Leveraging Developmental Trajectories of Broadband Screening to Detect Autism Risk in Primary Care

Background: The gap between the typical onset of autism symptoms and the average age of diagnosis remains wide, demonstrating the need for novel screening methods that detect ASD more reliably and at younger ages. Infant-sibling studies suggest that many children later diagnosed with ASD show developmental deceleration in the first two years of life, sometimes before clear autism symptoms emerge. However, this has not yet been demonstrated in low-risk samples so its screening value is unknown. This study leveraged routine developmental screening in primary care to examine whether developmental deceleration is an early indicator of ASD that can contribute to universal screening.

Objectives: Test the hypotheses that (1) developmental deceleration can be detected in a subset of low-risk primary care patients and (2) this pattern confers elevated risk for ASD.

Methods: The Children’s Hospital of Philadelphia has conducted universal screening for 10+ years across 31 primary care sites. The Survey of Well-Being in Children (SWYC) Milestones (Sheldrick & Perrin, 2013) is administered at 9, 18, and 24-30 months, according to American Academy of Pediatrics (AAP) guidelines. All patients with at least one SWYC screening and follow-up diagnostic data at ≥4 years were included in this epidemiological cohort, identified from electronic health records (N=32,280). The ASD prevalence rate in this cohort was 2.4%.

Results: Growth mixture models identified distinct developmental trajectories of SWYC scores from 9-30 months; a four-class model provided the best fit. Class 1 (67% of the cohort) had 9-month scores that met age expectations and significantly increased from 9-30 months. This class (average posterior probability [APP]=.92) had a lower probability of a later ASD diagnosis (0.5%) compared to the entire cohort (2.4%). Class 2 (19%, APP=.78) also met age expectations at 9 months, but showed a more modest increase from 9-30 months; the rate of ASD in this class was 1.6%.

In contrast, Class 3 (10%, APP=.81) showed developmental deceleration from 9-30 months and had an elevated rate of ASD (7.1%). Class 4 (4%, APP=.89) had lower 9-month scores and more significant developmental deceleration, with a very elevated rate of ASD (27.0%). Two-thirds of children with ASD were classified into Class 3 or 4 (i.e., sensitivity=74%). Specificity of these developmental profiles was 88%, positive predictive was 13%, and negative predictive value was 99%. Additional analyses will combine SWYC developmental trajectories and M-CHAT/F results to determine if the combination yields more accurate screening than either alone.

Conclusions: These data demonstrate that clear developmental deceleration is detectable through routine developmental screenings, and when this pattern is present, it confers elevated risk for ASD. As such, screening for developmental deceleration in the first two years of life may improve the status quo of universal screening in primary care. One important benefit of this approach is that it leverages tools pediatricians are already administering as a part of routine clinical care. Though this study used screenings from 9-30 months (given guidelines for screening at these ages), developmental deceleration may be detectable even earlier than 24-30 months with more repeated developmental screenings.