Examining the Stability of Social Communication Measures for Young Children with Autism Spectrum Disorders

Background:

Identifying stable measures of social communication has important implications for evaluating the impact of early intervention. This may be especially important for the estimated 30% of children with ASD who are preverbal or minimally verbal because these children often show floor effects on standardized assessments of language (Anderson, et al., 2007; Kasari et al., 2013). Initiating joint attention (IJA) gestures are predictive of later language use (Charman 2003; Kasari et al., 2012) and frequently used to characterize emerging communication skills. Evaluating changes in these gestures, as well as language is used to indicate the effectiveness of interventions as well as to evaluate child progress.

Objectives:

The objective of this analysis was to (1a) use a generalizability study (G-study) to evaluate the stability of early measures of IJA and language across two different observational measurement contexts and (1b) use a decision study (D-study) to estimate the number of additional observations required to provide optimal estimates of IJA and language in preverbal children with ASD.

Methods:

Data were analyzed from pre-intervention assessments of 73 participants enrolled in a recently completed RCT (HRSA #R40MC27707) examining outcomes for young, preverbal children with ASD between the ages of 36 and 54 months (mean age=42.88 months) who were at risk for remaining minimally verbal. Occurrences of IJA were coded from two different observational measurement contexts: a standardized language sample (LS) and the Early Social Communication Scales (ESCS; Mundy et al., 2003). A G-study was conducted to assess the stability of IJA by evaluating the g coefficient, which provides an index of the amount of variance that is attributable to true variance in measurement rather than to sources of measurement error. A higher g co-efficient indicates less measurement error; a g co-efficient that ranges from 0.6 to 0.8 is considered to produce a stable estimate (Bakeman et al., 1997). After conducting the G-study, a D-study was conducted to estimate the number of different measurement contexts needed to achieve a stable estimate of IJA. An additional G-study is currently being conducted to evaluate the stability of a measure of language (number of different words produced) across two observational measurement contexts (LS and parent-child interaction). A D-study will be conducted to estimate the number of additional observations needed to achieve a stable estimate.

Results:

Results from the G-study indicated that two measures of IJA were not sufficient for achieving a stable estimate (g=0.43). Results from the D-study indicated that up to four observations of IJA may be required to achieve a stable estimate in preverbal children with ASD (g=0.60), and that up to 10 observations may be required to achieve an optimal stability estimate (g=0.80).

Conclusions:

The low value of the g coefficient across the two measures indicates a need for additional measures of IJA to increase the stability of existing measures. An aggregated measure of IJA may provide a better measure of the predictive value of these skills on spoken language. These results are important given use of gestures is often evaluated by a single measure.