Little is known about severe pulmonary embolism in children. We aimed to report pulmonary embolism outcomes, identify risk factors for unfavourable outcomes, and evaluate the discriminative ability of two clinical-severity indices in children. In this retrospective cohort study, we included consecutive patients aged 18 years or younger with acute pulmonary embolism, objectively diagnosed radiologically or pathologically, between Jan 1, 2000, and Dec 31, 2016, from two Canadian paediatric hospitals (The Hospital for Sick Children, Toronto, ON, and the Children's Hospital of Eastern Ontario, Ottawa, ON). Exclusion criteria were sudden death without radiological or pathological pulmonary embolism confirmation and non-thromboembolic pulmonary embolism. The primary outcome was a composite of unfavourable outcomes of pulmonary embolism-related death and pulmonary embolism recurrence or progression. Potential predictors of the composite unfavourable outcome (ie, age at pulmonary embolism diagnosis, sex, underlying cardiac disease, severity of the pulmonary embolism, presence of a central venous catheter, associated venous thromboembolism, family history of thrombosis, treatment modalities, thrombophilia, obesity, and recent surgery) were explored with logistic regression. We calculated pulmonary embolism severity index (PESI) and simplified PESI (sPESI) using age-adjusted parameters; we estimated the ability of PESI and sPESI to predict mortality using receiver-operating characteristic (ROC) curve analysis. Of the 170 patients included, 37 (22%) had massive, 12 (7%) submassive, and 121 (71%) non-massive pulmonary embolism. Patients with massive or submassive pulmonary embolism were younger (median age 12·5 years [IQR 0·6-15·1] vs 14·4 years [9·3-16·1], p<0·0001), more likely to have a cardiac condition (16 [33%] vs 17 [14%] patients, p=0·009), and had more central venous catheters (29 [59%] vs 48 [40%] patients, p=0·027) than patients with non-massive pulmonary embolism. Aggressive treatment modalities were more commonly used in massive or submassive pulmonary embolism (22 [45%] vs 7 [6%] patients, p<0·0001). Of the predictors tested, only pulmonary embolism severity was associated with the composite unfavourable outcome in the multivariable analysis (odds ratio 3·53, 95% CI 1·69-7·36; p=0·011). The area under the ROC curve for PESI to predict 30-day mortality was 0·76 (95% CI 0·64-0·87). Sensitivity of sPESI was 100% and specificity was 30%. Massive or submassive pulmonary embolism led to higher rates of unfavourable outcomes than non-massive pulmonary embolism in children. Further adaptations of PESI and sPESI are required to improve their clinical usefulness in paediatric patients. Trainee Start-Up Fund (The Hospital for Sick Children).