156 6 Chapter 6 Assessment of reporting biases The likelihood of publication bias was investigated by preparing a funnel plot only if ten or more studies were included in the meta-analysis. Based on visual inspection, the likelihood of publication bias was evaluated. When asymmetry was observed, we attempted to identify possible reasons by considering the quality of the studies, the particular interventions included, and the contexts in which interventions were implemented. Data synthesis We performed statistical analyses using Review Manger software 5.3 (RevMan 5) and RevMan Web 2019 (RevMan Web 2019). We pooled study results using the random-e ects model. For continuous data, we recorded mean change from baseline to endpoint and SD for each group and calculated the MD. For dichotomous data, we recorded the number of participants with each outcome event and calculated the odds ratio (OR). We used all results reported at short-, medium-, and/or long-term follow-up. Given that all interventions had a duration of 12 weeks at minimum, we analysed available data at 6 months for the short term. We analysed data measured most medial to the other time points (i.e. for medium term, we used results at 12 months when 9 and 12 months were given). When possible, we discussed the intervention e ect estimate in the context of its MCID. If the meta-analysis led to statistically signi cant overall estimates, we transformed these results back into measures that are clinically useful in daily practice, such as the number needed to treat for an additional bene cial outcome (NNTB). Subgroup analysis and investigation of heterogeneity To explain heterogeneity among study results, we planned the following subgroup analyses a priori (when data were available) to determine if outcomes di ered among: 1. settings of the IDM intervention (e.g. primary, secondary, or tertiary care); 2. study designs (individually randomised patients versus cluster-randomised patients); and 3. intervention groups, with regard to di erent components as listed by the EPOC classi cation (EPOC 2008). We performed an additional post-hoc subgroup analysis based on the region in which the study was conducted (i.e. North America, South America, Northwestern Europe, Southern Europe, East Asia, Central Asia) to account for regional di erences in usual care and customs regarding hospitalisation, which proved to be large in Kessler 2018. The previous review authors planned to include an additional subgroup on disease severity (Kruis 2013), but they were unable to do so due to the poor quality of reporting. Also, Kruis 2013 performed an additional subgroup analysis based on control group (i.e. no treatment, treatment with one healthcare provider, treatment with one component, other disease management interventions). In the past decade, regular care has evolved in such a way that multiple individual ‘intervention components’ (e.g. exercise advice,
RkJQdWJsaXNoZXIy MTk4NDMw