The human reliability analysis (HRA) was investigated intensively in the last decades and many methods have been developed. The most important feature of HRA is a good data, which can be obtained from full scope plant simulators. The objective of this study is to investigate how importance factors of human failure events obtained from results of probabilistic safety assessment (PSA) can be used in sense to improve human reliability. The study considers recently developed method: Institute Jozef Stefan (IJS)-HRA, which integrates some features of existing methods and some new features, and its application in a large PSA model. Two ways for quantification and application of importance factors are presented. The results show that only few human failure events dominate in the HRA analysis. The results show that only few parameters from human reliability database contribute significantly to the risk. Identification of dominating human failure events and identification of the significant parameters are a valuable input for determining the priorities of simulator training. (C) 2007 Elsevier Ltd. All rights reserved.