The paper introduces *random survival forests* (RSF), an ensemble tree method for analyzing right-censored survival data. RSF extends Breiman's random forests (RF) by incorporating survival time and censoring status into the splitting criterion. The authors propose new survival splitting rules and a missing data algorithm for imputing missing values. They define ensemble mortality, a measure of mortality that can be used as a predicted outcome, based on the conservation-of-events principle. The method is evaluated through a large experiment using real and simulated datasets, demonstrating its superior or comparable performance to competing methods. A case study on the prognostic implications of body mass index in patients with coronary artery disease is presented, highlighting the complex relationships uncovered by RSF. The paper also discusses variable importance and a new missing data algorithm for forests, which can handle both training and test data with missing values.The paper introduces *random survival forests* (RSF), an ensemble tree method for analyzing right-censored survival data. RSF extends Breiman's random forests (RF) by incorporating survival time and censoring status into the splitting criterion. The authors propose new survival splitting rules and a missing data algorithm for imputing missing values. They define ensemble mortality, a measure of mortality that can be used as a predicted outcome, based on the conservation-of-events principle. The method is evaluated through a large experiment using real and simulated datasets, demonstrating its superior or comparable performance to competing methods. A case study on the prognostic implications of body mass index in patients with coronary artery disease is presented, highlighting the complex relationships uncovered by RSF. The paper also discusses variable importance and a new missing data algorithm for forests, which can handle both training and test data with missing values.