Detecting and correcting for rater-induced differences in standardized patient tests of clinical competence