Do Referral Factors Predict a Probable Autism Spectrum Disorder Diagnosis? A DBPNet Study

To determine the proportion of children referred to academic medical centers with concerns about autism spectrum disorders (ASDs) who received a probable ASD diagnosis, identify factors predicting ASD diagnosis, and describe the children with ASD concerns who were not found to have autism.

A total of 55 developmental-behavioral pediatricians (DBP) at 12 academic sites in the DBPNet research network recorded data on ≤15 consecutive new patients. They coded presumed diagnoses after their first visit with the child.

Of 784 new visits, 324 (41%) had concern for ASD; of these, 221 (68%) were presumptively ASD+; 103 (32%) were ASD-. In a mixed model accounting for clustering within site and covariates significant in bivariate analysis, significant predictors of receiving a presumptive ASD diagnoses were socialization concerns, languages other than English spoken in the home, and coming for second opinion. Also concern for "other behavior problems" (not mood, oppositionality, anxiety, attention, or repetitive behaviors) predicted not receiving ASD diagnoses. This model was not clinically useful because it misclassified 26.9% of children. ASD- children <4 years old had more language delay and less cognitive impairment and socialization concern than their ASD+ age peers. ASD- children ≥4 years old were more likely to have attention-deficit /hyperactivity disorder (ADHD) and learning disability with normal cognition than their ASD+ age peers.

Two thirds of children referred to academic centers with concern for ASD received a presumptive diagnosis of ASD. While those with ASD were not easily distinguished from those without ASD at referral, virtually all children with ASD concerns had multiple DBP diagnoses made and required DBP follow-up care.