133 Realist review of Midwife-Led Continuity of Care implementation Duplicate articles were excluded using an automated deduplication tool (DedupEndNote, version 0.9.7.), followed by manual deduplication in Endnote (X20.0.3) by KAZ. The full search strategy used for each database is detailed in online supplemental material 2. Selection and appraisal of documents Studies were eligible for inclusion if they included regions or practices from highincome countries that had implemented, or considered implementing, some form of continuity of care by midwives, whether or not this had been successful. Studies were included if they addressed one or more of the research questions. Studies were selected in two stages. First, titles and abstracts were screened by the first 2 authors RS and EN. Second, the full text of potentially relevant studies was screened for full compliance with the criteria by 2 authors, always one being RS, and the second being either EN or AP. For this process, the data extraction tool Rayyan was used.28 Subsequently, a quality assessment of the included articles on relevance, richness, and rigour was conducted independently by RS and EN, and 8 by RS and AP. Relevance refers to providing relevant information to answer the research question(s) and the extent to which the article can contribute to the theory-building of our intervention, the implementation of MLCC.26 The degree of relevance was categorised on a 5-point scale: low, low/medium, medium, medium/high, and high. Articles with low relevance were excluded. Richness was defined as the method used to ensure that the included documents provided a significant level of depth to contribute meaningfully to theory building.27 Rigour was defined as the method used to generate data that were credible and trustworthy.26 Thus, relevance, richness, and rigour depended on the purpose of this specific realist review.26 To facilitate this assessment, a custom-made Excel file was developed to systematically document and categorise the quality criteria for each article. If there were discrepancies between the 2 reviewers, a third reviewer (AP) was consulted. Data extraction All documents were uploaded into MAXQDA, version 2020, to allow for detailed and systematic analysis. The first 2 authors (RS and EN) independently coded the first 5 of the included papers to identify programme theories, descriptive contexts (C), mechanism resources (MRc), mechanism responses (MRp), and outcomes (O). The coding process was both deductive and inductive. Codes were created deductively based on the initial programme theories. New codes were created inductively as new contexts, or mechanisms related to the outcome emerged. Disagreements about potential contexts or mechanisms were discussed until consensus was reached. A third reviewer (AP or AdJ) was consulted if consensus could not be reached. The remaining papers were analysed by the first author, with regular consultation with EN, AP, and AdJ. 6
RkJQdWJsaXNoZXIy MTk4NDMw