The Mean Delta Method: Quantifying Assessor Stringency and Leniency and Identifying Outliers in Workplace-Based Assessments.

Abstract

Assessor stringency and leniency (ASL)-an assessor's tendency to award low or high scores-has a significant effect on workplace-based assessments. Outliers on this spectrum have a disproportionate effect. However, no method has been published for quantifying ASL or identifying outlier stringent or lenient assessors using workplace-based assessment data. The authors propose the mean delta method, which compares the scores that an assessor awards to trainees with those trainees' mean scores. This novel, simple method can be used to quantify ASL and identify outlier assessors without requiring specialized statistical knowledge or software. As a worked example, the mean delta method was applied to a set of end-of-shift assessments completed in a large Canadian academic emergency department from July 1, 2017, to May 31, 2018, and used to examine the net effect of ASL on learners' assessment scores. A total of 3,908 assessments were completed by 99 assessors for 151 trainees, with a median (interquartile range) of 37 (12-39) completed assessments per trainee. Using cutoff values of 1.5 and 2 standard deviations, a total of 11 and 3 outlier assessors were identified, respectively. Moreover, ASL changed overall scores by more than the mean difference between years of training for nearly 1 in 4 learners. The mean delta method was able to quantify ASL and identify outlier lenient and stringent assessors. It was also used to quantify the net effect of ASL on individual trainees. This method could be used to further study outlier assessors, to identify assessors who may benefit most from targeted coaching and feedback, and to measure changes in assessors' tendencies over time or with specific intervention.