As evidence-based interventions (EBIs) become more widely disseminated, fidelity of implementation (FOI) often wanes. This study explores the association between FOI and malleable variables within classrooms that could be targeted to optimize resources without compromising FOI as school-based EBIs are disseminated across real-world settings. We utilized process evaluation data from a national dissemination project of the Botvin "LifeSkills" Training (LST) middle school program, a universal prevention intervention shown to reduce substance use. The sample included 1,626 teachers in 371 schools across 14 states. Hierarchical linear models examined the relationship between observational measures of implementation factors and three domains of fidelity (e.g., adherence, student responsiveness, and quality of delivery). Findings suggest that curriculum modifications, student misbehavior, and shortage of time to implement the LST middle school program were factors most associated with lower FOI. Class size, access to program materials, and whether LST was delivered in a traditional classroom setting that is well-suited for instruction (versus in a less structured environment such as the school cafeteria) are less predictive. In scale-up of classroom-based universal interventions targeting behavioral health outcomes, our findings indicate that carefully vetting modifications, supporting classroom management strategies, and ensuring sufficient class time for implementation of highly interactive EBIs such as LST are important considerations. Since changes to EBIs are inevitable, efforts are needed to guide facilitators in making adjustments that improve program fit without compromising the essential intervention activities deemed necessary to produce desired outcomes.