(PS2-82) Assessment of Pediatric Bipolar Spectrum Disorder Symptoms in Low-income and Historically Underserved Youths: differential Item Functioning Based on Sex

Abstract Text: Bipolar Spectrum Disorders (BPSDs) are a vital public health concern, severely reducing quality of life and functioning, and occur in children and adolescents as well as adults. They are often misdiagnosed, and errors are more common in low-income and minoritized groups, often resulting in treatment delays and under-utilization of health services. There is great need for unbiased, clinically valid, and widely accessible screening and outcome tools to measure mania and depression in youth. This study examined differential item functioning (DIF) based on sex using linear regression, multi-group confirmatory factor analysis (MG-CFA), and multi-group item response theory (MG-IRT) methods. We hypothesized that more than a dozen of the items were likely to show significant DIF.

N=806 youths receiving outpatient mental health services and their primary caregivers completed K-SADS-PL+ semi-structured diagnostic interviews, and caregivers completed the PGBI 10M, 10Da, and 10Db scales as measures of hypomanic, mixed, and depressive mood symptoms. Of note, the majority of the sample was Medicaid eligible, receiving services at a community health center, and 60% identified as African American. Scale level effects were compared to a priori minimal important difference (MID) values.

Preliminary analysis using linear regression detected uniform DIF in PGBI 10M items 2 (happy, energetic, irritable), 3 (happy, energetic, restless), and 8 (mood and energy swings). MG-CFA results indicated measurement invariance across 10M, 10Da, and 10Db scales, with model fit varying across scales and the 10Da demonstrating the best fit. MG-IRT results indicated significant (p < 0.01) uniform DIF in 10M item 3, and marginal (p < 0.05) non-uniform DIF in 10M item 9 (rapid speech). For 10M item 3, females scored an average of 0.215 points lower than males, given the same level of the latent trait. Among 10Da items, marginal uniform DIF for item 4 (periods of anhedonia) was detected but was small in magnitude, and the 10Db demonstrated invariant item functioning. For no scale did the summed effects reach MID thresholds.

Results found a few instances of statistically significant but not clinically meaningful sex DIF on parent-reported mood items. While we identified DIF in fewer items than hypothesized, these results broadly supported our broader expectation that these PGBI subscales would not show clinically concerning bias, supporting the utility of these subscales as screening measures for pediatric BPSDs. Future work should evaluate potential bias when other informants (i.e., youths, teachers) complete scales, as well as looking at more demographic factors beyond sex.