Data extraction and framework revision and validation

Full-text publications will be obtained and the reference list reviewed. Any relevant studies found in the reference list will be screened (RS) for inclusion against the same inclusion criteria. Additional uncertainties regarding eligibility for inclusion will be resolved through discussion with other reviewers (RT or PG). Two reviewers (RS and ZAM) will independently extract study data from 10% of included qualitative studies and 10% of included quantitative studies using a standardised data extraction form that will be piloted prior to use. Conflicts will be resolved by a third party as required. Once interrater reliability (κ) >0.8 is achieved for extracted data, one reviewer (RS) will undertake the remaining data extraction in a staged process, with this detailed below in the extraction sections. The same staged process will be used when extracting data from quantitative and qualitative studies. Queries will be resolved through discussion with a second reviewer (ZAM).

The methods used to extract and synthesise the results of qualitative and quantitative studies are based on the meta-analytic techniques described by Sandelowski et al,³³ Thomas and Harden³⁴ and Timulak.³⁵ Extracted data will include study characteristics (author, journal, year of publication, study country and setting), participant characteristics (number of participants, age, health condition) and quantitative or qualitative outcomes (consequences, impact, effects of the diagnostic label).

Data for thematic analysis will be extracted from the published study and include the authors abstracted themes and relevant, supporting quotes, reported in the primary study. Direct quotes will not be extracted in isolation to ensure data ‘retains its meaning’ and is not interpreted or extracted out of the context of the primary study. This qualitative meta-analysis technique has been described by Sandelowski et al,³³ Thomas and Harden³⁴ and Timulak.³⁵

For studies with quantitative outcomes, extracted data will include, the text and numerical data from the results section reporting primary outcomes.³⁶ Examples of potential quantitative measures include the Short Form Health Survey (SF-36),³⁷ General Health Questionnaire (GHQ)³⁸ or work absenteeism.

The coding framework developed from social media responses will be iteratively revised using eligible studies retrieved by the electronic database search. Qualitative data will initially be extracted from a random sample of one-third of included qualitative studies and mapped to the coding framework. This framework will be expanded as additional themes emerge. The second third of included qualitative studies will be randomly selected, data extracted and mapped to the updated coding framework until data thematic saturation has been achieved. If new themes are still emerging at this point, the remaining third of qualitative studies will be analysed against the developed framework. Data saturation will be defined using indicative thematic saturation, which states data saturation as the non-emergence of new codes or themes that will result in expansion or revision of the coding framework.³⁶

Quantitative data will be summarised narratively.³³ For example, we will collate data from studies that used the SF-36, GHQ or absenteeism and summarise the findings reported in the results section. Unlike the large volume of expected qualitative studies, fewer quantitative studies with comparators are expected. Therefore, outcomes from all of the included quantitative studies will be extracted and, if possible, tabulated by condition and outcomes.