5 - (IOP 3) Classification Accuracy of the PGBI-10M Scale for Ksads-defined Bipolar Disorders in the ABCD Study at Multiple Waves

Mood disorder rates have been rising, also more than doubling after puberty onset. Early detection is crucial for effective treatment, but diagnosis is often delayed by years. The Parent General Behavior Inventory-10 Mania (PGBI-10M) has shown high classification accuracy in multiple studies and has been endorsed in research and clinical settings. However, it has not been evaluated for screening in the general population. Prior work on the classification accuracy of the PGBI-10M also has included community health, Medicaid-eligible, and predominantly African American families, making it possible to test for systematic bias in accuracy. This study used data from two waves of the ABCD study (5.1 release) to evaluate the classification accuracy of the PGBI-10M in a population-based sample.

Secondary analyses were conducted using PGBI-10M item-level ABCD data from baseline and Year 2 follow-up (N = 11,862 and 11,203). Receiver operating characteristic (ROC) analyses were used to assess PGBI-10M’s ability to predict KSADS diagnoses of bipolar disorder, using Hanley & McNeil’s (1983) test to compare Area Under the Curve (AUC) estimates with those from academic outpatient (N = 617) and community mental health (N = 530) samples.

Bipolar disorder was diagnosed in 2.0% of youths at baseline (Mage = 9.92 years) and 1.2% at Year 2 follow-up (M = 12.03 years). PGBI-10M AUC was .68 (SE = .021) at baseline and .77 (.024) at Year 2 follow-up (both ps <.001). AUCs were lower than in outpatient and community mental health settings (.84 and .78, respectively), resulting in < 10% positive predictive value.

To our knowledge, this study is the first time a mania assessment has been evaluated as a screener for bipolar disorder in a large general population sample of children. The PGBI-10M can differentiate bipolar disorder from other conditions but shows lower accuracy in population samples than in clinical settings. Of note, the classification accuracy in ABCD was lower than in community health. This is the opposite of what would be expected if the scale showed systematic bias, and it is likely due to ABCD cases being “milder,” reducing scale sensitivity. Universal screening for bipolar disorder is not recommended; instead, PGBI-10M is best used in clinical contexts alongside additional information.