Stability of the ADI-R
algorithm cut-off scores for autism, domain and total level raw scores was good (Kappa = 1.0 for all
cut off scores and Intraclass Correlation Coefficients for domain and total level scores were >.86 ).
Procedure: T1 and FU assessments and interviews were carried out by three experienced research
psychologists (SD, T1 assessments only; JM, FU assessments only; IM, T1 & FU assessments). All T1
assessments were conducted at home; most FU child assessments were administered at school but 5
children were assessed at home at parental request. All parent interviews were conducted with the main
caregiver at both T1 and FU. Child assessments were conducted within 8 weeks following the parent
interviews.
Data analysis: The stability of ADI-R scores from T1 to FU was assessed on three levels: algorithm
cut-off, domain and item level scores. Nonparametric analyses were employed where data were not
normally distributed. McNemar tests were conducted to investigate changes in ADI-R algorithm status;
paired t-tests were conducted on each domain and the total ADI-R score. Wilcoxon Matched-Pairs
Signed-Ranks tests were conducted to assess the stability of the ADI-R at item level. T1 variables
associated with ADI-R scores at FU were identified using Pearson correlational analyses. Paired t-tests
or Mann-Whitney U tests (as appropriate) were conducted in order to explore any T1 differences
between individuals whose ADI-R classification changed from T1 to FU and those whose classification
remained stable and between those individuals who showed the most and the least improvement in
ADI-R scores. A conservative p value of ≤.01 was employed in all analyses.
Results
Stability of ADI-R diagnostic criteria
All 35 participants scored above the algorithm cut-off for autism on all three domains of the ADI-R at
T1. At FU 28 participants (80%) scored above the cut off for autism on all three domains; 32 (91.4%)