(PS3-A9) Are Sex Differences on the BIS/BAS Scale Reliable? An Examination of Measurement Invariance

Introduction: Sensitivity to rewards and punishments is important for individuals with mood disorders because hypo- and hyper-sensitivity confer risk for mood disturbances, even during euthymic periods, and may lower the threshold for mood episode occurrences. Additionally, there are differences in the prevalence of mood disorders between sexes, but the cause of these differences is poorly understood. Prior work has used self-report measures like the Behavioral Inhibition System and Behavioral Activation System Scales (BIS/BAS) to characterize sensitivity to punishments and rewards and found sex differences that may partially explain sex differences in mood disorder risk. However, these studies have not evaluated the measurement invariance of the BIS/BAS scales across sexes to determine if differences are reliable, which we do here.

Methods:  Adolescents (N=1686, mean age=15.06 years) from the greater Philadelphia area completed an online screening survey to determine eligibility to participate in a longitudinal study. This survey included the BIS/BAS scales and a demographics questionnaire that asked about age, race, sex, and ethnicity. All analyses were conducted using RStudio. First, differences in the BIS and BAS scales were separately assessed by sex using independent samples t-tests. To determine if these differences were reliable, we planned to conduct stepwise measurement invariance analyses of configural, metric, and scalar invariance. Data were cleaned to remove cases for individuals who completed the BIS/BAS scale but did not complete the demographics questionnaire that was used to determine participant sex.

Results: First, preliminary analyses tested for sex differences between male and female participants on the BIS scale and BAS scale. These analyses suggested that female participants had significantly higher BAS scores (t(2654) = 3.56, p < .001) and a trend toward higher BIS scores (t(2700) = 1.92, p = .054). Fit information only was obtained for the configural invariance models because the configural models failed to demonstrate baseline invariance. Specifically, the BAS model demonstrated poor fit (χ2(108) = 795.13, p < .001, CFI=.753, RMSEA=.087, 95% CI [.086, .098]) and the BIS model did as well (χ2(28) = 203.57, p < .001, CFI=.924, RMSEA=.087, 95% CI [.076, .099]). So, parameter estimates for the two groups for each model indicated that the BIS/BAS indicators were reflective of the different underlying constructs across groups.

Discussion: Our analyses revealed preliminary differences by sex on the BIS/BAS scales. However, the fit of the invariance models indicated that there was a lack of measurement invariance for both scales when comparing males and females. Thus, it is inappropriate to compare the BIS/BAS scores of these groups using the conventional model structure of dividing items between BIS and BAS scales. Future work should determine if the items can be combined in a theoretically meaningful way that demonstrates measurement invariance. This slight change in the way that the measure is used could produce a more reliable indicator for the risk of mood disorders for researchers and clinicians.