The diagnosis of autism in adults often involves the use of tools recommended by NICE guidance but which are validated in children. The purpose of the paper is to establish the strength of the association between the Autism Diagnostic Interview-Revised (ADI-R) scores and the final clinical outcome in an all intellectual quotients adult autism diagnostic service and to establish if this in any way relates with gender and intellectual ability.
The sample includes referrals to Leeds Autism Diagnostic Service in 2015 that received a clinical outcome. Sensitivity, specificity and positive and negative predictive values were calculated to evaluate ADI-R and final clinical outcomes. Logistic regression model was used to predict the effect of the scores in all the domains of ADI-R and the two-way interactions with gender and intellectual ability.
ADI-R has a high sensitivity and low specificity and is useful to rule out the presence of autism, but if used alone, it can over diagnose. Restricted stereotyped behaviours are the strongest predictor for autism and suggests that the threshold should be increased to enhance its specificity.
This is a single site study with small effect size, so results may not be replicable. It supports the combined use of ADI-R and Autism Diagnostic Observation Schedule and suggests increasing ADI-R cut-offs to increase the specificity.
The clinical team may consider piloting a modified ADI-R as suggested by the results.
To the authors’ knowledge this is the only study of ADI-R in an adult population of all intellectual abilities.
