Abstract. OBJECTIVE: To identify predictors of severe retinopathy of prematurity (ROP) in a large population-based cohort and to examine risk-adjusted variations across units. STUDY DESIGN: Retrospective analysis of Canadian Neonatal Network data on neonates with birth weight <1500 g who were screened for ROP between 2003 and 2010. Characteristics of infants with and without ROP were compared and a risk-adjusted model for severe ROP was developed. Rates of severe ROP were compared between sites. RESULTS: 1163 of 9187 (12.7%) infants developed severe ROP. Lower gestational age, male sex, small for gestational age, patent ductus arteriosus, late onset sepsis, more than two blood transfusions, inotrope use, and outborn status were associated with an increased risk of severe ROP. Severe ROP rates varied significantly between units CONCLUSION: Younger, smaller and sicker male infants had higher adjusted risks of severe ROP and rates varied significantly among sites.