Harmonized Cognitive Assessment Protocol in the English Longitudinal Study of Ageing: Contrasting approaches to evaluation of factor structure

OBJECTIVES: Cognitive theories and previous research on cognitive performance suggest a hierarchical pattern of interrelationships among cognitive tests, but the psychometric properties of the same tests may change when adapted for a different context or applied to a different population. We evaluated the factor structure of a cognitive battery of tests using an exploratory, data-driven approach and a confirmatory, theory-driven approach. DESIGN: We estimated exploratory factor analyses (EFA) and confirmatory factor analysis (CFA). Agreement between the results based on the EFA and CFA approaches was evaluated by contrasting the identified domains and their corresponding items. SETTING: Epidemiologic cohort study in England. PARTICIPANTS: Adults aged 65+ in the English Longitudinal Study of Ageing (ELSA). MEASUREMENTS: Harmonized Cognitive Assessment Protocol battery of cognitive tests, adapted for the English context. RESULTS: Both the EFA and CFA solutions yielded adequate model fit (RMSEA's < 0.05; CFI's > 0.92; SRMR's<0.06). However, only after multiple iterative steps, the EFA produced a factor structure that largely conformed to a priori theory of human cognitive abilities. CONCLUSIONS: This study provides an important cautionary tale for factor analytic approaches to evaluating domain structures when the tests available for factor analysis do not encompass a broad enough content of the construct: a factor solution is only as good as the bank of available items.