ADOS Severity Scores Predict Performance on a Classic Measure of Intentional Understanding in Preschoolers on the Spectrum

Background: ADOS severity scores are designed to assess the extent to which children are impacted by the primary deficits of autism. Unlike previous ADOS scoring metrics, these severity scores are based on the revised algorithm (Gotham et al., 2007), including repetitive and restricted behaviors as well as impairments in social interaction and communication, thereby capturing the comprehensive clinical picture more accurately. The present research asked whether this new measure of autistic symptom severity could predict naturalistic play, imitative, and helping behaviors, as measured by performance on a modified version of a classic behavioral reenactment task (Meltzoff, 1995; Parish-Morris et al., 2007). 

Objectives: Determine whether ADOS severity scores predict children’s performance in a behavioral reenactment paradigm. We hypothesized that ADOS severity scores would significantly predict performance in the non-canonical condition of the task because success in this condition required attention to the intentional cues of the experimenter. Performance in the canonical condition could be solved based on prior knowledge of typical toy use and thus would not be predicted by severity scores.

Methods: Twenty-three children (19 male) were diagnosed with an ASD using the ADOS, the ADI-R, and expert clinical judgment. Children ranged in age from 3-6 years (mean: 60.77 months, SD=11.69), with an average full-scale IQ of 87.09 (SD=25.49). The severity scores for our sample ranged from 1-10, indicating a good spread, with a mean score of 7 (SD=2.45). In the behavioral reenactment paradigm, children observed an experimenter repeatedly try (and fail) to perform actions that either coincided with canonical toy use (e.g., stack a ring on a post) or conflicted with canonical toy use (e.g., put a train in a pot instead of on the available track). Children were asked, “Can you do it for me?” The first action children performed on each set of objects was scored “1” if it completed the experimenter’s intended action, and “0” if it did not (4 sets in all). 

Results: Two linear regression analyses were conducted, one predicting average performance on the canonical set and one predicting performance on the non-canonical set. In each regression, sex, IQ, and chronological age were controlled in Step 1. ADOS severity scores were added in Step 2. In the canonical condition, ADOS severity scores did not account for a significant amount of variance in performance (R² change = .05, F change = 1.74, p = ns). In the non-canonical condition, ADOS severity scores significantly predicted children’s responses (R² change = .20, F change = 8.27, p < .01), with the complete model accounting for a full 57% of variance in task performance.

Conclusions: ADOS severity scores predicted performance on a task requiring an understanding of the intentional cues, but not performance on a task that could be solved based only on prior experience with toys. This suggests that ADOS severity scores are a potentially useful metric for predicting live-action play behaviors in children with ASD, and are sensitive to intention understanding in this population, even after controlling for such powerful predictors as IQ.