Over 600 RCTs have demonstrated the effectiveness of psychosocial interventions for children and young people’s mental health, but little is known about the long-term outcomes. This systematic review sought to establish whether the effects of selective and indicated interventions were sustained at 12 months.
We conducted a systematic review and meta-analysis focusing on studies reporting medium term outcomes (12 months after end of intervention).
We identified 138 trials with 12-month follow-up data, yielding 165 comparisons, 99 of which also reported outcomes at end of intervention, yielding 117 comparisons. We found evidence of effect relative to control at end of intervention (K = 115, g = 0.39; 95% CI: 0.30–0.47 I2 = 84.19%, N = 13,982) which was maintained at 12 months (K = 165, g = 0.31, CI: 0.25–0.37, I2 = 77.35%, N = 25,652) across a range of diagnostic groups. We explored the impact of potential moderators on outcome, including modality, format and intensity of intervention, selective or indicated intervention, site of delivery, professional/para-professional and fidelity of delivery. We assessed both risk of study bias and publication bias.
Psychosocial interventions provided in a range of settings by professionals and paraprofessionals can deliver lasting benefits. High levels of heterogeneity, moderate to high risk of bias for most studies and evidence of publication bias require caution in interpreting the results. Lack of studies in diagnostic groups such as ADHD and self-harm limit the conclusions that can be drawn. Programmes that increase such interventions’ availability are justified by the benefits to children and young people and the decreased likelihood of disorder in adulthood.
Citation: Pilling S, Fonagy P, Allison E, Barnett P, Campbell C, Constantinou M, et al. (2020) Long-term outcomes of psychological interventions on children and young people’s mental health: A systematic review and meta-analysis. PLoS ONE 15(11): e0236525. https://doi.org/10.1371/journal.pone.0236525
Editor: Stephan Doering, Medical University of Vienna, AUSTRIA
Received: March 19, 2020; Accepted: July 7, 2020; Published: November 16, 2020
Copyright: © 2020 Pilling et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files
Funding: SP was supported by funding from the Royal College of Psychiatrists (RCPsych), Higher Education Funding Council for England (HEFCE) and the UCLH Biomedical Research Centre; PF by NHSE, HEFCE, CLARHC and NIHR; TK by RCPsych and NHSE; HM, PB by RCPsych; EA, CC, MC, TG, NL, SS, AR, AT, TV and KW by PF’s NIHR Senior Investigator Award and grant number 510388 from AFNCCF.
Competing interests: The authors have declared that no competing interests exist.
The under-treatment of children and young people’s mental disorder is ubiquitous globally , yet problems at this age are harbingers of adult disorders. Fifty percent of all adult mental ill-health is diagnosable by 14 years of age, and 75% by 18–25 years [2, 3]. Many children and young people also experience significant sub-threshold symptoms which may be precursors to the development of a mental disorder [4–6]. Access to treatments associated with long-term benefits could both address the unmet need for children and young people and reduce adult rates of mental ill-health.
Universal prevention efforts to address children and young people’s mental health have not yet reached consensus on how to reduce the burden associated with mental health problems [7–9]. Despite considerable efforts, the evidence for universal programs is not robust and there is uncertainty about their long-term impact . The challenge of universal prevention is addressing the wide range of interrelated risk factors (individual, family, school, community) which require comprehensive multilevel approaches .
In general, selective and indicated prevention programmes appear more clinically and cost-effective . Given the complications of pharmacological interventions there is a natural preference for psychosocial treatments for children and young people . Psychosocial interventions for mental disorders in children and young people are known to be efficacious [13, 14]. A recent comprehensive meta-analysis reported medium end-of-treatment effect sizes based on 447 studies (13). However, there are no existing systematic reviews which report long-term treatment outcomes across a broad range of disorders, which is of particular importance given that the majority of mental disorders are identifiable before the age of 18 years. Understanding whether the benefits of treatment are sustained can inform policy priorities for children and young people’s mental health services and this review was undertaken in response to a request from United Kingdom’s English Department of Health to examine the overall long-term effects of psychological interventions. Further, while these reviews have focused on treatment modality as a predictor of outcome, other important parameters have not been explored, including the level of training of those offering interventions, the setting in which interventions are provided and the dose required to achieve long-term outcomes.
Materials and methods
Protocol and registration
This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines. A protocol was developed and registered on PROSPERO (CRD42017081290). The protocol was adhered to except for the following deviations: (1) we undertook additional, exploratory subgroup analyses to explore heterogeneity in the data, and (2) we placed a stronger emphasis on long term outcomes, as end-of-treatment data has been comprehensively summarised in a recent report. All end of treatment data is presented as per protocol.
This review was undertaken to guide a major UK policy initiative , in order to explore (1) whether the effects of selective and indicated interventions were sustained in the longer term, (2) what models of intervention for which disorders had the most promising long-term outcomes, (3) what level of training and support was required for effective provision of interventions (4), whether delivery site (school, community or health setting) moderates the impact of interventions, and (5) what conditions are required to ensure robust provision of evidence-based interventions.
All randomised controlled trials of psychological interventions for children or young people between 4 and 18 years old with or at risk of developing a mental health disorder, were potentially eligible for inclusion. Eligible mental health disorders comprised: anxiety disorders (including generalised anxiety disorder, obsessive compulsive disorder, panic disorder, social anxiety disorder and phobic disorders); conduct disorders (including oppositional defiant disorder and conduct disorder); depressive disorders (including depression and clinically significant sub-threshold symptoms); eating disorders (including anorexia nervosa, bulimia nervosa and binge eating disorder); post-traumatic stress disorder; substance misuse (including drug and alcohol misuse); self-harm; and attention deficit hyperactivity disorder. Studies eligible for inclusion were those where the mean age of the sample was between 4 and 18, interventions were compared against a no-treatment control, wait list, attentional control, treatment as usual or an active intervention control and reported outcomes at 9–18 months post-treatment. We chose this timeframe because (a) very few studies collect data beyond 18 months, (b) intercurrent treatments present a major challenge for interpreting outcomes beyond this and (c) the majority of relapse occurs within the first year following treatment completion [16, 17]. Studies were excluded if their participant sample were recruited from inpatient settings (as the severity of the disorders in in-patient populations were unlikely to initially treated in school or community settings), had only a solely pharmacological control arm (as we wanted these intervention to be deliverable in school settings where pharmacological interventions were not routinely available), evaluated universal preventive interventions (as evidence suggested they may not have lasting effects), were published only as dissertations, abstracts or conference proceedings or were from non-OECD countries (as we wanted to considered a range of contextual factors such which could only be explored in OECD countries).
The following data bases were searched: PsycINFO; EMBASE; MEDLINE; ERIC (Educational Resources Index); BEI (British Education Index); the Cochrane Library (all databases); Specialised Register of the Cochrane Common Mental Disorders Group (CCMD-CTR); Headspace Research Database (National Youth Mental Health Foundation, Australia. Searches were restricted to 1960–2017 and English language only. The date of the last search was 21stst May 2019. Reference lists of all included studies were also hand-searched to identify further relevant studies.
A comprehensive search strategy was developed and all relevant bibliographic databases were searched with terms modified for each specific database. Search strategies are included in S1 Fig.
Each paper was identified as eligible for inclusion by at least two reviewers. Three reviewers independently screened all abstracts identified in the initial search and excluded studies that did not meet inclusion criteria. Full-text articles were subsequently reviewed in duplicate, and in cases of disagreement consensus was achieved through referral to a senior reviewer (SP or PF).
Data collection process
Seven categories of data were extracted using a standardized data extraction form. All data items were double extracted.
The following data items were extracted: demographic and clinical characteristics of the sample; programme type (selective or indicated; we included treatment interventions in the indicated category because inclusion criteria for these two types of interventions are often very similar, e.g. scoring above a certain value on a symptom severity scale); programme content including manualization, mode of delivery, duration and intensity of the intervention (that is the time over which the intervention was provided and the total time spent in delivering the intervention); comparator type (treatment as usual/waitlist/attentional control/no treatment control or active comparator), content, mode of delivery and duration of the comparator; intervention location (US or non-US); intervention setting (school, community or clinic setting); intervention agent (teacher, professional or paraprofessional); and studies’ methodological characteristics (see quality assessment below). Based on expert consensus a hierarchy of preferred outcomes and a method for identifying outcomes in studies reporting multiple outcomes was specified for each disorder prior to data extraction of outcome measures (see S2 File). This determined the extraction of outcomes at baseline 12-month follow up, and at post-intervention where available.
Risk of bias
The Cochrane Risk of Bias tool was used to assess the methodological quality of the eligible studies . The impact of publication bias and heterogeneity was assessed by visual assessment and statistical analysis of funnel plots . We also assessed the impact of date of publication on the study outcome. All methods were considered in the interpretation of the results.
We calculated overall summary estimates and 95% CIs with a random-effects meta-analysis, which is to be preferred when there are high levels of heterogeneity , using Comprehensive Meta-Analysis software (CMA V3). Hedges’ g was used as a summary statistic to facilitate comparisons within and between disorders. The majority of trials reported continuous outcomes (123/138 at follow-up, 99/99 at end of intervention); where this was not the case dichotomous outcomes (odds ratios) were converted to Hedges’ g values.
We conducted subgroup analyses by performing a series of separate meta-analyses to explore the associations between each of a range of moderators alone and by disorder, (see Table 2A for a complete list of all moderators) and ESs at post-intervention and 12-month follow-up. Subgroup analyses were conducted using a random-effects ANOVA, which partitions the variance (Q) into within-study (QW) and between-study (QB) components using random-effects weights, and is equivalent to the meta-regression approach with binary indictors (Ref: https://www.meta-analysis.com/downloads/Meta-analysis%20Subgroups%20analysis.pdf). We did not assume a common within-study variance across levels of the moderator/subgroups because of the likelihood of substantial heterogeneity. We used the QB variance component (equivalent to QM omnibus test in meta-regression) to determine whether the effect size was differentially associated with different levels of a moderator and compared the direction of significant between levels using confidence intervals.
We reported change scores (K = 111 from 99 studies at end of intervention and K = 165 from 138 studies at follow up) and adjusted for baseline scores inserting a correlation of 0.75. We considered CIs that did not overlap the line of no effect to be statistically significant and a Hedges’ g of 0.2 or greater to be of clinical importance . The heterogeneity between studies was calculated using the heterogeneity I2 statistic where an estimate above 40% suggests presence of heterogeneity .
All analyses were done using CMA V3. We chose to use Egger’s test of bias rather than Orwin’s failsafe N because Orwin’s test is not available for a random effects meta-analysis in CMA V3.
A total of 19,781 reports were identified in the initial search from which 3,811 were removed as duplicates. 15,970 titles and abstracts were then reviewed, identifying 863 potential studies for inclusion. The reviewers independently screened the full text of these and excluded 735 that did not meet inclusion or met exclusion criteria. This resulted in 128 treatment trials of psychosocial interventions where 12-month follow-up data were available. This search was supplemented with an update search conducted 21/05/19, which retrieved an additional 2800 records, of which 134 studies were screened, with ten additional studies meeting inclusion criteria, resulting in a total of 138 studies included in the review. The systematic review process is depicted in Fig 1.
Summary study characteristics are presented in Table 1. At baseline the studies included a total of 14,954 participants. Sample sizes varied widely (min 20, max 1,730). The 138 included studies yielded 165 comparisons containing 12-month follow-up data which were the focus of this analysis.
58 (35%) interventions had a significant CBT component, 48 (29%) were family or parenting based, 12 (7%) were psychoeducation or psychotherapeutic, 28 (17%) were combined interventions, and 19 (11%) were ‘other’. 113 (68%) were led by mental health professionals, 51 (31%) by paraprofessionals (school professionals or non-mental health professionals with intervention-specific training). Length of programmes varied from 1 to 144 sessions (median 12). Over 80% of outcomes measures were either self or parental report. 101 (61%) studies reported a method for assessing treatment fidelity. The most common disorders were conduct disorder (44 studies or 27%) and anxiety disorders (43 studies or 26%). Depressive disorders (29 studies or 18%) and substance misuse (27 studies or 16%) were also relatively common. Less common were eating disorders (12 studies or 7%) and PTSD (9 studies or 5%). The distribution of each study variable differed across disorders (see S1 Table).
Risk of bias within studies
The methodological quality of the studies as assessed by the Cochrane Risk of Bias tool varied considerably (see S2 Table). Generally, there was a high risk of bias, only 28 studies (20%) had relatively low risk of bias (i.e. high risk of bias in no more than one domain) though a further 70 had high risk of bias estimates in 2 domains. Almost half (47%) of all studies achieved low risk of bias ratings in only 2 or less domains.
Results of individual studies
Fig 2 presents the forest plots for each disorder, showing Hedges’ g with 95% confidence intervals for the intervention and control groups at 12-month follow-up.
Synthesis of results
Meta-analyses were conducted to compare intervention and control groups across all disorders at post-intervention and 12-month follow-up. Overall effect size (ES) post-intervention was moderate (K = 115, g = 0.39; 95% CI: 0.30–0.47 I2 = 84.19%, N = 13,982). The overall ES was small to medium at 12 months follow-up (K = 165, g = 0.31, CI: 0.25–0.37, I2 = 77.31%, N = 25,652) (see Table 2 and Fig 2). A number of studies only reported 12-month follow-up data (K = 39). Excluding these studies, the ES at 12-month follow-up was slightly but not significantly higher (K = 115, g = 0.36, CI: 0.28–0.43 I2 = 78.88%). Across diagnostic groups there were small to medium statistically and clinically important effects at end of intervention (range from g = 0.19, 95% CI:0.01–0.38, I2 = 78.47% for substance misuse to g = 0.66, CI: 0.28–1.03 I2 = 67.75% for PTSD). These effects were largely maintained at 12-month follow-up with no overall statistically significant decline (range from g = 0.21 CI:0.10–0.32, I2 = 61.36% for depression to g = 0.51 CI: 0.34–0.68 I2 = 81.29% for anxiety disorders) although there was a more marked decline in the case of depression and PTSD. An overall effect of date of publication was identified with ESs declining for more recent publications for end of intervention (Q = 10.08, df = 2, p = 0.006) and follow-up (Q = 16.92, df = 2, p<0.001; see Table 2). It should also be noted that the I2 statistic was generally high throughout these analyses which probably reflects heterogeneity in trial populations and interventions types and supports the exploratory approach we took to sub-group analyses in this review. We also explored whether the heterogeneity in the analyses could be explained by risk of bias by comparing low risk of bias studies (that is, those with 2 or less ratings of “high risk of bias”) with those with higher risk of bias. Across disorders heterogeneity generally remained high, between 60% and 86% in analyses of low risk of bias studies which suggests that risk of bias is not a substantial contributor to heterogeneity in this review. We identified 5 potential studies which might include data on self-harm, of which only 2 reported relevant outcomes at 12 months. These studies were however excluded as the populations in the studies were outside the scope of the review.
Analyses of between-group differences identified a number of potential associations (see Table 2). In particular, at follow-up interventions for under 12 years of age, anxiety and eating disorders and interventions of moderate intensity had higher, but not significantly so, ESs.
Greater specificity was achieved when studies of specific diagnostic groups were analysed separately. Analyses at 12-month follow-up are shown in Tables 3 and 4. Analyses at end of treatment are provided in S3A and S3B Table. For conduct disorders outcomes were maintained at follow-up (g = 0.23 95% CI 0.14–0.33, see Table 2). Group-based CBT was associated with negative outcomes (g = -0.27, 95% CI -1.87–1.33) and mixed group and individual interventions were somewhat worse than individual treatments QB(1) = 6.93, p = .008). For CD professionals may do better, although not significantly, than paraprofessionals (professional: g = 0.32, 95% CI 0.18–0.47) paraprofessional: g = 0.15, 95% CI 0.01–0.37; QB(1) = 3.03, p = .220)).
The outcome at follow-up for substance abuse interventions appears promising as there is no observed decline in ES (g = 0.19, 95% CI 0.01–0.38 at end of intervention and g = 0.26, 95% CI 0.15–0.36 at follow up). In substance misuse disorders; family-based interventions (g = 0.53, 95% CI 0.06–1.00) appear to be most effective and effects also appear somewhat stronger for those of moderate intensity (g = 0.53 95% CI 0.25–0.82) and those delivered by professionals (g = 0.32 95% CI 0.18–0.46).
Interventions for anxiety disorders hold up well from end of treatment (g = 0.61 95% CI 0.34–0.89) to follow-up (g = 0.51 95% CI 0.34–0.68) (Table 4). At follow-up individual CBT/BT (g = 0.67 95% CI 0.30–1.05) appears to be associated with larger effects. Moderate intensity interventions (g = 0.71 95% CI 0.41–1.01) appear more effective than interventions of low or high intensity. Effects for interventions delivered by paraprofessionals (g = 0.83 95% CI 0.36–1.30) had a greater but not significant than those delivered by professionals (g = 0.44, 95% CI 0.26–0.62).
For depressive disorders ESs declined post intervention (g = 0.38 95% CI 0.24–0.53) to follow-up (g = 0.21 95% CI 0.10–0.32) but a clinically important effect was still present. With regard to setting, interventions provided in schools (g = 0.32, 95% CI 0.12–0.51) and clinic settings (g = 0.21, 95%CI 0.05–0.37) may be more effective than community settings (g = 0.02, 95%CI -0.19–0.23).
No sub-group analyses were performed for eating disorders or PTSD due to limited study numbers.
The funnel plot for all disorders at follow-up showed evidence of considerable asymmetry indicating publication bias (see Fig 3), which was confirmed by an Egger’s test of bias  (1.65, p < .001, 95% CI [0.99, 2.30]). It should be noted that the considerable heterogeneity in our analyses may also be a major contributing factor to the asymmetry . When we produced funnel plots for each disorder separately, the asymmetry was less pronounced (Egger’s range: 0.34–1.57, all p > .05), with the exception of anxiety (2.74, 95%CI 0.99–4.50, p = 0.003) and depressive disorders (1.40, 95%CI -0.06–2.86, p = 0.060;). Correction for this bias using the trim-and-fill method did not alter the estimates.
This is the first meta-analysis to examine the long-term outcomes of psychosocial interventions for children and young people across most common mental health disorders. The meta-analysis included 138 studies representing 165 comparisons with 12-month follow-up continuous data on psychological interventions. The benefits we identified were typically obtained against standard care or other active treatments and therefore represent additional benefits over that gained from no care, which remains the experience of many children and young people with common mental disorders .
Notwithstanding the variability in ES, the heterogeneity in outcomes and the limited number of studies, a broadly consistent picture emerged of sustained, longer-term, and generally small to medium-size benefits against active control interventions. Younger children (under 12) may obtain greater benefit than older children at follow up. There is some indication that interventions delivered by paraprofessionals may be more effective in anxiety disorders equivalent for depression but less effective than those delivered by professionals for conduct disorder and substance misuse. Paraprofessional effectiveness is likely to be enhanced when training programmes are focused on specific interventions, targeted on less severe disorders and supported by appropriate training, continuing supervision and outcome monitoring . Parent training for conduct disorders and family-based interventions for substance misuse appeared effective. There was some evidence to suggest that both family and parenting interventions might be effective in depression and anxiety disorders; given the preponderance of CBT interventions for these disorders consideration should be given to further research and development of these interventions for children and young people with depression and anxiety disorders. Group-based approaches may be effective for depressive and anxiety disorders but may be contra-indicated for conduct disorders. Moderate intensity of intervention appears to be associated with larger effects across all disorders. This resonates with Mulley and colleagues’ view that more care does not necessarily mean better care . Like previous investigations , we found that in the school setting indicated interventions appeared as effective as other settings across all disorders. Unlike Brunwasser and colleagues  we found no evidence to suggest there may be consistent differences between programmes delivered in schools and those delivered in other settings. The lack of relationship between intervention fidelity to predefined protocols and outcome may be due to the fact that such measures are common to more recent studies, which also have lower ESs associated with improved design. It should also be noted that over 80% of studies included a supervision component which is seen as an essential part of effective psychological practice .
This review’s positive picture of long-term benefits is supported by Kodal and colleagues’ recent cohort studies  which assessed young men with a range of anxiety disorders for a mean of 3.9 years post treatment and demonstrated maintenance of treatment effects. Some of our included studies reported outcomes beyond 12 months, suggesting that effects were maintained beyond this point, but there were too few to incorporate in the meta-analysis and the likely increased use of intercurrent treatments beyond 12 months complicates both the design and interpretation of long-term follow up studies. Here there is a contrast with psychological and pharmacological interventions for a number of adult disorders, where the effectiveness of treatments across a range of disorders (e.g. depression ) show a relapsing and remitting course which is evident at 12-month follow-up.
This review suggests that a modest, persistent effect likely reflects meaningful improvements at population level in ameliorating and preventing the onset of disorders in young people and adults. Meta-analytic studies of prevention programmes support this view . Whilst we know of no other studies that explore the long-term outcome of selective or indicated interventions, the ESs observed are broadly comparable to those in similar reviews focused on short-term outcomes for depression and anxiety [11, 32–34]. This review reinforces the importance of providing effective interventions for children and young people; doing so offers potential long-term benefits which may reduce the burden of mental disorders in adulthood and better enable children and young people in their educational and social worlds which are important in ensuring better mental and physical health. The potential long-term benefits identified by this review provided support for a major national initiative to increase the availability of psychological interventions for children and young people in the English National Health Service (15).
The review has a number of limitations. The high level of heterogeneity in most analyses is a limitation that reflects variability in populations and methods that our exploration of intervention parameters did not capture. It may also reflect some studies’ use of less robust diagnostic measures and inclusion of participants with comorbid disorders. These factors, along with the moderate to high risk of bias characterizing most studies and the evidence of potential publication bias, mandate caution in interpreting the results and greater rigour in the design and reporting of future studies. Baseline severity could not be established due to the wide range of measures and in some cases lack of standardization and again limits the interpretation of these studies. The exclusion of drug interventions led to the exclusion of ADHD and studies for other diagnostic groups which only included drugs as the active comparator. The limitation of studies to those from OECD countries warrants some caution in the interpretation of the results particularly those concerning service delivery systems which might be differently configured in low- and middle-income countries.
Our analyses identify a number of important findings which could be the focus of further research. These include that the interventions could be provided in varying settings, including schools, and that interventions for anxiety and depression may be delivered by professionals or paraprofessionals without diminishing the magnitude of effect, although this may not hold true for substance use and conduct disorders. Importantly, our review suggests that younger children may obtain a greater benefit and that effective parent and family involvement is an important component of effective care. However, it should be noted that these interventions have been provided in the context of protocol-driven and well-supported and supervised care. These are essential aspects of any future research or implementation programme. We did not review any health economic outcomes but further research, and in particular any implementation studies, should consider cost-effectiveness. The absence of sufficient long-term data on self-harm is of particular concern given the high prevalence of this problem in young people, high-quality studies with long-term outcomes are urgently needed. The findings of our review suggests interventions should be provided early, under 12 if possible. It is also important to follow a well-described manual as was the case for most of the studies in this review. As almost all of the studies included supervision of implementers, ensuring effective support and supervision for the interventions may be necessary to achieve the outcomes observed. Future research across all disorders should report long-term outcomes (at least 1 year), including for self-harm and suicide prevention, and given that the effectiveness at end of treatment and follow-up has been established the use of waitlist controls should be discouraged.
Few, if any, systems with these characteristics commonly exist in routine practice and none have been robustly tested. Establishing new models of care and testing these models in large-scale implementation studies would be an important first step.
S2 File. Extraction and data analysis guidelines.
S4 File. List of reports of studies included in the review.
S1 Table. Observed frequencies for each study variable by disorder and associated chi-squared tests.
S2 Table. Risk of bias for studies included in the meta-analysis.
a. Subgroup analysis at end of intervention for conduct and substance disorders. b. Subgroup analysis at end of intervention for depressive and anxiety disorders.
- 1. Costello EJ, Copeland W, Angold A. Trends in psychopathology across the adolescent years: what changes when children become adolescents, and when adolescents become adults? Journal of Child Psychology & Psychiatry. 2014;52(10):1015–25.
- 2. Kessler R, Berglund P, Demler O, Jin R, Merikangas K, Walters E. Lifetime Prevalence and Age-of-Onset Distributions of DSM-IV Disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry. 2005;62(6):593–602. pmid:15939837
- 3. Jones PB. Adult mental health disorders and their age at onset. British Journal of Psychiatry. 2013;202:5–10. pmid:23288502
- 4. Briggs-Gowan MJ, Owens PL, Schwab-Stone ME, Leventhal JM, Leaf PJ, Horwitz SM. Persistence of psychiatric disorders in pediatric settings. Journal of the American Academy of Child & Adolescent Psychiatry. 2003;42(11):1360–9. pmid:14566174
- 5. Barkmann C, Schulte-Markwort M. Prevalence of emotional and behavioural disorders in German children and adolescents: a meta-analysis. Journal of Epidemiology and Community Health. 2012;66:194–203. pmid:20889591
- 6. Merikangas KR, He JP, Burstein M. Lifetime prevalence of mental disorders in U.S. adolescents: results from the National Comorbidity Survey Replication—Adolescent Supplement (NCS-A). Journal of the American Academy of Child and Adolescent Psychiatry. 2010;49:980–90. pmid:20855043
- 7. Ahlen J, Lenhard F, Ghaderi A. Universal Prevention for Anxiety and Depressive Symptoms in Children: A Meta-analysis of Randomized and Cluster-Randomized Trials. J Prim Prev. 2015;36(6):387–403. pmid:26502085
- 8. Hodder RK, Freund M, Wolfenden L, Bowman J, Nepal S, Dray J, et al. Systematic review of universal school-based 'resilience' interventions targeting adolescent tobacco, alcohol or illicit substance use: A meta-analysis. Prev Med. 2017;100:248–68. pmid:28390835
- 9. Le LK, Barendregt JJ, Hay P, Mihalopoulos C. Prevention of eating disorders: A systematic review and meta-analysis. Clin Psychol Rev. 2017;53:46–58. pmid:28214633
- 10. Dray J, Bowman J, Campbell E, Freund M, Wolfenden L, Hodder RK, et al. Systematic Review of Universal Resilience-Focused Interventions Targeting Child and Adolescent Mental Health in the School Setting. J Am Acad Child Adolesc Psychiatry. 2017;56(10):813–24. pmid:28942803
- 11. Werner-Seidler A, Perry Y, Calear AL, Newby JM, Christensen H. School-based depression and anxiety prevention programs for young people: A systematic review and meta-analysis. Clin Psychol Rev. 2017;51:30–47. pmid:27821267
- 12. Neufeld SA, Jones PB, Goodyer IM. Child and adolescent mental health services: longitudinal data sheds light on current policy for psychological interventions in the community. Journal of Public Mental Health. 2017;16(3):96–9. pmid:29721033
- 13. Weisz JR, Kuppens S, Ng MY, Eckshtain D, Ugueto AM, Vaughn-Coaxum R, et al. What five decades of research tells us about the effects of youth psychological therapy: A multilevel meta-analysis and implications for science and practice. Am Psychol. 2017;72(2):79–117. pmid:28221063
- 14. Fonagy P, Cottrell D, Phillips J, Bevington D, Glaser D, Allison E. What Works for Whom? A Critical Review of Treatments for Children and Adolescents. New York: Guilford Press; 2014.
- 15. Department of Health, Department of Education. Transforming Children and Young People’s Mental Health Provision: a Green Paper: Crown copyright; 2017.
- 16. Dunn V, Goodyer IM. Longitudinal investigation into childhood‐and adolescence‐onset depression: Psychiatric outcome in early adulthood. British Journal of Psychiatry. 2006;188:216–22. pmid:16507961
- 17. Ginsburg GS, Becker-Haimes E, Keeton C, Kendall PC, Iyengar S, Sakolsky D, et al. Results From the Child/Adolescent Anxiety Extended Long-Term Study (CAMELS): Primary Anxiety Outcomes. J Am Acad Child Adolesc Psychiatry. in press. pmid:29960692
- 18. Higgins JP, Green S. Cochrane Handbook for Systematic Reviews of Interventions. Chichester: Wiley; 2008.
- 19. Duval S, Tweedie R. Trim and Fill: A Simple Funnel-Plot–Based Method of Testing and Adjusting for Publication Bias in Meta-Analysis. Biometrics. 2000;56:455–63. pmid:10877304
- 20. Hedges LV, Vevea JL. Fixed-and random-effects models in meta-analysis. Psychological Methods. 1998;3(4):486.
- 21. Hedges LV. Distribution theory for Glass' estimator of effect size and related estimators. Journal of Educational Statistics. 1981;6(2):107–28.
- 22. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta‐analyses. British Medical Journal. 2003;327(7414):557–60. pmid:12958120
- 23. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. Bmj. 1997;315(7109):629–34. pmid:9310563
- 24. Sterne JAC, Sutton AJ, Ioannidis JPA, Terrin N, Jones DR, Lau J. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. Bmj. 2011;343:d4002. pmid:21784880
- 25. Costello EJ, He JP, Sampson NA, Kessler RC, Merikangas KR. Services for adolescents with psychiatric disorders: 12-month data from the National Comorbidity Survey–Adolescent.. Psychiatric Services. 2014;65(3):359–66. pmid:24233052
- 26. Mulley A, Coulter A, Wolpert M, Richards T, Abbasi K. New approaches to measurement and management for high integrity health systems. British Medical Journal. 2017;356:j1401. pmid:28360140
- 27. Clark DM. Realizing the mass public benefit of evidence-based psychological therapies: the IAPT program. Annual Review of Clinical Psychology. 2018 May 7;14. pmid:29350997
- 28. Brunwasser SM, Garber J. Programs for the Prevention of Youth Depression: Evaluation of Efficacy, Effectiveness, and Readiness for Dissemination. J Clin Child Adolesc Psychol. 2016;45(6):763–83. pmid:25933173
- 29. Herschell AD, Kolko DJ, Baumann BL, Davis AC. The Role of Therapist Training in the Implementation of Psychosocial Treatments: A Review and Critique with Recommendations. Clinical Psychology Review. 2010;30(4):448–66. pmid:20304542
- 30. Kodal A, Fjermestad K, Bjelland I, Gjestad R, Ost LG, Bjaastad JF, et al. Long-term effectiveness of cognitive behavioral therapy for youth with anxiety disorders. J Anxiety Disord. 2018;53:58–67. pmid:29195188
- 31. Vittengl JR, Clark LA, Dunn TW, Jarrett RB. Reducing relapse and recurrence in unipolar depression: a comparative meta-analysis of cognitive-behavioral therapy's effects. Journal of Consulting and Clinical Psychology. 2007;75(3):475–88. pmid:17563164
- 32. Stockings EA, Degenhardt L, Dobbins T, Lee YY, Erskine HE, Whiteford HA, et al. Preventing depression and anxiety in young people: a review of the joint efficacy of universal, selective and indicated prevention. Psychol Med. 2016;46(1):11–26. pmid:26315536
- 33. Merry SN, Hetrick SE, Cox GR, Brudevold-Iversen T, Bir JJ, McDowell H. Psychological and educational interventions for preventing depression in children and adolescents. Cochrane Database of Systematic Reviews. 2011;7(12):CD003380. pmid:22161377
- 34. Fisak BJ Jr., Richard D, Mann A. The prevention of child and adolescent anxiety: a meta-analytic review. Prev Sci. 2011;12(3):255–68. pmid:21437675