Multilingual validation of the short form of the Unesp-Botucatu Feline Pain Scale (UFEPS-SF)

Introduction

Historically, cats have their pain underestimated, and therefore undertreated, when compared to their companion animal counterpart, the dogs (Capner, Lascelles & Waterman-Pearson, 1999; Lorena et al., 2014; Steagall & Monteiro, 2019). The prototype Unesp-Botucatu Feline Pain Scale (UFEPS) was refined in 2011 (Brondani, Luna & Padovani, 2011) and the full scale validated in 2013 (Brondani et al., 2013b). Afterward, the development of other feline specific pain assessment tools (Calvo et al., 2014; Reid et al., 2017; Evangelista et al., 2019; Belli et al., 2021) improved recognition of pain-related behavior and veterinary health care attitudes towards the provision of analgesia (Simon et al., 2017; Steagall & Monteiro, 2019;).

According to the most recent revised definition of the International Association for the Study of Pain (IASP), pain is ‘an unpleasant sensory and emotional experience associated with, or resembling that associated with, actual or potential tissue damage’ (Raja et al., 2020). Pain is a multidimensional experience defined not only by sensation and intensity, it includes qualitative and temporal attributes involving the affective-emotional and individual cognitive dimensions (Steagall & Monteiro, 2019; Raja et al., 2020). According to IASP ‘verbal description is only one of several behaviors to express pain; inability to communicate does not negate the possibility that a human or a nonhuman animal experiences pain’ (Raja et al., 2020, page 2). Because humans are still not able to understand animals’ verbal expression, the recognition and the magnitude of animal pain rely on objective and/or subjective clinical pain assessment methods (Gaynor & Muir III, 2009; Steagall & Monteiro, 2019).

Objective methods are observer-independent and less prone to observer bias. They include nociceptive tests (Dixon et al., 2007), physiological parameters (Brondani, Luna & Padovani, 2011; Brondani et al., 2013b), and locomotor activity, like force plate, accelerometer, and gait analysis (Moreau et al., 2014; Klinck et al., 2015). Although they are important for certain types of studies, they cannot be used solely, have not been fully validated, and are generally non-specific. They may be invasive, require equipment, demand physical contact and the evaluators’ presence. Their correlation with acute pain is questionable (Steagall & Monteiro, 2019; Nicholls et al., 2021). Endocrine changes require laboratory analysis, which is time-consuming, costly and not accessible in real-time (Smith et al., 1996; Cambridge et al., 2000). Previous use of drugs, interference of the emotional state, and low correlation with acute pain restrict the use of simple and clinically applicable physiological measurements like heart and respiratory rates, salivation and pupil diameter (Höglund et al., 2018; Cambridge et al., 2000). In cats, systolic arterial blood pressure seems to be the only good indicator of acute pain (Smith et al., 1996; Brondani, Luna & Padovani, 2011; Brondani et al., 2013b). These characteristics make these methods applicable only to specific experimental conditions but difficult to perform in clinical research.

Subjective pain assessment methods are usually based on pain-related behaviors, are minimally invasive and do not rely on equipment use. Some instruments allow remote evaluation and facilitate pain assessment in the research and clinical setting (Steagall & Monteiro, 2019). However, before clinical implementation, a pain assessment instrument should be valid and reliable. Validity is the tool’s effectiveness to measure what has been proposed to and involves the three ‘C’s’ (criterion, content and construct validity) (Streiner, Norman & Cairney, 2015). Reliability demonstrates how much this measure is error-free by assessing its internal consistency, test/re-test stability, intra (repeatability) and inter-reliability (reproducibility) (Jensen, 2003). These attributes standardize the evaluations to guarantee reproducibility among scientific studies. In veterinary practice, a pain assessment instrument must indicate the analgesic intervention score to guide decision-making (Brondani et al., 2013b; Reid et al., 2017; Steagall & Monteiro, 2019).

Unidimensional (visual analogue, simple descriptive and numerical) scales are more straightforward when compared to composite scales. However, they are subjective, have limited reproducibility (Holton et al., 1998; Martinez-Martin, 2010; Belli et al., 2021), and may have ambiguous meaning (Martinez-Martin, 2010) since they do not encompass the multiple dimensions of pain (Robertson, 2018). On the other hand, composite and multidimensional scales have better consistency and accuracy than unidimensional ones because they evaluate different dimensions of pain (Martinez-Martin, 2010). In cats, four composite multi-items, weighed scales have undergone psychometric testing and some reported validity for acute pain assessment. They are chronologically the UFEPS (Brondani et al., 2013b), Glasgow Feline Composite Measure Pain Scale (CMPS-Feline) (Calvo et al., 2014; Reid et al., 2017), Feline Grimace Scale (FGS) (Evangelista et al., 2019), and the UFEPS-short form (UFEPS-SF) (Belli et al., 2021).

The UFEPS shows ‘evidence of validity, reliability and sensitivity at the level of a randomized control trial’ (Merola & Mills, 2016, page 60). It is available for online training at http://www.animalpain.org. However, the UFEPS is long and time-consuming, as it includes three subscales and 10 evaluation items. Using it in its entirety requires blood pressure measurement, which is not always feasible and may disturb cats. Appetite is another difficult variable to assess in real-time or when cats are being fasted, for example. Because the UFEPS is a biological and statistical multidimensional instrument, these physiological measurements may be excluded since each dimension (subscale) may be assessed separately and have their independently calculated intervention analgesia score.

To encourage widespread use of pain scales, they should be short, simple and easy to score. Our recently developed UFEPS-SF, together with UFEPS, underwent clinical validation against a control group, clinical pain, soft tissue and orthopedic surgery, overcoming some of the previous limitations of UFEPS (Belli et al., 2021). However, only real-time pain assessment was performed, evaluators were not blinded to the pain condition, and intra-observer reliability was not calculated (Belli et al., 2021). Therefore, the UFEPS-SF requires further validation via video-scoring in a blinded manner to calculate intra- and inter-observer reliability and a more in-depth statistical validation using a larger number of observers.

Finally, a crucial gap that limits the use of an instrument is language and culture. Simple literal translations do not assure the same consistency and accuracy when the instrument is translated to different languages (Guillemin, Bombardier & Beaton, 1993; Beaton et al., 2000; Sousa & Rojjanasrirat, 2011; Streiner, Norman & Cairney, 2015). Thus, to ensure scientific rigor, the instrument must be validated in the language and culture of use (Beaton et al., 2000; Sousa & Rojjanasrirat, 2011; Streiner, Norman & Cairney, 2015) to assert the same semantics as the original scale (Sperber, 2004), like reported for McGill questionnaire in people (Maiani & Sanavio, 1985; Boureau, Luu & Doubrère, 1992; Kim et al., 1995; Lázaro et al., 2001; Varoli & Pedrazzi, 2006; McGlone, Guay & Garcia, 2016). In feline medicine, the UFEPS is the only instrument for assessing acute pain-related behaviors with reported validation in five languages: English (Brondani et al., 2013b), Portuguese (Brondani et al., 2012; Brondani et al., 2013a), Spanish (Brondani et al., 2014), French (Steagall et al., 2017) and Italian (della Rocca et al., 2018).

This study aims to test the psychometric properties of the UFEPS-SF in cats in eight languages, based on the evaluation of intra- and inter-rater reliability, responsiveness, content, construct and criterion validities, principal components analysis, item-total correlation, internal consistency, specificity, sensitivity, and score for indication of analgesic intervention.

Materials & Methods

This study followed the Consensus Based Standards for the Selection of Health Measurement Instrument (COSMIN) checklist and terminology for assessing the methodological quality of studies (Mokkink et al., 2010a; Mokkink et al., 2010b) and the ARRIVE guidelines 2.0 (Percie du Sert et al., 2020).

Results

The English and the translated versions of UFEPS-SF is presented in Table 2 and Table S1 (http://www.animalpain.org).

Distribution of scores

For each item of UFEPS-SF, the score 0 prevailed before surgery (M1) and after rescue analgesia (M3). Scores 2 and 3 predominated at M2, a moment that scores were expected to be the highest after surgery, and scores 0, 1 and 2 were present at the moment of moderate pain (M4—24 h after the end of surgery) (Fig. 1).

Multiple association

According to principal component analysis only one-dimension resulted in eigenvalue >1 (3.38) and variance >20 (84.34) (Table S4). Loading values were 0.95 for posture and miscellaneous behaviors, 0.94 for attitude and 0.83 for reaction to palpation. Therefore all items had loading values above 0.50 and were included in dimension 1 (Fig. 2). Only reaction to palpation had a loading value <−0.5 (−0.56) in dimension 2, however eigenvalue of dimension 2 was below the acceptable limit 1 (0.39). Therefore UFEPS-SF is unidimensional and confirmatory factor analysis seemed redundant.

Responsiveness and construct validity

The scores of all scales, of all items of UFEPS-SF and of the need for rescue analgesia increased after surgery (pain) and decreased to baseline after rescue analgesia. Scores at 24 h after surgery (moderate pain) were intermediate between before and after surgery, showing that the scales were responsive to pain, to analgesic treatment and differentiated severe from moderate pain, therefore confirming responsiveness and construct validity hypothesis (Table 3; Figs. 3, 4 and 5). In all cases for each evaluator and gender, the UFEPS-SF results were the same as described above (Figs. 4 and 5). At all time-points, scores from female observers were higher than those of male individuals. The decrescent order of total scores were: German¹, Chinese^1,2, French/Italian/English^2,3, Spanish/Portuguese^3,4 and Japanese⁴ (1 > 2 > 3 > 4). According to these results, there was an effect of gender and evaluator in the model (Figs. 4 and 5).

Table 3:

Pain scores of the UFEPS-SF, CMPS-Feline, unidimensional pain scales and rescue analgesia indication before and after surgery, after rescue analgesia and 24 h after surgery in cats (n = 30).

Scales	Before surgery		After surgery		After rescue analgesia		24 h after surgery
	Median	Range	Median	Range	Median	Range	Median	Range
Rescue analgesia	0c	0–0	1a	0–1	0c	0–1	0b	0–1
Numerical rating rate	1c	1–4	8a	1–10	1c	1–4	2b	1–8
Simple descriptive scale	1c	1–2	4a	1–4	1c	1–3	2b	1–4
Visual analog scale	0d	0–42	76a	0–100	0c	0–44	12b	0–95
Item 1 (posture)	0c	0–2	3a	0–3	0c	0–2	0b	0–3
Item 2 (miscellaneous)	0c	0–2	3a	0–3	0c	0–1	0b	0–3
Item 3 (attitude)	0c	0–1	2a	0–3	0c	0–1	0b	0–3
Item 4 (palpation)	0c	0–2	2a	0–3	0c	0–2	1b	0–3
UFEPS-SF	0c	0–5	9a	0–12	0c	0–4	2b	0–12
CMPS-Feline	0c	0–5	14a	1–20	0c	0–6	3.5b	0–14

DOI: 10.7717/peerj.13134/table-3

Notes:

UFEPS-SF

Unesp-Botucatu Feline Pain Scale—Short form

CMPS-Feline

Glasgow Composite Multidimensional Pain Scale (Reid et al., 2017)

Different letters express significant differences between moments where a > b > c > d, according to the mixed linear model (Silva et al., 2020).

Determination of the intervention point for rescue analgesia

The suggestive cut-off score suggesting the administration of analgesics according to the UFEPS-SF ROC curve was ≥4 out of 12 (Table 6; Fig. 6). Based on the two methods used to calculate the diagnostic uncertainty zone, the low and high confidence interval according to the bootstrap method was 3.5 and 3.5, respectively, and the interval between the sensitivity and specificity values 0.90 was between 2.5 and 4.7, therefore the last measure corresponding to the largest interval was used to define the diagnostic uncertainty zone for the UFEPS-SF and the other scales, as in most cases the last interval was the largest one (Fig. 6). According to this result, the diagnostic uncertainty zone scores ranged from 3 (≤2 indicates true negative pain-free cats) to 4 (≥5 indicates true positive—cats suffering pain). The same reasoning was adopted for the other scales. When palpation is not possible while using UFEPS-SF, the suggestive score for indication of rescue analgesia is ≥3 (Table 6).

Table 6:

Cut-off scores, specificity, sensitivity, Youden index and diagnostic uncertainty zone corresponding to intervention analgesia indication of the UFEPS-SF, UFEPS, CMPS-Feline and unidimensional scales.

Scale	Cut-off score	Specificity	Sensitivity	Youden index	Diagnostic uncertainty zone (scores)
					True negatives (pain-free)	True positives (pain)
UFEPS-SF (0–12) based on indication of rescue analgesia	4	96	96	0.91	≤2	≥5
UFEPS-SF (0–12) based on UFEPS ≥7	4	96	96	0.95	≤2	≥5
UFEPS-SF excluding palpation (0–9)	3	95	92	0.87	≤1	≥4
UFEPS (0–27) excluding blood pressure	7	98	98	0.97	≤4	≥12
UFEPS (0–24) excluding blood pressure and appetite	7	98	97	0,95	≤4	≥11
UFEPS pain expression only (0–12)	2	84	98	0.83	≤2	≥3
UFEPS psychomotor activity only (0–12)	3	89	97	0.85	≤3	≥6
CMPS-Feline (0–20)	5	93	98	0.91	≤4	≥7
NRS (1–10)	4	97	97	0.94	≤2	≥5
SDS (1–4)	3	99	94	0.93	≤2	≥3 ≥
VAS (0–100)	31	96	95	0.92	≤27,5	≥34,5

DOI: 10.7717/peerj.13134/table-6

Notes:

UFEPS-SF

Unesp-Botucatu Feline Pain Scale—Short form

UFEPS

Unesp-Botucatu Feline Pain Scale (Brondani et al., 2013b)

CMPS-Feline

Glasgow Feline Composite Measure Pain Scale (Reid et al., 2017)

NRS

numerical rating scale

SDS

simple descriptive scale

VAS

visual analog scale

The same cut-off point described above (≥4) was calculated for UFEPS-SF based on the UFEPS cut-off point ≥7 (excluding the physiological variables appetite and blood pressure) and so were the lower and upper scores of the diagnostic uncertainty zone (Table 6).

The cut-off score for the UFEPS excluding blood pressure is ≥7 out of 27, and by excluding the subscale physiological variables (blood pressure and appetite), it is ≥7 of 24. The cut-off point of the subscales pain expression and psychomotor activity were ≥2 and ≥3, respectively (Table 6) (Brondani et al., 2013b).

The AUC for all scales and subscales was above 97% (confidence interval = 95–100%). The AUC for the UFEP-SF was 99.2% indicating the high discriminatory capacity of all instruments (Streiner & Cairney, 2007). The percentage of cats showing scores inside the diagnostic uncertainty zone (scores 3 and 4) was low both when they were pain-free (≤5% for six evaluators and 16.7% for two evaluators) and when they were probably suffering the most intense pain (0% for six evaluators and 1.7% for two evaluators).

As presented before for sensitivity and specificity data, the percentage of cats where rescue analgesia was indicated before surgery, according to the Youden index (YI) of the UFEPS-SF (total score ≥4) was 0 for five evaluators, 1.7 for one evaluator and 3.3 for two evaluators. After surgery, rescue analgesia would be administered to all cats according to seven evaluators and to 98% of cats according to one evaluator. Therefore predictive criterion validity was confirmed for all evaluators as in most cases, cats would not be treated with rescue analgesia in the time-point they were pain-free (baseline) and would receive analgesia in the time-point they were possibly suffering pain after surgery.

Discussion

The UFEPS-SF is reliable and valid for acute pain-related behavior assessment in cats. Additionally, the cut-off for rescue analgesia guides the feline practitioner with the decision-making process on when to administer analgesics in the clinical setting. Both psychomotor change and pain expression subscales were summarized from four to two items, and physiological variables were removed. The original UFEPS was time-consuming and complex. These limitations have been addressed with the short-form making the instrument feasible to use and more practical while eliminating the need for appetite and blood pressure assessments (Belli et al., 2021). Appetite may not be easy to assess during fasting or postoperatively, for example, in cats undergoing gastrointestinal surgery or with a presence of an esophageal tube. Cats also tend to eat less on the first day of hospitalization (Zeiler et al., 2014). Blood pressure assessment is intrusive, demands equipment and may be inaccessible in feral or restless cats. However, because UFEPS is a multidimensional instrument, assessment of these items may be excluded, and the form can still be used since independent cut-off points are available for each subscale. Other limitations of both UFEPS and UFEPS-SF were addressed in the clinical validation study (Belli et al., 2021). In the latter, comparisons of these instruments with a negative control group and their use for assessing clinical pain and postoperative pain (including soft tissue and orthopedic surgery) were performed with reported validity. Additionally, the current study involved a large number of evaluators who were masked to the cat’s clinical condition and pain state, the calculation of intra-observer reliability, and the assessment of validity and precision of the instrument in other languages (Guillemin, Bombardier & Beaton, 1993; Beaton et al., 2000; Sperber, 2004; Streiner, Norman & Cairney, 2015).

Validation was performed for each item separately, and rescue analgesia score was calculated without response to palpation, therefore, if necessary for convenience, the evaluator may choose to omit response to palpation of the affected area to avoid physical contact and/or disturbing the cat that may already be in pain, without compromising the psychometrical properties of the scale. To our knowledge, the UFEPS-SF is the only animal health care instrument available in the literature simultaneously assessing and validating its psychometrical properties in several languages and following the COSMIN checklist, taxonomy, terminology, and definitions (Mokkink et al., 2010a; Mokkink et al., 2010b). Previously, the UFEPS underwent cross-language validation in five languages (Brondani et al., 2012; Brondani et al., 2013b; Brondani et al., 2013a; Brondani et al., 2014; Steagall et al., 2017; della Rocca et al., 2018); however, this was performed with different studies rather than simultaneously, as presented herein. It is important to ensure that the results of studies performed in different geographic, cultural, language, and ethnic regions worldwide are reproducible and interchangeable. In the present study, the study design followed current guidelines for health care instrument translation with some adaptations (Sousa & Rojjanasrirat, 2011).

Content validation of the English version of UFEPS-SF was performed beforehand. All items were relevant and approved. Furthermore, the lowest confidence interval of inter-rater reliability was above 80% for all language versions, so a pilot test for cognitive debriefing would not be necessary (Sousa & Rojjanasrirat, 2011).

Intra and inter-observer reliability of UFESP-SF were very good, ensuring that the instrument is respectively repeatable and reproducible like reported for UFEPS (Brondani et al., 2012; Brondani et al., 2013a; Brondani et al., 2013b; Brondani et al., 2014; Steagall et al., 2017; della Rocca et al., 2018). In this context, a minimum of 30 heterogeneous samples and three different evaluators should be used to assess the reliability (Koo & Li, 2016). In the current study, we used eight subjects and 120 videos (samples). Video-assessment results were slightly better than those obtained in the clinical study with both UFEPS and UFEPS-SF, where evaluators knew whether cats were possibly feeling pain because the assessment was performed in real-time and in-person, and evaluators were aware of the cats’ pain state (Belli et al., 2021). In the current methodology, observer bias was avoided by randomizing videos on each phase (Kaufman & Rosenthal, 2009). However, the estimation of whether the cat required rescue analgesia before scoring any scale, may have still biased the evaluators to record higher scores to cats they judged to require analgesia or vice-versa. Evaluators did not receive any training but were individuals with experience in veterinary anesthesia and pain management. In rats, training improves the capacity to discern distinct degrees of pain regardless of observers’ previous experience (Roughan & Flecknell, 2003). In terms of distribution of scores, they were proportionally increased according to pain intensity, confirming the importance and representativeness of each score level to reflect the degree of pain.

Principal component analysis provides an overview of the instrument by showing how the items associate with each other to determine how many dimensions compose the scale (Gracely, 1992). As expected for an instrument that measures pain, all items’ eigenvectors were pointed to the moments of pain (after surgery and 24 h) and in the opposite direction to moments cats were supposedly pain-free (before surgery) or experiencing mild pain (after rescue analgesia). According to the Kaiser criterion, one component was selected, hence UFEPS-SF is unidimensional (Streiner, Norman & Cairney, 2015) as reported for other pain scales in cattle (de Oliveira et al., 2014), pigs (Luna et al., 2020), sheep (Silva et al., 2020), horses (Taffarel et al., 2015; Barreto da Rocha et al., 2021) and donkeys (de Oliveira et al., 2021). The only statistically defined multidimensional animal pain scale in cats is UFEPS (Brondani et al., 2013b); however, this classification is based on multivariate analysis and not on biological terms. Pain is a multidimensional phenomenon in nature and, in biological terms, components of both UFEPS-SF and CMPS-Feline represent not only pain intensity, but qualitative and temporal features, like sensory, motor, emotional, and cognitive dimensions (Brondani et al., 2013b). Except for the physiological variables, UFEPS-SF incorporates psychomotor (posture, comfort, activity and attitude), sensorial (reaction to palpation of the painful area), pain expression (miscellaneous behavior) and temporal features (response to analgesia); therefore, because UFEPS-SF covers many aspects of pain, it could be considered biologically multidimensional.

There are different methods to explore dimensionality and multiple associations between variables. In the present study, the structural models, observed according to principal component analysis, could have been verified by confirmatory factor analysis by comparing a one-dimensional model (all items together in one dimension) versus a two-dimensional model (some items in one dimension and others in the second dimension) (Bollen, 1989). This is particularly relevant because the UFEPS-SF was derived from the UFEPS, a multidimensional scale. Usually, the selection of items for each dimension is based on loading values, which is the correlation between the item and the dimension. In our study, all four items showed loading values >0.80 in the first dimension and item 4 had a higher loading value in dimension 1 (0.83) than in dimension 2 (−0.53). Thus, there is no mathematical rationale to include item 4 alone in a second dimension and apply confirmatory factor analysis.

Criterion validity comprises concurrent and predictive validity. Concurrent criterion validity is the measurement of the strength of an instrument simultaneously compared to a previously validated instrument considered adequate to measure the target attribute. UFEPS was chosen as one of the ‘gold standard’ instruments since it has undergone a robust validity protocol (Merola & Mills, 2016). However, as UFEPS-SF was derived from UFEPS, the CPMS-Feline was also used for comparison (Reid et al., 2017), and so were the unidimensional scales, as previously reported in cats and other species (Barreto da Rocha et al., 2021; Brondani et al., 2013b; Luna et al., 2020; de Oliveira et al., 2014; Taffarel et al., 2015; Silva et al., 2020; de Oliveira et al., 2021). The correlations equal to or above 0.9 in all comparisons confirmed concurrent criterion validity.

Predictive criterion validation defines how predictable the practical results would be when using the instrument. In this case, the selected variable was the percentage of cats that should receive rescue analgesia, and therefore would have benefited, when they were suffering postoperative pain, and the percentage of cats that would receive unnecessary analgesia when they were pain-free before surgery. According to the sensitivity and specificity, 100% of cats would receive rescue analgesia when the highest pain scores were expected (true positives) and 1% of cats would be administered analgesics when they were supposedly pain-free (false positives), confirming adequate predictive criterion validity of UFEPS-SF. Even when considering the data of each evaluator and the Youden index, only one of the eight evaluators would not provide analgesia in 2% of the cats at the time-point they were possibly suffering pain, and one and two evaluators would provide analgesia in 1.7 and 3.3% of the cats when they were pain-free, respectively. Only a few cats had scores within the diagnostic uncertainty zone, safeguarding decision-making about whether rescue analgesia should be provided in the clinical setting.

All scales and all items of UFEPS-SF were responsive to pain and rescue analgesia and were capable of discriminating intense from moderate pain, which confirms responsiveness as reported with UFEPS (Brondani et al., 2013b) and its clinical validation (Belli et al., 2021). Construct validity was corroborated by the three-hypothesis testing: (1) pain-related behaviors increased after surgery and (2) decreased after analgesia and (3) changed over time as expected. These changes confirm that the UFEPS-SF measures the construct pain. Other COSMIN approaches to assess construct validity included: i. internal relationships, given by the results of internal consistency, item-total correlation and principal component analysis; ii. relationships to scores of other instruments (as reported for criterion validity); and iii. crosscultural validity (Mokkink et al., 2010a; Mokkink et al., 2010b). The assessment of construct validity based on differences between relevant groups (Mokkink et al., 2010a; Mokkink et al., 2010b), has been performed in a previous publication where cats possibly experiencing pain in different clinical and surgical conditions showed higher UFEPS-SF pain scores than a negative control group of supposedly pain-free cats (Belli et al., 2021).

The exclusion of each item of UFEPS-SF influenced little and similarly item-total correlation, except reaction to palpation, suggesting that items have a first-hand association. The finding that the correlation was above 0.3 means that all items contribute to the scale. However, the correlation above 0.7 for its first three items may indicate that the scale is too specific, where one item may reiterate others (Streiner, Norman & Cairney, 2015).

Internal consistency was excellent for UFEPS-SF. When each item correlates well with the total score it is expected that internal consistency reduces by excluding the target item. This was the case for all items, except reaction to palpation, as mentioned for item-total correlation. According to the principal component analysis, this might be explained by the fact that reaction to palpation fits in two dimensions. Otherwise, the other items fit in only one dimension.

Any health instrument used for a specific diagnosis should correctly identify true positive (sensitivity) and negative individuals (specificity). The UFEPS-SF is specific and sensitive for acute pain-related behavior assessment in cats. The narrow diagnostic uncertain zone guarantees that cats suffering pain will likely receive analgesia and pain-free individuals will not be unnecessarily treated with analgesics. The high AUC of all scales and items of UFEPS confirmed the high discriminatory capacity of the tests and items.

This study presented some limitations. Only one observer for each language version was used, but because both intra and inter observer reliabilities were very good, their results were similar. Observers were experienced and, because in laboratory animals, training and experience enhanced pain recognition (Roughan & Flecknell, 2006), the next step is to validate the reproducibility of these results from naive or less experienced observers. Observations based on video analysis do not correspond to real-time assessment. In the current study, behaviors were condensed in short-edited videos, which might favor the observation of behavior exhibits and occurrences, especially of less frequent behaviors. The observer could pause and rewatch videos, possibly improving the detection of clinical signs of pain and reliability. A positive side of remote analysis is that real-time observation assessed by the in-person evaluator might affect pain expression as reported in rabbits (Pinho et al., 2020). However, Feline Grimace Scale scores were not different between real-time and video assessment (Evangelista et al., 2020) and were not affected by the presence of a caregiver during pain assessment (Watanabe et al., 2020). This limitation has already been addressed in the previous clinical validation study of UFEPS-SF, based on real-time comparisons (Belli et al., 2021). This previous study has shown that the UFEPS-SF may be used for different types of pain, including medical pain and postoperative assessment of soft tissue and orthopedic surgery (Belli et al., 2021).

Although the differences in time-point results were the same for evaluators, gender and when data were grouped, the influence of gender and evaluator is a potential bias in pain assessment. Three observers were male and five were female, which may be considered a small number of evaluators to assess gender effect. Women have more compassion towards pain than men (Sadeghiyeh, Khorrami & Hatami, 2012; Christov-Moore et al., 2014). Female veterinarians attribute higher pain scores (Williams, Lascelles & Robson, 2005) and consider that dogs and cats experience more pain after surgery than male individuals (Beswick et al., 2016). Therefore based on the current study and the gender-related empathy towards pain, the effect of gender should be considered in the future when validating pain scales. In practical terms, the evaluator effect was not sufficient to modify the percentage of cats that would (after surgery) or would not (baseline) receive analgesia, according to the Youden index.

Although sample size was not calculated, the number of cats was based on the guidelines for reliability studies (Koo & Li, 2016) and on the minimum subject to item ratio of 5:1 for exploratory factor analysis (Hair et al., 2014). In the current study subject to item ratio was 7.5 (30 cats:4 items).

Conclusions

The UFEPS-SF and its items, assessed by experienced evaluators, demonstrated very good intra-rater and inter-rater reliability, responsiveness, appropriate content, criterion and construct validity, item-total correlation, internal consistency, excellent sensitivity and specificity in Chinese, French, English, German, Italian, Japanese, Portuguese and Spanish languages. A cut-off score indicating the need to administer rescue analgesia was defined. Future studies are necessary to validate the reproducibility of these results when naive or less experienced observers assess the UFEPS-SF.

Supplemental Information