Objective: Evaluate whether a brief diagnostic aid for pediatric bipolar disorder is comparable across changes in administration format or clinical setting.
Goals: Examine portability of a 10-item diagnostic aid from an academic setting to community mental health. 2) Examine changes due to extracting 10 items from the full measure. 3) Examine diagnostic efficiency of the 10 items in community settings.
Methods: Primary caregivers at either an academic (N=813) or community site (N=489) completed the parent-report General Behavior Inventory and semi-structured KSADS interviews with highly trained raters. Additional parents at the community site completed the 10 item screen only (N=799), with 159 overlapping with the group that completed the longer research version. Raters and diagnoses were blind to the questionnaires.
Results: Item response theory indicated that instrument total scores functioned similarly across sites. However, individual items’ characteristics changed; e.g., elated mood was slightly more discriminating in the Community than the Academic site. Items functioned similarly in embedded versus extracted formats. ROC analyses indicated the 10 items discriminated PBD from all other cases, AUROC=.79, p<.0005. Scores of 5 or lower substantially decreased probability of bipolar disorder, whereas scores of 17 or higher substantially increase probability of bipolar disorder. Scores between 6 and 17 increase the probability of bipolar disorder mildly.
Conclusions: The 10 item PGBI functions similarly across formats and settings. Although some items showed bias, the net effects cancelled out for the total scale score. Future studies should examine whether race, socio-economic status, or reading level cause symptom level differences.