A Bifactor Model of the Autism Spectrum Disorder Phenotype

Abstract

I read with interest the newly published work by Kim et al. describing the covariance structure of parent-reported autism spectrum disorder (ASD) symptoms.¹ The authors compared many different latent class, factor analytic, and factor mixture models in a large, clinically referred sample, concluding that the ASD phenotype is best described by three continuous latent factors of social interaction, communication, and repetitive behavior. I found the study to be methodologically rigorous, employing robust estimation techniques, testing a wide range of categorical-dimensional hybrid models, and even replicating the rank-order of model choices in a separate sample. However, given the large reported interfactor correlations in the final model (r = 0.78-0.83; Kim et al., Figure 2), I was disappointed that the authors did not explore a bifactor model² of ASD phenotypic traits. In this correspondance, I aim to demonstrate the ways in which bifactor models provide further insight into the structures of complex psychopathological constructs.