Psychosocial Interventions for School Refusal with Primary and Secondary School Students: A Systematic Review

1.1 THE PROBLEM, CONDITION, OR ISSUE

School refusal is a psychosocial problem characterized by a student's difficulty attending school and, in many cases, substantial absence from school (Heyne & Sauter, 2013). The literature reveals ongoing discussion about the conceptualization and definition of school refusal. Some authors avoid differentiating school refusal from truancy (e.g., Kearney, 2003) while others argue for the value in doing so (e.g., Elliott, 1999; Heyne, Sauter, & Maynard, in press).

Drawing on the work of Berg and colleagues (Berg, 1997, 2002; Berg, Nichols, & Pritchard, 1969; Bools, Foster, Brown, & Berg, 1990), researchers commonly differentiate school refusal from truancy according to the following features. First, students who exhibit school refusal typically display severe emotional distress associated with school attendance, which is not common among truanting youth (e.g., Bools et al., 1990; Egger et al., 2003). Second, when not at school, school refusers are usually at home with their parents' knowledge. Truanting youth, on the other hand, stay away from school without their parents knowing about the absence (Kearney, 2002) or their whereabouts (Berg, Casswell, Goodwin, Hullin, McGuire, & Tagg, 1985). Third, school refusers seldom exhibit characteristics of conduct disorder. Truancy, on the other hand, is commonly associated with severe antisocial behavior (e.g., Egger et al., 2003; Vaughn, Maynard, Salas-Wright, Perron, & Abdon, 2013). This review will adopt Berg and colleagues' conceptualization of school refusal as a type of school attendance problem, which is different from truancy.

The emotional distress associated with school refusal is often in the form of fear or anxiety. In representative samples of clinic-referred school refusers (i.e., participants are included irrespective of the presence or absence of an anxiety disorder), it is commonly found that around 50% of participants have an anxiety disorder (e.g., Baker & Wills, 1978; Bools et al., 1990; McShane et al., 2001; Prabhuswamy et al., 2007; Walter et al., 2010). A broad range of anxiety disorders has been observed, including separation anxiety disorder, specific phobia, social phobia, generalized anxiety disorder, and panic disorder with agoraphobia. Even when full diagnostic criteria for a particular anxiety disorder are not met, prominent symptoms may lead to a diagnosis of Anxiety Disorder Not Otherwise Specified (e.g., Heyne, King, Tonge et al., 2002; McShane et al., 2001). Other school refusers experience fear or anxiety related to school attendance at a level below the diagnostic threshold. In Egger and colleagues' (2003) community study, many more school refusers reported anxiety-related symptoms than were diagnosed with an anxiety disorder. For example, 36% reported a fear specific to school, and 27% reported headaches or stomachaches associated with separation or school attendance. Indeed, the physiological symptoms commonly reported among school refusers are headaches and stomach aches (Berg, 1980; Torma & Halsti, 1975). As many as 80% of referred school refusing children and adolescents in Honjo and colleagues' (2001) study expressed somatic complaints.

The relationship between school refusal and depression is not clear. According to Elliott (1999), the depression seen in school refusers may be specific to school refusal or a reflection of the relationship between anxiety and depression. King, Ollendick, and Tonge (1995) noted large variability in the incidence and severity of depression among school refusers. We identified four representative samples of clinic-referred school refusers in which depressive disorders were found to be less common than anxiety disorders (Baker & Wills, 1978; Bools et al., 1990; Walter et al., 2010; Wu et al., 2013), two samples in which the disorders were fairly equally represented (Buitelaar, Van Andel, Duyx, & Van Strien, 1994; McShane et al., 2001), and just one study in which depressive disorders were a little more common (Prabhuswamy et al., 2007) ¹ . At the symptom level, Buitelaar et al. (1994) reported that depressive symptoms among school refusers were less common than anxiety-related symptoms. In a study of depressive symptoms and school refusal, Honjo and colleagues (2001) concluded that “the typical case of school refusal appears to exhibit somatic complaints in the foreground rather than depression” (p. 629). If depressive symptoms are present, they are more typically seen in adolescent school refusers relative to children (Baker & Wills, 1978).

Externalizing behavior, such as anger and argumentativeness, may also be seen when parents attempt to get the young person to school (Berg, 2002). In some school refusal cases, the presence of multiple externalizing behaviors over time will lead to a diagnosis of oppositional defiant disorder, with rates ranging between 21% and 44% of referred school refusers (Heyne et al., in press). The school refusal criteria of Berg and colleagues specify the absence of severe antisocial behavior such as stealing and destructiveness, hence the finding that conduct disorder (CD) is usually not observed among samples of school refusers.

Twenty-five years ago the etiology of school refusal was described as “incompletely understood” (King & Ollendick, 1989, p. 213). Even today it is not well understood, probably attributable in part to the substantial heterogeneity in the presentation of school refusal and its associated risk factors (Heyne, 2006). Thambirajah, Grandison, and De-Hayes (2008) describe school refusal as multiply determined by a broad range of risk factors which interact with each other and change over time. Several authors (Heyne, 2006; Heyne & King, 2004; Thambirajah et al., 2008) summarize factors identified in the school refusal literature, specified according to individual factors (e.g., behavioral inhibition; fear of failure; low self-efficacy; physical illness), family factors (e.g., separation and divorce; parent mental health problems; overprotective parenting style; dysfunctional family interactions), school factors (e.g., bullying; physical education lessons; transition to secondary school; structure of the school day), and community factors (e.g., increasing pressure to achieve academically; inconsistent professional advice; inadequate support services). These may operate as predisposing, precipitating, and/or perpetuating factors. Precipitating factors can be difficult to identify, especially in cases where onset is not recent (Baker & Wills, 1978). A common perpetuating factor is the young person's access to items (e.g., television) and experiences (e.g., attention from a parent) associated with being at home rather than school (Heyne et al., 2001).

In the absence of treatment, most school refusers continue to display problematic school attendance and emotional distress (King, Tonge, Heyne et al., 1998). Significant adverse consequences may occur in the short- and long-term. Research suggests that non-attendance negatively affects learning and achievement (Carroll, 2010) and places youth at risk for early school dropout (Christle, Jolivette, & Nelson, 2007). Problems in social adjustment may also occur. Coolidge, Brodie, and Feeney (1964) described moderate or severe limitations (e.g., few or no friends; primarily home-centered activities) in approximately 70 percent of school refusers followed up in adolescence or early adulthood. Berg, Butler, and Hall (1976) reported that one-third of school refusers treated as inpatients three years earlier continued to be socially impaired and had severe symptoms of emotional disturbance. Valles and Oddy (1984) compared successfully and unsuccessfully treated school refusers and found that those who did not return to school displayed more antisocial behavior and a trend towards poorer social adjustment at 7-year follow-up (Valles & Oddy, 1984). Other follow-up studies indicate a risk for ongoing mental health problems in late adolescence (Buitelaar et al., 1994) and adulthood (Berg & Jackson, 1985; Flakierska-Praquin, Lindstrom, & Gillberg, 1997; McCune & Hynes, 2005).

Family members are also affected by school refusal. Parents are likely to experience distress associated with the crisis-like presentation of school refusal or the challenge of resolving the problem (Heyne & Rollings, 2002; Kearney, 2001). In some cases parents will have to take time off work to supervise a ‘home-bound’ child. Another consequence of school refusal is family conflict (Kearney & Bensaheb, 2006; McAnanly, 1986; Ollendick & King, 1990). Valles and Oddy (1984) observed more conflict in the families of non-successfully treated cases of school refusal relative to successfully treated cases. One possible explanation is that the ineffective efforts of parents trying to get their child to attend school may exacerbate tension and conflict between the parents themselves and between the parents and their child. School refusal also places a strain upon school staff who may be uncertain about how to deal with the problem and may incur stress displaced onto the school by family members (McAnanly, 1986).

The prevalence of school refusal is difficult to ascertain due to the discrepancy in how school refusal is defined and the lack of national reporting. A qualitative analysis (Heyne & King, 2004) and a large community-based study (Egger et al., 2003) suggest that school refusal occurs for about 1–2% of young people. Estimates among clinic-referred youth range between 5 and 16% (Heyne & King, 2004). Hersov (1985) suggested that school refusal is more prevalent among secondary school students. The age distribution observed in studies, which included children and adolescents, supports this contention (e.g., Last & Strauss, 1990; McShane et al., 2001). According to Heyne and Sauter (2013), the higher referral rates for adolescents might be explained by the complexity of the problem in adolescence and not by a higher prevalence in this period. School refusal is equally common among males and females (Egger et al., 2003; Heyne & King, 2004) and there appears to be no direct association with socioeconomic status (Heyne, King, Tonge, & Cooper, 2001).

In sum, school refusal is not highly prevalent but it can have serious consequences for a young person's development and well-being. It appears that the emotional distress associated with school refusal is commonly in the form of anxiety and less consistently in the form of depression. From a developmental perspective, efficient and effective treatment is imperative.

1.2 THE INTERVENTION AND HOW THE INTERVENTION MIGHT WORK

There is a long history of psychosocial treatments for school refusal. Blagg (1987) provided a detailed review of studies describing the psychodynamic approach, family therapy, and behavioral approaches. The behavioral approaches were based on classical conditioning, operant conditioning, social skills training, or a combination of these. Blagg's review exposed the extent to which clinicians who intended to operate within a specific behavioral paradigm conducted interventions which can be conceptualized within an alternative behavioral paradigm, and sometimes within another paradigm altogether. Similarly, it showed that clinicians work on personal and family factors not specific to the behavioral paradigm(s) they intentionally draw upon. The complexity of specifying and determining the active ingredients of an intervention was well illuminated. Blagg concluded as follows: “In many ways, the search for a definitive behavioural technique or group of techniques to suit all cases is inappropriate. … every case is unique and will require slightly different emphases in treatment” (p. 91). The behavioral interventions commonly discussed in his review were exposure-based interventions, relaxation training, and/or social skills training with the student, and contingency management procedures with the parents and school staff.

Exposure-based interventions stemming from the classical conditioning paradigm (e.g., imaginal and in-vivo systematic desensitization; emotive imagery) are intended to reduce the young person's anxiety associated with school attendance and thereby make it easier for him or her to attend school. As noted by Chu and Harrison (2007), the specific mechanisms through which exposure facilitates approach behavior in anxious youth are still unclear. Occasionally, with younger children, exposure is implemented in the form of flooding whereby parents block the young person's avoidance and enforce attendance at school (e.g., Kennedy, 1965).

Relaxation can help young people manage the stress (negative emotional responses and physiological arousal) that occurs in situations associated with school attendance (e.g., getting ready to go to school; taking tests; giving a class talk; being around other children at school). Procedures include those to relax the body (breathing retraining; progressive muscle relaxation training), the mind (autogenic relaxation training; guided imagery), or the world around them (soothing activities such as listening to music) (Weersing, Gonzalez, Campo, & Lucas, 2008). During systematic desensitization, relaxation may be employed as an anxiety-inhibitor.

Social skills training addresses social-related difficulties that may be a cause, consequence, or correlate of school refusal, anxiety, or depression (Heyne & Sauter, 2013). Difficulties might be experienced in forming and strengthening friendships, dealing assertively with bullying, and answering peers' or teachers' questions about absence from school. The young person is helped to generate ideas for responding to difficult social situations, evaluate these ideas, and select a preferred response. The clinician models variations on the response and engages the young person in practice, with attention to micro-skills (e.g., body language to show interest in another), macro-skills (e.g., keeping conversations going; assertiveness), and social-cognitive skills (e.g., perspective-taking; gauging the receptiveness of another) as needed.

Contingency management draws on operant conditioning principles. Interventions derived from this paradigm are indicated when school refusal has become positively reinforced (e.g., Meyer, Hagopian, & Paclawskyj, 1999) or negatively reinforced (e.g., Hagopian & Slifer, 1993). Parents are helped to manage the antecedents and consequences of their child's behavior to increase desirable behaviors (e.g., use of coping skills; school attendance; doing homework) and reduce undesirable behaviors that get in the way of school attendance (e.g., tantrums; arguments; excessive reassurance-seeking). Adaptive changes may also occur in the parent–child relationship (Kearney & Silverman, 1995), such as a reduction in parent over-involvement with a child who ordinarily seeks and receives an excessive amount of reassurance. Procedures include positive reinforcement, planned ignoring, reducing the young person's access to items and experiences at home which might otherwise positively reinforce the refusal to attend school, and improved instruction-giving (Heyne & Sauter, 2013). School staff members are also encouraged to employ contingency management befitting the school setting.

The beginning of the cognitive-behavior therapy (CBT) approach to treating school refusal is clearly evidenced in the case reports of Mansdorf and Lukens (1987). Based on the notion that attitudes and beliefs interfere with the effectiveness of behavioral interventions, Mansdorf and Lukens used self-instruction techniques to help two school-refusing children employ coping self-statements to guide positive behavior. A cognitive restructuring process was used with parents to challenge distorted beliefs about their child's problem and the management of school refusal. Since then, various authors have suggested that unhelpful cognition is linked with the development and maintenance of school refusal (e.g., Kennard, Ginsburg, Feeny, Sweeney, & Zagurski, 2005; McNamara, 1988; Okuyama et al., 1999; Place, Hulsmeier, Davis, & Taylor, 2000), and two systematic investigations showed a relationship between school refusal and maladaptive cognition (Heyne et al., 1998; Maric, Heyne, de Heus, van Widenfelt, & Westenberg, 2012).

Today, treatment for school refusal typically incorporates both cognitive and behavioral interventions. Five CBT manuals have been developed (Heyne & Rollings, 2002; Heyne, Sauter, & van Hout, 2008; Kearney & Albano, 2007 ² ; Last, 1993; Tolin et al., 2009). All manuals involve: treatment delivery per case (as opposed to group treatment) and usually in a clinical setting; some level of involvement with parents (from parents as consultants to parents as co-clients); consultation with school staff; and an emphasis on the completion of between-session tasks. Graded exposure to school attendance is commonly advocated. Most of the manuals incorporate psychoeducation, problem-solving training with the young person as a matter of routine or as required, and family work on communication and problem-solving. Cognitive therapy (CT) interventions are employed with all school refusers or as required, and there is variation across manuals in the type of CT interventions employed. Just two of the five manuals explicitly refer to cognitive interventions with parents. The number of sessions and treatment duration vary considerably (e.g., 15 sessions of 90-120 minutes across 3 weeks; 8 sessions of 50 minutes across 4 to 8 weeks; 12 sessions of 45-60 minutes across 12 weeks). The earliest CBT manual (Last, 1993) was standardized, with all cases receiving the same intervention, but the newer manuals advocate individualized treatment based on the main function(s) served by the young person's behavior (e.g., Kearney & Albano, 2007) and/or a broader case formulation including predisposing, precipitating, perpetuating, and protective factors (e.g., Heyne & Rollings, 2002; Heyne et al., 2008; Tolin et al., 2009).

In a trial conducted by Last and colleagues (1998), CBT was compared with an educational-support therapy (ES) to control for the nonspecific effects of CBT. In short, ES comprised educational presentations and supportive psychotherapy. More specifically, it included handouts with questions for the school refuser to consider; a daily diary to record feared situations and associated thoughts, feelings, and responses; encouragement for the young person to talk about their fears; and instruction in identifying maladaptive thinking. There was no instruction or encouragement for the young person to confront feared situations and no instruction about how to modify maladaptive thinking.

Medications have been trialed in studies of tricyclic anti-depressants (Berney et al., 1981; Bernstein et al., 2000; Gittelman-Klein & Klein, 1971), a tricyclic antidepressant or benzodiazepine (Bernstein, Garfinkel, & Borchardt, 1990), or a selective serotonin reuptake inhibitor (Wu et al., 2013), all combined with psychosocial treatment. The effects of inpatient treatment combined with psychosocial interventions have also been reported (e.g., Blagg & Yule, 1984; McShane et al., 2004; Walter et al., 2010). For this review we were interested in examining intervention types that could be implemented by clinical or counseling professionals in clinics, schools, or community facilities thus we focused on psychosocial interventions.

Research examining mediators and moderators of school refusal can contribute to better understanding of the development, maintenance, and treatment of school refusal. To date, just one study has investigated the role of a purported risk factor in mediating the outcome of treatment for school refusal. Maric, Heyne, MacKinnon, van Widenfelt, and Westenberg (2013) found that increases in self-efficacy at posttreatment significantly mediated posttreatment increases in school attendance and decreases in fear about attending school. The finding corroborated Pina and colleagues' (2009) suggestion that improved school attendance would be mediated by increased self-efficacy for handling school situations. According to Maric et al. (2013), cognitive or behavioral interventions may promote a school refuser's self-efficacy for facing the anxiety-provoking situation of attending school, which in turn may lead to increased school attendance. There have been numerous studies of individual or family factors predicting treatment outcome. The most consistent finding to date is that treatment response is inferior for school-refusing adolescents with social anxiety disorder (Bernstein et al., 2001; Heyne et al., 2011; McShane et al., 2004). Problem severity–in the form of school absenteeism or co-occurring psychopathology–is another negative prognostic indicator (Bernstein et al., 2001; Last et al., 1998; Layne et al., 2003; McShane et al., 2004; Walter et al., 2013). Very little attention has been paid to school factors predicting outcome (Heyne et al., in press).

In sum, a broad range of factors may be associated with the development and maintenance of school refusal. Earlier, therapists purported to operate within a specific theoretical paradigm, but the various interventions they employed can be conceptualized according to various paradigms. Currently, treatment for school refusal commonly integrates behavioral and cognitive interventions conducted with the young person, their parents, and school staff. The overarching aim of these interventions is the reduction of the young person's emotional distress and an increase in school attendance to help the young person follow a normal developmental pathway (Heyne & Sauter, 2013). Some time ago Valles and Oddy (1984) raised the question of whether reductions in emotional distress primarily facilitate increased school attendance or vice versa. Contemporary theorizing echoes the uncertainty associated with school refusal and its treatment. When discussing mediation models of treatment outcome, Heyne et al. (in press) suggested that “school attendance and internalizing problems might best be regarded as separate variables in causal chains, sometimes as a mediating variable and sometimes as an outcome variable, depending on the proposed relationship with the other variables under discussion. ³ ”

1.3 WHY IT IS IMPORTANT TO DO THE REVIEW

Although there have been systematic reviews and meta-analyses conducted on the effects of interventions for children and adolescents with anxiety disorders (Brendel & Maynard, 2014; Ishikawa, Okajima, Hirofumi, & Sakano, 2007) and prior narrative reviews of school refusal (see Maynard et al., 2013 for a summary of prior reviews), we have not located a meta-analysis of interventions focused specifically on school refusal. Prior reviews on school refusal focused on what is known about school refusal in terms of etiology, prevalence, assessment, and treatment; however, few reviews have focused specifically on intervention outcomes. Reviews that have focused more specifically on outcomes have included studies with little or no criteria of study design or internal validity. Only one prior review has employed systematic review methodology (Pina et al., 2009), and no reviews have quantitatively synthesized effects using meta-analytic methods.

King and colleagues (King et al., 2005; King, Tonge, Heyne, & Ollendick, 2000) conducted two narrative reviews of school refusal intervention outcome studies. The 2000 review included eight published studies of CBT interventions using a range of research designs. Their report concluded, “At first glance, our review of research suggests empirical support for cognitive-behavioral therapy in the treatment of school refusal…” (p. 501). “However, since very few controlled studies have been reported at this stage in treatment research, it would be premature to extol the clinical virtues of cognitive-behavior therapy” (p. 506). King et al.'s 2005 review focused on a broader topic of anxiety and phobic disorders and included seven studies examining effects of behavioral or CBT interventions with school refusal. Although the authors used substantially the same studies in both reviews, the two reviews came to different conclusions, likely due to the inclusion of an additional controlled study in the latter review. In the 2005 study, the authors concluded, “Overall, school refusal has responded to CBT programs as demonstrated in a number of controlled studies, with general maintenance of gains” (p. 249). A more recent review by Pina and colleagues (2009) included both single-subject and group-design studies examining effects of psychosocial interventions for school refusal. The authors presented effect sizes for the included group design studies, but did not synthesize effects across studies. Conclusions from this review indicate that behavioral strategies alone and behavioral strategies in combination with cognitive strategies are promising interventions to reduce school refusal.

Prior reviews of school refusal intervention research also have been limited to published research and have primarily employed either qualitative or vote-counting methods for synthesizing study outcomes. Moreover, most of the existing reviews of school refusal have tended to focus on CBT interventions. Taken together, the past reviews are important to our understanding of school refusal and provide some evidence to guide interventions. However, they do not systematically or quantitatively address the question of whether and which interventions are effective for increasing school attendance and decreasing anxiety for children exhibiting school refusal.

This systematic review expands and improves upon prior work in several ways. First, this review applied a systematic and transparent process for searching, retrieving, and coding studies, including the search for unpublished studies. Using a systematic method to conduct the review of outcome research limits bias and reduces chance effects, leading to more reliable results (Cooper, 1998). Searching for unpublished studies could produce additional studies that have not been included in prior reviews. Further, explicitly and transparently describing the review process allows for others to replicate and expand the review to include new studies or criteria.

Second, this review sought to include evaluations of a variety of interventions operating in a broad set of geographical contexts, including programs across the United States and other countries. This broader reach allowed for the possibility of identifying studies that may not have been included in prior reviews.

Lastly, prior reviews have largely been limited to a narrative approach, presenting a description of programs or using a vote-counting method to categorize outcomes of programs as significantly positive, significantly negative, or null. The vote-counting method, however, disregards sample size, relies on statistical significance, and does not take into account measures of the strength of the study findings, thus possibly leading to erroneous conclusions (Glass, McGaw, & Smith 1981). Meta-analysis, on the other hand, represents key findings in terms of effect size, rather than statistical significance. Thus, meta-analysis provides information about the strength and importance of a relationship, the magnitude of the effects of the interventions, and the characteristics of effective interventions. Moreover, because there are a relatively small number of studies in this area employing small sample sizes, quantitatively synthesizing effects across studies can more accurately answer questions of effectiveness than the narrative or vote counting methods employed by prior reviews. Given the small sample sizes in several of the included studies, it is possible that individual primary studies could fail to demonstrate a significant effect because they were underpowered. One of the strengths of meta-analysis over other synthesis methods is the capability to find effects that are obscured or not readily apparent when using less sophisticated approaches (Lipsey & Wilson, 2001). By pooling effect size estimates across studies, we can combine the results of underpowered studies, thus producing a synthesized effect estimate with considerably more statistical power to discover meaningful effects that may otherwise be missed in low-powered individual studies (Card, 2012).

3.1 CRITERIA FOR CONSIDERING STUDIES FOR THIS REVIEW

3.2 SEARCH METHODS FOR IDENTIFICATION OF STUDIES

3.3 DATA COLLECTION AND ANALYSIS

3.4 DEVIATIONS FROM THE PROTOCOL

4.1 DESCRIPTION OF STUDIES

4.1.1 Results of the search

Electronic searches of bibliographic databases and searches of other sources identified a total of 8,798 citations. Titles and abstracts were screened for relevance, and 8,759 were excluded as duplicates or deemed inappropriate. The full texts of the remaining 39 potential studies were reviewed and screened for eligibility by two independent coders. Six studies passed full-text screening and were included in the primary review. An additional two studies were identified that assessed the comparative effects of psychotherapy with and without medication and met all other criteria. These studies were retained for a separate analysis that was not part of the original protocol. See Figure 4.1 for the flow chart of the study selection process.

OPEN IN VIEWER

4.1:

FLOW CHART OF STUDY SELECTION PROCESS

4.2 RISK OF BIAS IN INCLUDED STUDIES

4.2.4 Reporting Bias

Although no protocols for the included primary studies were located, we classified all eight included studies as low risk of bias with regards to selective outcome reporting as they appear to report expected outcomes and provided adequate data on outcomes of interest to this review.

See Figures 4.2 and 4.3 for a summary of risk of bias within and across included studies.

OPEN IN VIEWER

4.2

RISK OF BIAS ITEMS BY STUDY

OPEN IN VIEWER

4.3

RISK OF BIAS ACROSS STUDIES

4.3 SYNTHESIS OF RESULTS OF PSYCHOSOCIAL-ONLY INTERVENTIONS

4.3.1 Mean Effects on Anxiety Outcomes

Four of the included psychosocial studies assessed effects on anxiety. Results indicated that the overall mean effect at post-test on anxiety, assuming a random effects model, was 0.06 (95% CI [-0.63, 0.75], p = .86), demonstrating a non-significant effect. The precision of the point estimate should be interpreted with caution, as there was significant heterogeneity between the studies (Q = 11.13, p = .01; I ² = 73.05; τ ² =.36). The mean effect size and confidence intervals for each study are shown in the forest plot in Figure 4.4 below.

OPEN IN VIEWER

4.4

FOREST PLOT OF MEAN EFFECTS ON ANXIETY

4.3.2 Mean Effects on Attendance Outcomes

The mean effect size at post-test of the six psychosocial studies that assessed effects on attendance, assuming a random effects model, was g = 0.54 (95% CI [0.22, 0.86], p = .00), demonstrating a positive and significant effect. Results of the Q-test were not significant (Q = 8.82, p = .12) and values for I ² (43.32) and τ² (.06) indicate a small amount of heterogeneity. The mean effect size and confidence intervals for each study are shown in the forest plot in Figure 4.5 below.

OPEN IN VIEWER

4.5

FOREST PLOT OF MEAN EFFECTS ON ATTENDANCE

4.4 SYNTHESIS OF RESULTS OF PSYCHOSOCIAL INTERVENTION + MEDICATION

4.4.1 Mean Effects on Anxiety Outcomes

The overall mean effect on anxiety outcomes from the two studies examining effects of CBT with medication versus CBT only, assuming a random effects model, was -0.05 (95% CI [-0.40, 0.31], p = .80), demonstrating negative, but not statistically significant effect on anxiety. Results of the Q-test were not significant (Q = .30, p = .58) and values for I ² and τ ² were .00. The mean effect size and confidence intervals for each study are shown in the forest plot in Figure 4.6.

OPEN IN VIEWER

4.6

FOREST PLOT OF MEAN EFFECTS ON ANXIETY

4.4.2 Mean Effects on Attendance Outcomes

The overall mean effect on attendance outcomes of the two studies examining effects of CBT with medication versus CBT only, assuming a random effects model, was 0.61 (95% CI [0.01, 1.21], p = .046), demonstrating a positive and statistically significant effect. Results of the Q-test were not significant (Q = 1.93, p = .17) and values for I ² (48.23%) and τ ² (.09) indicate a small amount of heterogeneity. The mean effect size and confidence intervals for each study are shown in the forest plot in Figure 4.7 below.

OPEN IN VIEWER

4.7

FOREST PLOT OF MEAN EFFECTS ON ATTENDANCE

4.5 SENSITIVITY AND MODERATOR ANALYSES

Due to the small number of studies and homogeneity across the medication studies and the psychosocial studies on attendance outcomes, performing a number of sensitivity and moderator analyses was not warranted. However, we did conduct a sensitivity analysis to examine the effects of our selection of the treatment and comparison groups for the Richardson study on effect size and a sensitivity analysis to assess the impact of the Blagg and Yule (1984) study on attendance. We also conducted sensitivity analysis to compare the effect sizes from published and unpublished studies. In terms of moderator analyses, C2 guidelines require RCTs be separate from QED studies in the meta-analysis or that research design be examined as a moderator; therefore, we examined study design as a moderator variable on attendance and anxiety with the psychosocial intervention studies. Because we found significant heterogeneity between studies on anxiety, we also examined the type of comparison group condition and publication status (i.e., alternative intervention versus a wait list control condition) as a moderator.

5.1 SUMMARY OF MAIN RESULTS

The present review found relatively few rigorous studies of interventions for school refusal. All but one included study assessed effects of a variant of CBT, thus there appears to be a lack of rigorous evidence of other types of interventions for school refusal. Findings of the current review point to different effects on anxiety and attendance. While the psychosocial-only and CBT-plus-medication interventions found, on average, positive and significant effects on attendance compared to control, effects on anxiety at post-test across both sets of studies were not significantly different from zero. Moreover, the magnitude of treatment effects varied across the psychosocial-only studies on anxiety, and thus current estimates of treatment effects on anxiety should be evaluated with caution.

The mixed findings of effects on anxiety and attendance are quite interesting and appear to be somewhat counterintuitive. Anxiety is often viewed as contributing to the development and maintenance of attendance problems exhibited by school refusers, so one would expect that interventions targeting school refusal would decrease anxiety, which would then lead to improved school attendance. The results here, however, demonstrate a more immediate and significant difference in improvement of attendance and no significant difference between treatment and control groups on anxiety outcomes. Although this seems somewhat counterintuitive, the mechanisms of the specific interventions point to a reasonable explanation for these mixed results, particularly at post-test. That is, the increased exposure to school, which is a key component to some of the interventions in this review, may result in more immediate improvement in attendance. However, the increased exposure to school could, at least in the short term, result in an increase in anxiety. A decrease in anxiety may follow from a child's continued attendance at school. In the discussion of Last's (1998) treatment outcome study it was also argued that the emphasis in CBT on increasing school attendance may have heightened anxiety levels. Only Heyne et al. (2002) examined outcomes of both attendance and anxiety at a follow-up time period, so we could not examine longer-term effects of school refusal interventions on anxiety.

There are two main caveats that must be considered when interpreting the findings of this review. First, all of the psychosocial only intervention studies, except for two (King et al., 1998; Sahel, 1989), compared effects of two interventions. Authors examining comparative effects reported improvement across both groups on at least one (Richardson, 1992) or both (Heyne et al., 2002; Last et al., 1998) outcomes of interest to this review. In two of the six psychosocial only intervention studies, the control group was also provided a variant of a behavioral or CBT intervention (Heyne et al. 2002; Richardson, 1992), which could lead to a downward bias in the overall mean effect. While there is not consistent evidence that CBT is more effective than other types of interventions, CBT does appear to have some evidentiary support as an intervention for school refusal.

Second, there are few rigorous trials available to support any one CBT intervention for school refusal. There was some variability in the CBT interventions examined in this review (e.g., number of sessions; amount of contact with parents) and while some researchers used similar CBT interventions, none of the studies were replication studies assessing the same manualized intervention by independent researchers. Replication studies are an often-cited requirement for classifying specific interventions as empirically supported (Chambless, 1998; Flay, 2005). Although the Heyne et al. (2002), King et al. (1998), and Wu et al. (2013) studies assessed effects of very similar youth CBT plus PTT interventions, there was variation in the number of sessions and duration of treatment, and two of those studies were conducted by the same research group (Heyne et al., 2002; King et al., 1998). The intervention delivered in the Heyne et al. study was comprised of eight child therapy and eight PTT sessions over four weeks whereas the intervention in the King et al. study was comprised of 6 child therapy and five PTT sessions, and Wu offered 12 sessions across 12 weeks. Last et al. (1998) and Bernstein et al. (2000) also implemented similar CBT interventions; however, there was some variation in terms of number of sessions. In sum, CBT may have some evidentiary support, particularly for attendance outcomes; however, it may be premature to classify any specific form of CBT as empirically supported at this time.

5.2 OVERALL COMPLETENESS AND APPLICABILITY OF EVIDENCE

Six studies examining effects of psychosocial interventions on school refusal and two studies examining effects of medication with CBT versus CBT with placebo (Bernstein et al., 2000) or CBT alone (Wu, 2013) were included in this review. All but one study examined a variant of CBT intervention, thus there is an overall lack of evidence of other types of interventions for school refusal. This review was limited to more rigorous study designs, thus there are fewer studies in this review than some prior reviews of school refusal. While having more studies can be beneficial, we believe that having a higher standard of rigor and internal validity for studies included in this review provides more credible evidence of school refusal interventions than prior reviews. Most of the studies did not report information related to socio-economic status, and only half of the studies reported information on race or nationality. Of those four studies that did report race, ethnicity or nationality, participants in all but one were primarily Caucasian, thus there appears to be a lack of studies examining effects of interventions with more diverse participants.

5.3 QUALITY OF THE EVIDENCE

The studies included in this review were comprised of mostly randomized trials, with two quasi-experimental studies. None of the studies provided a description of the method of random allocation and only one study blinded participants and outcome assessors, thus the studies in this review were assessed as high risk for performance and detection biases. Most of the included studies assessed effects on only the completers, which may result in an overstatement of treatment effects. Many of the studies in this review were older, and thus may explain to some extent the lack of use of ITT analysis, as ITT analysis has become more common in the recent decade. Unfortunately, there were too few studies to conduct sensitivity analyses to provide a systematic evaluation of the impacts of quality indicators or risk of bias.

5.4 LIMITATIONS AND POTENTIAL BIASES IN THE REVIEW PROCESS

The current review is limited to mainly published research, and while efforts were made to search the gray literature, only two unpublished studies were included in this review. Thus, our review may be upwardly biased due to the possible threat of publication and reporting bias, particularly for anxiety outcomes. Our searching of conference abstracts was weighted toward U.S. conferences, and thus we may have missed unpublished conference papers presented in non-U.S. conferences. As described in the search strategy, we decided post-hoc to forgo the searches in Google and Bing, and thus we may have missed some published and unpublished studies. This study is also limited by the small number of studies included in this review and thus low statistical power and limits to the analytic techniques that could be employed.

5.5 AGREEMENTS AND DISAGREEMENTS WITH OTHER STUDIES OR REVIEWS

The current findings are generally consistent with prior narrative reviews in that behavioral and cognitive interventions were found to be the most frequently studied interventions for school refusal. Although several prior reviews describe positive effects of cognitive and/or behavioral interventions for school refusal, many also point to mixed results and the overall lack of evidence for any one treatment modality (Elliot, 1999; King & Bernstein, 2001; King et al., 2000), which is consistent with the results of this review. This review, however, represents a more rigorous extension of existing work in that a more comprehensive and systematic search process was undertaken than in prior reviews and more rigorous inclusion criteria was used to improve the credibility of the review for causal inference. Only one prior review used systematic search procedures (Pina et al., 2009) and none have included unpublished studies. Moreover, prior reviews have used less stringent inclusion criteria and none have employed meta-analytic techniques to quantitatively synthesize results of included studies. This is especially important for this body of studies, as the sample sizes are small. Quantitatively synthesizing results of underpowered studies can produce a pooled effect estimate with considerably more statistical power to discover meaningful effects that can be missed in low-powered individual studies (Card, 2012).

6.1 IMPLICATIONS FOR PRACTICE AND POLICY

The current evidence provides tentative support for CBT in the treatment of school refusal, and for the supplement of medication to CBT, at least for the improvement of attendance. The effects on anxiety are somewhat less clear as we found no differential effects on anxiety outcomes at post-test. However, the current review highlights important caveats to this conclusion. Most of the studies in this review compared effects against other, and sometimes very similar, interventions and the body of studies included in this review also suffer from some significant biases that could result in an upward bias of the effect. Furthermore, the evidence shows only immediate effects of interventions, as there was only one study that reported comparative longer-term effects on both attendance and anxiety, thus there is insufficient evidence to indicate whether or not treatment effects sustain or whether anxiety was indeed reduced with greater exposure to school. While the interventions show promise in the treatment of school refusal, there is an overall lack of sufficient evidence to draw firm conclusions of the efficacy of CBT as the treatment of choice for school refusal.

6.2 IMPLICATIONS FOR RESEARCH

Overall there are few rigorous trials of interventions for school refusal, particularly studies examining effects of interventions other than behavioral interventions or CBT. Moreover, most of the studies that met inclusion criteria for this review were published more than 10 years ago and most were assessed at high risk of bias. The few rigorous studies found for this review, the age of included studies and the risks of bias present in most of the included studies indicate a need for better-controlled studies. Study design and analytic methods have progressed over the past decade, with more rigorous designs being expected and intent-to-treat analysis becoming more common since the time that most studies in this review were conducted. Future research in this area could benefit from more sophisticated and rigorous designs and analytic techniques. Moreover, independent replications of the manualized interventions examined in this review are needed, as are longer-term evaluations of effects of interventions. Assessing long-term effects could provide additional answers and insights as to the mixed findings of the effects of interventions on attendance and anxiety. Future studies should also consider other types of interventions for rigorous evaluation. While the functional treatment of school refusal has been identified as a promising approach (Kearney & Silverman, 1990; 1999), to date, this intervention lacks rigorous evaluation.

To further our theoretical understanding of school refusal and its treatment, studies should include measures of factors held to mediate outcome. Factors currently regarded as important include maladaptive cognition and problem-solving at the individual level, parenting style and family functioning at the family level, and the quality of the student-teacher relationship at the school level (Heyne et al., in press). To understand the temporal precedence of changes in anxiety and changes in school attendance, both should be measured at various points during treatment. Pertinent pre-treatment measures of the severity and chronicity of school refusal would also permit robust predictor and moderator analyses which are especially important given the common notion that the longer a school refuser is away from school, the harder it is for him or her to return. Furthermore, future studies could benefit from larger sample sizes and attention to mitigating potential biases to improve statistical power and causal inference. Larger samples would permit analyses that can account for the heterogeneity so characteristic of school refusal, such as the influence of the presence of different types of anxiety disorders. For example, outcomes may be inferior for school-refusing youth with social anxiety disorder relative to those with alternate anxiety disorders (Blote, Miers, Heyne, & Westenberg, in press).