This protocol follows the Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) guidelines  (see Additional file 1) and is registered on PROSPERO, an international register of systematic reviews  (CRD42016052961). Any changes to the protocol will be recorded on PROSPERO.
Participants will be school-aged children and adolescents. Although research into school attendance typically focuses on children aged 5–17 , we expect the exact age range to vary between studies, and we will include any age ranges applicable for the country of study. We will exclude retrospective reports collected during adulthood, and studies where the sample is a specific population not comparable to other study samples, for example, those focusing on children with a particular health condition.
Main study variables
We intend to include any study that investigates the relationship between emotional disorder and school attendance. Based on a previous systematic review on a similar topic  and an initial scoping search, we anticipate that studies will vary in the way they investigate the relationship between these variables, with some treating emotional disorder as the exposure and school attendance as the outcome and others considering poor attendance the exposure and emotional disorder the outcome. Emotional disorder must be assessed by diagnosis with a standardised diagnostic measure or by measures of symptoms using a validated scale. This will allow for inclusion of studies reporting on children with a diagnosed emotional disorder, as well as those with higher levels of emotional symptoms compared to their peers. We anticipate that a broad range of methods will be applied to assess and define school attendance and will take an inclusive approach. Methods of assessing attendance can include, but will not be limited to: a quantitative measure of actual attendance, or assessment of attendance behaviours using structured interviews or self-report measures. We will include any terminology used for attendance including, but not restricted to: attendance, non-attendance, school refusal, school phobia, absenteeism and truancy.
Types of studies
Quantitative studies of any design that report on the relationship between emotional disorder and school attendance will be eligible, including population surveys and case-control, cross-sectional, longitudinal and cohort studies. Studies where the primary aim was to evaluate the effectiveness of an intervention will be excluded, as will case studies, case series, editorials or opinion pieces and papers published in languages other than English. No restrictions will be placed on country of study. Publications that do not provide enough information to allow for quality appraisal, for example, conference abstracts, will be excluded.
Studies will be identified from the following information sources:
Electronic database searching: we will search electronic databases from date of inception to present, as we do not have reason to believe the concepts of emotional disorder or school attendance are likely to have changed substantially over time. The following databases will be searched using a pre-specified search strategy (see ‘search strategy’):
MEDLINE (via OvidSP), including Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE
PsycINFO (via OvidSP)
Education Resources Information Centre (ERIC) (via EBSCOhost)
Applied Social Sciences Index and Abstracts (ASSIA) (via ProQuest)
Education Research Complete (via EBSCOhost)
British Education Index (via EBSCOhost)
Australian Education Index (via ProQuest)
ProQuest Dissertations and Theses (PQDT) (via ProQuest)
Grey literature: grey literature is usually understood to mean literature that is not formally published in sources such as books or journal articles , and its inclusion in a systematic review minimises the impact of publication bias. We will search Health Management Information Consortium (HMIC), Conference Proceedings Citation Index, ProQuest Dissertations and Theses and the website OpenGrey (via http://www.opengrey.eu/), all of which index grey literature.
Citation searching: citation lists of included studies will be checked (backward citation searching), as will papers that have cited included studies, according to Web of Science (forward citation searching).
Experts in the field: experts in the field and corresponding authors of included papers will be contacted for any further information, such as unpublished material or papers in press.
A comprehensive search strategy has been developed following a scoping review of the topic area and consultation with an information specialist as well as experts in the fields of child mental health and education. The search strategy will include both free-text searching and controlled vocabulary searching (e.g. MEDLINE Medical Subject Headings (MeSH) terms). Terms will be grouped according to four concepts:
Child terms (e.g. child, adolescent, student, pupil)
Setting terms (e.g. school, education, nursery)
Attendance terms (e.g. attendance, non-attendance, refusal)
Emotional disorder terms (e.g. emotional disorder, anxiety, depression, internalizing)
The master search strategy for MEDLINE (OvidSP) can be found in Additional file 2.
An initial screening of titles and abstracts of identified papers will be performed by two independent reviewers, who will assess each study for relevance according to the pre-specified eligibility criteria. Studies that cannot be conclusively excluded from title and abstract screening will be taken forward to full text screening, at which stage the full text will be obtained and a second screening process performed, again by two independent reviewers. This will result in a final set of papers to be included in the review. Discrepancies between the two reviewers at any stage will be resolved through discussion and, if required, referral to a third reviewer. The number of studies identified, included and excluded at each stage will be reported using a PRISMA flow diagram  together with reasons for exclusion at the full-text stage.
EndNote X7 software will be used to manage references throughout the review process. The results of searches will be exported to EndNote and duplicates will be removed.
A data extraction form will be developed specifically for the current study, guided by the full-text screening stage and following templates provided by experts in systematic reviewing. The data extraction form will be pilot tested on three studies selected from those to be included in the review. Data extraction will be completed by two independent reviewers for all included studies. Where there are multiple publications from one study, we will treat all sources as one and extract the data into one form. Data items to be extracted will include: study design, age and gender of participants, country of study, sample size, setting, method of assessing emotional disorder, measures of school (non-)attendance and relevant findings including effect sizes, confidence intervals and p values, where provided.
Risk of bias in individual studies
Assessing risk of bias in included studies is important in determining the validity of the results and interpretation of findings. Risk of bias of included studies will be assessed by two independent reviewers using the Newcastle-Ottawa scale (NOS), which is a widely used scale designed for use in epidemiological systematic reviews . The risk of bias for each included study will be taken into consideration during data synthesis.
Study characteristics will be summarised using means and standard deviations (or medians and interquartile ranges) for continuous variables, and numbers and percentages for categorical variables. A narrative report of study characteristics will also be provided. Where at least two studies have used the same method to analyse the relationship between emotional disorder and school attendance, random effects meta-analysis (using the DerSimonian and Laird method ) will be used to pool the quantitative results across studies. The pooled “effect” estimate (e.g. mean difference, standardised mean difference, risk ratio etc) will be reported with 95% confidence interval and p value. Heterogeneity will be quantified using the I-squared statistic, and we will assess the presence of heterogeneity using Cochran’s Q test. If there are sufficient data, we will perform subgroup analyses to investigate the effects of the following potential moderator variables on the relationship between emotional disorder and poor school attendance: mean age of children, type of emotional disorder, somatic symptoms, measurement source (e.g. child-report, parent-report), assessment method (diagnostic tool or measures of continuous symptoms), study setting and type of school. If subgroup analyses show statistically significant evidence that two or more of these variables are moderators, and there are at least ten studies per moderator, then we will consider using meta-regression to examine the moderators in a single model. If there are sufficient studies, we will also examine the possibility of publication bias using funnel plots and Egger’s regression test . We will use the assessment of risk of bias in individual studies to indicate the strength of the body of evidence for each main outcome variable.