Outcome	Missing data and approach
Primary outcome: number of reported cases of child abuse and neglect (self‐report)	Hazzard 1984: percentage of participants in each group who self‐reported making reports provided in‐text, but no report of participant attrition, so unable to use existing formulae/calculators to calculate an effect size from proportions and group sizes. Approach: exclude from analyses, include in study summaries   Kleemeier 1988: no data reported for this outcome, aside from a statement of non‐significance. Approach: exclude from analyses, include in study summaries
Secondary outcome: knowledge of the reporting duty, processes, and procedures	McGrath 1987: tables in the text report the percentage of each group who answered each questionnaire item correctly, without any other summary statistics (e.g. means, standard deviations). There are existing formulae and calculators to permit effect size calculations using proportions and group size (www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-OR2.php); however, we deemed it inappropriate to calculate and report effect sizes for this study. Firstly, whilst calculating a composite effect size would be possible, formulae to adjust the standard error require the correlation between effect sizes, and we could not locate any data that provided an estimate of the correlation between the items. Assuming that the correlation is the same between multiple effect sizes may bias the calculated composite effect size and its standard error (Borenstein 2009). Secondly, whilst selecting an individual item most aligned with the outcome domain would be appropriate, there was little detail reported for the exact items to guide our decision‐making. Approach: exclude from analyses, include in study summaries
Secondary outcome: knowledge of core concepts in child abuse and neglect	McGrath 1987: as above for secondary outcome: knowledge of the reporting duty, processes, and procedures.   Dubowitz 1991: no means or standard deviations reported, only t‐test and P value between experimental and control groups at post‐test. Approach: utilised David B Wilson's suite of effect size calculators to calculate a Cohen's d, 95% confidence intervals (CI), and variance (www.campbellcollaboration.org/escalc/html/EffectSizeCalculator-SMD7.php). The standard error was calculated from the 95% CI and was used, along with the standardised mean difference, in a generic inverse variance meta‐analysis for this outcome. No data were reported for the follow‐up time point for this outcome, aside from a statement of non‐significance.