Effects of interventions for social anxiety and shyness in school-aged children: A systematic review and meta-analysis

In school, shyness is associated with psychosocial difficulties and has negative impacts on children’s academic performance and wellbeing. Even though there are different strategies and interventions to help children deal with shyness, there is currently no comprehensive systematic review of available interventions. This systematic review and meta-analysis aim to identify interventions for shy children and to evaluate the effectiveness in reducing psychosocial difficulties and other impacts. The methodology and reporting were guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement and checklist. A total of 4,864 studies were identified and 25 of these met the inclusion criteria. These studies employed interventions that were directed at school-aged children between six and twelve years of age and described both pre- and post-intervention measurement in target populations of at least five children. Most studies included an intervention undertaken in a school setting. The meta-analysis revealed interventions showing a large effect in reducing negative consequences of shyness, which is consistent with extant literature regarding shyness in school, suggesting school-age as an ideal developmental stage to target shyness. None of the interventions were delivered in a classroom setting, limiting the ability to make comparisons between in-class interventions and those delivered outside the classroom, but highlighting the effectiveness of interventions outside the classroom. The interventions were often conducted in group sessions, based at the school, and involved activities such as play, modelling and reinforcement and clinical methods such as social skills training, psychoeducation, and exposure. Traditionally, such methods have been confined to a clinic setting. The results of the current study show that, when such methods are used in a school-based setting and involve peers, the results can be effective in reducing negative effects of shyness. This is consistent with recommendations that interventions be age-appropriate, consider social development and utilise wide, school-based programs that address all students.

Introduction Shyness is commonly experienced by school-aged children [1]. Despite being a frequently used term, there is a diversity of constructs that underpin 'shyness', including behavioural inhibition, social reticence, social withdrawal, anxious solitude and social anxiety [2]. There have been several approaches to defining shyness in the past. Some conceptualisations theorise shyness as either behavioural inhibition to the unfamiliar (i.e., wariness in unfamiliar situations) or social withdrawal [i.e., elevated rates of solitary behaviour or symptoms of social anxiety disorder; [3][4][5][6][7]. In contrast, substantial literature has investigated shyness as encompassing individual differences in wariness or anxiety in novel situations, embarrassment or self-conscious in anticipation of social evaluation and reticence in social situations [7]. Shyness has also been considered from a developmental perspective, proposing an interactional child-by-environment model. By this model, behavioural inhibition and social withdrawal are considered risk factors for further social anxiety. Interactions between the child and the environment, and the child and their parents and peers, can either promote or diminish the risk of later anxiety [4,8,9].

Taxonomy of shyness
In order to organise and operationalise the various concepts of shyness in use, Rubin, Coplan [7] proposed a taxonomy of shyness. This taxonomy places behavioural solitude (i.e., lack of interaction in presence of peers) as the over-arching, observable behaviour of shyness. The source of this solitude is either internal, termed social withdrawal (i.e., removing oneself from social interaction) or external, termed active isolation (i.e., being excluded by others). If the source is internal (i.e., social withdrawal), the motivation for withdrawal is either by preference, termed social disinterest, or a result of fear or wariness. The source of fear is then split into four categories: 1) behaviour inhibition (i.e., fear of novelty); 2) anxious solitude (i.e., wariness in familiar social situations); 3) shyness (i.e., wariness of social novelty and/or perceive evaluation); and 4) social reticence (i.e., observed display of onlooker behaviours). In this taxonomy, these fears and behaviours can become clinically significant over time and manifest as a social anxiety disorder. This taxonomy provides a clear conceptualisation of shyness and social anxiety, and outlines observable behaviours, sources, motivations and specific fears.

Shy children in school
In addition to the potential manifestation of social anxiety disorder theorised by Rubin, Coplan [7], children with shyness may also experience a range of other difficulties that, although not clinically diagnosable, can vastly impact their wellbeing, social networks and academic performance [10]. Many of these difficulties are experienced at school, where peer interactions are an integral component of the environment. Shy children are often quiet across a range of situations in school, both in the classroom and in social situations [11]. Talking, in or outside of class, can make a child the centre of attention and open to social evaluation, which sits at the centre of the taxonomy of shyness. Shy children have fewer in-class interactions and respond less often to direct or class-wide questions than their non-shy peers [12]. Research has shown that shy children often have lower academic attainment, poorer performance on tests of language development, and are more likely to have difficulty adjusting at school [10].
Shyness is also associated with psychosocial challenges in school. Shy children often have a limited number of friends and are at risk of peer victimisation and exclusion [7,13]. They may also use social withdrawal as a way to avoid or cope with peer victimisation [14]. Shyness is positively associated with somatic complaints, school-related stress, anxiety and depressive symptoms [15,16]. Shyness can increase over time, predicting difficulties later in adolescence [17]. Shy children often have poor social skills and high levels of anxiety and depression symptoms in early adolescence [17]. Longitudinal studies show that shyness and social withdrawal are significant risk factors for social anxiety disorder [8,18]. These results are aligned with the Rubin, Coplan [7] taxonomy of shyness and social anxiety, demonstrating the theorised pathway to social anxiety disorder.

School-based interventions for shy children
Given the short-and long-term psychosocial and academic outcomes for shy children, there have been multiple attempts at buffering the impacts of shyness. In the classroom, teachers can use concepts, such as shyness, as a tool to tailor how they work with an individual child [19]. Teachers at a Norwegian elementary school broadly categorised shy children in their classroom as either, 1) withdrawn, 2) anxious, and/or 3) having poor self-esteem. These categories then informed the support given to the individual child, including cognitive support and feedback and encouraging active learning [19]. Informal, teacher-facilitated support or intervention is a common response to shyness within the classroom, as teachers recognise shy children and the potential problems they encounter [20][21][22]. Teachers report employing social learning strategies, such as verbal encouragement, praise and modelling behaviour, as well as peerfocused strategies to promote inclusion, such as encouraging joint activities [20]. However, the effectiveness of these individual attempts is limited to within the classroom and may not impact poor psychosocial outcomes for shy children in broader contexts.
Beyond classroom support, there are many different structured interventions targeting shyness in school-aged children. Clinical interventions are typically conducted in non-naturalistic settings with homework-style practice in naturalistic settings, and comprise of social skills training, psychoeducation, cognitive restructuring and exposure tasks [8]. Criticisms of this approach are that such interventions do not consider nor change the environment itself and focus on treating social anxiety disorders, ignoring shyness more broadly [8]. Clinical interventions need to be age-appropriate and consider cognitive and social development, social context and parent involvement [23]. As shy children are often excluded or victimised by their peers, interventions need to consider the environment and peer interaction. Developmental interventions include peers in the intervention itself, aiming to increase the use of successful social skills in naturalistic settings [8]. However, this approach requires school resources and willingness of peers to be involved. Crozier [1] suggests that a focus on individual screening and pathologising shyness may not lead to effective intervention, as not all shy children develop anxiety disorders. Wider, school-based programs that address all student's social confidence, instead of targeted interventions, may be more suitable intervention for shyness [1]. Given the wide range of intervention approaches and intervention programs themselves, there is no clear best-practice for interventions for shy children. This is further complicated by inconsistent use of terminology related to shyness [1].
To reduce academic and concomitant psychosocial difficulties in school for shy children, there is a need for effective, feasible interventions. To date, there is no comprehensive systematic review of the available interventions for shy children. This systematic review and metaanalysis aim to provide an overview of the available interventions for shy children aged six to twelve years, describe the characteristics of the interventions, summarise intervention strategies being used, and determine their overall effectiveness, as well as effectiveness of interventions in relation to the following domains: 1) setting where the interventions is delivered; 2) mode of delivery; 3) intervention focus; and 4) rater of outcome measures.

Method
The methodology and reporting on this systematic review were guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and checklist. The PRISMA statement and checklist supports researchers in the critical and transparent reporting of systematic reviews in areas of health care [24,25].
[The PRISMA checklist is provided as Supporting Information].

Eligibility criteria
To be eligible for inclusion in this systematic review, studies were required to describe an intervention in school-aged children (between six and twelve years old) for social anxiety and shyness. Only studies describing both pre-and post-intervention measurement in target populations of at least five children were included. Only original articles published in English were considered for eligibility. Conference abstracts, case reports, reviews, student dissertations and editorials were excluded.

Data sources and search strategies
Literature searches were conducted in five electronic databases: CINAHL, Embase, Eric, Psy-cINFO and PubMed. All publication dates up to 23 rd December 2020 were included. The search strategies per database are listed in Table 1.

Methodological quality and level of evidence
The Qualsyst critical appraisal tool by Kmet [26] and the National Health and Medical Research Council (NHMRC) Evidence Hierarchy Levels of Evidence [27] were used to assess the methodological quality of the included studies: I (systematic review of level II studies); II (randomised controlled trial); III-1 (pseudo-randomised controlled trial); III-2 (comparative study with concurrent controls); III-3 (comparative study without concurrent controls); IV (case series with either post-test or pre-post outcomes). The Qualsyst tool provides a systematic, reproducible and quantitative means of appraising the methodological quality of research across a broad range of study designs. The Qualsyst consists of 14 items. All items have a Database and search terms (subject headings and free text words) CINAHL: ((MH "Shyness") OR (MH "Social Isolation") OR (MH "Social Isolation (Saba CCC)") OR (MH "Impaired Social Interaction (NANDA)") OR (MH "Social Isolation (NANDA)")) AND ((MH "Clinical Effectiveness") OR (MH "Treatment Outcomes") OR (MH "Effect Size") OR (MH "Outcome Assessment") OR (MH "Outcomes (Health Care)+") OR (MH "Intervention Trials") OR (MH "Program Evaluation") OR (MH "Evaluation+") OR (MH "Course Evaluation") OR (MH "Evaluation Research+")) three-point ordinal scoring (yes = 2, partial = 1, no = 0). A total score can be converted into a percentage score. A score above 80% is considered strong quality, a score of 60 to 79% considered good, a score of 50 to 59% considered adequate, and a score below 50% considered poor quality. Studies with poor study quality were excluded from further analysis in this review.

Data extraction
A data extraction form was created to extract data from the included studies under the following categories: study design (according to NHMRC level), methodological quality (Qualsyst), participants (numbers, groups), age (range, mean, standard deviation), gender, intervention, inclusion criteria of the individual study (if stated), outcome measures and treatment outcomes. To ensure the meta-analysis focused on factors that impact on shyness, authors identified and extracted only data collected using the main outcome measure related to shyness (see Table 2). Due to the lack of dedicated shyness outcome measures in literature, the most suitable outcome measure related to shyness was chosen. Data including means, standard deviations, and sample sizes were extracted from the included studies to enable the calculation of the overall effect of shyness interventions (within-group pre-post intervention comparisons), and comparisons between shy children and control groups (between-group experimental vs. control intervention group comparisons).

Data items, risk of bias and synthesis of results
Risk of bias in the included studies was assessed at an individual study level using the Kmet appraisal checklist [26]. Risk of bias was minimised in this process by having a full overlap                  between independent abstract and article reviewers, and by two independent assessors independently scoring 100% of the methodological quality of included studies. Final study selection and quality assessment were the result of consensus-based ratings. Discrepancies were resolved by involving a third reviewer. No author of this review was affiliated with any of the included studies. Extracted data were synthesised in relation to the methodological characteristics of each included study and the findings of individual studies with regards to the treatment outcomes of shyness interventions.

Meta-analysis
Using the extracted data from the main outcome measure related to shyness, estimates were calculated of pooled effect sizes weighted by sample size using random-effects models for summary statistics. To determine potentially confounding variables, effect sizes of shyness interventions were grouped by setting (school, clinic and/or home), focus (child and/or parents), mode of delivery (individual and/or group sessions), and rater of outcome measures (child, parents, clinician and/or teacher). The Hedges-g formula for standardized mean difference (SMD) with a 95% confidence interval (95% CI) was used to report effect sizes. A test for overall effect for each intervention setting, mode, focus and outcome rater produced a weighted effect size (z). Tests for heterogeneity were conducted to identify inconsistency in treatment effects, included I 2 and chi-square (Q). All statistical analyses were performed using software package Comprehensive Meta-Analysis Version 3.3.070 (Biostat; Englewood, NJ, USA). Within-groups effects were examined by analysing the pre-post data for studies both with and without control groups. The benefit of within-groups analyses is that it allows the examination of the effect of an intervention in and of itself, without controls. Between-groups analyses (comparing results of control group to that of intervention group) were also conducted. This allows comparison of different forms of interventions against each other.

Systematic review
Study selection. A total of 4,864 articles were identified (CINAHL: n = 605, Embase: n = 1158, ERIC: n = 1849, PsycINFO: n = 968 and PubMed: n = 929). After the removal of duplicate articles, 5299 abstracts were screened. A total of 149 studies were assessed at a full text level for eligibility. Of these, 129 were excluded and 20 were included (see Fig 1). No studies were excluded due to poor quality. An additional five studies were included through searching the reference lists of the 20 studies that met the inclusion criteria. This resulted in a total of 25 included studies.
Participants of studies included in the systematic review. The total number of participants across the 25 included studies was 1,895, with the average participants across studies 75.8. Griffin, Caldarella [28] had the largest sample of 388 participants and Cook, Xie [29] the smallest sample of 5 participants. The average age of total participants across the studies was 9.1 years (SD = 5.4), with the average age of the total sample not reported in nine studies. Of the 25 studies, only five had more male than female participants, with four studies not reporting the gender of the total or sub-samples. While a range of diagnoses were reported across some studies, 13 studies reported the sample to be typically-developing and five studies did not report diagnosis. Studies were conducted across nine countries, with the highest number conducted in the USA (n = 10), followed by Australia (n = 4). Additional details on participant characteristics are reported in Table 2.
Study design, methodological quality and risk of bias of studies included in the systematic review. Most studies were randomised or pseudo-randomised control trials, with only three employing a multiple baseline design (see Table 2). The methodological quality for each study according to Kmet criteria is reported in Table 2. The average methodological quality rating across all studies was 83.4% (SD: ±8.7, range: 64-96%), indicating "strong" methodological quality. Of the studies, 17 were rated as "strong", with all others rated to have "good" methodological quality. No study was rated to have adequate or poor methodological quality.
Shyness outcome measures. While studies reported several outcome measures, only those relevant to shyness and/or social anxiety were the focus of this review. Across categories of self-report, parent-report, teacher-report, clinician-rating and observation measures of shyness, self-reported (n = 13) and parent-reported (n = 13) shyness outcome measures were most frequently used and clinician-rating was used least across studies (n = 7; see Table 2). Using the categories of outcome measures above, nine studies used two different types of outcome measures, seven studies used only one type of outcome measure, and nine studies used three or more types out outcome measures.
Interventions. The majority of studies included an intervention that was delivered weekly (n = 15), in a child group format (n = 14), in the school setting (n = 10). Only four studies reported session durations of 40 minutes, with 14 reporting sessions for 60 minutes or longer. Intervention delivery was reported to be at least 7 weeks in 17 studies (see Table 3).
Descriptions of active intervention components reported in the included studies were reviewed and categorised. In terms of active intervention components, the studies used psychoeducation (n = 11), in-vivo exposure (n = 11), SST (n = 9), therapist modelling (n = 9), cognitive restructuring (n = 8), behaviour modification (n = 6), peer-mediation (n = 6) and videomodelling (n = 1). Across the studies, 12 used only one or two intervention components, while only five studies used a combination of 4 or more intervention components (see Table 3).

Sessions included psychoeducation and teaching and coaching of each Child-Direction Interaction, Bravery-Directed Interaction and Parent-Direction
Interaction. Parents learn to adopt a "step behind" approach, provide praise for their child's behaviours, apply skills in anxiety-provoking social situations for their child, and distinguish between anxiety and oppositional behaviours. Coaching sessions involved dyadic parent-child coaching. Parents received instructions for out of session exposures. The final session was a "graduation party" were parents were coached to use their skills Mentors were any educational professional at the student's school that participated in a 60 min training session. Prior to intervention, mentors held 2 x 40 min sessions. The first session was to build rapport and present life bus metaphor, used to normalise emotion and provide language to talk about emotions. The second session comprised a brief review of content and "courage tools". The intervention consisted 0f a) assignment of a mentor with unconditional positive regard; b) morning meetings for positive interaction, words of encouragement and pre-correction of problems; c) daily mentoring of performance and d) afternoon meetings for positive interaction and performancebased feedback. During meetings, students would provide daily ratings of distress.

Cognitive-Behavioural Group Therapy for Adolescents (CBGT-A)
Involves two phases with eight sessions each: 1) Educative and skills building and 2) Exposure. In first phase, therapist provides information about the program and delivers presentation on social phobia. The skills building unit involves teaching social skills, problem solving and cognitive restructuring. The second phase involves behaviour rehearsals and in vivo exposures within the session and as homework.

Buddy Bench Social involvement
Observers, teacher and principal were trained in the intervention by principal investigator. Two specially decorated benches are placed in two playgrounds in the school. Teachers instructed students on how to use the bench, posted rules in every classroom and issues a daily school-wide announcement reminding students the use the bench. Students were instructed, if they felt alone, to sit at the bench. If someone invites them to play, say "yes" or "no thank-you". If they saw a student at the bench, students were instructed to invite them to play.

Cognitive bias modification training Interpretation bias
Each session consisted of 10 ambiguous scenarios; each scenario consisted of 3 short sentences. The last word of the last sentence was missing one letter.
In the positive training group, all final words made the story end positively. In the neutral group, all words made the story end in an irrelevant way.
Children were asked to read each scenario and image themselves as the central character. Children pressed a button and the missing last word appeared with one letter missing. Children had to fill in the missing letter as quickly as possible.
Children were asked a 'yes' or 'no' question that measured comprehension of the story. All children performed first training session in clinical with parent and trained research assistant. The remainder were completed at home.  Initial individual sessions addressed key relationships with peers, education about four interpersonal problem areas: 1) role disputes; 2) role transitions; 3) interpersonal deficits; 4) role insecurity. The first three group sessions included psychoeducation, didactics on communication skills and role-play to practice communication skills. The following five group sessions includes didactics on communication and problem-solving skills, communication analysis to identify problematic communication, role-play and practice interpersonal skills. Interpersonal events in participants lives determined the group content of that week. The last two group sessions focused on reviewing skills and generalising to other scenarios. An individual mid-intervention session allowed for review of participants progress.

Second Life
Self-expression Lee [42] Second Life is a virtual world in which participants can interaction. To maintain anonymity, participants create avatars and are only referred to by screenname. Two Second Life classes were held that the same time so no participants could be identified by other group members. Participants were assigned speaking topics through a chat function, which they presented to peers in the virtual Second Life class through microphone. Participants were allowed 15 lines of notes in preparation. Members sat in chairs while the speaker stook at a podium. After all participants spoke, they participated in a 10min question and answer session to discuss unclear points, good points and points to around. A moderator helped participants to manage environment and audio equipment Group sessions involved 1) a 5 min free play session; 2) self-presentation speeches; 3) circle time and 4) leader-facilitated free play. Self-presentation speeches gave children opportunity to speak freely about a familiar topic (i.e. a new toy). Circle time involved didactic content, focusing on a set of social skills each week. These included initiating/ maintaining interactions, understanding/expressing feelings and regulation of negative affect. Puppets and songs were used to convey content. During facilitated play, group leader guided participation, using prompting, modelling, encouragement and reinforcement of social skills taught.  Treatment consisted of group discussions between participants and psychologist, and individual sessions to establish a dialogue and increase nonverbal interaction skills. In individual sessions, participants were counselled on over-sensitivity and unrealistic fears. Counselling involved discovering early trauma that may contribute to shyness. Parents were encouraged to give children confidence and role play. Individual sessions involved conditioning eye contact through reinforcement. Group sessions involved peermodelling assertive responses, and desensitisation procedures. Participants were encouraged to increase exposure by communication with dolls and increasing to unfamiliar children.  significant improvement was found for play skills (n = 2) and aspects of social functioning (n = 8); social competence, social skills, social interaction, social adjustment, interpersonal skills, peer victimisation, perceived social support from peers, and pro-social behaviour. Further, four studies reported treatment gains to be maintained at follow-up periods between 6 months and 5 years (see Table 3).

Meta-analysis
Effects of interventions. Twenty of the 25 studies were included in the meta-analysis. Five could not be included in the analysis as the data required were not reported [28,31,32,34,36]. Authors were contacted to collect the required data, but no responses were received.
Overall treatment effects were calculated for shyness interventions on within-group prepost outcome measures. Sub-group analysis was conducted to compare the effect as a function of intervention characteristics: 1) setting (i.e., clinic, home, school, online or a combination); 2) mode of delivery (i.e., group interventions, individual interventions or both); 3) intervention focus (i.e., parent focused, child focused or both); and 4) rater of outcome measure (i.e., clinician-rated, parent-rated, self-report, teacher-rated or a combination).
Between groups analysis was also conducted to compare experimental groups post-interventions scores with those of the control groups. A further 3 studies were excluded from this analysis as they did not include control groups. The following four control condition types were included: 1) waitlist control groups where participants served as an untreated comparison group who eventually went on to receive the intervention; 2) control groups that received no intervention; 3) alternative treatment controls where participants received an intervention that did not have the approach of the intervention being tested; and 4) medication control groups, where participants received medication instead of the behavioural intervention. Overall effect of shyness interventions. Effect sizes ranged from 0.04 to 3.18 in the within-group pre-post intervention without groups analysis, as shown in Fig 2. Of the 20 studies included 75% (n = 15) produced a large effect size and 15% (n = 3) produced a moderate effect. An effect size of < 0.2 was measured in 10% (n = 2) of the studies. The overall intervention effect was large and statistically significant (z(20) = 7.03, p < .001, Hedge's g = 1.21, 95% CI = 0.87-1.54). The between-study heterogeneity was significant Q(19) = 137.16, p < 0.001) and 86.2% of true variability (I 2 ) could be explained by individual study characteristics.
Effect size as a function of intervention characteristics (within-group). Table 4 shows the effect sizes of shyness interventions grouped by delivery setting, focus of the intervention, mode of delivery, and rater of outcome measures.
Setting. Interventions that were delivered within a clinic demonstrated the largest effect size of those calculated as a function of setting (1.38), indicating a large, significant effect (z(9) = 10.50, p < .001, Hedge's g = 1.38, 95% CI = 1.12-1.63). Interventions delivered online (z(1) = 4.36, p < .001, Hedge's g = 1.21, 95% CI = 0.67-1.76) and those delivered in schools (z(9) = 3.91, p < .001, Hedge's g = 1.03, 95% CI = 0.51-1.55) both produced a significant, large effect size. However, caution is needed when interpretation this results as only one study involved an online intervention. Interventions set in a combination of the home and a clinic produced the lowest effect size of all settings, showing a moderate, significant effect size (z(1) = 2.74, Hedge's g = 0.62, 95% CI = 1.07-2.74). However this should be interpreted with caution as only one study used an intervention set in both a clinic and the home [40].
Focus. Interventions focused on the children alone produced the largest effect size of 1.33 of those calculated as a function of recipient of the intervention (z(13) = 5.93, p < .001, 95% CI = 0.89-1.78). Interventions that focused on both parents and children produced the lowest effect size, as demonstrated by a moderate but non-significant effect (z(3) = 1.67, Hedge's g = 0.73, p = 0.1, 95% CI = -0.13-1.59).
Mode of delivery. Interventions that includes individual sessions, group sessions or both were all significant and large in effect. Those that utilised a combination of both individual and group sessions produced the largest effect (z(6) = 5.29, Hedge's g = 1.6, p < .001, 95% CI = 0.88-1.5).
Rater of outcome measures. Interventions that used outcome measured rated by the children themselves, teachers, clinicians, parents or a combination of clinician and parents all produced large and significant effect sizes. Those that used measures completed by parents alone produced the largest effect size, however, this included only one study (z(1) = 5.2, Hedge's g = 2.5, p < .001, 95% CI = 1. 55-3.44). Those that used measures completed by clinicians and parents produced the lowest effect size, however, the effect size was still large and significant (z(2) = 2.44, Hedge's g = 0.97, p < .05, 95% CI = 0.69-2.15).
Effect size as a function of intervention characteristics (between-group). Table 5 shows the effect sizes of shyness interventions grouped by delivery setting, focus of the intervention, mode of delivery, and rater of outcome measures when compared to control groups.
Setting. When compared to a control group, interventions delivered in a clinic produced the largest effect size of those calculated as a function of setting z(9) = 3.69, Hedge's g = 1.05, p < .001, 95% CI = 0.5-1.61). Interventions delivered in a combination of the clinic and home, and those delivered online, produced small and non-significant effects. However, these only comprised of one study each. Interventions delivered in school produced a moderate, significant effect size (z(7) = 2.93, Hedge's g = .76, p < .01, 95% CI = 0.25-1.27).    Mode of delivery. Interventions that used group sessions (z(13) = 4.31, Hedge's g = .92, p < .001, 95% CI = 0.49-1.33) or a combination of individual and group sessions produced large effect sizes when compared to control groups (z(3) = 1.98, Hedge's g = .88, p < .05, 95% CI = 0.1-1.75). Interventions using only individual sessions produced a small and non-significant effect when compared to control groups.
Rater of outcome measure. Interventions that used measures rated by parents demonstrated a large but non-significant effect size when compared to a control group, however, this included only 2 studies. Interventions that used measures rated by clinicians showed a large, significant effect size (z(9) = 3.76, Hedge's g = .95, p < .001, 95% CI = . 45-1.44).
Publication bias. The Begg and Mazumdar rank correlation procedure produced a tau of 0.588 (two-tailed), indicating there is no evidence of publication bias. This finding was supported by Duval and Tweedie's trim-and-fill procedure using the fixed-effect model; the point estimate for the combined studies is 0.433 (95% CI: 0.319, 0.546). Using trim and fill, these values are unchanged. Under the random-effects model the point estimate for the combined studies is 0.819 (95% CI: 0.499, 1.138). Using trim and fill, these values are unchanged. Both of these procedures indicate the absence of publication bias (see Fig 4 for funnel plot).

Discussion
This study systematically identified available interventions for shy children and evaluated the effectiveness of these interventions in reducing psychosocial difficulties in school. Using systematic review and meta-analysis procedures, all study designs were included when identifying the available interventions. Both RCTs and quasi-experimental studies were included in the meta-analysis to broaden the scope and examine the effectiveness of all possible intervention studies for shy children.
The systematic review revealed that 25 studies met the inclusion criteria, comprised of 24 different interventions aiming to address shyness. All the included studies and the employed interventions were directed at school children, aged between six and twelve years. School is identified as a primary setting where shyness and its associated difficulties manifest or be noticed for the first time, as it is often a child's first social environment away from parents. School often presents many different social situations for a child to navigate, such as classroom interactions, playgrounds and social cliques. Therefore, schools are suitable contexts for delivering 'early' intervention.
The results of the systematic review and meta-analysis support the association between intervention and reduction in shyness for this age group. As such, school-age may be an ideal developmental and social stage in life to target shyness to lessen the impact of shyness in school-age and later in life. However, the systematic review excluded any children outside of the age range, thus the systematic review cannot confirm that interventions at younger or older age groups, such as pre-school children, adolescence or young adults, are more or less effective. However, it is possible that shyness could be identified and addressed at earlier developmental stages or need intervention later into adolescence. A longitudinal study of fifth grade boys showed that, when children had better peer relationships, their shyness was more likely to decrease or remain stable over four years [53]. Those who were described as having poor peer relationships often increased in shyness.
Shy children may experience a wide range of difficulties in school that may impact their academic performance, social interactions and overall wellbeing [10]. A population-based, longitudinal study of children showed from ages 1.5 to 12.5 years, parent-reported shyness increased steadily over time [17]. Shyness that remained stable and increasing shyness also predicted poor social skills and higher levels of anxiety at the end of the follow-up [17]. The results of the current study suggest that by promoting protective factors and introducing intervention, shyness can change as a child matures into adolescence and young adulthood, but that without such protective factors, shyness can remain a hindrance. However, how adolescents and young adults experience shyness and the required composition of active intervention ingredients to affect change in shyness in this age group are not well understood. Further research is needed into the effectiveness of interventions for shyness for younger children and for adolescents, as well as long-term impacts of interventions into adulthood.
Most studies included interventions that were delivered in a school setting. The withingroup meta-analysis revealed interventions in this setting showed a large effect in reducing shyness, which is consistent with extant literature regarding shyness in school. Historically, the classroom has been the setting for implementing shyness interventions, as teachers often notice and informally attempt to address reticent behaviour [20]. Such informal interventions often included tailoring material to accommodate a noticeably shy child, individualised support within the classroom, and using social learning strategies such as modelling and positive reinforcement [19,20,22]. Of the included interventions that were set in a school, none were set in the classroom, suggesting that extending interventions beyond the classroom can have a large impact on shyness. These school-based interventions often involved clinical methods such as social skills training, psychoeducation, and exposure. The interventions were often conducted in group sessions, based at the school, and involved activities such as play, modelling and reinforcement by the facilitator [28,29,32,34,36,43,46,47,51,52]. These methods have previously been criticised for not considering the social environment and peer interaction within which shyness manifest [8]. However, traditionally such methods have been confined to clinic settings and clinic-based interventions demonstrated the highest effect-sizes. However, the advantage of delivering interventions in a school setting, rather than a clinic setting, is the added value of ecological validity [54]. As such, the burden is less on school-based intervention, compared with clinic-based interventions, for treatment effects to generalise to a natural social context within which treatment strategies are applied [54,55]. The results of the current study show that, when such methods are used in a school-based setting and involve peers, the results can be effective in reducing shyness. This is consistent with recommendations by Mychailszyn, Cohen [23] and Crozier [1] that interventions should be age-appropriate, consider social development, and utilise school-wide programs that address all students, rather than targeted, clinic based interventions.
Findings from the within-group meta-analysis indicated that interventions that focused solely on the child produced the largest effect size when compared to other interventions that focused on parents alone or a combination of child and parent. Interventions focussing on both parents and children, often in the form of parent training and education, produced the lowest, non-significant effect size. This is contrary to previous recommendations that advocated for implementing interventions for shyness that involve both parents/carers as well as children themselves [23]. Wider literature regarding interventions for children with developmental disorders, such as autism spectrum disorder, have found that involved parent training and coaching alongside interventions for children is most effective in improving language and communication outcomes, compared to sole child or parent training [56]. This finding suggests that shyness may be unique to other conditions or disorders affecting social communication. This finding may be explained by how shyness develops. Early interactions between the child, their environment, parents and peers are believed to either promote or diminish the risk of later anxiety and shyness for the child [4,8,9]. It may be possible that a parent's role in early development and supporting interactions between their child, environment, other adults, and peers may be more important than at an intervention level once shyness has developed. Therefore, parental involvement in shyness interventions may be more important when delivering interventions to children before they start school. The taxonomy of shyness proposes that shyness can stem from peer exclusion or different sources of fear within the child, including fear of novel social situations, fear in familiar situations and fear of perceived evaluation [7]. Given the results of the current meta-analysis, it may be possible that such internal (fear) and external (exclusion) sources of shyness are best addressed with the children, to resolve internal fears and promote inclusion with peers.
Overall, the between-groups meta-analysis revealed that all interventions of shyness demonstrated a large, significant effect size when compared to control groups of either no intervention, treatment as usual or medication interventions. When examining this effect as a function of setting, focus, mode, and rater of outcome, the results closely mirror that of the withingroups analysis. That is, clinic-based, child-focussed, and a combination of individual and group delivered interventions produced the largest effect sizes. Within-groups results should be interpreted with caution due to the lack of control group. However, the only difference between within-groups analyses and between-groups containing a control group was that, for between-group meta-analysis, group delivered interventions were slightly higher than a combination of individual and group delivered interventions.
The findings from this study builds on the evidence for effectiveness of interventions for shyness of school-aged children, by improving their social interactions with peer and reducing reticent behaviour. However, this review found no evidence of long-term benefits of reducing later development of social anxiety disorders or long-term impacts on educational and wellbeing outcomes. Further research with longitudinal follow-up is necessary to establish the longterm effectiveness of shyness interventions.

Limitations
There was variation in how shyness was defined, conceptualised, and operationalised across the included studies. Some studies required a diagnosis of social phobia for inclusion into the intervention, whereas others relied on parent or teacher report of shy behaviours. This is reflective of definitional variation in the literature regarding shyness and limits the generalisability of the results found between studies. The children included in individual studies had a range of diagnoses that may have impacted the effectiveness of the included interventions. Further research is needed to examine effects of interventions for children with and without clinical levels of social anxiety and wider diagnoses. The current review focused on school-age children aged between six and twelve years. As such, no conclusions can be drawn about the effectiveness of interventions for younger children and adolescents or the long-term impacts of interventions. When examining settings of interventions, two categories only included one study each. Therefore, the results of these categories need to be interpreted with caution. This review was unable to ascertain which individual intervention components contributed most to the effectiveness of interventions. Further research is needed to isolate the active ingredients of the interventions and determine which contributes most to the effectiveness of interventions.

Conclusion
Shyness impacts many school-aged children and can have lasting effects on peer interactions, wellbeing, psychosocial and academic achievement. The current study provides a comprehensive review of interventions for shyness, identifies the most commonly used strategies and intervention effectiveness. Of the 25 studies included in the review, most interventions were delivered weekly, to a group of children in a school-based setting. They employed strategies such as psychoeducation, exposure, modelling, cognitive restricting, and peer mediation to address shyness. Across all included studies, reductions were reported in anxiety, social phobia, and internalising behaviours. The meta-analysis revealed that clinic-based, child-focussed, and a combination of individual and group interventions wielded the most benefits in reducing shyness. However, school-based interventions also produced large effect-sizes and have ecological validity as an advantage. This systematic review and meta-analysis provide an evidence-based for the most effective interventions for shy children that must utilise clinical strategies, such as modelling and exposure, that should ideally be delivered in a school-based setting that facilitates interactions with peers.