50 Chapter 3 Assessment of risk of bias in included studies Two reviewers (G.H. and J.v.H.) independently rated methodological quality. Studies were assessed using the suggested risk of bias criteria for Cochrane Effective Practice and Organisation of Care (EPOC) reviews [25]. We assessed studies for generation of allocation sequence, concealment of allocation, similar baseline outcome measurements, similar baseline characteristics, incomplete outcome data, blinding of participants, blinding of outcome assessors, protection against contamination, selective outcome reporting and other risks of bias. The decision on whether the criteria were fulfilled was resolved by discussion, or consulting a third researcher (Y.S.). As suggested by Davey et al [26], we scored each study for risk of bias as ‘Low’ if all criteria were scored as ‘Low risk’, ‘Moderate’ if one or two criteria were scored as ‘Unclear’ or ‘High risk’, and ‘High’ if more than two criteria were scored as ‘Unclear’ or ‘High risk’. Interrater agreement for the individual domains of the risk of bias was calculated by between-group Kappa agreement, using the assessments from each reviewer before resolution of disagreements. Data synthesis and analysis Study outcomes were organized in tabular form and a qualitative assessment was made based on the methodological quality, intervention characteristics, outcomes, statistical significance, anddirectionof effects observed. Core elements of interventions were identified in a similar approach to Sinha et al. [27] and Fan et al. [20]. One primary reviewer (G.H.) reviewed each study and listed all characteristic elements included in the studied interventions. Individual studies were further examined to determine whether they were using the same terms to describe different elements or different terms to describe the same elements. This process ultimately enabled a set of core elements. Subsequently, each individual intervention was again reviewed, determined as adhering or not to a particular category. If possible, summaries of intervention effects for each study were provided by calculating risk ratios (RR) for dichotomous outcomes and standardised mean differences (SMD) for continuous outcomes. We performed a meta-analysis using the Review Manager 5 data analysis programme when two or more studies were randomized controlled trials and the outcome measures and type of intervention could be compared. If no significant heterogeneity was present (I² < 70%) [28], a random-effects meta analysis was performed of binary (e.g. ED revisits) and continuous (e.g. ED LOS) outcomes. Statistical significance was set at p<0.05.
RkJQdWJsaXNoZXIy MjY0ODMw