Table 2. Statistical analyses used for validation of the short (UFEPS-SF) and long (UFEPS) versions of the UNESP-Botucatu multidimensional feline pain assessment scales.
| Analysis Type | Description | Database | Test |
|---|---|---|---|
| Inter-rater reliability | A matrix was generated to assess the agreement of the total score of each scale among the observers. | All observers (3), cats (n = 50), groups and time points were used (7 time points in Surgery group—before surgery, 1, 2, 3, 4, 5 and 6 h after recovery from anesthesia; 2 time points in Clinical group—before and 20 min after rescue analgesia; and 5 time points in the Control group—0 (30 min after admission), 30, 60, 90, and 120 min. | For the UFEPS, UFEPS-SF and Glasgow CMPS-Feline, intraclass coefficient (ICC) ”consistency” type was used and its 95% confidence interval (CI). For the NS and SDS the weighted kappa coefficient was used. The 95% CI kw (“cohen.kappa” function of the “psych” package) was estimated. For the VAS, intraclass coefficient (ICC) “agreement” type was used and its 95% CI (”icc” function of the ”irr” package). Interpretation of values: <0.5 poor; 0.5–0.75 moderate; 0.75–0.9 good; >0.9 excellent (Koo & Li, 2016). |
| Concurrent criterion validity | Correlation of the total score between all scales at all time points. | Spearman rank correlation coefficient (r; “rcorr” function of the “Hmisc” package). Interpretation of the degree of correlation: <0.19 very weak; 0.2–0.39 weak; 0.4–0.59 moderate; 0.6–0.79 strong; 0.8–1—very strong (Evans, 1996). | |
| Construct validity (responsiveness to control group) | The responsiveness of each scale was determined by testing the hypothesis that animals in the SG and ClinG groups have higher pain scores than in the CG. | The time point with the highest UFEPS score for each cat before rescue analgesia was selected from the main evaluator. The database included the scores of all evaluators at these same highest UFEPS score time points [3 evaluators × 40 time points/cats (20 cats from SG; OrthG, n = 12 and SoftG, n = 8 and 20 cats from ClinG) = 120)] were compared to the scores of all evaluators assessed at 120 min in CG (3 evaluators x 1 time point × 10 cats = 60). | Analyses were performed for the OrthG, SoftG, and ClinG separately, as well as for the SG and ClinG together. Unpaired Wilcoxon test was used to compare scores (“wilcox.test” function of the “stats” package). |
| Construct validity (responsiveness to rescue analgesia) | Responsiveness and effect of time (sensitivity to change) was determined for all scales by testing the hypothesis that scores after analgesia are lower than those before analgesia. | Only one-time point before and one-time point after rescue analgesia of cats that received analgesia were selected from SG (SG, surgery; OrthG, orthopedic, n = 11 and SoftG, soft tissue surgeries, n = 5), and those that did not receive sedation from ClinG were included (clinical, n = 13) (3 evaluators × 29-time points = 87). | Analyses were performed for the OrthG, SoftG, and ClinG separately, as well as for the SG and ClinG together. The paired Wilcoxon test was used to compare the scores before and after analgesia (“wilcox.test” function of the “stats” package). |
| Sensitivity of the scale | Based on true positives—cats with pain (surgical and clinical groups). | 272 and 244 time points of grouped SG and ClinG when UFEPS scores were ≥ 7/24 and Glasgow CMPS-Feline were ≥ 5 for the three evaluators were used as database respectivelly. From these time points, the number of time points at which each scale had their score ≥ the cut-off point was filtered (cats that were supposedly feeling pain—true positives) and divided by these time points | Sensitivity = True positives/Total number of time points (“ci.coords” function of the “ROCR” package). Interpretation: excellent 100–95%; good 94.9–85%; moderate 84.9–70%; not sensitive <70% (Streiner & Norman, 2008). |
| Specificity of the scale | Based on true negatives—cats without pain from the control group (CG). | All time points of all evaluators of CG (5 time points × 10 cats × 3 evaluators = 150). The calculation was based on the number of time points when each scale presented a score lower than the cut-off point (cats that were supposedly not feeling pain - true negatives) divided by the total number of time points. | Specificity = True negatives/Total number of time points (“ci.coords” function of the “ROCR” package). Interpretation: excellent 100–95%; good 94.9–85%; moderate 84.9–70%; not specific <70% (Streiner & Norman, 2008). |
Notes.
Groups: SG, surgery group (OrthG, orthopedic; SoftG, soft tissue); ClinG, clinical; CG, control group. Cut-off points for calculation of sensitivity and specificity: numeric (NS) ≥ 4/10, simple descriptive (SDS) ≥ 2/4, visual analog (VAS) > 28/100 (Brondani et al., 2013a), the short version of the UFEPS (UFEPS-SF) ≥ 4/12, UNESP- Botucatu multidimensional feline pain assessment scale (UFEPS) ≥ 7/24 and Glasgow feline multidimensional pain scale (Glasgow CMPS-Feline) ≥ 5/20 (Reid et al., 2017).