We propose a method for path analysis of survival data with recurrent events. By applying an additive model for the intensity, concepts like direct, indirect and total effects may be defined in an analogous way as for traditional path analysis. The focus is on understanding how to analyze the effect of a dynamic covariate, e.g. the number of previous events, and at the same ensuring that the effect of a fixed covariate is unbiasedly estimated. Theoretical considerations as well as simulations are presented. A dataset on recurrent tumors in rats is used for illustration.