An Initial Evaluation of the Validity of the Gilliam Autism Rating Scale-Third Edition (GARS-3) in a Clinical Sample

Background: Third party rating scales are often used to support diagnostic decision making in Autism Spectrum Disorder (ASD) diagnostic evaluations.  One such measure is the Gilliam Autism Rating Scale-Third Edition (GARS-3; Gilliam, 2014).  The GARS-3 is a standardized norm-referenced instrument designed to assist in the diagnosis of ASD.  The GARS-3 is keyed to correspond with DSM-5 diagnostic criteria and was standardized with a sample of 1,859 individuals with ASD 3 to 22 years of age.  Gilliam (2014) provides psychometric support for the GARS-3 as evidenced by internal consistency of .79 - .94 for GARS-3 subscales and .93 - .94 for GARS-3 Autism Index scores.  Gilliam also reports criterion-related validity in support of the GARS-3 as evidenced by significant correlations between the GARS-3 Autism Index and (a) the Childhood Autism Rating Scale, Second Edition (CARS-2) and (b) Autism Diagnostic Observation Schedule.

Objectives:  The authors conducted an initial evaluation of the validity of the GARS-3 by correlating scores with well-established measures of autism symptomatology.  Investigators also contrasted GARS-3 scores between individuals who met DSM-5 diagnostic criteria for ASD and those individuals who did not. 

Methods:  Participants were 20 individuals (M age = 8.23 yr; 80% Male; 80% Caucasian; 12 diagnosed with ASD, 8 diagnosed with another DSM-5 disorder) who participated in comprehensive diagnostic evaluation at a training clinic at a land grant university in the Southeast U.S.  Diagnostic evaluation was established using results from the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) and CARS-2.  Caregivers completed the GARS-3 as part of the diagnostic evaluation; GARS-3 ratings were collected for the purposes of research and were not used in diagnostic decision making. 

Results:  Preliminary results indicate significant relationship between ADOS-2 Module 3 Total Scores and CARS-2 T Score, r(3) = .87, p < .05.  ADOS-2 Module 2 Total Score also correlated highly with the CARS-2 T Scores, r(5) = .74, p < .05. ADOS-2 Module 3 Total Score and the GARS-3 Autism Index were not correlated, r(3) = -.04, ns. CARS-2 T Score and the GARS-3 Autism Index were also uncorrelated, r(18) = .06, ns.  CARS-2 T Scores differed across ASD and non-ASD groups, t(18) = 3.42, p < .05.  Individuals with ASD diagnosis (n = 12) earned similar GARS-3 Autism Index scores (M = 79.92; SD = 18.62) to individuals without ASD diagnosis (n = 8; M = 93.75; SD = 10.63); t(18) = -1.90, ns. 

Conclusions:  Preliminary results suggest weak relationships between the GARS-3 Autism Index Score and the ADOS-2, and the GARS-3 Autism Index Score and CARS-2 T Score.  ADOS-2 and CARS-2 T scores differed across diagnostic groups, with individuals with ASD earning significantly higher scores.  GARS-3 Autism Index scores, however, did not differ between groups with ASD and those without ASD.  Initial findings warrant caution in using the GARS-3; however, larger samples are needed to fully document the utility of the GARS-3 in diagnostic evaluation.