Identifying autism in young children

Early Detection of Infants and Young Children with Autism


Associate Professor Kylie Gray
Professor Bruce Tonge

The use of screening tools by primary care physicians is currently the best method we have for the early identification of children with Pervasive Developmental Disorders. Screening of at risk children who have developmental problems in order to channel referrals to specialist clinical assessment services for autism is a potential way of ensuring that children who are likely to have autism are referred for specialist diagnostic services at as early an age as possible.

Work at the Monash University Centre for Developmental Psychiatry and Psychology has resulted in the development of a screening tool for Pervasive Developmental Disorders (autism) in children with developmental delay aged 18-48 months, Developmental Behaviour Early Screen (DBC Early Screen). An evaluation of its performance within a community population of children with developmental delay has also been undertaken.

Due to previous work which highlighted the potential of the Developmental Behaviour Checklist (DBC) as a screening tool for autism in young people aged 4-18 years with intellectual disability, the efficacy of the DBC as a screening tool for autism in children with developmental delay aged 18 - 48 months was evaluated. Subjects consisted of 60 children with autism and developmental delay and 60 children with developmental delay without autism. Parents of the children completed the 96 item DBC rating the behaviour of their child within the past 6 months.

Analyses aimed to identify those items of the DBC which best predicted the diagnosis of autism. Univariate logistic regressions were performed to establish which items of the DBC differentiated the autism and control groups. A confirmatory factor analysis was performed with the 30 items identified by the univariate logistic regressions. Factor loadings were then used to develop the DBC screening algorithm. Receiver Operating Characteristics (ROC) analysis was used to evaluate the overall performance of the DBC algorithm as a screening tool for autism. Using a cut point of 0.60 or greater, 17 DBC items were selected to create a DBC autism screening algorithm. Analyses revealed that with a cut-off score of 11 this 17-item version of the DBC-P performed well as a potential screening tool, with an Area Under the Curve of 0.874, sensitivity of 0.8750, and specificity of 0.6909.

The evaluation of the DBC Early Screen involved the screening of a community sample of children referred to a community service for children suspected of developmental delay using the DBC Early Screen. In a sample of 22 children (aged 23 - 49 months), 15 screened positive using the DBC Early Screen. Examining the sample in terms of those who received the a diagnosis of a Pervasive Developmental Disorder (Autistic Disorder, PDD NOS, and Asperger's Disorder) and those who did not, resulted in a sensitivity of 0.82, specificity of 0.80, overall classification efficiency of 0.82, predictive value of a positive test of 0.93, and predictive value of a negative test of 0.57.

When screening for developmental problems in infants and young children sensitivity, specificity, and positive predictive values of 70 - 80% are regarded as acceptable. Work to date has shown that the DBC Early Screen has good sensitivity and specificity in terms of identifying cases of Pervasive Developmental Disorder from samples of infants and young children with developmental delay. A larger community field trial is planned to establish its efficacy as a population screening tool.