Chapter 7: Discussion

The objective of the present project was partially to compile a battery of cognitive tests suitable for computer-assisted administration and scoring and applicable for screening populations at risk of mild degrees of cognitive impairment. Subsequent aims were to estimate the effect of subject-related and situation-related factors on the involved test parameters and to estimate the relationship between test performance and associated symptoms and to establish a reference material on the basis of the normal population covering the age span of occupational activity. Finally one objective was to estimate the sensitivity and specificity of the composed test battery. To estimate the effect of the subject-related and situation-related factors on the involved test parameters, a model comprising 16 independent variables representing these factors was applied. However, although a large number of factors was included, the model cannot be considered to cover the complete universe of significant components, and it is not known if other factors would be of similar or more relevance. Univariate analyses showed that the response range for a number of test parameters was rather wide. For the bivariate and multivariate statistical analyses of the error-count parameters correction for overdispersion had to .be introduced because of different levels of difficulty between the 'elementary' tasks of some of the tests (see p 38). The bivariate analyses of the test parameters in relation to the factors often showed significant outcomes, which, however, were reduced in number when a multivariate method was applied. This occurrence was ascribed to confounding, i.e. dependence between the factors.