Youth with Autism Spectrum Disorder (ASD) are at elevated risk for internalizing symptoms such as anxiety and depression (Bellini, 2004; Kim et al., 2000; Matson & Williams, 2014). These internalizing problems can affect self-esteem, social competence, academic performance, and physical health; thus, it is critical to accurately identify internalizing symptoms in order to provide appropriate intervention to those in need (Michael & Merrell, 1998). One of the most common ways to screen for internalizing symptoms is through use of rating scales completed by youth, parent, and/or teacher informants. However, inconsistent inter-rater agreement findings across studies of youth with ASD have rendered the literature difficult to summarize and in need of more systematic investigation. No prior meta-analysis has examined inter-rater or cross-informant ratings agreement concerning different internalizing constructs in youth with ASD specifically-despite its relevance to a multi-method and multi-informant approach to assessment typically recommended as best practice (Taylor et al., 2018). The present meta-analysis (a) closely examined the level of agreement across different rater-pairs (i.e., parent vs. youth, teacher vs. youth, and parent vs. teacher) assessing internalizing problems (i.e., anxiety, depression, and broad internalizing) in youth with ASD, (b) investigated both inter-rater correlations and cross-rater mean differences, (c) assessed potential moderator variables that could impact the magnitude or direction of correlations or mean differences, and (d) systematically summarized findings and trends.Results indicated that across the three constructs (i.e., anxiety, depression, and broad internalizing), the mean r ranged from 0.399 to 0.430 (moderate range) for parent vs. youth self-report ratings and 0.256 to 0.296 (small range) for parent vs. teacher ratings. In the case of teacher vs. parent ratings, the observed mean inter-rater correlations ranged from 0.229 to 0.342 (small to moderate range) but were non-significant for all three constructs. Moderator analyses within the parent vs. youth self-report inter-rater correlations indicated that method of youth self-report administration moderated correlations for anxiety, while mean age of the youth moderated correlations for depression. No significant moderators were noted for other inter-rater correlations across the three rater-pairs. For parent vs. youth self-report standardized mean differences, mean effect size g was 0.220 for anxiety, 0.788 for depression, and 0.090 for broad internalizing. However, evidence of possible publication bias and associated re-estimation yielded non-significant bias-adjusted mean g estimates in the small to negligible range for both constructs. For parent vs. teacher ratings, mean g values ranged from 0.015 to 0.176, but all were deemed negligible. In the case of teacher vs. youth self-report ratings, mean g varied considerably, ranging from -0.033 to 0.670-but all mean g values were non-significant and based on only a small number of studies. No significant moderators were found for any of the standardized mean differences across all rater-pairs and constructs.These results suggest that covariation across informants regarding internalizing symptoms in youth with ASD tends to be small to moderate, depending on the rater-pairs, and typically involves negligible mean differences between rater types. Additional inter-rater studies are needed, in general, to improve precision of effect size estimates and provide additional power for moderator analyses, but are needed, in particular, for teacher vs. youth self-report ratings-where overall estimates are based on too few studies.
Read
