Using Incorrect Cut-Off Values in Autism Screening Tools: The Consequences for Psychological Science

Abstract

The 10-item Autism-Spectrum Quotient (AQ10) is frequently used to screen for high autistic traits in clinical practice and research. For the past decade, however, the National Institute for Health and Care Excellence (NICE) has recommended the use of an incorrect ≥7 cut-off value, instead of the correct ≥6 value specified in the original research on the AQ10. Our inspection of the literature suggests that this discrepancy has proliferated across research, generating confusion over the past decade. After examining use of the AQ10 cut-offs in previous research, we drew on five datasets (overall N = 7612) to empirically test the consequences of applying different AQ10 cut-off values for the interpretation of research. Our analyses showed that a one-point difference in the AQ10 cut-off – the error made in the NICE guidelines – significantly changed the effect sizes of autism-related associations in 3 of the five datasets by up to 59.75% and made the relationships non-significant in two datasets. This demonstrates that the (mis)use of the AQ10 cut-off can be consequential for research, hence we discuss the urgent need to establish and apply appropriate autism screening cut-off values in future research.

Using network analysis to improve understanding and utility of the 10-item Autism-Spectrum Quotient

“…Waldren et al (2021) recently reported that the National Institute for Health and Care Excellence (NICE) had, for almost a decade, recommended the incorrect AQ10 cutoff value to screen for autism (≥7 instead of ≥6, based on Allison et al, 2012). This error has generated confusion among researchers and clinicians who administer the AQ10 in their work, with evidence of several studies (e.g., Payne et al, 2020;Rahman et al, 2021) using the incorrect cutoff value (see Waldren et al, 2022).…”