Psychosocial Interventions for School Refusal with Primary and Secondary School Students: A Systematic Review
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). It is often distinguished from truancy, in part because of the severe emotional distress associated with having to attend school and the absence of severe antisocial behavior. Truancy, on the other hand, is not typically associated with emotional distress and is commonly associated with severe externalizing behavior. The emotional distress associated with school refusal is often in the form of fear or anxiety, and sometimes in the form of depression. School refusal occurs for about 1-2% of young people, and estimates among clinically referred youth are considerably higher.
There is substantial heterogeneity in both the presentation of school refusal and its associated risk factors. Significant adverse consequences may occur in the short- and long-term, including school dropout and problems with social adjustment. Family members and school staff are also affected by school refusal. The most commonly studied interventions for school refusal are behavioral approaches and cognitive-behavioral therapy (CBT). 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). Behavioral interventions include exposure-based interventions, relaxation training, and/or social skills training with the student, and contingency management procedures with the parents and school staff. CBT manuals additionally focus attention on the identification and modification of maladaptive cognition that may maintain the young person’s emotional distress and absenteeism. In some instances parent cognition is also targeted. Other interventions have been used to treat school refusal (e.g., psychodynamic treatment, family therapy, medication) but CBT has been the most studied intervention and most prior reviews have focused on CBT and/or behavioral interventions.
While prior reviews have found some support for CBT and behavioral interventions for reducing anxiety and/or improving attendance, the reviews have been mixed (Maynard et al., 2013). Prior reviews have also been limited to published research, have not adequately assessed the quality of evidence, and have primarily employed either qualitative or vote-counting methods for synthesizing study outcomes. No prior meta-analysis of interventions targeting school refusal has been located.
The purpose of this review was to inform practice and policy by evaluating the effects of psychosocial interventions for school refusal. The following research questions guided this study:
1) Do psychosocial interventions targeting school refusal reduce anxiety?
2) Do psychosocial interventions targeting school refusal increase attendance?
Electronic searches were conducted in 15 databases and 4 research registers, and internet searches were conducted for conference proceedings and other grey literature. Searches were conducted using the following keywords: (anxiety OR “school refus*” OR “school phobia”) AND (attendance OR absen*) AND (evaluation OR intervention OR treatment OR outcome OR program) AND (student* OR school* OR child* OR adolescen*). Reviews of reference lists of included studies and prior reviews and personal contact with authors of prior studies of school refusal were also conducted to identify potential studies for this review.
Published or unpublished studies assessing effects of psychosocial interventions to improve attendance or reduce anxiety with school-age youth who met criteria for school refusal were included in this review. To be eligible for inclusion in this review, studies must have been conducted or reported between January 1980 and November 2013 and employed an experimental or quasi-experimental design. In addition, studies must have used statistical controls or reported baseline data on outcomes regardless of study design. Studies that assessed effects of medications only or studies conducted in residential treatment centers were excluded from this review.
DATA COLLECTION AND ANALYSIS
Titles and abstracts of the studies found through the search procedures were screened for relevance, and those that were obviously ineligible or irrelevant were screened out. Documents that were not obviously ineligible or irrelevant based on the abstract review were retrieved in full text for final eligibility screening. Two reviewers independently screened the full-text articles for inclusion. Studies that met eligibility criteria were coded independently by two coders. Two review authors also independently assessed the risk of bias in each study using the Cochrane Collaboration’s ‘Risk of Bias’ tool (Higgins et al., 2011). Coders met to review the coding agreement and any discrepancies were discussed and resolved by consensus.
Effect sizes were calculated in Comprehensive Meta-Analysis (CMA) version 2.0 (Borenstein, Hedges, Higgins, & Rothstein, 2005). We adjusted for differences at baseline by computing
the pre-test effect size and subtracting it from the post-test effect size. The standardized mean difference effect size statistic, employing Hedges’ g to correct for small sample size bias (Hedges, 1981), was used. When an author used more than one measure of an outcome, an effect size was calculated for each measure and a mean ES was calculated so each study contributed only one effect size per study for each outcome. Four meta-analyses were performed; two meta-analyses were performed to synthesize studies assessing effects of psychosocial interventions on anxiety and attendance and two were performed to synthesize effects of studies assessing effects of medication in combination with psychotherapy on anxiety and attendance. A weighted mean effect was calculated by weighting each study by the inverse of its variance using random effects statistical models. We assessed statistical heterogeneity using the Q-test and I2 statistic.
Several moderator and sensitivity analyses were planned, but due to the small number of studies included in this review and lack of heterogeneity, we limited additional analyses performed.
A total of eight studies examining effects of interventions on anxiety or attendance with 435 school-age participants exhibiting school refusal were included in this review. Six studies examining effects of psychosocial interventions and two studies assessing comparative effects of psychosocial interventions with and without medication met inclusion criteria for this review. Six of the included studies were randomized controlled trials (RCT) and two were quasi-experimental design (QED) studies. The majority (75%) of the studies were published in peer-reviewed journals. Five of the interventions took place in a clinic setting, one in the school, one in the school and home and one in an undisclosed setting. All but one of the six psychosocial intervention studies in this review assessed the effects of a variant of cognitive-behavioral therapy (CBT) compared to no treatment control (k = 1), an unspecified control (k = 1) or alternative treatment control group (k = 4). For the two studies assessing effects of medication, the same CBT intervention was applied across treatment and control groups with either Fluoxetine or imipramine as the treatment and placebo or no placebo as the control.
The mean effect of the psychosocial interventions at post-test on anxiety was g = 0.06 (95% CI [-0.63, 0.75], p = .86), demonstrating a non-significant effect. The homogeneity analysis indicated a moderate degree of heterogeneity (Q = 11.13, p = .01, I2 = 73.05%, τ2 = .36). Effects on attendance were significant (g = 0.54 (95% CI [0.22, 0.86], p = .00). The homogeneity analysis indicated a small degree of heterogeneity (Q = 8.82, p = .12, I2 = 43.32%, τ2 = .06). Similar results were found for the mean effects of medication + CBT studies, with effects on anxiety being not significant (g= -0.05, 95% CI [-0.40, 0.31], p = .80) and effects on attendance being positive and statistically significant (g = 0.61 (95% CI [0.01, 1.21], p = .046). Studies were homogenous for the medication + CBT studies for both anxiety (Q = .30, p = .58; I2 = 0.00% and τ2 = .00) and attendance (Q = 1.93, p = .17, I2 = 48.23%; τ2 = .09).
The present review found relatively few rigorous studies of interventions for school refusal. Seven of the eight included studies assessed effects of a variant of cognitive behavioral therapy (CBT), thus there appears to be a lack of rigorous evidence of non-CBT interventions for school refusal. Findings of the current review were mixed. While both the CBT 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 on anxiety varied across the psychosocial only studies, and thus current estimates of treatment effects should be evaluated with caution.
The current evidence provides tentative support for CBT in the treatment of school refusal, but 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. Most of the studies in this review compared effects against other, and sometimes very similar, interventions that could mask larger effects if compared to wait list control or other disparate interventions. Furthermore, most studies only measured immediate effects of interventions; only one study reported comparative longer-term effects on both attendance and anxiety. Thus, there is insufficient evidence to indicate whether or not treatment effects sustain, and whether or not anxiety might further decrease over time with continued exposure to school.
Several risks of bias were present in most studies included in this review, particularly related to blinding of participants and assessors, which must be considered when interpreting the results of this review. Performance and detection bias resulting from inadequate blinding of participants and assessors to condition could upwardly bias the mean effects. In addition, insufficient details related to random sequence generation and allocation concealment were provided to adequately assess selection bias in most studies, and two studies reported non-random allocation to condition. While most studies in this review reported to use random assignment procedures, it is uncertain whether selection bias is present due to inadequate generation of randomization or concealment of allocation prior to assignment.
The few rigorous studies found for this review and the risks of bias present in most of the included studies indicate a need for better-controlled studies. 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. Furthermore, future studies could benefit from larger sample sizes and attention to mitigating potential biases to improve statistical power and causal inference.