The Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000) is widely accepted as a “gold standard” diagnostic instrument, but its downward limits restrict utility in research samples of very young children at risk for ASD.

Objectives:

The objective of the present investigation was to modify the ADOS for use in very young children. 

Methods:
A modified ADOS, the ADOS-Toddler module (ADOS-T), was used in 272 evaluations.  Participants included children 12-30 months of age with either best estimate diagnoses of ASD, non-spectrum developmental delay or typical development.  A final set of protocol and algorithm items was selected based on their success in discriminating the diagnostic groups. 

Results:

For all children ages 12-20 months and nonverbal children 21-30 months, a new ADOS-T algorithm yielded a sensitivity and specificity both of 95%.  For verbal children 21-30 months, the revised “Some Words” Module 1 algorithm (Gotham et al., 2007) yielded a sensitivity of 93% and a specificity of 95%. 

Because of the variability in early development, we proposed to use scores on the algorithm to indicate a range of concern, rather than employ the traditional cutoffs.  For all children ages 12-20 months and nonverbal children 21-30 months, a score of 12-15 on the ADOS-T algorithm indicates mild concern, and scores >15 indicate significant concern.  Similarly, for verbal children 21-30 months, scores of 7-10 on the “Some Words” algorithm indicate mild concern, and scores >10 indicate significant concern.

Conclusions:

The ADOS-T appears to be a useful addition to the existing ADOS.  It is appropriate for use in children as young as 12 months and successfully discriminates between diagnostic groups, yielding high sensitivity and specificity.  Given the variability of behavior in very young children, the proposed “range of concern” may be a more useful interpretation of algorithm results than the traditional “cutoffs” approach.