Key points are not available for this paper at this time.
Higher EPV is needed when low-prevalence predictors are present in a model to eliminate bias in regression coefficients and improve predictive accuracy.
Ogundimu et al. (Tue,) studied this question.