Addi van Bergen

Summary and general discussion 147 7 An important outcome of this dissertation is the utility of the SEI-HS for identifying high-risk/high-need population segments (Chapter 6). As we had hoped for at the start of this study, measuring SE can thus help identify and quantify at-risk groups and gain better insight into their characteristics and health risks. This information is important for guiding public health policy and resource allocation. Embedding the SEI-HS in the PHM is a good choice, as data on SE can now be collected every four years, analysed by GGD epidemiologists, and presented and discussed with local policymakers. Some municipalities, such as Delft, use the SEI-HS in their municipal (omnibus) survey that takes place every two years. The findings of this dissertation are also relevant outside of the Netherlands. Although the main focus of this dissertation was on Dutch local health monitoring and policy, we firmly positioned our research in a broader international context, with particular reference to the theoretical framework of the World Health Organization (WHO) on the social determinants of health [4] and the definition of SE developed by the WHO Social Exclusion Knowledge Network [1, 5]. This work bridges social sciences and health research, which we further enhanced by publishing only in open access journals indexed in PubMed. The articles in this dissertation are regularly cited by health researchers from countries all over the world, such as the United Kingdom [6], Spain [7], Finland [8], Croatia [9], Switzerland [10], Czech Republic [11], Ukraine [12], Cameroon [13], Brazil [14], Hong Kong [15], the United States [16] and Lebanon [17]. In particular, reference is made to the use of nonlinear canonical correlation analysis, to the results of our systematic review, and the definition, operationalisation and measurement of SE. STRENGHTS AND LIMITATIONS Strengths A major strength of this dissertation is that we were able to build on many years of theoretical and empirical research conducted by the SCP. We adopted its definition and operationalisation of SE and used the SCP social exclusion index of Hoff & Vrooman [2, 3], as the standard for measuring SE in the Dutch adult population. Another strong point of this study is that we had three large datasets at our disposal: 2008 PHM data for the G4 (N=20,877), 2012 PHM data for 19 GGDs nationwide (N=258,928) and 2016 PHM data for the G4 (N=33,285). Not only were we able to adapt and improve the SE index based on 2008 data in the 2012 dataset, but our results were also stable and reliable, likely replicable not due to coincidence or p-hacking [18]. The use of nonlinear canonical correlation analysis for the construction of the SEI-HS is a strong point as well. In comparison with, for example, factor analysis, nonlinear canonical correlation analysis yields scales with fewer items and a broader scope, resulting in a more concise measure with higher content validity [2].