195, *p* = 0.318, η2 = 0.001, ω<sup>2</sup> = 0.000. Because of sphericity violation (Greenhouse–Geisser ε = 0.732), we used Greenhouse–Geisser correction to interpret the analyses.

In the SEM analyses, the free model best fitted the data, as shown in Table 2, with significant mean slope change (*b* = −0.312, SE = 0.145, *p* = 0.030) and interindividual variability (*b* = −1.066, SE = 0.411, *p* = 0.010). This suggests that treatment control was perceived as weaker after DBS than before. Yet, slope estimations were not significantly di fferent from 0 at any specific measurement session; in addition, Bayesian analyses showed that the level and step models were better adjusted to the data than the free. Overall, it implies that representations of treatment control did not undergo significant changes over DBS surgery.

**Hypothesis 2.** *Recourse to instrumental coping was stable over the pre- and post-DBS period, as showed by a rANOVA, F (23, 51)* = *2.774, p* = *0.051,* η2 = *0.032,* ω<sup>2</sup> = *0.007. Because of sphericity violation (Greenhouse–Geisser* ε = *0.851), we used Huynh–Feldt correction to interpret the analyses*.

In line with this and as summarized in Table 3, the three models designed through SEM did not di ffer significantly from one another and were very similar in terms of fit indices. However, the Bayesian solution for the free model was not adequate, contrary to these of the level and the step. Detailed analyses of the step model showed that the global mean slope did not change significantly (*b* = −0.562, SE = 0.501, *p* = 0.262).


Note: SD = standard deviation, SE = standard error, *p* = probability, χ*2* = chi-squared, df = degrees of freedom, CFI = comparative fit index, pclose = probability that the RMSEA value is below .05, TLI = Tucker-Lewis index, RMSEA = root mean square error of approximation, CI = confidence intervals, DIC = deviance information criterion, enp = effective number of parameters




*J. Clin. Med.* **2020**, *9*, 1186
