Addi van Bergen

Summary and general discussion 145 7 on data from the 2008 PHM, which was completed by 20,877 adults. As the content of the questionnaires differed among cities, three different indices were constructed using nonlinear canonical correlation analysis. The psychometric properties of the constructed indices were adequate to good. The content validity, however, was only moderate. Our study showed that a measure for social exclusion could be constructed with available health questionnaires. Recommendations were made to enhance content validity by adding extra items from the SCP social exclusion index to the PHM. In chapter 4, we followed the recommendations made in the previous chapter. Nineteen of the 26 GGDs, covering over 70% of the Dutch population, included extra SCP items on material deprivation, access to basic social rights and normative integration in their 2012 PHM. Data from 258,928 respondents aged 19 years or older were thus obtained. The dataset was randomly divided in half: a development sample and a validation sample. Nonlinear canonical correlation analysis in the development sample produced an overall index and four dimension scales, the SEI-HS, containing 9 PHM items and 8 SCP items. The internal consistency, internal structure and construct validity were satisfactory to good and in line with the original SCP social exclusion index, and the content validity was good. Replication of the SEI-HS in the validation sample confirmed its generalisability. Both index and dimension scores were trichotomised into ‘moderate to strong’, ‘some’ and ‘no’ exclusion based on the 95th and 90th percentiles in the Dutch adult population to facilitate their application in public health monitoring and policy. The SEI-HS enables researchers to take the next step in advancing our much needed knowledge on SE and health. In chapter 5, we presented the results of a cross-cultural validation study of the SEI- HS. In the four cities, Amsterdam, Rotterdam, The Hague and Utrecht, particularly high levels of SE were found among non-Western immigrant groups, e.g., 20%, 21% and 29% of adults of Surinamese, Moroccan and Turkish origin, respectively, were found to have moderate to strong SE; only 4% of adults of native Dutch origin were found to have the same. A sequential explanatory mixed methods design was used to explore the possible cultural bias of the SEI-HS. Data from the 2012 PHM were used to evaluate the structural validity and differential item functioning of the SEI- HS in three major immigrant groups in the G4. For each SEI-HS item, semantic, conceptual and contextual connotations were compared between the three immigrant groups and native Dutch based on semi-structured interviews with 11 Surinamese, 9 Moroccan, 10 Turkish and 22 Dutch respondents with high scores on the SEI-HS. Confirmatory factor analysis corroborated the 4-factor structure of the SEI-HS in all three immigrant groups, and no substantial differential item functioning was found for migration background. The interviews uncovered some methodological shortcomings, but these did not substantially impact the excess of social exclusion observed in the immigrant groups. Our study confirmed the cross-cultural validity of the SEI-HS in three major immigrant groups in the Netherlands. The high levels of SE among non- Western immigrants in the G4 proved to be real and not a methodological artefact. Our conclusion was that policy measures to enhance social inclusion and reduce exclusion are urgently needed.