Modeling issues in bivariate disease mapping