A Comparison of Diagnostically Relevant Item Response Characteristics Between Females and Males with Autism

Background: Males are 4-5 times more likely to be diagnosed with Autism Spectrum Disorder (ASD) than females. This powerful observation implies, not only that males are at greater risk, but that there may be real differences in the expression of ASD between males and females. Previous work from our group and others has primarily focused on broad indicators (IQ, social and repetitive behavior composite scores, etc.). No prior studies have evaluated measurement factors that may lead to sex differences in specific item characteristics relevant to diagnosis.

Objectives: To compare diagnostically-relevant item characteristics (discrimination and thresholds) between females and males with ASD.

Methods: We analyzed data from 2,418 probands with ASD (304 females and 2114 males) included in the Simons Simplex Collection. A series of multi-group confirmatory factor analyses (MG-CFA) were conducted using items measuring four constructs hypothesized to show sex differences in item endorsement - interpersonal relationships, social avoidance, need for sameness, and restriction of interests. Items measuring each of these constructs were selected from the Autism Diagnostic Interview-Revised (ADI-R), Social Responsiveness Scale (SRS), and Repetitive Behavior Scale-Revised (RBS-R). Chi-square difference tests compared item discrimination and threshold differences between males and females with ASD.

Results: Restricted interests items showed highly significant differences in item thresholds between males and females; item discrimination was consistent across sexes. Restricted interests items were less likely to be endorsed (higher thresholds) for females relative to males across all three instruments. There were also two interpersonal relationships items that showed nominally significant sex differences, reflecting that females with ASD are teased less frequently than males and that females with ASD are more likely to share enjoyment. However, these differences did not meet false discovery rate correction.

Conclusions: No sex differences in the measurement characteristics of social avoidance and need for sameness items were observed. Less frequent teasing and increased endorsement of shared enjoyment, while not meeting false discovery rate correction, may suggest a small influence of early socialization experiences and greater peer support in providing some protection for females with ASD from interpersonal relationship difficulties. Additional research is needed to directly evaluate early learning or peer relationship protective factors as modifiers of the expression of ASD in females. Most striking was the observation that restricted interest items from three separate instruments showed higher thresholds in females but equal discrimination. Equal discrimination implies that restricted interests items are an important and relevant aspect of restricted/repetitive behavior in females and males. Sex differences in item thresholds could be interpreted as implying measurement artifact resulting from the use of male-centric items within existing instruments. However, the present analyses suggest this explanation is unlikely because the restricted interests item from the SRS (#39) had no obvious sex bias in wording/exemplars. Instead, fewer restricted interests in females with ASD appears to be a real phenomenon. Additional research is needed to completely rule out the possibility of measurement or rater bias. If lack of bias is confirmed, future investigations should evaluate biological and environmental protective factors that lead females to express fewer restricted interests.