Psychiatr Q. 2025 Oct 14. doi: 10.1007/s11126-025-10221-y. Online ahead of print.
ABSTRACT
Children’s mental health issues have always been a focus of our attention. However, there has been limited exploration on how to conduct rapid, low-cost, and effective screening in large samples of children.This study collected data from all fifth-grade students in a district of Beijing, China, from 2018 to 2022. First, information from the Child Behavior Checklist (CBCL) was gathered for all samples. For children whose scores on any CBCL dimension exceeded the normal range, semi-structured interviews were conducted to clarify diagnoses. Subsequently, a random forest model was constructed to assess screening effectiveness.Among 104,276 participants, only 2.24% of children were diagnosed with a mental disorder following the two-step screening procedure, which is lower than the prevalence reported in epidemiological surveys. Among all diagnosed mental disorders, the highest prevalence was observed for Attention-Deficit/Hyperactivity Disorder (ADHD), at 1.48%. Additionally, 5.49% of the patients exhibited comorbidities, with the most common being ADHD comorbid with ASD. Regarding specific disorders, the comorbidity rate for tic disorders (TD) was the highest at 34.78%. A random forest model was constructed to predict ADHD, depressive disorder, and ASD. The model performed exceptionally well in predicting normal children with specificity greater than 0.9; however, its predictive performance for patients was poor, with sensitivity below 0.2 and precision below 0.3.The two-step CBCL screening underestimates the true prevalence of mental disorders but shows high predictive accuracy for negative cases. Applying random forest to pre-exclude negatives in large samples can reduce screening costs and improve efficiency by focusing interviews on predicted positives.
PMID:41085931 | DOI:10.1007/s11126-025-10221-y
