Item Level Differences in the Modified Checklist for Autism in Toddlers with Follow-up (M-CHAT/F) by Race/ethnicity

Background: The American Academy of Pediatrics recommends universal screening for Autism Spectrum Disorder (ASD), yet despite recent improvements in the ASD prevalence gap between White, Black and Hispanic children, White children are still diagnosed earlier. A recent study at Children’s Hospital of Philadelphia (CHOP) identified racial/ethnic disparities in the accuracy of the Modified Checklist for Autism in Toddlers with Follow-Up (M-CHAT/F).

Methods: The objective of this follow-up study was to examine whether individual M-CHAT/F items differentially predict ASD by race/ethnicity. Children who were seen in CHOP primary care between January 2011 and 2015, were aged between 16 and 26 months, and had a completed M-CHAT/F, were included in this study, yielding a cohort of 20,375 children.

Results: Chi-square analyses revealed that minoritized children had endorsed many items at significantly higher rates than their White, Non-Hispanic counterparts: Black, Non-Hispanic children had higher endorsement rates for 20 of 23 items, Hispanic children 18 items, and Asian, Non-Hispanic children 14 items. Individual logistic regressions then examined whether the relationship between item endorsement and diagnostic outcome differed by race/ethnicity. Results revealed 4 items that were differentially predictive of ASD: items 8 (X2 = 20.83; P < 0.001), 9 (X2 = 19.75; P < 0.001), 18 (X2 = 31.89; P < 0.001) and 22 (X2 = 27.69; P < 0.001). Follow-up analyses indicated that all four items were less predictive of ASD in Black, Non-Hispanic compared to White, Non-Hispanic children. For example, White Non-Hispanic children who endorsed Item 9 (bringing object to show) were 41.21 times more likely to be diagnosed with ASD, while Black, Non-Hispanic children were only 9.88 times more likely to be diagnosed with ASD. Follow-up analyses also indicated that items 18 and 22 were not predictive of ASD in Asian, Non-Hispanic children (OR = 0.22, 0.77, respectively).

Conclusion: The results of this study have identified potential areas of improvement for a widely used ASD screening tool and emphasize the importance of the follow-up interview to identify false positives. Studies should continue to utilize real-world, representative samples to better understand how the M-CHAT/F behaves across all groups. Items that are less predictive must be reconsidered for inclusion in subsequent revisions. Future screening tools should optimize for equity and accuracy at the onset, rather than after widespread implementation.