The factor structure of the BDI-II and BAI and the longitudinal relationship between depression and anxiety: evidence from a two-wave latent variable study

Abstract

A number of studies have evaluated the latent factor structures of both the Beck Depression Inventory 2nd edition and the Beck Anxiety Inventory. However it is notable that for both of these measures there has been difficulty identifying a replicable and theoretically coherent factor structure. Further, no study has evaluated the factor structure of the BDI-II and BAI in a New Zealand sample. Therefore, the first general aim of the present thesis was to evaluate the factor structure of the BDI-II and BAI using a large sample of New Zealand university students. The second general aim of this thesis was to investigate the longitudinal relationship between depression and anxiety. Previous prospective studies have found most support for unidirectional causal progression from anxiety to depression over time. This pattern suggests that anxiety acts as a risk factor for the development of subsequent depression but not vice versa. To evaluate this relationship a powerful multivariate statistical technique called latent variable path modelling was used. The BDI-II and BAI were administered to a large student sample on two occasions with a three month interval between waves. Using both exploratory and confirmatory factor analyses, the present study replicated a two factor model of the BDI-II that was comprised of cognitive-affective and somatic dimensions. The present study was the also the first to a replicate a factor model for the BAI in a non-clinical sample, namely a four factor model. This research also suggested that the factor structures of the BAI and BDI-II in a New Zealand sample paralleled those identified in American samples. The results of these factor analyses are discussed in light of previous factor analytic research with both clinical and non-clinical samples and in relation to the continuity of dysfunction hypothesis. The results of the latent variable path modelling provided clear support for a bi-directional longitudinal relationship between depression and anxiety. This model suggests that neither depression or anxiety have a more dominant precipitating influence on the other over time. These results are discussed in light of previous longitudinal research that has employed latent variable path modelling and in the context of more general models of the relationship between depression and anxiety.