Rumination in Children with Social Anxiety Disorder: Effects of Cognitive Distraction and Relation to Social Stress Processing

The study employed a 2 × 2 × 2 mixed repeated measures design, consisting of two experimental groups (SAD vs. HC), two experimental conditions (distraction vs. uninstructed post-event processing) and two repeated measurement points (Stress Task 1 [T1], Stress Task 2 [T2]). All participants took part in both experimental conditions. The order was counterbalanced within each group.

Children aged 9 to 13 years were recruited through information letters, flyers, and advertisements. Following a telephone screening, eligible families were invited to take the Kinder-DIPS (Margraf et al., 2017), a modified German version of the Anxiety Disorders Interview Schedule for Children–Revised (Silverman & Nelles, 1988). Inclusion criterion for the group with SAD was a primary diagnosis of SAD according to the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013). Children in the HC group did not have any lifetime diagnosis of a mental disorder. Exclusion criteria for all children were health conditions that could alter psychophysiological assessment (e.g., asthma, cardiac arrhythmia, methylphenidate) and past or current psychological treatment. Initially, 62 participants were included in the study but two children with SAD had to be excluded due to irregularities in the study procedure. Participant characteristics are found in Table 1. All participants spoke German fluently, and only one participant in the HC group did not report German as their first language. The following comorbid disorders were present in the group with SAD: specific phobia (n = 11), general anxiety disorder (n = 6), child separation anxiety disorder (n = 2), elective mutism (n = 1), sleeping disorders (n = 1), depressive disorders (n = 1), and dyslexia (n = 1).

This study was granted ethical approval by the University’s Research Ethics Committee. Parents and children were both provided with written and verbal information about the study. To participate in the study, written parental consent and child assent were both required. All participants received a child-appropriate voucher (30€) as compensation and children in the clinical group were offered treatment in the department’s outpatient clinic.

After children arrived at the laboratory, psychophysiological measuring devices were attached and a 5-min baseline was taken. Next, all children participated in a social stress task consisting of a social performance situation during which the children had to answer questions regarding a short story in front of two unknown female observers (Schmitz et al., 2011, 2012). The children completed the task twice. Prior to each task, a 5-min uninstructed anticipatory processing phase was implemented during which children were told to wait while the experimenter checked the psychophysiological measures before the experiment could proceed. No further instructions were given, to allow participants to ruminate spontaneously. During the following social stress task, children answered four standardized questions posed by two female adult observers about a previously heard short story. Children had 1 min to hear and answer each question and were instructed to do their best, because the observers and peers would later rate their performance based on video recordings (Schmitz et al., 2011; Spence et al., 1999). All observers received a briefing and were trained to give standardized neutral verbal and nonverbal feedback while maintaining a friendly attitude. After the first social stress task, participants were randomly assigned to one of two conditions: In the uninstructed post-event processing condition, children waited 5 min for the experiment to proceed without further instructions, allowing for spontaneous post-event processing. In the distraction condition, participants played a noncompetitive memory game (Majong 3 Free.Ink, 1C Wireless, download: 14.06.2017) on a tablet. Tablet-based virtual games have been proven successful in inducing cognitive distraction (e.g., Hagenaars et al., 2017). After a 10-min break, the described sequence was repeated for the second implementation of the social stress task (see Fig. 1).

Electrodermal activity was recorded with the VU-AMS Ambulatory Monitoring System (Geus et al., 1995; Willemsen et al., 1996). Two silver/silver chloride electrodes were attached to the medial phalanges of the index and middle finger of the participant’s nondominant hand. SCL was measured throughout the whole experiment and data were preprocessed using automatic artifact detection and low-pass filtering (for a detailed description, see http://​www.​vu-ams.​nl/​). Means of SCL relative to baseline were calculated for relevant periods (e.g., anticipatory processing phase).

Statistical analyses were conducted using the open statistics software R (R Core Team, 2018). Hypotheses were evaluated via mixed linear models (MLMs) to account for the hierarchical nature of the data. The mixed-models packages lme (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2017) were used and the level of significance was set at α = 0.05 for all statistical analyses. As proposed by Luke (2017), models were fitted using restricted maximum likelihood and p values were derived with Kenward–Roger approximation to account for small sample sizes. To analyze Hypothesis 1a and b, the MLM was fitted with one between-subjects factor, Group (SAD, HC), and two within-subject factors, Scale (positive anticipatory processing, negative anticipatory processing) and Task (T1, T2), and all possible interaction terms as fixed effects. To analyze Hypothesis 1c and d, the MLM was fitted with one between-subjects factor, Group (SAD, HC), and one within-subject factor, Task (T1, T2), and all possible interactions as fixed effects. To analyze Hypotheses 3 and 4, corresponding MLMs were built, adding Condition (distraction, uninstructed post-event processing) as a third within-subject factor. All models included a random intercept to control for subject effects. To analyze Hypothesis 2, stepwise multiple regression models were conducted. To analyze Hypothesis 2a, trait social anxiety, and negative and positive anticipatory processing were included as predictors of subjective anxiety during the social stress tasks. Because of possible developmental influences (Alfano et al., 2002) and group differences in depression scores, age as well as depression scores were included as control variables. To analyze Hypothesis 2b, the regression model additionally included state anxiety levels as predictor of subjective performance ratings as these differed significantly between groups during the social stress tasks (see Manipulation Check). Averaged values of both social stress tasks were used for the analyses. Adjusted R², the Akaike information criterion, and significant increases in predictive validity are reported for each regression model. An a priori power analysis (Faul et al., 2007) indicated a sample size of n = 27 participants per group based on a medium effect size (Schmitz et al., 2010) and a power of (1 – β) = 0.95.

Children with SAD and the HC group did not differ in their subjective anxiety scores assessed directly after the 5-min baseline, t(58) = -1.41, p = 0.164, d = 0.36, 95% confidence interval (CI) [-0.16, 0.89]. The manipulation check revealed a robust increase in subjective anxiety from baseline to the social stress tasks in both groups at both tasks, group with SAD: all ts ≥ 9.24, all ps < 0.001, all ds ≥ 1.94; HC group: all ts ≥ 3.98, all ps < 0.001, all ds ≥ 0.70. Children with SAD reported higher subjective anxiety than the HC group during T1 and T2, all ts ≥ 6.85, all ps < 0.001, all ds ≥ 1.77. There was a significant increase in SCL between baseline and the stress task in both groups in T1 and T2, all ts ≥ 5.25, ps < 0.001, ds ≥ 0.54, but no difference in SCL between groups during T1 and T2, all ts < 0.80, ps > 0.428.

In line with our first hypothesis and cognitive models of SAD (Clark & Wells, 1995; Hofmann, 2007), children with SAD in our study reported significantly more negative anticipatory processing than children in the HC group. Importantly, these findings extend previous studies that used nonclinical community samples (Vassilopoulos et al., 2014). However, partly contradicting our first hypothesis, we found no evidence of elevated autonomic arousal during anticipation of the social stress tasks, and subjective anxiety in the group with SAD was elevated only in anticipation of the second social stressor, probably owing to stress sensitization in our clinical group (Asbrand, Heinrichs, et al., 2019a). In contrast, Vassilopoulos et al. (2014) demonstrated an association between an instructed anticipatory processing condition and elevated anxiety levels in a community sample aged 10 to 11 years. While Vassilopoulos et al. (2014) used an instructed negative anticipatory processing condition, we did not specifically instruct children to engage in anticipatory processing, to capture this process with higher ecological validity. It is likely that instructed anticipatory processing would lead to a significant amplification of ruminative processes and related qualities such as subjective anxiety, particularly in children whose cognitive processes are still developing and are not as distinctive as in adolescents or adults (Alfano et al., 2002).

In accordance with our second hypothesis, negative anticipatory processing statistically predicted elevated subjective anxiety experienced during the social stress tasks and lower subjective self-evaluations of performance. Importantly, negative anticipatory processing explained variance beyond trait and state anxiety, suggesting a unique contribution of this process to disorder maintenance in childhood SAD (Halldorsson & Creswell, 2017). In their cognitive model, Clark and Wells (1995) argued that anticipatory processing may activate a complex dysfunctional processing pattern consisting of negative interpersonal self-beliefs and self-directed attention. This results in a negative cognitive-emotional processing of the feared social situation, for example, negatively biased self-evaluations and heightened subjective anxiety during social stress, as found in our study. The consideration of negative anticipatory processing as part of a complex maintenance process seems to be especially important in light of the small to medium amount of explained variance found in our study.

Positive anticipatory processing was comparable between the experimental groups and did not predict anxiety levels or subjective performance ratings in the social stress tasks. This indicates that SAD in children may be characterized more by elevated levels of negative anticipatory processing than by an absence of positive thoughts. This corresponds to the notion that a higher frequency of negative thoughts, as opposed to the presence of positive thoughts, is associated with psychopathology (Kendall & Chansky, 1991).

In accordance with our third hypothesis and previous research in socially anxious children (Asbrand, Schmitz, et al., 2019b; Schmitz et al., 2010), children with SAD reported more negative post-event processing than children in the HC group. Thus, children with SAD experienced more dysfunctional cognitions regarding their performance, the observers, and their appearance during the social stress task. Since previous research has shown that post-event processing is highly related to the maintenance of negative self-perceptions and biased self-evaluations in affected children (Miers et al., 2014; Schmitz et al., 2011), effective psychotherapeutic interventions to target and alter negative post-event processing are needed (Asbrand, Schmitz, et al., 2019b). In this context, our study is to our knowledge the first to investigate the effects of a cognitive distraction intervention on post-event processing in childhood SAD. Partly confirming our fourth hypothesis, distraction was associated with reduced anxiety levels in the group with SAD and reduced autonomic arousal in all participants (e.g., Wong & Moulds, 2009). However, in contradiction to our expectations and results from previous studies with highly socially anxious undergraduates (Blackie & Kocovski, 2016; Kocovski et al., 2011), the implemented distraction condition did not affect the frequency of negative or positive post-event processing. Several explanations are conceivable for these differences. First, most of the previous studies on the effects of distraction on post-event processing used an instructed post-event processing condition, thus specifically amplifying negative cognitions probably beyond their natural occurrence (e.g., Blackie & Kocovski, 2016). By contrast, our study used an uninstructed post-event processing condition, which may have resulted in a higher ecological validity but probably also smaller effects of distraction than when compared to instructed post-event processing. Rowa et al. (2014) reported results similar to ours, that is, a positive influence of a distraction condition compared to a post-event processing condition on anxiety levels, but no reduction of negative post-event processing in an adult SAD sample. They suggested that a distraction task may provide anxiolytic effects even alongside the presence of naturally occurring negative post-event processing. A similar process may have occurred in our sample, further indicating that negative post-event processing may be associated with stable negative self-schemata that may be reduced only with more extensive cognitive interventions. Previous developmental research has shown that children experience negative emotions and related cognitions more intensely (e.g., no mixed states of both positive and negative cognitions; Alfano et al., 2002). As a consequence, children have more difficulties profiting from cognitive interventions such as distraction due to an intense subjective experience of negative cognitions and emotions. Children may consequently need training and instructions beyond the relatively mild form of a single-session distraction condition used in our study (Volkaert et al., 2020).

There are several limitations to the current study. First, it is unclear to what extent our participants were distracted by the tablet game. We chose not to ask children about the extent of distraction because a major goal of our study was to measure anticipatory and post-event processing without strong reactivity and social desirability effects. Still, given the reduced levels of subjective anxiety and SCL in the distraction condition, we assume participants were overall able to successfully engage in distraction. Second, although we found positive short-term effects of distraction, further research is needed to study possible detrimental long-term effects. This is particularly important as distraction may be conceptualized as covert avoidance or safety behavior in SAD depending on its specific implementation (Clark & Wells, 1995). Third, although our sample size was sufficient according to an a priori power analysis, future studies may want to replicate our results in larger or more diverse samples by including children covering a larger age range and by comparing children with SAD to other clinical groups. Fourth, the study’s cross-sectional design prevents the determination of causal effects and our results may not be directly transferable to social interactions, as these are often more ambiguous and require a different skill set than social performance situations (Voncken & Bögels, 2008).

Future studies are needed to continue to examine the predictors and consequences of negative anticipatory and post-event processing in children with SAD. Both ruminative processes are likely part of a complex SAD-maintaining network (Hirsch et al., 2006; Wong, 2016); hence, experimental research focusing on several key aspects of cognitive models and their interconnections, for example, self-focused attention and its associations with rumination, is essential. Continuing research efforts are further needed to refine interventions to target anticipatory and post-event processing and to integrate them in a comprehensive intervention concept, for example, SAD-specific CBT (Leigh & Clark, 2018). Future studies should therefore compare the effectiveness of different therapeutic strategies (e.g., cognitive reappraisal; Helbig-Lang et al., 2015; Shikatani et al., 2014), methodological approaches, and intervention intensities in altering ruminative processes in childhood SAD.

In conclusion, the present findings suggest that SAD in children is associated with elevated levels of negative anticipatory and post-event processing, which are thought to be maintenance factors of the disorder in cognitive models of adult SAD (Clark & Wells, 1995). Negative anticipatory processing was further associated with elevated anxiety and reduced subjective performance ratings, suggesting possible detrimental effects of said process. The implemented distraction condition did not reduce negative post-event rumination in our sample but had a positive effect on emotional and physical poststress arousal.