Standardized ADOS-2 Toddler and Module 1 Severity Scores in a Clinical Sample

Background:  

The ADOS-2 is a comprehensive assessment instrument which forms part of the recommended “gold standard” for the assessment of Autism Spectrum Disorder (ASD).  While the instrument is primarily designed to classify cases of ASD based on raw scores, it is also possible to derive an algorithm of symptom severity from total scores.  Standardized calibrated severity scores (CSS) provide a metric of the severity of autism-specific symptoms and behaviors and, unlike raw scores, are reasonably independent of factors such as IQ (particularly Verbal IQ), language and age.  Severity scores were formally introduced for Modules 1-4 in the ADOS-2 revision.  Recently, Esler et al. (2015) calculated total and domain CSS for the ADOS-2 Toddler Module, but the algorithm has yet to be verified in an independent sample. 

Objectives:  

Our aim was to examine the reliability and clinical utility of the ADOS-2 CSS for the Toddler Module in an independent clinical sample.  A sample of children assessed with Module 1 were included for comparison purposes.  Specifically, we focused on the application of ADOS-2 CSS algorithms in children aged 3 years and under. 

Methods:  

The sample consisted of 176 children (male = 88.1%; M_age = 32.21 months, SD = 6.23, range = 17-46 months).  Of the 176 assessments, 125 received a diagnosis of ASD (71%), 44 (25%) received non-spectrum diagnoses and seven (4%) children did not receive any medical diagnosis.  Participants received extensive developmental and diagnostic assessments.  Final diagnostic status was based on best estimate clinical (BEC) judgement involving DSM-5 criteria, parent interview, and assessment results including the ADOS-2.  The final sample included 80 Toddler Module assessments (non-Verbal, NV = 62; some words, SW = 18) and 107 Module 1 assessments (NV = 70; SW = 37). 

Results:  

The CSS was more stable than raw scores across age and language as evidenced by fewer significant differences between scores across modules.  Younger non-verbal participants scored higher than other groups, however effect size was reduced for the CSS compared to total and domain raw scores.  The CSS was highly correlated with the clinician completed CARS2-ST, but not with the parent completed Autism Spectrum Rating Scale (ASRS).  Neither gender nor chronological age were found to be associated with raw scores or the CSS.  Linear regression indicated that verbal DQ was a significant predictor for raw scores but not the CSS.  Auditory comprehension and Child Behavior Checklist (CBCL) externalizing behavior were significant predictors of total and domain raw scores and the CSS.   Raw and CSS RRB scores were significantly influenced by nonverbal DQ and auditory comprehension. 

Conclusions:  

Our findings are consistent with previous research suggesting that, in clinical populations, ADOS-2 standardized scores are less influenced by participant characteristics than raw scores.  These results offer support for the introduction of the Toddler Module CSS into clinical practice.