Mathews 2017.

Study characteristics
Methods	Study design: randomised controlled trial Unit of allocation: participant Unit of analysis: participant Adjustment for clustering: not applicable
Participants	Location: Pennsylvania, USA Setting: licensed childcare facilities (Mathews 2017, p 4) Sample size calculation: yes (Mathews 2017, p 7) Sample size: 765 early education and care providers over 18 years of age, working with children under 5 years of age; intervention group n = 388, control group n = 374 (Mathews 2017, p 4) Mean age (SD): not reported; largest age category: (i) intervention group = 18 to 29 years (40.6%), (ii) control group = 18 to 29 years (40.1%) (Mathews 2017, p 9) Gender: (i) intervention group = 98.9% women, (ii) control group = 96.5% women (Mathews 2017, p 9) Race/ethnicity: (i) intervention group = 84.7% non‐Hispanic white, (ii) control group = 83.7% non‐Hispanic white Years of experience: largest group > 15 years: (i) intervention group = 25.2%, (ii) control group = 25.2% (Mathews 2017, p 9) Previous child protection training: (i) intervention group = 79% yes, (ii) control group = 78.1% yes (Mathews 2017, p 9) Previous experience with child maltreatment reporting: not reported Baseline equivalence: analysis by Chi² tests showed that each group had comparable baseline demographics (Mathews 2017, p 9)
Interventions	Name: iLookOut for Child Abuse ‐ Online Learning Module for Early Childcare Providers Contents: (1) cognitive aspects of mandated reporter training: (i) definitions of abuse, (ii) signs of abuse, (iii) legal requirements for reporting; (2) affective aspects of mandated reporter training: (i) empowering participants to contact CPS when there was a reasonable suspicion, and (ii) developing attitudinal dispositions to help protect children (Mathews 2017, p 5) Processes and teaching methods: immersion in real‐life simulations via "interactive, video‐based storyline with films shot in point of‐view (i.e. the camera functioning as the learner's eyes), with the learner taking the role of a teacher of 4 year olds at a childcare centre" (Mathews 2017, p 5) Delivery mode: online, designed "to allow for mobile access" and "to accommodate individuals with sensory disabilities" (Mathews 2017, p 5) Trainers and qualifications: not applicable, as training was online. Module developed by team of experts in child abuse, instructional design, paediatrics, early childhood education, online learning, mandated reporter training, law, ethics, and child advocacy. Duration: 4 weeks (Mathews 2017, p 5) Intensity: self‐paced Intervention integrity: not reported; online delivery offers the possibility of uniformity Comparison condition: waitlist control
Outcomes	Eligible measures (outcome domain) iLookOut Knowledge (secondary outcome: knowledge of reporting duty, processes, and procedures), comprising 21 items with response options true/false iLookout Attitudes (secondary outcome: attitudes toward the duty to report child abuse and neglect), comprising 13 items with responses on 7‐point Likert‐type scale Ineligible measures (reason): nil Timing of outcome assessment: pre‐test (before training), post‐test (after training), follow‐up (after 4 months) (Mathews 2017, p 6)
Notes	Funding: intervention was funded by Penn State University Center for the Protection of Children; research was funded by Penn State REDCap, Penn State Clinical and Translational Research Institute, NIH/NCATS Grant Number UL1 TR000127 Author contact: yes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: selection bias unlikely due to adequate description of the generation of the randomised sequence Quote: "the host website randomised participants using an automated scheme" (Mathews 2017, p 6)
Allocation concealment (selection bias)	Unclear risk	Comment: study protocol states the trial was open and unmasked, yet published report of Phase I of the trial states that "participants were blinded to treatment allocation" (Mathews 2017, p 6). Unclear whether blinding refers to allocation concealment up until point of assignment
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: study protocol states the trial was open and unmasked, yet published report of Phase I of the trial states that "participants were blinded to treatment allocation" (Mathews 2017, p 6). It is likely that participants would have been able to identify if they were taking part in the intervention, as this was self‐administered online. Due to subjective nature of measures (pre‐post self‐report) and their close alignment with intervention content, performance bias is likely.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no blinding of outcome assessment, detection bias likely due to the subjective nature of measures and their close alignment with intervention content (pre‐post self‐report)
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: loss of 21 participants after allocation was not evenly distributed across conditions (5/374 (1.34%) control condition; 16/388 (4.12%) experimental condition). Reasons for losses were not explicitly reported, but were described as "failure to complete" (Figure 1, p 8). The overall rate of attrition is low (2.76%), and analyses excluded lost participants (Mathews 2017, p 8). A total of 406 participants agreed to be contacted to gather follow‐up data (n participants by group not reported), yet only 201 participants who had received the intervention completed the 4‐month follow‐up measure. Consequently, no between‐group comparisons could be performed by study authors.
Selective reporting (reporting bias)	Low risk	Comment: study protocol is available, and all outcomes that were prespecified were reported in the prespecified way
Other bias	Unclear risk	Comment: additional potential sources of bias related to the specific study design have been identified
Reliability of outcome measures (measurement bias)	Unclear risk	Comment: outcome measures were developed specifically for the study. No report of coefficient alphas, but authors report actions to enhance reliability and validity of the measures (e.g. use of experts in development, piloting to "improve content validity and reliability, including test‐retest reliability" (Mathews 2017, p 6).
Group comparability (selection bias)	Unclear risk	Comment: group comparability at baseline was supported by statistical testing on demographic variables (Table 1, p 9). Statistically significant difference in gender between conditions (13/369 control; 3/372 experimental). No other statistically significant differences were found. Authors do not report statistical assessment of group comparability of primary and secondary outcomes at baseline. For knowledge, mean = 13.54 for both conditions at baseline. For attitude, mean = 5.78 for control condition and 5.80 for experimental condition at baseline (Mathews 2017, p 9). Quote: "baseline data showed each group had almost identical knowledge and attitudes" (Mathews 2017, p 8)
Contamination (contamination bias)	Unclear risk	Comment: unclear whether control and experimental participants worked in the same settings, thereby potentially leading to contamination. Although the potential participant pool was large ‐ educators at 1900 childcare facilities ‐ it is possible that a small proportion of participants in intervention and control groups may have been from the same workplaces.