Objective: To investigate the reliability of 3 scales used for assessment of pain in dogs.
Design: Prospective study.
Animals: 50 dogs that had surgery.
Procedure: Dogs were allocated into 3 groups (group 1, 25 dogs assessed 1 hour after the end of surgery; group 2, 41 dogs assessed between 21 and 27 hours after the end of surgery; group 3, 16 dogs assessed on the day of surgery and on the subsequent day). Each dog was scored for pain 4 times by 3 (groups 1 and 3) or 4 (group 2) veterinarians, using all 3 scales (i.e. simple descriptive, numerical rating, and visual analogue) during each scoring period. Analysis of data was performed using ANOVA, log-linear modeling, calculation of reproducibility coefficients, and Cohen's kappa statistic.
Results: Significant variability existed among observers for use of all 3 scales. Variability among observers and between observers and dogs accounted for 29 to 36% of the total variability (group 1, 36.1 and 32.3% and group 2, 35.1 and 29.7%, for visual analogue scale and numerical rating scale scores, respectively). Kappa statistic values calculated for data obtained by use of the simple descriptive scale indicated that agreement was fair for the observers (group 1, 0.244 to 0.299; group 2, 0.211 to 0.368; group 3, 0.233 to 0.321).
Clinical implications: Analysis of pain score data in dogs must incorporate observer variability when more than 1 observer is used. Comparative analysis of data accrued from pain studies in various hospitals must account for this variability.