Browsing by Author "Pulkkinen, Tero"

Background: In recent years, research on dimensional models of psychopathology has substantially increased. Some of the most promising models have included a dimension of general psychopathology that is often called the p factor. However, the research on the practical uses of such models is still scarce, as is the evaluation of models beyond their goodness-of-fit indexes. This study aims to contribute to filling this gap by examining the underlying structure of preadolescent psychopathology and answering an important practical question: whether such models of preadolescent psychopathology can be used in predicting substance use in adolescence, which is a critical period in the development of substance use disorders. In addition, the model containing the p factor is evaluated with methods that have been seldom used but that can provide a more nuanced picture of psychopathology. Methods: Using the data from the UK Household longitudinal study, the underlying structure of psychopathology was first modelled with three confirmatory factor analyses (n = 3437). The models in question were a two-factor model consisting of internalizing and externalizing; a one-factor model consisting of the p factor; and a bifactor model consisting of internalizing, externalizing and the p factor. The bifactor model was also evaluated with several bifactor model indices. Then the models that exhibited at least acceptable statistical fit were used to predict the use of several substances in adolescence (n = 1610). Results and conclusions: The two-factor model and the bifactor provided an acceptable fit to the data. Both the externalizing factor in the two-factor model and the p factor in the bifactor model were also able to predict the use of all substances. However, the bifactor model had issues with interpretation and contrary to theorizing, it could be characterized as primarily unidimensional. Future research is encouraged to compare the practical utilities of different models of psychopathology and to evaluate bifactor models in more detail.