There are indications that the emphasis on early diagnosis of autism spectrum disorder has produced important positive changes in screening and diagnostic practices. Universal autism screening leads to earlier referral for evaluation, earlier age of diagnosis, and earlier initiation of appropriate, autism-specific interventions (Valicenti-McDermott, 2016).
Along with these changes have come questions about how early it is possible to reliably detect and diagnose autism, including specific concerns about the impact of the DSM-5 definition on early identification (Barton et al., 2013; Christiansz et al., 2016)
The Toddler Module of the Autism Diagnostic Observation Schedule – 2nd Edition (ADOS-2) has shown considerable promise at helping clinicians make the diagnosis of autism spectrum disorder before age 30 months. However, even with the best tools available, there is continuing concern that some young children’s ASD diagnoses are missed, only to be picked up at a later evaluation.
A recent paper (Bacon et al., 2017) offer reflections on the state of the art, as well as new data further illuminating those concerns. Bacon and colleagues first summarize the literature supporting the following point:
- “. . . during the first 12 months of life, babies who eventually develop ASD appear largely indistinguishable from those who do not: they exhibit similar levels of social engagement, babbling, and visual attention patterns. . . Then, between 12 and 24 months, reduced levels of social attention and social communication as well as increased repetitive behavior with objects emerges.” (p. 1)
They then go on to report on “a large sample of 1- to 2-year-olds identified by failure of universal developmental screening procedures or parent or physician concern from the general pediatric population” in which “early developmental patterns were examined longitudinally until age 3 years.” They prospectively examined four groups of toddlers: early-onset ASD (toddlers who were identified as at risk for ASD at initial evaluations and all subsequent evaluations). late-onset ASD (toddlers who were identified as nonspectrum at initial evaluations, but then later identified as at risk for ASD), and non-ASD language delay (toddlers with language delay identified at the initial evaluation) as well as typically developing (toddlers who were identified as typical at all assessments and did not have a history of autism in the family). All ASD diagnoses were confirmed at age three years.
Results of the study indicated that, of the 107 toddlers eventually identified with ASD:
- 69 met diagnostic criteria for ASD at their initial evaluation and continued to have an ASD diagnosis at all subsequent evaluations. At their initial evaluation, 5 fell in the mild-to-moderate concern level on the ADOS-T[oddler Module] and 64 fell in the moderate-to-severe concern level.
- 38 toddlers were initially identified as nonspectrum at ages <24 months but met diagnostic criteria for ASD at a subsequent evaluation. At the initial evaluation, 21 of the 38 fell in the little-to-no concern level on the ADOS Toddler Module (ADOS-T).
Thus, in one out of three toddlers eventually diagnosed with ASD, the diagnosis was missed by experienced clinicians in an evaluation that included the ADOS-2 Toddler Module.
The Toddler Module worked reasonably well for the 52 toddlers with non-ASD language delay although some exceeded the “concern” cutoffs (4 in the mild-to-moderate concern, and 5 in the moderate-to-severe concern level) on the ADOS-2; none were given the clinical diagnosis of ASD. None of the typically developing toddlers exceeded the ADOS-2 cutoffs. The figure below from the Bacon et al. (2017) paper summarizes the ADOS-2 results for the four groups.
Bacon and colleagues’ paper also includes data related to eye-tracking, exploration, and a parent–child interaction (PCI) task, each of which showed promise for distinguishing toddlers with ASD, even those not given the clinical diagnosis until later. However, the use of such fine-grained behavioral coding tasks is still at the experimental stage and the procedures are not ready for clinical application.
The authors concluded that:
- “Use of the ADOS gold-standard diagnostic instrument did not prevent missed early diagnoses despite its use by clinicians with specialized experience in the assessment of ASD and its administration to all participants at equally young ages” (p. 13); they did not feel this was attributable to autistic regression, at least not in cognitive and language domains;
- “Differences in age of diagnosis may be due to severity of early language impairment [the later diagnosed group started out with slightly less impaired expressive and receptive language], applicability of diagnostic criteria at young ages, and/or sensitivity of standard diagnostic tests.” (p. 14)
The later-diagnosed group “received a provisional ASD diagnosis an average of 8.7 months later than the [early-diagnosed] group and therefore inevitably began receiving autism-focused treatment later as well”; the authors concluded that there is a need for “more sensitive and accurate early screening methods that can detect even the [later-diagnosed group] individuals at early ages.” (p. 14)
Further, they concluded that “. . . if autism is not detected, then it may be because it was not clinically observed by experts instead of not yet present. The basic research and clinical concern is that 1 out of 3 toddlers with ASD missed at early ages is a substantial percentage, especially considering that this occurred despite use of state-of-the-field diagnostic tools and highly experienced clinicians and despite the early presence of social and stereotyped behavior abnormalities on nonstandard specialized tests.” (p. 14; emphasis added)
The take-home message from this study is not to question the value of the ADOS-2 Toddler Module; it is still the best available observation tool for use with children under 30 months. And these data may serve to reduce some clinicians’ concerns about specificity; none of the typically-developing toddlers exceeded the ADOS-2 cutoffs. Nonetheless, it is clear that even in the best of circumstances, the Toddler Module misses early indicators of autism in some children who subsequently clearly meet the diagnostic criteria for ASD. The greater sensitivity of other behavioral assessment procedures offers hope that the next generation of diagnostic instruments will reduce the frequency of overlooked ASD diagnoses in very young children.
Bacon, E. C., Courchesne, E., Barnes, C. C., Cha, D., Pence, S., Schreibman, L., … Pierce, K. (2017). Rethinking the idea of late autism spectrum disorder onset. Development and Psychopathology, 1–17. https://doi.org/10.1017/S0954579417001067
Barton, M. L., Robins, D. L., Jashar, D., Brennan, L., & Fein, D. (2013). Sensitivity and specificity of proposed DSM-5 criteria for autism spectrum disorder in toddlers. Journal of Autism and Developmental Disorders, 43, 1184–1195. https://doi.org/10.1007/s10803-013-1817-8
Christiansz, J. A., Gray, K. M., Taffe, J., & Tonge, B. J. (2016). Autism Spectrum Disorder in the DSM-5: Diagnostic Sensitivity and Specificity in Early Childhood. Journal of Autism and Developmental Disorders, 46, 2054–2063. https://doi.org/10.1007/s10803-016-2734-4
Valicenti-McDermott, M. (2016). Age of Diagnosis of Autism Spectrum Disorder in an Ethnically Diverse Population Before and after the 2007 AAP Recommendation for Universal Screening. Paper presented at the Pediatric Academic Societies Annual Meeting, May, 2016. Baltimore, MD.