Diagnostic scoring choices change the prevalence and correlates of mental illness in the ABCD Study

<p>Objective: Mental disorders are the leading cause of disability in youth. Identifying which youth have, or will develop, mental illness relies increasingly on large, shared cohorts whose diagnostic data are analyzed by hundreds of investigators. Deriving psychiatric diagnoses from these data, however, requires analytic decisions that are rarely reported or standardized. Here, we tested whether these largely invisible decisions can change, or even reverse, conclusions about youth mental illness in the largest and most frequently analyzed cohort, the Adolescent Brain Cognitive Development (ABCD) Study.Method: We first reviewed the ABCD literature to document current practices. Then, using KSADS-COMP diagnoses from ABCD Release 7.0 (N=11,858 at baseline), we held the data fixed and, for every combination of timeframe, threshold, and informant, we recomputed prevalence for 13 DSM-5 categories and re-estimated associations of six categories with demographic, psychosocial, and neuroimaging variables.Results: In the same children, these choices changed the prevalence of a single disorder from 1.9% to 56.8%. Caregiver and youth reports agreed barely above chance and identified nearly independent sets of children; the choice of informant alone thus substantially determined who was classified as a case. Across demographic (sex, income, race/ethnicity), psychosocial (screen time, family conflict), and neuroimaging (resting-state functional connectivity) correlates, only 26% of associations held under every decision, with brain-diagnosis relations the least stable.Conclusions: These findings suggest caution in interpreting existing results and underscore the need to report diagnostic choices and their influence on findings. We provide a harmonized dataset and reporting standards to support these efforts.</p>