Psychological markers in the detection of autism in infancy in a large population.

BACKGROUND Investigation to see if there are key psychological risk indicators for autism in a random population study of children at 18 months of age; and to assess how well these discriminate children who receive a diagnosis of autism from other forms of developmental delay. METHOD Sixteen thousand children in the southeast of England were screened for autism by their health visitor or GP, during their routine 18-month-old developmental check-up, using the CHAT (Checklist for Autism in Toddlers). From a previous high-risk study we predicted that children at 18 months of age who failed three items ('protodeclarative pointing', 'gaze-monitoring', and 'pretend play') would be at risk for receiving a diagnosis of autism. From other evidence, we further predicted that those 18-month-olds who failed one or two of the key items (either pretend play, or protodeclarative pointing and pretend play) would be at risk for developmental delay without autism. RESULTS Twelve children out of the total population of 16,000 consistently failed the three key items. Of these, 10 (83.3%) received a diagnosis of autism. Thus, the false positive rate was 16.6% (2 out of 12 cases), and even these 2 cases were not normal. When the 10 children with autism were reassessed at 3.5 years of age, their diagnosis remained the same. Thus the false positive rate among the cases diagnosed with autism was zero. In contrast, of 22 children who consistently failed either protodeclarative pointing and/or pretend play, none received a diagnosis of autism, but 15 (68.2%) received a diagnosis of language delay. CONCLUSIONS Consistent failure of the three key items from the CHAT at 18 months of age carries an 83.3% risk of autism; and this pattern of risk indicator is specific to autism when compared to other forms of developmental delay.