If the 2020 PRISMA guidelines28 are not available shortly, then the 2009 PRISMA guidelines29 will inform our systematic review. The development of this protocol was informed by Moher et al.30 and Shamseer et al.31 This review protocol has been registered in PROSPERO (CRD42020176664).
Regarding study characteristics, only original studies (human or animal) will be included. The human studies are inclusive of foetuses (i.e., conception to birth), children (i.e., new-born to nine years of age), and adolescents (i.e., 10 to 24 years; see Sawyer et al.25) from any country. The animal studies are inclusive of animals exposed to pollutants at the same developmental stages. In instances where the same data set is reported in multiple articles, the article that provides the clearest evidence of pollution impacts on/associations with child and adolescent mental health will be included. Studies will be excluded if:
- reference to pollution or mental health is tangential (e.g., only in the introduction or recommendations);
- mental health was not formally measured (e.g., through a clinical interview, scale/checklist, self- or caregiver-report, animal behaviour task, or biomarker [animal studies only]);
- intervention is the focus (e.g., interventions designed to reduce pollution); or
- solvents or pesticides are reported in relation to using these for suicide or substance-abuse.
Regarding report characteristics, only peer-reviewed, indexed journal articles (published and pre-print, on-line) will be included. To fit the profile of the reviewer team, publications in languages other than Cantonese, English, Mandarin, Portuguese, or Spanish will be excluded.
Peer-reviewed articles, published on or before 10 April 2020, will be retrieved using the following databases: Africa-Wide, CINAHL, PsycARTICLES, PsycINFO (all via EBSCOhost platform); Medline (via Web of Science Clarivate Analytics); PubMed; Scopus (which includes contents of Embase); Web of Science Core Collection; SciELO Citation Index; and ERIC. The database search will be supplemented by a manual search of the reference lists of well-cited articles identified in the database search. The reviewer team (of which researchers from the Global South comprise the majority) will be sensitive to the inclusion of indexed Global South studies given their historic under-representation in scholarly literature32.
In order to identify relevant article publications, the search strategy listed in Table 1 will be followed. It was developed by author LS1, a librarian working in South Africa, and tested by LS2, a research assistant working in the United Kingdom. The search will be limited to the title, abstract, and topic fields for the databases that allow this limiting.
Articles meeting the search strategy will be populated into EndNote and then exported to Zotero to annotate (i.e., to formulate as citations that include the title and abstract).
Annotated citations containing the search terms in the title and/or abstract will be independently screened for eligibility by 11 reviewers (LT, YAV, CB, MECL, GPA, MAO, LG, LL, IM, AT, KH). Initially, the first 50 titles and abstracts will be independently screened by all 11 reviewers to determine the clarity of the inclusion/exclusion criteria. Once the criteria have been refined (if/as necessary), the remaining annotated citations will be randomly divided into five sets. Each set will be independently reviewed by groups of 2 reviewers (with one group of 3, due to the odd number). Each of these groups includes at least one reviewer from the Global South. Those articles that meet the inclusion criteria and to which the exclusion criteria do not apply—as well as those where this information is not clear from the title or abstract—will be selected for a full text review. Consensus discussions (see Saldana33) will be used to resolve any disagreements regarding article selection for full assessment. Should consensus not be reached, LT or KH will evaluate the disputed citation.
Data collection process
The same 11 reviewers will extract relevant data from the identified articles. Reviewers will again work as five groups. Identified articles will be divided between the groups. Within groups, reviewers will independently extract data from the articles assigned to that group. Data extraction will be guided by a data-charting form that will be developed by LT and KH and calibrated by all reviewers (using 10 of the identified articles). The data-charting form will correspond to the items for which data will be sought (see data items). Once a reviewer group has reached consensus on the accuracy of the extracted information, another reviewer group will verify the data extracted from 10% of the completed group’s articles.
We will extract data for the following categories of information: study descriptors [country context; design/methods (including how mental health was measured); sample (size, age range of human participants or animal subjects, sex of participants/subjects); study limitations]; pollution factors [type, level/intensity, duration of exposure]; and factors/processes that enable/sustain child and adolescent mental health despite exposure to pollution.
Outcomes and prioritization
We will extract data that relate to mental health outcomes in the face or aftermath of exposure to pollution. This includes data relating to symptoms of: (i) neurodevelopmental disorders (e.g., intellectual disability, communication disorders, autism spectrum disorder, attention-deficit and/or hyperactivity disorder, or specific learning disorders); (ii) disruptive, impulse-control, and conduct disorders; (iii) depressive disorders; (iv) anxiety or post-traumatic stress disorders; (v) substance disorders; or schizophrenia. Where possible, the extent of symptoms (e.g., no, limited/mild, severe) will be noted too.
Rigour and risk of bias in individual studies
Three review authors will independently assess the risk of bias in the included quantitative studies by considering the following characteristics34: 1. Completeness of data: were participant exclusions, attrition, and incomplete data adequately addressed in the published report? 2. Selective reporting: is there evidence of selective reporting of analyses (quantitative studies) and might this have affected the published results? 3. Reliability and validity: do the included measures have internal consistency? Have measures used been validated for use in the populations or languages in which they were administered? 4. Participant selection: were the participants sampled in such a way that the results of the study can be generalized to the population? To assess risk of bias in animal studies, SYRCLE’s risk of bias tool for animal studies35 will be applied by three review authors with animal study expertise. SYRCLE’s tool is similar to the Cochrane risk of bias tool specific to randomized control trials.
There is little consensus on which quality appraisal tool is preferred for judging the rigour of qualitative studies and so the decision is generally informed by the objective of the systematic review and reviewer experience with qualitative methodologies36. Given that the author team includes experienced qualitative researchers and that the eligible qualitative studies are likely to include a broad range of qualitative methodologies, the Standards for Reporting Qualitative Research (SRQR37) is preferable36. The SRQR is a 21-item checklist that evaluates all aspects of qualitative studies, including problem formulation, research design, findings and interpretation of findings.
Disagreements between the review authors over the risk of bias and quality limitations in individual studies will be resolved by discussion, with involvement of two additional review authors where necessary. Following Hughes-Morley et al.38, we will not omit any studies that demonstrate bias or limited quality. Instead we will de-emphasise the results of these studies in the conclusions of our synthesis.
Because we anticipate that studies will not be sufficiently homogenous to accommodate meta-analyses, the results will be tabulated and narratively synthesized. An advantage of narrative syntheses is their juxtaposition of quantitative and qualitative evidence in order to provide a detailed response to the question directing the review39. This juxtaposition will yield a detailed account of what is currently known about how pollution impacts on, or relates to, child and adolescent mental health worldwide, as well as what facilitates child and adolescent mental health resilience in the face or aftermath of pollution exposure. This will be used to signpost limitations and silences in current understandings of child and adolescent mental health during/following exposure to pollution and to advocate for specific research and practice agendas. To ensure replicability of the narrative synthesis, we will make public the completed data-charting forms that informed the synthesis (e.g., as supplemental, on-line files when the review is published). LT and KH will lead the synthesis, with input from the remaining reviewers.
Meta-bias includes both the selective reporting of outcomes due to their significance, magnitude, or direction and publication bias31. To asses meta-biases due to selective outcome reporting, we will: 1) evaluate whether studies have associated protocols, and whether those protocols were published prior to the recruitment of participants; 2) look for discrepancies between the published article and protocol (for those studies with a protocol); and 3) contact authors of the study, where additional information is needed. We do not plan any assessment of meta-biases due to publication bias.
Confidence in cumulative evidence
To assess confidence, we will apply the Grading of Recommendations Assessment, Development and Evaluation (GRADE40) and GRADE-CERQual41–42.