What package?  It is likely related to the fit of the model with that variable versus without that variable, and likely the package has made it so that with some assumptions it follows a chi-sq distribution.  These won't tell you whether one variable is more important than another though, that will require substantive knowledge about the area.