Evaluation of the Factors Associated with at-Risk for Autism Designation During Developmental Screening, and Agreement Between at-Risk Designation and Subsequent Clinical Diagnosis In Young Children: A Retrospective, Observational Study

Background: Autism is a life-long disability that is prevalent in as many as 1 in 90 children. Early-age screening, diagnosis, and service provision are critical to achieving optimal outcomes.  Children's Specialized Hospital created a Developmental Screening Clinic (DSC) to improve early screening to identify developmental delays, autism and autism spectrum disorders, and to improve access to care, inter alia.

Objectives: This analysis focused on examining pre-specified factors for why a child is given an at risk/not at risk for autism designation during screening, and separately, the factors that explain agreement between the at risk/not at risk designation and subsequent clinical diagnosis of autism/no autism during the first two years of the DSC.  The factors examined for both questions include: age and sex (male as referent category) of the child; the site at which the screening occurred (site A as referent), and; the year of screening.   

Methods: The number of children screened, mean and median ages of children screened for the 2-year period, and percentage male/female were calculated.  For the confirmatory analysis, the proportion at risk of autism and the proportion of cases for which there was agreement were both calculated.  Logistic regression was employed to estimate the adjusted odds ratios of the factors associated with an at-risk designation and agreement with clinical diagnosis.

Results: 879 children ages 8 months to 79 months were screened for Autism/ASD.  Mean and median ages for the children were 32.6 and 32 months, respectively, with 74.7% (657/879) being male. Just over 58% (513/879) of children were at risk of autism. Of the 556 children for whom a clinical diagnosis was available, 58.8% (327/556) were diagnosed with autism. Adjusting for age and sex, children screened at site B and C were 58% less likely (Odds Ratio [OR], 0.42) and 7% more likely (OR, 1.07), respectively, to be designated as at risk than children screened at site A (the referent site).  In addition, children screened during year two were 85% (OR, 1.85) more likely to be designated as at risk of autism than children screened in the first year.   Agreement model results indicated that after again adjusting for age and sex, children screened at site B and C were 38% less likely (OR, 0.62) and 51% less likely (OR, 0.49), respectively, to have agreement between the screening and final diagnosis than children screened at site A (the referent group).  In addition, children screened during year two were 2.3 times more likely (OR, 2.3) to have agreement between the screening and final diagnosis than children screened in the first year of the DSC.

Conclusions: Our results suggest that spatial (by site) and temporal (across years) variation existed regarding the likelihood of at-risk for autism designation and agreement between this designation and agreement with the subsequent clinical diagnosis.  The findings suggest that unmeasured factors resulted in a difference in the patient selection mechanism across sites and time, the quality of the screening/clinical evaluation processes, or both.