Addi van Bergen

Chapter 4 84 5. Generalisability No marked differences in the canonical functions were found between the analysis in the development and validation samples. The eigenvalues of the index and subscale Social Participation were similar in the two samples. The eigenvalues of the subscales Material Deprivation, Social Rights and Normative Integration were almost similar: 0.50, 0.52 and 0.44 respectively in the validation as opposed to 0.49, 0.53 and 0.43 in the development sample. The same holds true for component loadings and weights. DISCUSSSION The findings of this study show that we succeeded in developing a reliable and valid multidimensional measure for SE, the Social Exclusion Index for Health Surveys or SEI-HS. The OVERALS analyses empirically confirmed our multidimensional model with SE as the underlying latent construct. The limitations we encountered in previous retrospective research with regard to content validity and generalisability were successfully tackled in this nationwide prospective study. Content validity was enhanced by the addition of extra items. Instead of three dimensions in our previous study, the SEI-HS measured all four dimensions of SE. Generalisability was enhanced by successful replication of the SEI-HS in a representative validation sample. Other psychometric properties were found to be satisfactory to good and in line with the original SCP Index. Low to moderate intercorrelations between index and subscales confirmed the internal structure of the SEI-HS and construct validity was established through hypothesis testing. The internal consistencies of two of the SEI-HS dimension scales were found to be weak. Both the Social Rights and Normative Integration scales had Cronbach’s alpha coefficients lower than 0.70. By using canonical correlation analysis to construct a measure for SE, we selected those elements from the underlying theoretical dimensions that interrelate with one another and form a coherent construct. Of course, social participation, material deprivation, access to basic social rights and normative integration are broader concepts than the dimension scales resulting from these analyses. Access to basic social rights, for example, also comprises e.g. access to other public and private services such as education, legal aid, acceptance for insurance and banking and help with finding a job. When empirically tested, these forms of access proved not relevant to the concept of SE, at least not in the general population in the Netherlands [33, 34]. These aspects of basic social rights were therefore not included in the Social Rights scale. The SEI-HS dimension scales are thus relevant and of value only in the context of the concept SE. One of the study’s strengths is the use of a sound and validated instrument to supplement items on domains where the Netherlands PHM fell short. The SCP items were originally selected by the SCPwith nonlinear canonical correlation analysis from an item pool of 232 items derived from extensive literature and empirical research, focus groups and cognitive tests [5, 33, 34, 49]. Thus, the selected items not only