Addi van Bergen

Chapter 2 26 between often large general population studies and smaller studies in specific groups, mostly at high risk of SE. The purpose of this study is to systematically summarise the evidence on the association between multidimensional SE and health and to evaluate six hypotheses i.e. that high SE/low SI is associated with: 1) adverse mental health, 2) adverse physical health, and 3) adverse general health outcomes in a) the general population and b) populations at high risk of SE. METHODS We followed PRISMA guidelines for reporting this systematic review [22, 23]. The review protocol is registered on the PROSPERO database (registration number CRD42017052718) and is available at https://www.crd.york.ac.uk/PROSPERO. Electronic search We developed and executed, with the help of a qualified librarian, a search strategy to identify all studies that reported the association between SE/SI and health. The following three major electronic health databases were searched up to January 2018: PubMed, EMBASE, and CINAHL. Searches were conducted in March 2015 and January 2018. The terms ‘social exclusion’ and ‘social inclusion’ were searched in title and abstract. Search strategies can be found in Supplementary file 1. Study selection Two authors (AvB, plus MB or KS) independently screened all records identified by the electronic search on title and abstract using a sequentially applied algorithm previously introduced by Curran et al. [14] (Figure 2). First, records without an abstract and inconclusive title were moved to a separate database to be assessed on the basis of full text. Next, language, study population, country and type of publication/study design were checked. Studies had to be written in English, Dutch, German, Spanish, or French; involve an adult population; and be set in EU-countries or OECD-countries [https:// www.oecd.org/about/members-and-partners/] , because of their relatively similar welfare regimes. Only research articles published in peer-reviewed journals describing quantitative studies were included. Articles had to test the relationship between SE/SI and a health measure and report statistical results. Next, we excluded studies not using a multidimensional construct of SE/SI (minimum two of four dimensions), studies in which health formed part of the SE/SI measure and studies using an ecological measure of SE/SI. We did not exclude studies on sample size criteria.