Stability of the ADI-R
17
autism population (cf. Edelson, 2006). Thus, at T1 30% of participants had IQ scores of ≤50.
Furthermore, since the inclusion criteria also required them to have received a formal ASD diagnosis
prior to initial assessment at mean age 3.5 years, it is likely that the sample may include a higher
proportion of children with more severe difficulties than is typical of the general autism population.
(Many children in the UK do not receive a confirmed diagnosis of autism until after this age; Howlin &
Moore, 1997).
It might also be argued that analysis using ratings of 0-3 on the ADI-R increases the chances of finding
significant change over time (changes in scores from 3 to 2 on the ADI-R algorithm are considered to
be less significant than change from 2 to 1 or 1 to 0; Starr et al., 2003). Nevertheless, our results are
consistent both with those of Fecteau et al. (2003), who used the same scoring method and those of
Starr et al. (2003), who employed the more conservative method of collapsing scores 2 and 3. Given
the relative severity of intellectual ability in the current sample compared to the Starr et al. (2003) and
Fecteau et al. (2003) studies, the consistency with which improvements in autistic symptomatology
have been observed over time suggests that this profile of change is not an artefact of measurement but
is likely to reliably represent the typical course of development of the disorder.
Summary
The findings of the current study suggest that ADI-R classification of autism at a pre-school age
remains stable throughout the elementary school years, although some children initially meeting
criteria for autism went on to meet ASD criteria or fell below this threshold at FU. However, the
severity of autism-specific symptoms improves over time. As found in previous research,
improvements are more likely to occur within the areas of communication and social interaction than in
the repetitive behaviour domain. Overall, children who make the most improvements in symptom