Using Multilevel Modeling to Identify Trauma-Informed Mechanisms of Change for Preschool Expulsion Risk

Background and Purpose: Preschool expulsion continues to disproportionately impact children of color and trauma-exposed children. Trauma-informed (TI) school interventions improve children’s academic outcomes; however, it is unclear through what mechanisms TI interventions influence school discipline rates nor whether they benefit all students equitably. One potential mechanism is TI self-efficacy, which includes confidence in responding to trauma-related behaviors. The current study uses a multilevel framework to test whether perceptions of children’s behavior and expulsion risk are influenced by teacher-level differences, namely teachers’ trauma-informed self-efficacy, as well as teacher-child racial/ethnic match.

Methods: The current data is from the first wave of a pilot RCT of TI training among preschool staff. In the current wave, 22 teachers completed an online survey of TI attitudes and general stress and surveys on inhibitory control and expulsion risk for 4 randomly selected students in their class. TI self-efficacy was measured using the Attitudes Related to Trauma-Informed Care (ARTIC) measure, child behavior using the inhibitory control subscale of the Child Behavior Questionnaire (CBQ), and expulsion risk using the Preschool Expulsion Risk Measure (PERM). Unconditional and conditional linear slope models were estimated using HLM software. Initial models included child sex, race/ethnicity, gender, and teacher-child racial/ethnic match, which have been linked to expulsion risk. Non-significant covariates were trimmed from the final models for parsimony.

Results: In the unconditional model, 20% of the variance in expulsion risk was due to classroom-level differences. In the random slope model, a negative association was found between CBQ and expulsion risk (t = -.54, SE = .08, p<.001); one SD increase in inhibitory control was associated with a .54 decrease in expulsion risk. The variability in the CBQ expulsion slope was statistically significant (p = .016), suggesting classroom-level differences. Child race/ethnicity, teacher-child racial/ethnic match, and child gender were not significantly associated with inhibitory control or expulsion risk.

In a conditional random slope model, teacher’s TI self-efficacy was not significantly associated with expulsion risk (p = .425), however was significantly associated with the effect of CBQ on expulsion risk (t = -.20, SE = .07, p= .007). A one unit change in TI self-efficacy was associated with a .20 weaker slope, predicting 63% of the class-level variance in the CBQ effect (i.e. the CBQ to Expulsion risk slope).

Conclusions and Implications: As expected, higher inhibitory control was related to lower expulsion risk, and as teacher’s TI self-efficacy increased, the effect of CBQ on expulsion risk lessoned, indicating that TI self-efficacy may buffer the effects of behavior on expulsion risk. These findings highlight teacher’s confidence in managing trauma-in the classroom as a meaningful mechanism through which to decrease preschool expulsion through social work interventions. Contrary to prior work, child race and teacher-child racial/ethnic match was not related to inhibitory control or expulsion risk. Prior work has demonstrated that Black children at particularly high risk for expulsion, demonstrating the need to examine differential pathways to expulsion among students of color, as well as the differential impact of TI interventions on expulsion risk.