Date created: | Last Updated:

Category: Project

Description: The reliability of measurements is crucial in psychology research. The validity of the study conclusions reflects the validity of measurements used [Flake and Fried, 2020]. Using valid measurements is especially important when working with large datasets. Such datasets are increasingly popular in neuropsychiatry [Thompson et al., 2020, Turner, 2014] and allow researchers to answer questions with higher confidence [Thompson et al., 2020]. One example of a massive dataset primarily aimed at neuropsychiatric research is the Adolescent Brain Cognitive Development study (ABCD) [Casey et al., 2018]. There is an increasing volume of depression research that uses ABCD data (https://abcdstudy.org/publications/). Adolescent depression is a popular direction of research because of its high prevalence in society (https: //www.nimh.nih.gov/health/statistics/major-depression). This brings relevance to the measures used for depression estimation in the ABCD. In our previous pre-registered analysis (https://osf.io/8s7bv), we have looked at the only continuous measure of depression available in the ABCD - the Child Behavior Checklist (CBCL-Dep, further referred to as CBCL) [Achenbach and Edelbrock, 1991], a parent report measure, and compared it against the commonly accepted gold standard - the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS) [Townsend et al., 2020]. ABCD contains both parent and child report K-SADS data, but it does not contain information on consensus diagnosis. We used the area Under the Curve (AUCROC) to evaluate the predictive performance of the CBCL on the child report KSADS. The results were quite poor: AUCROC for differentiating adolescents with depression from adolescents without depression was 0.63, which corresponds to ”poor” performance [Ferdinand, 2008]. CBCL was not a good predictor of K-SADS depression scores, as measured on ABCD data. For the specificity hypothesis, the results were even poorer. CBCL could differentiate adolescents with depression from adolescents without depression but with another form of psychopathology with AUCROC equal to 0.48, and adolescents with depression but without another form of psychopathology from adolescents without depression but with another form of psychopathology with AUCROC equal to 0.46. These results are below chance. As a next step, we would like to extend our conclusions beyond the ABCD data. We intend to look into two datasets that contain psychiatric evaluations of adolescents matching the age of the baseline wave of the ABCD data: the Healthy Brain Network (HBN) [Alexander et al., 2017] and the Brazilian High Risk Cohort Study for the Development of Childhood Psychiatric Disorders (BHRC) [Salum et al., 2015]. Since these datasets did not have enough subjects with depression in the 9-11 years old range, we extended our consideration to 9-13 years old for both datasets. Both of those datasets use CBCL as well as a clinician diagnosis (HBN: K-SADS; BHRC: the Development And Well-Being Assessment (DAWBA)) [Goodman et al., 2000]).