Optimization of self‐ or parent‐reported psychiatric phenotypes in longitudinal studies

Background The Adolescent Brain Cognitive Development (ABCD) study is a longitudinal study of US adolescents with a wide breadth of psychiatric, neuroimaging and genetic data that can be leveraged to better understand psychiatric diseases. The reliability and validity of the psychiatric data collected have not yet been examined. This study aims to explore and optimize the reliability/validity of psychiatric diagnostic constructs in the ABCD study. Methods Parent‐and‐child‐reported psychiatric data for 11,876 children (aged 9.5 ± 0.5 at first assessment) were examined over 4 years to derive specific constructs for psychiatric diagnoses using longitudinal information. Rates of psychiatric disorders were calculated and compared to those reported in the epidemiological literature. Results The rates of self‐reported psychiatric disorders at any single time point (broad diagnostic construct) were higher than indicated by epidemiological studies. Narrow diagnostic constructs, which required the endorsement of psychiatric disorders at a majority of longitudinal assessments, demonstrated a better rate approximation of literature‐reported prevalences for most disorders (e.g. the prevalence of broad obsessive‐compulsive disorder (OCD) was 13.3% compared to narrow OCD at 2.6% and a literature‐reported prevalence of 2.3%). Analysis of comorbidity, using OCD as a representative example, also showed a better approximation of literature‐reported comorbidity rates using the narrow construct, with some exceptions. Conclusions Self‐ or parent‐report‐based assessments tend to overestimate prevalences of psychiatric disorders in the ABCD Study, particularly when longitudinal data are summed to create lifetime prevalences. Such assessments should be accompanied by more in‐depth assessments or clinician‐administered structured interviews if using data where accurate disorder classifications are paramount.