A PRISMA-P checklist is included in the Additional file 1, and this protocol is registered on Prospero (ref CRD42021233924).
Children and young people age 3–12 where ADHD or any of the core symptoms (impulsivity, inattention or hyperactivity) are mentioned in the title or abstract. Studies solely including “typically developing” populations will be excluded, as will studies of populations with acquired brain injury.
Any non-pharmacological intervention or strategy, for example interventions or strategies may be psychological, cognitive, behavioural, or aim to change the learning environment. Interventions that intervene with diet or nutrition (including dietary supplements) will be excluded. Interventions requiring specialist equipment not commonly available to schools (e.g. EEG or specialist animals such as horses) will be excluded. Studies will be excluded if the intervention clearly requires a trained therapist to deliver it. Studies will not be excluded if young people are taking medication (for ADHD or other conditions). Studies that report solely on parent training, with no child-delivered component or school-relevant outcomes will be excluded, as will studies of the overarching effect of “summer treatment programmes” as these are intensive extended intervention programmes. A separate review of summer treatment programmes for ADHD is planned.
For group designs, the comparator will be no intervention or treatment as usual (including medication). For within-subject designs, the comparator will be each participants’ baseline data.
Eligible studies will provide pre- and post- measures that capture a facet of at least one of the following child-focussed outcomes. Some outcomes may overlap, for example planning is an executive function, but also an organisational skill. Outcome measures may be recorded by parents, teachers, the young person or through observational measures or computer-assisted technology. We will include studies of both (i) interventions that directly target the outcomes, and (ii) studies that measure any of these outcomes although they were not the explicit focus of the intervention.
Inattention (and attention but not “joint attention” referring to eye-gaze in autism studies)
Organisational skills, including organised actions, time management and task planning 
Teacher conflict—or relationships with teachers
Peer conflict—or relationships with peers, and other indicators of social impairment such as friendships, peer relationships and bullying
Executive functioning, including working memory, inhibitory control, shifting between tasks or cognitive flexibility, emotional self-regulation, initiating, planning and problem solving, self-monitoring 
Classroom functioning, e.g. task-relevant behaviours, communication, engagement, social skills, leadership skills, learning problems, problem behaviours, academic competence, classroom physical environment [35, 36]
Global functioning and quality of life
Types of study to be included
For research question 1, eligible studies will report changes in any of the outcomes listed above following non-pharmacological intervention and include control or comparison data. Systematic reviews that include studies meeting the above criteria, randomised and non-randomised controlled trials will be included for the quantitative synthesis. In addition to these study designs, controlled before and after studies, case-control studies or case series (including multiple baseline designs) are eligible study designs for research question 2. Individual study participants can be their own controls, e.g. using pre-intervention baseline data. Studies that include one intervention group that has the same measure pre- and post-treatment will be eligible for inclusion, as long as there are data available for a comparative ‘control’ period pre-treatment. Qualitative studies and individual case studies will be excluded from the synthesis, although they may inform the intervention mapping analysis. Process evaluations will be eligible study designs for research question 2 where they report on an intervention study that meets inclusion criteria. Where there are existing high quality systematic reviews that match our inclusion criteria, we will include these reviews and only synthesise primary research studies published since the review search dates. It is anticipated that most existing systematic reviews will not match our aims precisely, in this case individual eligible studies within reviews will be included.
Given the ultimate aim of extracting detailed intervention information and plans to request intervention manuals from authors, we have made the pragmatic decision to include only studies published in English from 2000 to the search date (late 2020). This is based on experience from prior efforts to contact authors of school-based ADHD interventions; authors publishing more than two decades ago were often not traceable, or reported no longer having intervention manuals or further information. Prior to final analysis, searches will be re-run to identify any further recent publications. Conference abstracts will be excluded, and dissertations will be included. Forward and backward citation searching of included articles will also be conducted.
Although the ultimate aim is to translate evidence to a school setting, eligible studies can be in any setting (home, school, after-school, hospital school, or clinical settings). As part of the scoping of literature, searches and title and abstract screening will include children and adolescents up to the age of 18; however, given the volume of eligible literature and that studies conducted with older age groups may have limited translatability to primary school settings, full-text screening will include only studies of children aged 3–12 years as the eligible study population. Following full-text screening, if there are insufficient studies in this younger population (n < 3 for any of the outcomes listed above), full texts will be re-screened and studies of secondary school pupils (age 12–18 years) will be included. Depending on the volume of evidence available for each outcome, studies based in the school setting will be prioritised over those in other settings.
The following databases and registries will be searched: Medline, PsycINFO, Australian Education Index, ERIC, Education research complete, British Education Index, Embase, Health Management Information Consortium, Social Policy and Practice (via OvidSP); ASSIA, Social Sciences Citation Index, Conference Proceedings Citation Index; Conference Proceedings Citation Index–Social Science and Humanities.
In addition, the following databases will be screened for relevant studies: The Cochrane Library, Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE), Cochrane Central Register of Controlled Trials (CEN-TRAL), Cochrane Methodology Register (CMR), Health Technology Assessment (HTA) and The Campbell Library ISRCTN Registry. An example draft search strategy for PsycINFO can be seen in Table 2. Searches will have two steps, the first step will include ADHD terms, AND at least one of the other outcome terms (combined with OR), AND intervention terms. The second search will include ADHD terms AND intervention terms and exclude those results already present from step 1. This second step will capture studies that measure only the ADHD-specific symptom outcomes but none of our other outcomes of interest. Records from the two steps will be combined to form the search results. Scoping searches revealed that these terms are very broad: to focus the screening, additional NOT terms were applied to the records following all database searches (see Table 3).
Additional search strategies include forward and back citation chasing, contacting experts and searching through key journals (where journals or authors are highly recurrent in included studies). Potentially eligible trials identified through trial registrations will be followed-up to identify relevant publications, and where trials appear ongoing authors will be contacted to enquire if they have relevant data to contribute.
Two reviewers will screen titles and abstracts, and then full-text records to assess study eligibility. Disagreements will be resolved through discussion, with a third reviewer (DM) if necessary. Screening and search records will be kept following a PRISMA diagram, with reasons recorded for studies that are excluded at full-text screening. Because scoping searches indicated a very high number of records would be retrieved from database searches (i.e. > 130,000), title and abstract screening will utilise Swift-Review software to minimise burden and facilitate the screening process . This uses a ‘literature prioritisation’ algorithm, text mining and machine learning in order to prioritise screening for abstracts that appear eligible based on decisions made thus far. Piloting will involve reviewers independently making decisions on 100 titles and abstracts (that include eligible studies measuring the full range of outcomes of interest and studies that are ineligible). Reviewers will then screen from the top of the prioritised list, and the library will be re-prioritised frequently throughout screening to enhance the prioritisation algorithm (approximately every 100 records). Title and abstract screening will continue until 100 consecutive records are ineligible, and then after reprioritisation, a further 100 are not eligible. A random 100 records from those remaining will then be screened to ensure all are ineligible prior to concluding title and abstract screening.
Full-text screening will be conducted by three reviewers. All reviewers will screen 10% of the records to assess concordance. Providing reviewers have a high level of agreement (> 95%) on studies meeting inclusion criteria, the remaining full texts will be each screened by one individual. If not, double-screening will continue until concordance is reached. Screening will result in a series of studies meeting inclusion criteria across all outcomes.
Following screening, additional searches will be conducted and authors contacted in order to identify additional data relating to the implementation of included interventions, for example process evaluations, intervention manuals and qualitative analyses.
A data extraction form will be developed and piloted with 10 studies from a range of designs. A separate data extraction form will be used for systematic reviews. Data will be extracted by one reviewer, with data from the first 10% of included studies extracted and checked by a second reviewer. The second reviewer will also scrutinise extraction of data used for calculating effect sizes, intervention descriptions, and categorisation of outcome measures for at least an additional 10% of studies, as these are considered aspects of data extraction where accuracy is critical. If discrepancies are identified, further studies will be double-extracted until consistency is reached.
For individual intervention studies (including eligible studies within systematic reviews), descriptive data on the study sample will be extracted (number of participants, age, sex, clinical characteristics), along with detail of the study year, author, funding source, country, study design, setting, intervention and control group conditions. As much detail as possible will be extracted regarding the nature, content and implementation of the intervention, following the template for intervention description and replication (TIDieR) for describing interventions, which include items on reporting fidelity and implementation . Information will be extracting regarding who delivers the intervention and formats of delivery, dosage and duration of baseline, intervention and follow-up periods. As much detail as possible will be extracted relating to theoretical approaches used, implied or explicitly referred to. Where participants represent a selected group (e.g. those with ADHD), information on diagnostic assessment and current treatments will be extracted where these data are available.
Components of interventions in each study will be paired with the estimate of effect on the outcomes of interest. These are likely to be challenging to disentangle, and for this aspect of the analysis the identified intervention components will be paired with any pertinent outcomes measured. For example, if an intervention includes four components and two relevant outcomes, the effect size estimate will be recorded for each pairing of component and outcome; eight effect sizes in this example (of which the four effect estimates for each outcome will be the same). It is acknowledged that this approach will be somewhat reductionist where interventions are poorly described; however, synthesising these effects across studies should allow for identification of promising components of interventions, to be further evaluated through intervention mapping.
For each intervention-outcome pairing, participant numbers, means and standard deviations (SDs) for outcome measures will be extracted, or a measure of intervention effect with an estimate of precision such as confidence intervals. These will be converted to Hedge’s g, a standardised measure of study effect size that corrects for overestimation of the true population effect . For non-randomised studies of interventions, we will follow Cochrane guidance and extract estimates of effect and precision as well as information about how the effect estimate was derived (e.g. the confounders controlled for) . For high-quality systematic reviews that fully meet our inclusion criteria, similar data will be extracted including intervention descriptions in included studies, as well as lists of included studies.
Risk of bias (quality assessment)
There are multiple tools that can be used to assess risk of bias. The Joanna Brigg’s Institute tools for critical appraisal will be utilised because it has separate quality appraisal tools available for appraisal of systematic reviews, RCTs and non-randomised experimental studies . There are no guidelines of which we are aware to appraise the quality of interventions and implementation, so to appraise the quality of interventions, and the quality of implementation of the intervention, we will develop an appraisal tool that reflects the components of the TiDieR checklist and best-practice recommendations on assessing implementation and fidelity .
Strategy for data synthesis
Data will be synthesised separately for the two research questions. For research question 1, eligible study designs will be systematic reviews, randomised and non-randomised trials (i.e. between-group studies). For research question 2, within-group studies such as case-series and multiple baseline studies will also be eligible for inclusion, along with process evaluations and intervention manuals.
Screening will result in a series of studies meeting inclusion criteria for each of the outcomes. The evidence for each outcome will be synthesised separately, where constructs overlap substantially these will be merged and discussed as pertinent to both outcomes.
Depending on the level of evidence and volume of research available for each outcome, data synthesis will vary. Should there be a large number of eligible systematic reviews, research question 1 may include a synthesis of reviews or a ‘review of reviews’ in addition to the analysis linking components of primary studies to intervention effect estimates; we will use the AMSTAR tool to assess review quality and follow best-practice guidance . For systematic reviews that meet inclusion criteria, we will conduct a separate synthesis to explore whether primary research that has been published since the review date fits the reported findings. Data from interventions included within systematic reviews will be extracted and synthesised in addition. Where it is not possible to include a systematic review, for example if many of the studies in the review are not pertinent to the research questions, individual study details will be extracted from reviews and checked for eligibility.
Where included studies are sufficiently homogeneous, random-effects meta-analysis will be used to assess the cumulative evidence for intervention effectiveness. Heterogeneity will be assessed using the I2 statistic , and funnel plots and Egger’s regression will be used to investigate potential effects of publication bias . We will prioritise quantitative analysis of existing meta-analyses and randomised controlled trials for each outcome, and conduct meta-analysis or quantitative synthesis of other study designs where appropriate, when sufficient systematic reviews (at least 1) or RCTs (at least 5 without a high heterogeneity estimate) are unavailable. Moving beyond the component-outcome analysis, subgroup analyses and meta-regression will be used to examine the influence of specific intervention or study features associated with effectiveness, for example by examining the impact of setting, duration, delivery agent, or type of intervention, the population studied, the country and region of the study setting and the gender of participants on outcomes. The composition and nature of control conditions such as “treatment as usual” will vary depending on the location and design of included studies, and we intend to report a descriptive table detailing clinical guidance for treatment of ADHD for the countries in which included studies are set.
Given that interventions and measures for the pre-school aged population may differ from those who are of school age, we will also assess whether the effectiveness of non-pharmacological interventions varies by the age of the study population. The strength of the body of evidence for each outcome in research question 1 will be assessed using GRADE .
Synthesis of non-randomised group-based study designs can be complex; these studies are subject to a higher risk of bias than randomised designs , and the direction of bias between studies may vary due to different features of study design . However, evidence from non-randomised studies will complement evidence from RCTs and may be the best available evidence for some outcomes, and so if the study designs and measures are sufficiently homogeneous, meta-analyses and meta-regression will be undertaken. Findings from within-subject study designs where each participant provides their own control data will be used to inform the intervention mapping analysis for research question 2.
The findings from the data synthesis will identify and quantify the current evidence relating to whether outcomes of interest can be affected by non-pharmacological interventions for children and young people, and begin to identify broad features of interventions that are associated with effectiveness. These findings will be discussed with a stakeholder group comprised of people with ADHD, parents, school staff and experts in education and psychology in order to discuss the context and implications of the results, and the feasibility of delivering these in primary schools.
Given the anticipated difficulty of establishing whether individual intervention components of multi-component complex interventions have an effect on outcomes, following the initial evidence synthesis to quantify effect sizes, intervention mapping  will be applied to the most promising strategies with the strongest evidence for effectiveness, and studies meeting inclusion criteria that report using these components will be scrutinised. Intervention mapping is a theory- and evidence-based approach to understanding which components of interventions are likely to result in behaviour change, so the theoretical pathway linking each intervention component with the outcome will be elicited. Logic models of behaviour change will be constructed for each promising intervention-outcome component, underpinned by evidence and theory, and components that align with empirical and theoretical evidence will be selected for inclusion in the toolkit . In order to achieve this, intervention manuals, process evaluations and other publications relating to interventions will also be retrieved by contacting authors or via searches, findings and implications from relevant systematic reviews will be integrated, and literature relating to identified theories such as behaviour change and complex systems theories will also be drawn upon.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.