Aim: The variability in quality of CPR provided during cardiac arrest across pediatric institutions is unknown. We aimed to describe the degree of variability in the quality of CPR across 9 pediatric institutions, and determine if variability across sites would be affected by Just-in-Time CPR training and/or visual feedback during simulated cardiac arrest. Methods: We conducted secondary analyses of data collected from a prospective, multi-center trial. Participants were equally randomized to either: (1) No intervention; (2) Real-time CPR visual feedback during cardiac arrest or (3) Just-in-Time CPR training. We report the variability in median chest compression depth and rate across institutions, and the variability in the proportion of 30-s epochs of CPR meeting 2010 American Heart Association guidelines for depth and rate. Results: We analyzed data from 528 epochs in the no intervention group, 552 epochs in the visual feedback group, and 525 epochs in the JIT training group. In the no intervention group, compression depth (median range 22.2-39.2mm) and rate (median range 116.0-147.6min-1) demonstrated significant variability between study sites (p<0.001). The proportion of compressions with adequate depth (0-11.5%) and rate (0-60.5%) also varied significantly across sites (p<0.001). The variability in compression depth and rate persisted despite use of real-time visual feedback or JIT training (p<0.001). Conclusion: The quality of CPR across multiple pediatric institutions is variable. Variability in CPR quality across institutions persists even with the implementation of a Just-in-Time training session and visual feedback for CPR quality during simulated cardiac arrest.