Sex Differences In the Early Screening of Autism Spectrum Disorders

Background: Epidemiological studies of Autism Spectrum Disorders (ASDs) indicate 1% prevalence and a 4:1 male-to-female ratio. Research also demonstrates sex differences in these disorders throughout development. Because differences may exist very early on, the predictive value of current early screening tools may vary based on sex. 

Objectives: To examine whether the Modified Checklist for Autism in Toddlers (M-CHAT; Robins et al., 1999) identifies males and females with ASDs at similar rates and whether specific items better predict ASD diagnosis in males and females.

Methods: The sample included 7895 males and 7556 females (mean age=20.62 months, SD=3.09) screened for ASDs using the M-CHAT, a parent-report questionnaire, at toddler’s pediatric well-visits. Parents of children who screened positive on the M-CHAT completed the M-CHAT Follow-up Interview (FUI). Those who screened positive on the FUI completed an ASD evaluation.

Results: Of those who completed the M-CHAT (n=15,451), 10% (n=788) of males and 7.5% (n=564) of females screened positive. Of all of those who screened positive on the M-CHAT (n=1,352), 58% were male. Of those who completed the FUI (n=1,078), 17% (n=104) of males and 9.9% (n=46) of females continued to screen positive. Of those who screened positive on the FUI (n=150), 69.3% were male. Of the children who completed a diagnostic evaluation (n=140), 53.8% (n=56) of males and 36.1% (n=13) of females were diagnosed with an ASD. Of all of those with an ASD (n=69), 81.2% were male. Females were more likely than males to be false positives on the FUI, (χ² (1)=9.64, p=.002), and evaluation (χ² (3)=16.10, p=.001) stages. Therefore, positive predictive value (PPV) for M-CHAT was three times greater for males than females on M-CHAT alone (.071 and .023, respectively). PPV for M-CHAT + Follow-up Interview was nearly double for males than females (.538 and .283, respectively). A discriminate function analysis yielded seven items in boys and eight items in girls that best predicted ASD. Six of the items were the same for males and females, but two items (point to request and imitation) were better predictors in females and one item (response to name) was a better predictor of diagnosis in males.

Conclusions: Based on the population rates, we expected 80% of screen positives to be male. Rather, only 58.3% of screen positive cases were male.  Males comprised 69.3% of those positive on the FUI, and of the children diagnosed with ASD, the sex ratio was as expected based on epidemiological reports (81.2% male). These rates indicated that PPV is approximately threefold greater for males than females when only the paper-and-pencil M-CHAT is administered, and nearly twofold greater for males than females for M-CHAT + FUI, suggesting that screen positives results in males are more likely to predict an ASD diagnosis. Largely overlapping but distinct subsets of items best predict an ASD diagnosis in males and females; this may be useful for the development of sex-specific algorithms of the M-CHAT. Future research will further investigate the predictive power of individual items and subsets of items in males and females.