Measuring autism-associated traits in the general population: Factor structure and measurement invariance across sex and diagnosis status of the Social Communication Questionnaire

Abstract

Autism screening questionnaires are sometimes used as a measure of “autism-associated traits” in samples drawn from the general population, even though such tools are primarily developed and designed for use in samples of children diagnosed with or being assessed for autism. Here, we explore the psychometric properties of the Social Communication Questionnaire (SCQ) current version reported at age 8 in a large population-based sample. Using data from the Norwegian Mother, Father and Child Cohort study (MoBa), we perform exploratory ( N = 21,775) and confirmatory ( N = 21,674) factor analyses on items and compare our results with previously suggested factor structure models of the SCQ. Furthermore, we test for measurement invariance across sex and registry-ascertained autism diagnostic status ( N diagnosed = 636). A 5-factor model provided best fit to the data in both children with and without autism diagnoses, though with some qualitative differences in what the factors represent across these groups. This model performed largely consistently across boys and girls in the general population. Taken together, the SCQ’s measurement properties must be carefully considered when it is used in population-based samples and measurement invariance testing of other autism screening tools used in similar contexts is warranted. Lay abstract Using questionnaires in research relies on the expectation that they measure the same things across different groups of individuals. If this is not true, then interpretations of results can be misleading when researchers compare responses across different groups of individuals or use in it a group that differs from that in which the questionnaire was developed. For the questionnaire we investigated, the Social Communication Questionnaire (SCQ), we found that parents of boys and girls responded to questionnaire items in largely the same way but that the SCQ measured traits and behaviors slightly differently depending on whether the children had autism. Based on these results, we concluded that researchers using this questionnaire should carefully consider these differences when deciding how to interpret findings. SCQ scores as a reflection of “autism-associated traits” in samples that are mostly or entirely made up of individuals without an autism diagnosis may be misleading and we encourage a more precise interpretation of scores as a broader indication of social-communicative and behavioral traits.