Development and evaluation of a health‐related quality‐of‐life tool for dogs with Cushing's syndrome

Abstract Background Clinical signs and consequences of Cushing's syndrome are likely to impact upon a dog's life. Quantification of this impact on a dog's health‐related quality‐of‐life (HRQoL) could contribute to optimized disease management. Hypothesis/objectives To develop a novel HRQoL tool to aid assessment of dogs with Cushing's syndrome and to evaluate factors that impact upon dogs living with this disease. Animals Two hundred and ten dogs with Cushing's syndrome and 617 dogs without Cushing's syndrome. Methods Cross‐sectional study design. Dog owners answered questions relating to the HRQoL of their dogs which were refined to develop the final tool. The tool was analyzed for reliability, validity, and interpretability, including Cronbach's alpha and principal components analysis. Factors impacting upon the HRQoL of dogs with Cushing's syndrome were assessed using appropriate nonparametric tests. Results The tool was refined from 32 questions to 19 and showed good internal consistency (α = .83). Owners rated questions related to “owner impact” as more important and those related to demeanor as less important. There was a positive correlation between the tool score of dogs with Cushing's syndrome and owner's assessment of their dog's quality‐of‐life (r = .41, P < .001). Dogs currently on treatment with trilostane had a statistically better HRQoL (.33, interquartile range [IQR] .23–.44) than those not receiving trilostane (.36, IQR .33–.54, P = .04). Conclusions and Clinical Importance The developed tool quantifies the HRQoL of dogs with Cushing's syndrome and could assist clinicians in the clinical assessment of dogs with Cushing's syndrome.


| INTRODUCTION
Assessing the quality-of-life of animals is an integral role of a veterinarian and is required during decision-making on treatment and euthanasia to optimize the health and welfare of animals under their care. 1 In the current study, health-related quality-of-life (HRQoL) refers to the state of an individual animal's life as thought to be perceived by them at a point in time by their owner. This includes the physical, social, and environmental needs and impacts which are reflected by the animal's health and behavior. 2 A fundamental issue currently is that assessment of welfare or quality-of-life is not standardized or validated. The assessment is nearly always subjectively and compassionately inferred by veterinary professionals and animal owners. Consequently, quantification is increasingly promoted to optimize and standardize decision-making in this area. [2][3][4] The British Veterinary Association's Animal Welfare Strategy highlighted the use of welfare assessments as 1 of their 6 priorities, which includes the use of practice-based quality-of-life assessments. 5 In human medicine, formal HRQoL measures are commonly implemented in practice to provide additional information about the dogs without solely assessing laboratory results or clinical outcomes. 6 It is accepted that lack of assessment on a dog's HRQoL could result in inadequate relief from suffering and suboptimal clinical decision-making 4,7 with an awareness that the severity of the clinical signs affecting an individual might correlate poorly with results of routine blood tests. 8 Practicebased quality-of-life tools developed for veterinary medicine have followed the methodology produced in human medicine to measure HRQoL. 2,[9][10][11][12][13][14] Cushing's syndrome in dogs results from excessive circulating glucocorticoids. The disease is clinically characterized variably by polyuria and polydipsia, polyphagia, bilateral alopecia, muscle atrophy with generalized weakness, hepatomegaly, systemic hypertension, and lethargy. 15,16 These clinical signs can all impact upon dogs' as well as to their owner's lives. Currently no tool to quantify the impact or the long-term residual effects of Cushing's syndrome on a dog's life has been published. Therefore, such a tool is warranted to optimize disease management, taking the financial and emotional strain of Cushing's syndrome to the owner into consideration in its design. Any negative impact of disease and treatment on the owner could lead to cessation of treatment or even euthanasia.
The aims of this study were to develop a novel HRQoL tool to aid clinical assessment of dogs with Cushing's syndrome and to evaluate factors that might impact upon the quality-of-life of dogs with this disease. It was hypothesized that Cushing's syndrome cases not receiving treatment would have a poorer HRQoL.

| METHODS
HRQoL tool development followed a standard psychometric process of item identification, selection, and refinement. 9,10,17,18 An item was defined as any aspect of Cushing's syndrome and its management that could potentially impact on a dog's HRQoL. Health-related quality-of-life tools must be shown to be valid, reliable, and interpretable before recommending their use in a clinical context. 8,12,18  Owners were asked to describe the frequency of each specified item impacting on their dog's life over the previous week. Responses were assigned a score (all the time (3), often (2), occasionally (1), never (0) for negatively phrased questions). The scores were reversed in positively phrased questions.

| Item refinement
To develop the finalized "CushQoL-pet" tool, the questions included were refined based on statistical analysis of the responses.
1. Chi-squared analysis was performed on each item, comparing the results from dogs with and without Cushing's syndrome. Items with at least weak evidence of differences between the 2 groups were retained as these were deemed specific to the impact of Cushing's syndrome on HRQoL (P < .20).
2. Internal consistency of the questionnaire was measured by Cronbach's alpha, using only the responses of owners of dogs with Cushing's syndrome. Internal consistency indicates the reliability of the questions to measure the same latent concept. 20 In the context of this study, the latent concept was "HRQoL." Cronbach's alpha was calculated using a 1-way repeated measure analysis of variance model, with HRQoL question responses functioning as the repeated measure. Initially, correlations were examined in an inter-item correlation matrix to assess how much each individual question responses correlated with all included questions. Low correlations (r < .30) were deemed poor and those questions were removed if the overall Cronbach's alpha coefficient increased after removal. Correlations between pairs of questions were examined to check whether they were deemed highly correlated (r > .60), suggesting the same information is being captured twice therefore falsely raising the internal consistency of the tool. 21 The question with the smallest effect on the Cronbach's alpha was removed. An overall test Cronbach's alpha of α > .70 for the retained questions was deemed an appropriate internal consistency. 22 Internal validation using the dog's name, age, sex, and breed prevented duplication of responses relating to a single dog. If duplicates were found, the earliest response was used for analysis.

| Interpretation, validation, and reliability of CushQoL-pet
After refinement of the tool, the finalized questions were utilized to produce a combined score of HRQoL, rating between 0 and 1 (0 indicating the best possible HRQoL and 1 indicating the worst possible).
The scoring was calculated as follows: CushQoL− pet Score = Σ of the question scores=total maximum score Questions were included in the questionnaire to assess the validity of the tool. One question asked owners to describe their dog's current quality-of-life on a 7 point scale (from "as good as it could be" to "as poor as it could possibly be"), to assess construct validity. Correlations were analyzed with Spearman's rank correlation. Wilcoxon rank-sum and Kruskal-Wallis tests compared CushQoL-pet scores of dogs with and without Cushing's syndrome. Dogs with and without Cushing's syndrome were further categorized by (1) age group (<7, 7-11, >11 years), and (2) health status ("healthy" and "not-healthy" group, if a disease other than Cushing's syndrome was reported by the owner).
Principal components analysis (PCA) assessed the underlying structure and identified subsets within the CushQoL-pet tool. The principal components describing the largest amount of data variation were retained for further interpretation. Retention was based on visualization of the decreasing proportion of data variance described by each principal component, using a scree-plot. 21,23 For each retained principal component, the HRQoL question loading scores were analyzed and interpreted to observe those with the greatest influence on each principal component. Loadings closest to −1 or 1 for an item indicate a strong influence on the component. 21 A loading of ≥0.3 was selected as an appropriate cutoff. 10,24 The internal consistency of the identified subsets was examined with Cronbach's alpha. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was used postestimation to assess whether the patterns of correlation from the PCA were relatively compact and therefore the results were reliable. 25 KMO values >.50 indicate adequate sampling.
To assess the reliability of the tool, inter-rater measures of the questionnaire were assessed on 13 dogs, with pairs of owners of the same dog completing the questionnaire independently of each other.
Intra-rater reliability was carried out with 15 owners to examine the stability of the responses from the same person carrying out the questionnaire at an interval of 2 weeks, with no changes to the management of their dog's Cushing's syndrome. Paired scores were assessed with the intraclass correlation coefficient (ICC), with results interpreted as poor reliability (<.50), moderate (.50 to <.75), good (.75 to <.90), and excellent (≥.90). 26 Bland-Altman plots were analyzed to assess score agreement between 2 owners and repeat response at a 2 week interval. 27

| Evaluation of factors impacting the HRQoL of dogs with Cushing's syndrome
The online questionnaire also asked owners to provide some additional information as well as the core questions. Owners assessed the importance of each HRQoL question to themselves and their dog (1), not at all important (0)). 10 Inter-rater and intra-rater reliability assessments of owner reported importance were also carried out as described above. Additional demographic information relating to their dog included age, breed, sex, weight, insurance status, and other health concerns. Specific questions about owners included owner lifestyle, time spent with their dog, and whether they were the primary care-giver. Owners of dogs with Cushing's syndrome were asked disease-specific questions, including treatment currently received, time since diagnosis and how their dog's quality-of-life had changed since their diagnosis. Differences between dogs with and without Cushing's syndrome were analyzed using chi-squared analysis. Factors impacting upon the CushQoL-pet score were assessed using nonparametric analyses (either Wilcoxon rank-sum test or Krukshal-Wallis test). Statistical significance was set at <.05.
From the focus group discussions, interviews, and reviews of relevant literature, 32 HRQoL items specific to dogs with Cushing's syndrome were identified.

| Item selection
A questionnaire was developed incorporating the 32 items identified.
During pretesting, 6 questions deemed inappropriate or ambiguous were removed and 3 were reworded for clarification resulting in 26 HRQoL questions included in the online questionnaire. When comparing responses to the HRQoL questions by owners of dogs with or without Cushing's syndrome, no difference was observed in the responses to "medication stress" or "off food" so these were removed from the tool (P = .22 and P = .33, respectively). Based on correlations between an individual question and all other HRQoL questions, 3 questions were removed ("frequency of urination", "vet stress," and "begs for food") as they were poorly correlated to the other items (r = .29, .23, and .22, respectively), improving the internal consistency of the score. A number of HRQoL questions were found to be highly correlated with each other: "thirsty" with "emptying water bowl" (r = .68) and "weak" with "struggles to walk" (r = .63). "Emptying water bowl" and "weak" were removed from the score as these resulted in the least change in Cronbach's alpha coefficient. The process of item refinement reduced the number of items from 26 to 19 in the final CushQoL-pet tool, with a Cronbach's alpha of α = .83 (Table 1).

| Item refinement
All 32 items initially identified for inclusion in the online questionnaire which were subsequently retained or excluded from the final tool, CushQoL-pet, are outlined (Supporting Information).

| Interpretation, validation, and reliability of CushQoL-pet
The median HRQoL score for dogs with Cushing's syndrome using the  When assessing the reliability of the HRQoL questions, inter-rater (n = 13, ICC = .88, 95% CI .55-.97) and intra-rater agreement (n = 15, ICC = .78, 95% CI .49-.92) indicated good reliability. There was also a correlation with Bland-Altman plots, which suggest good agreement of paired owner and repeated owner responses for the HRQoL questions ( Figure 2).
Test-retest results showed a correlation between the difference in the 2 CushQoL-pet scores (Spearman's rho = .64, P < .001) and how owners described a change in their dogs quality-of-life (Table 3). Test-retest results also showed a significant correlation between owners assessment of quality-of-life and the CushQoL-pet score (rho = .69, P < .001). F I G U R E 2 Bland-Altman plots of inter-rater (n = 13) and intrarater (n = 15) owner scores of CushQoL-pet 9-13) and dogs without Cushing's syndrome was 7 years (IQR 4-10).   Table 4).

| Evaluation of HRQoL in dogs with Cushing's syndrome
T A B L E 3 Median test-retest CushQoL-pet scores at initial completion and at 3 month follow-up, stratified by owner assessment of change in their dog's quality-of-life in this time period (n = 71 These are deemed important qualities of a quality-of-life tool 8,12,29 ; however, they are infrequently assessed in published quality-of-life assessments in dogs. 29 The internal consistency of CushQoL-pet indicated good reliability to measure the same latent concept; "HRQoL of dogs with Cushing's syndrome" (α = .83). Inter-rater, intra-rater, and test-retest assessments of the tool further indicated reliability of owner completion of CushQoL-pet. The overall reliability of the interrater was slightly higher than the intra-rater reliability. This suggests . This indicates that the HRQoL described by CushQoL-pet is specific to Cushing's syndrome and suggests the score is not highly influenced by the dog's age or other morbidities which has been a concern regarding the application of disease-specific HRQoL tools. 29 When comparing the changes in the median test-retest scores with the owner assessment of the change in their dog's quality-of-life over the same time period, the tool was able to detect the direction of change (either improvement or deterioration). The median score differences suggested that the tool was better at indicating improvement in HRQoL than a deterioration in HRQoL. A decreased score of about −.10 indicated owner-assessed improvement and an increase of +.05 indicated deterioration. This study focused purely on owner-reported HRQoL, unaffected by veterinarian's opinions, as this type of reporting is currently lacking in the veterinary literature. Nevertheless, the lack of veterinarian assessment of health status is a potential limitation of this study. Future replication of results within a practice setting, alongside veterinarian evaluation of a clinical assessment, could provide further evidence of reliability and validity. 30 In particular, assessment of changes in owner questionnaire response behavior over time and evaluation of CushQoL-pet's responsiveness to changes in HRQoL would be of interest. 31 The design of the tool was intended for it to be quick for owners to complete, as well as being easy to interpret for veterinarians to encourage its uptake in primary-care practice. The tool is comparable in length to other HRQoL tools, 13,14,32 with some other published quality-of-life tools noticeably longer. 11,33,34 The average time to complete the questionnaire during this study was 6 minutes. However, this included a number of additional questions that will not be included in the final version used in practice and therefore completion of the CushQoL-pet in a clinical setting is likely much shorter than this. A suggested integration of the tool into practice would be during therapeutic monitoring consultations. During refinement of the tool, 7 questions were removed as they were either shown to be poorly correlated with the other questions, not specific to the Cushing's syndrome dog population or were highly correlated with another question, indicating repetition. A recent study found that the shortening of a much longer tool was valid and would likely increase its acceptability. 35 A scoring system on a 0-1 scale was used without weighting of the questions to ensure the final score was easy to calculate and interpret within primary-care practice. 8,36 Three principal components were retained for further analysis Areas described of highest importance to owners and their pets generally related to areas of "owner impact." This is interesting but perhaps unsurprising, reflecting similar findings in other studies. 10,39 Owner-related questions included within the HRQoL tool were those that impact upon the dog and potentially affect how Cushing's syndrome is managed. 40

| CONCLUSIONS
In conclusion, CushQoL-pet is the first tool to quantify the HRQoL of dogs with Cushing's syndrome. The validated tool can be used within practice and research to aid clinical assessment of dogs with the disease and could provide a supplementary tool to current monitoring methods.

ACKNOWLEDGMENTS
The authors thank Dechra Veterinary Products Ltd, Vets4Pets, Independent Vet Care, Bicester Vets, Dogs Trust, Veterinary Information Network, Bristol Dog Science Group, and VetCompass for helping with the promotion of the questionnaire used in this study.

CONFLICT OF INTEREST DECLARATION
Imogen Schofield is supported at the Royal Veterinary College by an award from Dechra Veterinary Products Ltd. Dechra Veterinary Products Ltd did not have any input in the design of the study, analysis and interpretation of data or in writing the manuscript. Dechra Veterinary Products Ltd did aid in the promotion of the questionnaire.

OFF-LABEL ANTIMICROBIAL DECLARATION
Authors declare no off-label use of antimicrobials.

Approval was granted by the Royal Veterinary College Ethics and
Welfare Committee (URN 2015 1373).