On the subjective–objective distinction for measures of memory and cognition: Theoretical and methodological issues in questionnaire development and validation

The aim of this thesis was to develop a questionnaire for cognitive functioning, which could possibly be used as a screening instrument for early signs of dementia in the future. The introduction discusses the often made distinction between subjective and objective measures. A background to the four articles is provided, focussing on findings of weak relationships between self-report- and laboratory measures of memory/cognition. Studies I and II provided results and conclusions that guided instrument development and validation in Studies III and IV. All studies were based on data from participants in the Betula Prospective Cohort Study. Study I investigated predictors of scores on an established self-report instrument for memory failures (PRMQ). Candidate predictors were memory performance on laboratory tests, age, depressive symptoms, and personality traits. There was no relation to age, and test performance did not predict self-reported memory, but depressive symptoms and personality did. Given the finding of a lack of a relation to age, and a bulk of research articles claiming that memory complaints are common in the elderly or increase with age, Study II used a global rating of problems with memory, and reports of perceived causes. In contrast to Study I, problems ratings were related to age, such that increasing age meant higher severity of problems. Furthermore, perceived causes of memory problems differed across age. The elderly reported aging while the young reported stress and multitasking as primary causes. With these results as a background, the purpose of Study III was to develop a new instrument (the Cognitive Dysfunction Questionnaire - CDQ) with the explicit aim that scores should be related to laboratory test performance. A global construct of cognitive functioning with an emphasis on memory systems was adopted, and an item pool was generated. Based on exploratory principal components analysis and correlations with criterion measures (laboratory test performance), twenty items in six domains were selected. Preliminary psychometric evidence showed that the CDQ was reliable, and related to age and objective measures, but not to depressive symptoms. In Study IV, twenty additional items were constructed, and the CDQ was responded to by participants in independent samples. Confirmatory factor analysis was used to test the factor structure derived from Study III, and refinement was undertaken by collapse of two domains and exclusion of items. The final factor structure was cross-validated. Competing models and measurement invariance across age and sex was tested. Psychometric properties were investigated for the final 20-item version.