Participants were 318 youths aged 5 to 10 (50 percent with bipolar diagnoses) and 324 youths ages 11 to 17 (41 percent with bipolar diagnoses). Both the youth and the primary caregiver were include in the assessments. All participants and their families completed a semistructured diagnostic interview using the Schedule for Affective Disorders and Schizophrenia for School-Age Children. Index tests used were the Parent Young Mania Rating Scale, the General Behavior Inventory, the Parent General Behavior Inventory, the Child Behavior Checklist, the Youth Self-Report, and the Achenbach Teacher Report Form. To maximize the usable sample size and to avoid the possible introduction of bias due to missing data, the study used the multiple-imputation procedure developed by Graham and Schafer, imputing 10 sets of value separately for the younger and older samples. Areas under the curve (AUC's) from receiver operating characteristic analyses and multilevel likelihood ratios quantified test performance. Findings indicate that in identifying bipolar disorders, the Parent Report (AUC's from 0.78 to 0.84 in both age groups) outperformed the Teacher Report (AUC of 0.57 in the younger sample and 0.70 in the older sample) as well as the General Behavior Inventory (AUC's 0.67) and the Youth Self-Report (0.71). Combining tests did not produce clinically useful classification improvement. The findings thus suggest that the Parent report on these instruments could facilitate differential diagnosis of bipolar disorder in youths ages 5-17, especially by decreasing the rate of false-positive diagnoses. 4 tables and 73 references