Combining self- and informant-reports to estimate true correlations between psychopathology constructs – and how dimensional models (e.g., HiTOP) can facilitate this

Background: While psychopathology research increasingly favors dimensions over categorical diagnostics, nosology, and treatment, much about the dimensions’ assessment is still unknown, particularly for children and adolescents. With most research based on single-source, cross-sectional data, findings are contaminated by random error and systematic biases that make up much of scale scores’ variance and make their correlations uninterpretable. We describe a scale score variance decomposition model that leverages multi-informant data to estimate scale scores’ – and even individual items’ – true correlations among themselves and with other variables.Methods: After presenting the model, we reviewed relevant literature for its application in personality and psychopathology domains. Also, since the model treats cross-rater correlations as the degrees of valid scale score variance, we reviewed literature on cross-rater agreement for normal and maladaptive psychological traits in children and adults.Results: True correlations based on multi-informant data are consistently stronger than correlations based on single-source data, validating our model and showing its benefits. Cross-rater agreement tended to have similar magnitudes for both adaptative and maladaptive traits, 1) suggesting similar degrees of valid variance, systematic method biases, and error and 2) validating our models’ use for psychopathology ratings. However, agreement tends to be somewhat lower in children than adults, implying that multi-informant child research requires larger samples for similar precision when estimating true correlations. Discussion and Conclusions: We discuss how multi-informant data and our model can be especially useful for understanding child psychopathology and conclude by offering suggestions for how to leverage the model to explore dimensional models of psychopathology.