Table 2.

Risk of bias in the published studies included.

Study	Risk of bias
Bush et al, 2017 [50]	Small sample size (attrition bias) Authors did not report how the knowledge of allocated condition was blinded to participants and researchers during the study (performance and detection bias)
Tighe et al, 2017 [46]	Owing to the changes in inclusion criteria after the commencement of the trial, one-fourth (26.2%) of the included participants did not meet the criterion for frequency of suicidal thoughts (attrition bias) Participants, clinical personnel, and outcome assessors were not blind to treatment allocation Data on usage were available for 65.6% of the included participants. Of these, 15% did not complete treatment Small sample size Sensitivity of measures was poor because of number of items (performance bias)
Franklin et al, 2016 [48]	Neither participants nor clinical personnel were blind to treatment allocation (performance bias) Participants were recruited online; therefore, motivation and ability to engage were potentially higher than recruitment through the community (selection bias) Participants paid to complete study Amount of app usage not specified—up to the user, level of engagement was not controlled for
Stallard et al, 2018 [47]	Participants were identified by their Child and Adolescent Mental Health Services clinician. Participants themselves decided whether or not to take part (selection bias) No control group present No blinding reported. All participants in the study received an intervention as there was no control condition (performance and detection bias)