ADOS Module 4: Increases in Sensitivity and Comparability to Other Modules Using a Revised Diagnostic Algorithm

Background: Given the life-long course of autism spectrum disorders (ASD), it is a priority for the field to increase our understanding of how symptoms of ASD manifest across the lifespan. Many longitudinal studies of ASD have included the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000), making it possible to use ADOS scores to examine developmental trajectories of ASD symptoms. Diagnostic algorithms for ADOS Modules 1-3 have been revised to improve diagnostic validity and increase item overlap across modules, thereby facilitating comparisons of scores across childhood and early adolescence (Gotham et al., 2007). Algorithms for the recently published ADOS-Toddler module (Luyster et al., 2009) also follow a similar structure to that of Modules 1-3, making it possible to examine symptom trajectories using the ADOS from a very young age. However, the algorithm for Module 4, used with verbally fluent adolescents and adults, has not yet been revised. This impedes comparison of ADOS scores across the lifespan.

Objectives: To make the Module 4 algorithm consistent with the ADOS-Toddler algorithm and revised algorithms for Modules 1-3.

Methods: A total of 327 Module 4 administrations were collected from 288 individuals (M age=22.31 years). For these 288 individuals, Best Estimate Clinical Diagnosis (BEC) included Autism (33.7%), ASD (31.9%), Nonspectrum (28.8%) and Typical/No Diagnosis (5.6%). Following procedures used to create revised algorithms for Modules 1-3, “preferred” items were chosen for their ability to successfully discriminate Autism and Nonspectrum groups. Exploratory multi-factor item response analysis was used to organize items into new domains and goodness-of-fit was verified using Confirmatory Factor Analysis. After summing items to create an algorithm total, Receiver Operating Curves (ROC; Siegel et al., 1989) were calculated using BEC as the reference standard to determine appropriate diagnostic cut-offs. Sensitivity and specificity were compared for the original algorithm (based on cut-offs for the Communication, Social and Combined domain totals) and the newly proposed algorithm.

Results: Several of the Module 4 preferred items overlapped with items comprising the revised Module 3 algorithm. Results of exploratory and confirmatory factor analyses supported a 2-factor solution, consistent with other modules. Using the original algorithm, 66.1% of individuals with a BEC of Autism met criteria for an ADOS classification of Autism, while 70.5% of individuals with a BEC of ASD met or exceeded thresholds for an ADOS classification of ASD. Specificity of the original algorithm ranged from 81.9 to 93.9%.  In contrast, the revised algorithm yielded improved sensitivity, with 86.2% of individuals with Autism meeting the Autism cut-off and 83% of individuals with ASD scoring above the threshold set for ASD. Specificity using the revised algorithm ranged from 85.1 to 90.5%.

Conclusions: In addition to promoting comparability with algorithms used for the ADOS-Toddler and Modules 1-3, revisions to the Module 4 algorithm yielded substantial gains in sensitivity while maintaining similar levels of specificity. Availability of a revised Module 4 algorithm will facilitate future efforts to examine trajectories of ASD symptoms across the lifespan.