Influence of Rater Training on Inter- and Intrarater Reliability When Using the Rat Grimace Scale

Abstract

Rodent grimace scales facilitate assessment of ongoing pain. Reported rater training using these scales varies considerably and may contribute to the observed variability in interrater reliability. This study evaluated the effect of training on interrater reliability with the Rat Grimace Scale (RGS). Two training sets (42 and 150 images) were prepared from acute pain models. Four trainee raters progressed through 2 rounds of training, scoring 42 images (set 1) followed by 150 images (set 2a). After each round, trainees reviewed the RGS and any problematic images with an experienced rater. The 150 images were then rescored (set 2b). Four years later, trainees rescored the 150 images (set 2c). A second group of raters (no-training group) scored the same image sets without review with the experienced rater. Inter- and intrarater reliability were evaluated by using the intraclass correlation coefficient (ICC), and ICC values were compared by using the Feldt test. In the trainee group, interrater reliability increased from moderate to very good between sets 1 and 2b and increased between sets 2a and 2b. Action units with the highest and lowest ICC at set 2b were orbital tightening and whiskers, respectively. In comparison to an experienced rater, the ICC for all trainees improved, ranging from 0.88 to 0.91 at set 2b. Four years later, very good interrater reliability was retained, and intrarater reliability was good or very good). The interrater reliability of the no-training group was moderate and did not improve from set 1 to set 2b. Training improved interrater reliability, with an associated reduction in 95%CI. In addition, training improved interrater reliability with an experienced rater, and performance was retained.