Autism Screening: Where Do We Go From Here?

The lengthy diagnostic odyssey experienced by families of children with autism spectrum disorder (ASD) remains a major barrier to early access to ASD therapeutic services. Although accelerating this process has been identified as a top priority,¹  there is a polarized debate about whether ASD screening can accomplish this. The American Academy of Pediatrics recommends that all children be assessed by using an ASD screening tool at 18 and 24 months old.²  In contrast, the US Preventive Services Task Force concluded that there was “insufficient evidence to recommend screening for ASD in children aged 18 to 30 months for whom no concerns of ASD have been raised.”³  The issue has been debated passionately,^4–6  but without resolution, in part because of limited community-level data regarding screening feasibility, accuracy, and outcomes. The current study by Guthrie et al⁷  contributes important findings that further inform this debate.

The authors demonstrated that universal screening was feasible, with >90% of toddlers screened through a large primary care network by integrating the Modified Checklist for Autism in Toddlers (M-CHAT) With Follow-up (M-CHAT/F) into routine primary health care. Physicians were prompted to administer the questionnaire through an electronic health record system, into which the results were incorporated and used, providing an excellent example of a “Learning Healthcare System.”⁸  However, the timing of screening was variable, with only 47.8% of children screened at the recommended ages of 18 and 24 months,²  and the follow-up interview was completed for only 41.2% of children whose initial M-CHAT/F scores warranted it. Given that ASD diagnostic rates were lower among children who received the follow-up interview, the authors suggested that this next step may have been completed selectively “to confirm a clinical opinion of not [having] ASD.” However, the positive predictive value (PPV) was only 9.6% among children whose questionnaire scores should have, but did not, lead to a follow-up interview. Overall PPV, taking into account all “screen-positive” children, was only 14.6%. One of the unique strengths of the M-CHAT/F is the 2-step assessment process, which is designed to maximize both sensitivity and specificity. Inconsistent administration of the follow-up interview undermined this strength, yielding a PPV that is questionably too low to trigger an automatic referral for specialized assessment. Whereas the PPV increased to 72.6% when outcomes included other developmental concerns, the sensitivity dropped to 11.7%, limiting the utility of M-CHAT/F screening for this purpose.

Another important finding was that child and family sociodemographic factors influenced screening rates and accuracy. It has been argued that ASD screening may be especially useful for at-risk families who experience disparities in health care. Rates of screening for all groups exceeded 80%, contributing significant progress toward leveling the playing field.⁵  However, despite comparable sensitivities, the specificity and PPV estimates were lower for at-risk groups, and the highest screening rates were reported for children from white, English-only–speaking, high-income, and privately insured families.

Finally, screening may have accelerated the diagnostic process. The authors reported that the mean time to diagnosis was 7.5 months shorter for children who screened positive than for those who screened negative. Notably, Robins et al⁹  found that an earlier version of the M-CHAT identified ASD symptoms more consistently than general inquiry about parental concerns, and similar findings have been reported for the Infant-Toddler Checklist.¹⁰  Although the reported M-CHAT/F sensitivity of 38.8% seems low, this estimate was based on children diagnosed with ASD up to ages 4 to 8 years. Children who are diagnosed later tend to have less severe symptoms and developmental delays,¹¹  so ongoing monitoring, which includes multiple assessments over time, may be needed to accurately identify children across the continuum of clinical expression. It is likely that some of the false-negative cases from the current study would not have been diagnosed even if the case patients had been referred shortly after the initial screen. Recent data from prospective cohorts of high-risk infant siblings indicate that as many as 45% of those diagnosed with ASD at age 3 years were not diagnosed at age 2 years despite comprehensive assessments.¹² 

Ultimately, the potential added value of ASD screening must be considered relative to what would occur in its absence. Identification of the early behavioral signs of ASD varies by the knowledge and experience of the primary care physician as well as disparities related to family sociodemographic factors, contributing to delays in ASD diagnosis. Although it is difficult to object to the guidance from the US Preventive Services Task Force³  to listen carefully to parents’ concerns, we must acknowledge the false dichotomy between screening and surveillance in this context. Developmental surveillance has been defined as a flexible, continuous process whereby knowledgeable professionals collect relevant information from multiple sources (including parent report and direct observation) toward the goal of identifying and addressing developmental concerns.¹³  To that end, there is no reason to default to open-ended inquiry. Why not use the best available measurement tools to identify developmental concerns with the highest possible accuracy? Arguably, the M-CHAT/F remains a strong candidate in that regard.

ASD

autism spectrum disorder

M-CHAT/F

Modified Checklist for Autism in Toddlers With Follow-up

PPV

positive predictive value