Methods

Search Strategy

PubMed, Embase, PsychINFO, Medline, Web of Science, and International Pharmaceutical Abstracts databases were searched systematically from the date of database inception until 3rd October 2020 to identify relevant literature. MeSH, Emtree and truncated terms were used where applicable (E-table 1). Key search terms were: asthma, child, Intervention, adherence and randomized. All authors were contacted via email or, if not reachable via this route, by ResearchGate messaging for further details about the studies.

Study selection

Authors CP and TJ reviewed the abstracts, followed by the full texts against the inclusion/exclusion criteria. Where there were differing opinions a third, opinion was sought (RH). Inclusion criteria were based on the Participant-Intervention-Comparison-Outcome-Study Design (PICOS) framework. Any interventions that focused on adherence to ICS with at least one outcome measure of adherence and used a randomised control trial (RCT) design were included. The comparison group was either usual treatment or basic education arms. Articles were included where the full text was written in English, and where the population of interest was patients aged 0-18 years old with a diagnosis of asthma. Although many preschool children with wheeze do not respond to ICS16, studies often recruit younger children and therefore this age-range was intended to avoid missing relevant articles. If they do not have the treatable trait of airway eosinophilia likely to respond to ICS 17 this will be highlighted in the section regarding reliability of the criteria for asthma diagnosis. Studies were excluded if they did not meet the above criteria or if they were an RCT comparing two medications only, or where the majority of participants were not children (e.g. the mean age of participants was over 18 years or only adults were recruited).

Data Extraction and Synthesis

Following full-text review, CP and TJ extracted details of: study characteristics (setting, number of participants, diagnosis criteria, intervention and control descriptions and the outcome of interest); Effectiveness; Risk of Bias (RoB); and Behaviour change techniques (BCTs); target of the BCTs; relationship to PAPA.
Intervention Content
Intervention content were coded for PAPA as follows: Level 1 (intervention only targeted perceptions or only practicalities and not tailored); Level 2 (both perceptions and practicalities targeted but not tailored or only targeting one component (perceptions or practicalities) and tailored) and Level 3 (both perceptions and practicalities targeted and tailored to the individual).
Specific components within the intervention for changing adherence (Behaviour Change Techniques; BCTs) were also coded independently using BCT taxonomy V1 app 18. Any differences in the selected BCTs were discussed until consensus was reached (Table 1).
Risk of Bias
RoB was assessed independently using the Cochrane Risk of Bias Handbook19 by CP, AC and HF using the Covidence platform (www.covidence.org) to record coding decisions and consensus discussions. The RoB score was based on the adherence outcome. Each study was scored across five domains: selection bias; performance and detection bias, attrition bias and reporting bias, and was scored as either low, high or unclear risk for each study. Authors were contacted for clarity when information relating to domains seemed unclear.
Study Reliability
To ascertain which interventions were truly effective, study reliability was considered. Although other validated tools have been used to assess quality such as the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) tool 20 the authors felt that there were several crossovers between risk of bias, the reliability scores, the 3CBC approach and in particular, the indirectness section of the GRADE tool and based on the aim of this review the reliability measurements would be more useful when considered with RoB. Both diagnosis and adherence measures can range from being subjective to objective, therefore considering the reliability of the approaches used is key for determining study reliability. Through multidisciplinary team discussions (including with respiratory physicians, pharmacists and a health psychologist) a coding hierarchy that considered the reliability of the asthma diagnosis and adherence measurement used was created and applied to the specific studies within this review (E-table 2).
Based on the RoB, the reliability of the asthma diagnosis, and the objectivity of the adherence measurement, the most reliable and least biased studies were used to ascertain what components constituted an effective intervention. Previous literature suggests that optimising the context, channel of delivery, and content of the intervention is important for intervention effectiveness 11, and thus the 3CBC 11 was also applied to this review.
Studies were summarised by a narrative synthesis. Meta-analysis was not conducted due to the wide study heterogeneity in terms of setting, asthma diagnosis criteria, and outcome measures used. The study protocol is published on PROSPERO (https://www.crd.york.ac.uk/prospero/#searchadvanced) (ref: CRD42016029213).