UCLA

Most social communication tests used to measure change in young children with autism spectrum disorder have not undergone rigorous psychometric evaluation. Notably, and most relevant to the early intervention literature, there is little information on the sensitivity of these measures to change over time, despite their frequent use. Further, most syntheses of intervention studies combine data based on constructs (e.g. language) without accounting for the potential error introduced when different measures are combined together. While this may be appropriate, more information is needed on whether there is substantial heterogeneity across commonly used social communication outcomes. The aims of this study were to examine the sensitivity to change and expected change over time of social communication measures in ASD clinical trial and longitudinal research studies. A systematic review and meta analyses was conducted to generate pooled effect sizes within each identified outcome measure, rather than pooling multiple measures together. Meta regression was used to determine whether the length of the measurement period was related to the magnitude of change over time and whether this relationship was influenced by factors such as cognitive ability, children’s age and year of publication. The average length of the included studies ranged from 3 months to 20 months. Overall, the expected change over time, measured using standardized mean differences, was small to medium, although there was considerable variability. For example, ADOS Severity scores had an expected change of 0.114 standardized mean units and an average measurement period length of 16 months whereas Vineland socialization scores had an expected change of 0.3581 standardized mean units over a shorter average measurement period of 10 months. Most outcomes were not sensitive to change over time; the expected change over time was independent of the length of the study in 9 of the 42 measures. Further, change in some measures was influenced by factors like cognitive ability, chronological age, whether children were receiving behavioral interventions and how the outcome was reported (e.g. age equivalent versus standard score). These data suggest that careful consideration is needed when selecting an appropriate outcome measure and tests that measure similar constructs can vary considerably in their expected change over time. Some recommendations based on these data include: use caution when reporting standard scores to measure change over time, use both parent report measures and clinician administered measures to track progress in behavioral interventions, and use caution when combining different outcome measures to synthesize intervention data.