The current study explored the extent to which a common measure of perceived implementation of critical features of School-wide Positive Behavior Support (SWPBS) predicted fidelity of implementation 3 years later. Respondents included school personnel from 261 schools across the United States implementing SWPBS. School teams completed the "PBIS Self-Assessment Survey" to self-assess fidelity of implementation in different SWPBS settings (school-wide, non-classroom, classroom, individual). These scores were then analyzed to assess whether certain items predicted the fidelity of SWPBS implementation, as assessed through another fidelity of implementation measure, the "School-wide Benchmarks of Quality," 3 years later. Regression analyses indicated that self-reported fidelity of implementation of Classrooms Systems significantly predicted both sustained implementation and student outcomes, as assessed through levels of Office Discipline Referrals. Within Classroom Systems, regular acknowledgement of expected behaviors, matching instruction to student ability, and access to additional support were the strongest predictors of sustained implementation. Results are discussed in terms of critical areas for focusing SWPBS training to increase the likelihood of sustained implementation. [This paper was published in "Journal of Positive Behavior Interventions" (EJ1041152)."]