Background/Context: Childhood trauma can increase preschooler's risk for behavioral problems, poor student-teacher relationships, and expulsion (Loomis, 2020; Zeng et al., 2019). Efforts have been undertaken to address the gaps in knowledge and confidence of ECE teachers by integrating trauma-informed approaches into ECE settings, who by virtue of their wide reach across many populations, combined with high levels of contact with young children, are well-positioned to reach trauma exposed children. The empirical support for such approaches in schools tends to have mixed outcomes and lack rigorous research designs (Maynard et al., 2019), highlighting the need for a better understanding of the mechanisms through which trauma-informed approaches may influence teacher and child outcomes (Chafouleas et al., 2016). Despite increased focus on trauma-informed school approaches, little work has focused on child or teacher outcomes associated with different trauma-informed training content (Purtle, 2018; Sun et al., 2023). This gap makes it difficult to know which trauma-informed approaches are most effective at reducing trauma-related risks within preschool settings. Purpose/Objective/Research Question: The current study used a randomized design to examine whether two different trauma-informed trainings had differential influences on teacher attitudes/stress regulation and child expulsion risk. Based on prior work linking self-focused training to stronger trauma-informed attitudes, and stronger trauma-informed attitudes to lower expulsion risk (Loomis & Felt, 2020), we hypothesized that self-focused training would be related to greater decreases in child expulsion risk and associated indicators compared to skills training. Setting: Preschool programs across Utah were invited to participate in the study through email and personal contacts of the PI and research team. Teachers were ultimately recruited from seven sites over two years. Participating sites included three private preschools and four public preschools. Schools were located across the state, and served urban (n=5, 71%) and rural counties (n=2, 29%). Population/Participants/Subjects: A total of 73 teachers completed the trainings (29 skills and 44 self-reflection) and pre/post-training surveys from 2020-2022. Teachers were, on average, 38 years old (SD = 10.62) and represented a range of experience from less than one year (11.9%) to more than 20 years (10.4%). Teachers overwhelmingly identified as female (64.4%) and White (64.4%) or Latine/Hispanic (20.5%). Intervention/Program/Practice: Both trainings included didactic lecture, case studies, small group discussions, short videos, and journaling activities. Trainings were held online over Zoom and were each facilitated by two social work graduate students with prior training in education and trauma-informed teaching. The trainings, developed and facilitated by graduate students under the supervision of the PI, consisted of 3 x 2-hour monthly sessions. Teachers were randomized to one of two trainings. The Trauma-Informed Skills training was informed by a commonly used trauma-informed school framework (Wolpow et al., 2016) and focused on skills to promote strength-based, relationship-focused classroom environment and domains related to promoting safety, supporting child agency, and referrals for trauma-informed supports. The Trauma-Informed Self-Reflection training drew resources on self-care and vicarious trauma from existing frameworks and resources (Nicholson et al., 2020; Wolpow et al., 2009) focused on definitions of secondary, vicarious trauma, burnout, self-compassion, didactic content, and reflections. Additionally, this training included case studies focused on how a teacher's past trauma may "show up" in the classroom, and a discussion and development of a comprehensive self-care across a number of domains of well-being. Research Design: Teachers were randomized to one of two trainings; each followed the same format of three two-hour sessions that took place over a period of three months (approximately a month in-between each training). Teachers completed surveys before and after the training period. Before and after training, teachers completed a survey on their trauma-informed attitudes (Baker et al., 2018) and emotion regulation strategies (Gross & John, 2003) and measures for four randomly selected students from their class, including student-teacher relationship (Pianta, 2001) and expulsion risk (Gilliam & Reyes, 2018). When considering child outcomes, child gender and teacher-child racial/ethnic match was included as a covariate. Data Collection and Analysis: Data was collected using surveys completed in Qualtrics on a computer or smartphone. We used linear growth modeling to chart within person change in trauma-informed outcomes over time and conditional models to determine whether training group predicted teacher-level variance in changes. Models were estimated using Hierarchical Linear Modeling, Version (HLM 7; Raudenbush, Bryk, & Congdon, 2011). Findings/Results: For teacher-level variables, the Trauma-Informed Reactions subscale (secondary trauma responses) significantly increased from pre- to post-training (p = 0.032). There were significant increases in reappraisal emotion regulation strategies (p < 0.001) and marginal decreases in suppression strategies (p = 0.063) from pre- to post-training, which have implications for expulsion risk. Based on statistically significant variance in the slope for teacher outcomes, training group was added as a level-2 predictor but was not significant. There were significant increases in student-teacher closeness (p < 0.001) and decreases in classroom disruption (p = 0.025), an expulsion risk indicator, from pre- to post-training. Child gender and racial disparities in expulsion risk indicators were noted (e.g., higher expulsion risk reported for students of color and boys), and gaps did not decrease as a result of training. When there was a racial/ethnic match between teachers and students, teachers reported significantly higher closeness (p = 0.010), higher inhibitory control skills (p = 0.001), and lower classroom disruption (p = 0.035), and total expulsion trended in a direction suggesting lower overall expulsion risk (p = 0.066). No person-level variance in slope was indicated, so training content was not examined as a level 2 predictor for child outcomes. Conclusions: The current study illustrates the promise of trauma-informed training in addressing some teacher-level factors (e.g., emotion regulation) that may be linked to expulsion risk but not in reducing child expulsion risk directly or racial or gender disparities in expulsion risk. In the current study, the content of training did not differentially influence training outcomes; future research could include classroom observations to examine more nuances in trauma-informed expulsion prevention efforts.