The Use of the Developmental Behaviour Checklist as a Screening Instrument for Autism

Investigators:

Dr. Avril Brereton, Monash University
Professor Bruce Tonge, Monash University
Associate Professor Stewart Einfeld, University of New South Wales
Associate Professor Andrew Mackinnon, Monash University
Dr. Kylie Gray, Monash University

The effectiveness of the DBC as a screening instrument has been evaluated.  A recent study has determined that a subset of items within the DBC can be used as a reliable and valid autism screening tool (Brereton, Tonge, Mackinnon & Einfeld, in press).  A 29-item scale — the Developmental Behaviour Checklist-Autism Screening Algorithm (DBC-ASA) — was developed using items from the DBC and evaluated in a sample comprising 180 children meeting criteria for DSM-IV Autism and 180 controls matched for age, sex and IQ range.  The DBC-ASA has good validity in discriminating young people (4-18 years) with autism and IQ ranging from normal to severe intellectual disability from others using a cut-off score of 17.  These results suggest that the DBC-P may be useful in identifying children and adolescents who may have autism and who require a diagnostic assessment.

The identification of candidate screening items and evaluation of their performance involved three stages.  First, univariate logistic regression analyses were conducted to identify those DBC items which were predictive of group membership. Second, confirmatory factor analysis techniques were used to extract a single common factor from the predictive items which maximally aligned with the diagnosis of autism. loadings on this factor formed the basis for the final selection of items in the screening instrument. Third, the selected items with unit weighting were summed to form a scale which was subject to a receiver operating characteristic (ROC) analysis (Kramer, 1988) to assess the power of the DBC to discriminate between group membership and to assist determination of the most useful cutpoint on the scale.  The sensitivity and specificity of the screen at this cutpoint was also calculated.

Items with loadings greater than 0.70 in magnitude were included in the scale with the addition of three items (‘unaware of other’s feelings’, ‘repeats words and phrases’, and ‘upset and distressed over small changes in routine’). These additional items also loaded highly at greater than or equal to 0.62. From a purely statistical point of view there was no reason to add these items, however from a clinical point of view these items are specifically suggestive of autism and were therefore included.   Twenty-nine items formed the DBC Autism screening algorithm (DBC-ASA).

The 29 items with unit weighting were subject to ROC analysis to assess overall diagnostic performance.  Unit weights were chosen because they are likely to provide a more stable result across samples than using weights from the factor analysis itself.  The area under the ROC curve (AUC) of 0.80 (95% C.I.: 0.75 to 0.84) demonstrates that the DBC-ASA provides good differentiation between autism cases and non-cases.  Three alternative cut-points were evaluated.  Increasing the cut-point above 17 resulted in moderate decrements in sensitivity for only modest improvements in specificity.  Accordingly, the classification of scores of 17 or greater as cases was adopted. At this value, the test achieved a sensitivity of 0.86 (95% CI: 0.80 – 0.91) and a specificity of 0.69 (95% CI: 0.62 – 0.76).  Internal consistency a = .94.

The results suggest that the DBC-ASA effectively detects children who may have autism from other children with developmental delay. A clinical field trial is now necessary to test the application of the DBC-ASA as a useful screening instrument.

Further Information

Dr. Avril Brereton Tel: (03) 9905 1402