Effects of Age, IQ, and Sex on Autism Diagnostic Instrument Scores

Background: There is growing recognition that scores on standard measures of autism spectrum disorder (ASD) symptoms are affected not just by level of ASD symptomatology, but by factors such as age and IQ. Recently, sex has been introduced to this discussion as important to consider in interpretation of scores on ASD measures. Some groups have even called for sex-specific algorithms, citing concerns that “male-biased” diagnostic instruments may lack sensitivity for identifying females with ASD. However, previous investigations of how sex affects scores have been hindered by small groups of females available for analysis. Small samples have also prevented adequate investigations of whether and how sex, age, and IQ interact in affecting scores on ASD instrument scores.

Objectives: To determine if, after accounting for age, IQ, and language level, scores on ASD diagnostic instruments systematically differ as a function of sex.

Methods: Data were obtained and merged from eight registries. Inclusion criteria were: ASD diagnosis by clinician best estimate, aged 12 months to 17.99 years, nonverbal intelligence quotient (NVIQ) assessment, and Autism Diagnostic Observation Schedule (ADOS)/ADOS-2 assessment. The final study sample had 8,944 participants (including 1,458 girls).

A linear mixed-effects model, including a random effect for site, was fit to 10 different outcomes within the ADOS (Social Affect and Restricted and Repetitive Behavior Calibrated Domain Scores), Autism Diagnostic Interview—Revised (ADI-R; Algorithm A, B, and C raw total scores), and Social Responsiveness Scale (SRS; Social Communication/Interaction (SCI) and Mannerisms raw total scores). The sample was split for the ADI-R analyses into age<4 years and age ≥4 years, to reflect differential scoring. The SRS models were restricted to children who were at least 4 years old. Initial models included age, quadratic age (as appropriate), NVIQ, language level, and all interactions. Sex and all interactions were added to the model backwards stepwise. Final model selection was determined by interpretability of the results, considering statistical significance (p<0.05) of model variables and Akaike and Bayesian information criterion (AIC and BIC) fit statistics.

Results: The effects of age, IQ, and language level emerged in almost all models, occasionally with significant interactions. Adding sex to all models was justified by the fit criteria. Sex interactions were suggested by fit criteria for ADI-R C total (restricted and repetitive behaviors (RRB)), but the interactions were not statistically significant. Boys received more severe scores than girls on both ADOS and ADI-R RRB domain scores, and girls received more severe scores than boys on SRS SCI. Although fit criteria supported including sex in the models for ADOS social affect (SA), ADI-R A and B (for both age algorithms), and SRS Mannerisms, the coefficients were not statistically significant.

Conclusions:: In a large sample of children with ASD, previously identified developmental factors (age, IQ, and language level) were confirmed to impact ASD symptom scores. Additionally, measures from three diagnostic instruments differed by sex, primarily in the RRB domain with girls receiving lower (less severe) scores. However, effects of sex were small and likely of limited clinical significance.