Background: Retention of early-career teachers is a persistent problem in the United States, with 44% of teachers leaving the profession within five years nationally (Ingersoll et al., 2018). Similar trends are seen in Georgia as a whole (Henson et al., 2015), and among Atlanta Public Schools (2019). Retention of Black educators presents an even greater challenge, with the turnover rate nationally being about 50% greater (Carver-Thomas & Darling-Hammond, 2017). Addressing this problem requires foundational knowledge about conditions supporting success in retention. Intervention: In this work we examine the impact of the Collaboration-and-Reflection-to-Enhance-Atlanta-Teacher-Effectiveness (CREATE) program on the three-year teacher retention rate of novice teachers. CREATE is designed to increase teacher collaboration through mentoring and involvement in collaborative learning communities, reduce stress, increase teacher support, and improve novice teachers' instructional planning capacity. Participation begins in the preservice teaching year, while residents are completing their credential at Georgia State University's College of Education and Human Development (GSU-CEHD). Key components of the program are: (1) Progressive Core Classroom Roles: In Year 1, residents spend time in local schools with a Cooperating Teacher, completing their preservice teaching practicum. In Year 2, most CREATE residents are paired with another CREATE teacher in a single classroom. In Year 3, residents become the sole "teacher of record" in their own classroom; (2) Critical Friendship meetings that allow residents an opportunity to share and collaborate, discuss their work and dilemmas of practice, build classroom management skills; (3) Cognitively-Based Compassion Training® (CBCT), a form of mindfulness training, (4) Planned interactions with mentor teachers and the CREATE program team for support; and (5) Summer Resident Academies for developing social emotional competencies and pedagogical skills. CREATE has evolved since the beginning of the grant by incorporating equity-centered practices to develop critically-conscious, compassionate, and skilled educators who are committed to teaching practices that prioritize racial justice and interrupt inequities. BAU consists of the standard 1-year preservice program administered through GSU-CEHD. The current work presents findings from a multi-year external evaluation of CREATE, as part of the U.S. Department of Education's Investing in Innovation (i3) Development grant funds. In this work we focus on the impact of CREATE on teacher retention outcomes. The purpose of the study is to address the following questions: Was CREATE implemented with fidelity? What is the impact of CREATE on completion of the teacher preparation program at GSU-CEHD and teacher retention into the first and second year of teaching? Is there a differential impact on retention for Black and non-Black educators? Participants: The subgroup sample sizes used in the analyses were 38 CREATE teachers (19 from each cohort) and 83 comparison cases (50 from Cohort 1 and 33 from Cohort 2). All participants were registered in GSU-CEHD based in Atlanta. Study Design: The five-year quasi-experimental evaluation follows two staggered cohorts of study participants for three years each: in their preservice teaching year and then their first two years as full-time classroom teachers. CREATE members were matched on a series of baseline covariates to similar comparison cases graduating from the regular GSUCEHD within each of the two cohorts. Analysis: For the implementation study we examined whether CREATE achieved a-priori declared thresholds of fidelity of implementation (FOI). For analyzing impacts on teacher retention, we use "discrete-time survival analysis" (Singer & Willett, 1993) to evaluate the impact of CREATE on the three-year early career trajectory (graduation from GSU CEHD in Year 1, teaching in Year 2, and teaching in Year 3). The method produces maximum-likelihood estimates of key parameters in the model. We calculated the fitted hazards function, and the sample survival function, indicating the proportion of teachers expected to remain in the early career trajectory by the end of each time interval. The survival analysis tests the impact of CREATE on an uninterrupted teaching trajectory, which means graduating on time and proceeding to and remaining in teaching through the following two years. We assessed baseline equivalence on a series of covariates. Results: Was CREATE implemented with fidelity? Three of five components met fidelity thresholds each year. Cognitively-Based Compassion Training reached fidelity in just one of two years, and mentoring supports were just below threshold. It is noteworthy that thresholds were set very high. (Details will be available in the full paper.) What is the impact of CREATE on completion of the teacher preparation program at GSUCEHD and teacher retention into the first and second year of teaching? Is there a differential impact for Black and non-Black educators? The survival percentages for the CREATE and comparison groups for the full sample, the sample of non-Black educators, and the sample of Black educators are displayed in Tables 1-3. The tables display percentages of teachers who left their teaching career trajectory within each year, and the percentages who remained in each time interval. The percentages of teachers remaining are also displayed graphically in Figures 1-3. Full results of the survival models that show effect estimates and p values for addressing the research questions are shown in Table 4. We observe marginally significant positive impact of CREATE on on-time retention (p=0.038). (Table 4 Model 5; Survivor curve in Figure 1). There is a statistically significant differential effect of CREATE on retention favoring Black educators (Table 4 Model 7). A comparison of Figures 1-3 indicates that average positive impact of CREATE on retention is achieved through its effect on Black educators. (In the full study we show that baseline equivalence of <0.25 SD was achieved for the analytic sample on a series of covariates.) Conclusions: This work provides preliminary evidence showing that CREATE has positive impact on on-time graduation and teacher retention through the first two years of teaching, with the effect being specific to Black educators. Additional exploratory and qualitative work will help to clarify the mechanism. Surveys of teachers show that among the retained groups different mechanisms may be active among Black and non-Black educators. Specifically, impacts on intermediate outcomes show a trend of positive impact of CREATE on stress management and resilience (Figure 5), that is not observed among non-Black educators.