Autism Symptom Severity in Males and Females: An Exploration of Gender Differences Using Item Response Theory

Friday, May 12, 2017: 12:00 PM-1:40 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
A. Sturm1 and M. Kuhfeld2, (1)UCLA, Los Angeles, CA, (2)University of Texas Austin, Austin, TX

Studies examining the role of gender in autism have found inconsistent support for phenotypic differences between boys and girls. However, few studies have systematically tested gender differences in autism symptomatology using gold-standard measures in large samples.


This study aimed to systematically explore gender differences in the phenotype of ASD in a large sample of clinic- or research-referred youth and adults (N = 9697). Specifically, the present study explored symptom bias, mean and variance differences, and the potential moderating role of nonverbal IQ for each of the four ADOS modules.


Participants included cross-sectional data compiled from the National Database for Autism Research, Autism Genetic Resource Exchange, and Simons Simplex Collection. Multigroup unidimensional confirmatory factor models were fit to each ADOS module using revised scoring algorithm items to determine differences in factor means and variances between males and females. Fitted models included Module 1 no words (N = 1496) and some words (N = 1792), Module 2 less than age 5 (N = 1076) and greater than or equal to age 5 (N = 1162), Module 3 (N = 3170) and Module 4 (N = 1001). Items were then evaluated for significant item bias by gender using differential item functioning. Finally, to test whether gender differences are moderated by nonverbal IQ, each unidimensional model was estimated including gender, NVIQ, and their interaction as latent predictors.


None of the items currently included in the ADOS scoring algorithm showed significant DIF (wABC > .35) by gender. Only Module 2 less than 5 years (z = 19.1, p < .001) and Module 4 (z = 9.22, p < .001) showed significant differences between males and females on factor means, such that males scored significantly higher than females. In addition, only Module 2 less than 5 years (σfemale = 1, σMale = .18) and Module 4 (σfemale = 1, σMale = .4) showed pronounced differences in factor variability. Finally, the interaction between gender and nonverbal IQ significantly predicted ADOS severity for all Modules (βrange=-0.006 to – 0.002, p < .01) excluding Module 2 greater than or equal to age 5 (β = -0.001, p= .24).


Males and females who have the same autism severity are nearly equally likely to receive a particular score on a specific item of each module and show similar autism severity and variability in severity. While males scored significantly higher than females when assessed using Module 2 less than 5 years and Module 4, there were also the fewest number of individuals in these groups which highlights the importance of investigating symptom and phenotypic differences in large samples and thus these two populations should be further explored. Finally, although NVIQ impacted ADOS scores for males more so than females, parameter estimates were so small that these results are not likely substantively meaningful. In the absence of phenotypic differences, factors aside from ASD core features, such as associated features and comorbid disorders, must be explored to develop a better understanding of the disparity in prevalence.

See more of: Epidemiology
See more of: Epidemiology