Identifying Broader Phenotype Features: A Comparison of the ADOS and BPASS in School-Age Children

Background: The Autism Diagnostic Observation Schedule (ADOS) has often been used to detect broader autism phenotype (BAP) in high-risk (HR) and low-risk (LR) children at 3 years of age (Messinger et al., 2013; Ozonoff et al., 2014; Charman et al., 2016). Likewise, the BPASS – a parent questionnaire assessing features of the BAP – has been used to detect BAP in older children (Dawson et al., 2007). Although both measures assess a number of similar constructs, little research has explored their concordance in identifying BAP features.

Objectives: The main objective is to examine the convergent validity of the BPASS with the ADOS for identifying BAP features in a school-age sample of high and low-risk children. Specifically, we examine the correlation between ADOS and BPASS scores, and use ROC methods to identify a threshold on the BPASS that maximizes sensitivity and specificity when compared to the ADOS.

Methods: Given the focus on the broader phenotype, all participants diagnosed with autism spectrum disorder (ASD) were excluded from the sample. We administered both the ADOS module 3 and the BPASS to 59 school-age children: 36 high-risk (HR) participants– those with an older sibling with ASD, and 23 low-risk (LR) children– those with no known family history of ASD. Following prior research (Messinger et al., 2013, Ozonoff et al., 2014), an ADOS Social Affect + Restricted Repetitive Behavior (SARRB) score >= 4 was used as a cutoff indicative of possible BAP features. The BPASS has two sections, one based on parent report of BAP features, the other based on examiner judgments of eye contact, communication style, voice prosody, etc., after a brief conversational interaction. The variable used in analyses was the summed total of the examiner-scored BPASS items 8-13. To assess the convergent validity of the BPASS, we first examined the correlation between each participant’s ADOS SARRB and BPASS subtotals. Next, we used ROC analysis to identify the optimal cutoff for the BPASS in classifying the children identified by the ADOS BAP cutoff.

Results: Findings showed a .75 correlation between ADOS SARRB and BPASS subtotals. Using the ADOS cutoff of 4, 30.5% of children were classified with BAP (15 HR and 3 LR). ROC analysis yielded a BPASS threshold of 8.5 as the optimal cutoff, with an Area Under the Curve (AUC) of 0.9. Using this BPASS cutoff, 40% of children were classified with BAP (17 HR and 6 LR). Table 1 shows the contingency table of classifications; Figure 1 shows the ROC curve.

Conclusions: The BPASS has strong convergent validity with the ADOS. The ROC analysis revealed that the BPASS identified participants with BAP features, as indexed by the ADOS, with strong sensitivity, specificity, and negative predictive value, and adequate positive predictive value. These findings suggest that the BPASS is a valid tool in detecting BAP outcomes.