Autism Screening in Primary Care: Who Benefits and How?

Background: While the American Academy of Pediatrics recommends standardized screening for autism in primary care, the 2016 United States Preventive Services Task Force (USPSTF) statement found the evidence for this practice to be inadequate. While tools like the Modified Checklist for Autism in Toddlers (M-CHAT) are recognized as robust and early intervention is recognized as effective, the downstream impact of universal screening has not been well investigated, particularly for individuals facing known health disparities (i.e., reduced access to autism diagnostic and treatment services).

Objectives: To investigate the effects of correct identification of autism risk via standardized screening in primary care using a matched sample of children whose positive autism screens were overlooked by their providers as a naturalistic control group.

Methods: We extracted health records for 5771 children receiving care at an urban academic medical center who attended at least one 18 or 24/30 month well visit (Time 1; T1), as well as at least one second well visit between 4-6 years of age (Time 2; T2). For T1 visits, provider interpretation of the M-CHAT was compared to results from manual rescoring, yielding groups of children whose positive screens were either correctly identified or overlooked. To adjust for differences between the groups due to non-random assignment, we used 1:1 propensity score matching to select a sample in which the two groups were balanced on demographics, number of M-CHAT items failed (i.e., symptom severity), and other T1 variables. To investigate the effects of correct identification, we compared the groups on T2 diagnostic outcomes and services access using logistic regression.

Results: Of the 534 children with positive M-CHATs at T1, only 54 were correctly identified by their providers; of these, 47 had appropriate matches among the group whose positive screens were overlooked. Combined (n = 94), children in the matched samples reflect the significant racial/ethnic diversity of the clinic population (43.6% Black, 28.7% Latino, 17.0% White, 10.6% Other). Controlling for number of M-CHAT items failed at T1, correctly identified children were more likely to be diagnosed with autism (aOR = 29.42, p = .020) or another psychiatric condition (aOR = 2.90, p = .037) at T2. Notably, correctly identified and overlooked children with positive screens had similarly low rates of access to intervention at T2 (43% and 38%, respectively; aOR = 1.18, p = .719), with no differences related to demographics.

Conclusions: These results provide mixed support for universal primary care autism screening. While correct identification of a positive screen by 30 months of age was significantly associated with being diagnosed with autism by age 6, correct identification was not associated with improved access to intervention. There was little evidence that correct identification led to differential diagnostic outcomes or access to intervention based on race/ethnicity, sex, or socioeconomic status, so universal screening may mitigate health disparities in diagnosis and treatment access observed in previous research. However, the functional benefits of screening may be limited if the low rates of access to services observed here are representative for diverse populations.