On informative detection bias in screening studies