Cultural adaptation and validation of the Kidney Disease Quality of Life-Short Form (KDQOL-SF™) version 1.3 questionnaire in Uganda

Background: Chronic kidney disease is on the rise in sub-Saharan African countries such as Uganda, and patients often present with advanced disease requiring kidney replacement therapies. Health-related quality of life is a key outcome in chronic kidney disease management but, in Uganda, no tools have been validated to measure this. The aim of this study was to culturally adapt and validate the Kidney Disease Quality of Life-Short Form version 1.3 (KDQOL-SF™) questionnaire for use in the Ugandan setting. Methods: We conducted a four-phase, mixed-methods study which included translation, cultural adaptation, optimisation of face validity and field testing. Our participants included healthcare workers, and patients aged >18 years with an estimated glomerular filtration rate <15 mL/min/1.73 m 2 . Results: The tool was culturally adapted and translated into one of the Ugandan languages, Luganda, which, with an English version of the tool, was validated and field tested. Over 80% of the subdomains had less than 10% floor and ceiling effects. For reliability, Cronbach’s α coefficient scores ranged from 0.96 to 0.41, with 10 out of 18 subdomains scoring >0.70, indicating acceptable internal consistency. The tool demonstrated discriminant validity, with patients with comorbidities reporting lower quality of life scores, as postulated. conclusions: The Luganda and English versions of the KDQOL-SF questionnaire have sufficient face and content validity, reliability and acceptability to assess the quality of life of patients with kidney failure in Uganda.


INtrODUCtION
Non-communicable diseases are on the rise globally and in 2016 contributed to 40 million (71%) of the 56 million deaths reported [1]. Uganda, like most countries in sub-Saharan Africa, faces a dual burden of communicable and non-communicable disease, with 33% of the annual mortality in the country in 2016 attributable to non-communicable diseases, including kidney disease [2]. Chronic kidney disease (CKD) plays a major role as a cause and a consequence of other communicable and non-communicable diseases [3,4]. The true burden of CKD is not well documented in low and middle-income countries (LMICs), although estimates suggest these countries contribute close to 70% of the global burden of 700 million people living with CKD [5,6].
In LMICs, patients often present with advanced CKD requiring kidney replacement therapies, including haemodialysis and kidney transplantation [7,8]. Unfortunately, the costs of these therapies are too high for the majority of patients, necessitating the implementation of palliative care [4,9]. This includes symptom alleviation and providing spiritual, social and psychological support. The focus is on improving the quality of life (QOL) of patients and their families, rather than prolonging life [10,11].
The World Health Organisation defines QOL as the "individual's perception of their position in life in the context of culture and value systems in which they live and in relation to their goals, expectations, standards, and concerns". QOL is an important indicator of successful chronic disease care, and many tools have been developed to assess it. The Kidney Disease Quality of Life-Short Form version 1.3 (KDQOL-SF™) instrument is one of the most frequently used tools to assess QOL in patients with kidney disease (see Table 1). It scores patients in three domains: physical health, mental health, and kidney disease-specific domains. Patients with lower scores are identified as having a poorer quality of life than patients with higher scores.
This tool, and all others, are not culturally adapted or validated for use in the Ugandan setting. Culturally adapting interventions is one of the keys to successful implementation [12] and has been defined as "the systematic modification of an evidence-based treatment (or intervention protocol) to consider language, cultural, and context in such a way that it is compatible with the user's cultural patterns, meaning, and values" [13]. User's attitudes towards a tool, and therefore the likely utility of the tool, can be improved by incorporating cultural elements, including spiritual beliefs, social norms and routine activities of life in society [14].
Because the validity and cultural appropriateness of the KDQOL-SF are uncertain for use in Uganda, the purpose of the study reported here was to culturally adapt and test the instrument for use in Ugandan patients with end-stage kidney disease.

MEtHODS
We conducted a mixed-methods study to culturally adapt, translate into Luganda and evaluate the psychometric properties of the Ugandan version of the KDQOL-SF. The study was carried out in four phases, detailed in Table 2. It involved translation, evaluation and improvement of face and content validity, cultural adaptation and field testing with psychometric evaluation.
The KDQOL-SF version 1.3 is a publicly available tool which combines the generic SF 36 instrument with kidney disease-specific items. The questionnaire consists of 80 items divided into 19 subdomains (Table 1). Items are scored on a 0-100 point scale, with higher scores indicating better health-related quality of life (HRQOL), and summarised into a physical composite summary (PCS) score, mental composite summary (MCS) score and a kidney disease composite summary score (KDCS) [15,16]. A final item, the overall health rating item, asks respondents to rate their health on a 0-10 scale.

Phase 1: translation
The translation was conducted at the Makerere-Mulago Palliative Care Unit and Makerere University in accordance with the guidelines provided by Rand Health. The tool was translated into Luganda, the most-spoken language in central Uganda. This was done by three bilingual native Luganda speakers, working together. Back translation into English was then performed by a professional Luganda language specialist. Two of the original native speakers reviewed and adjusted the tool based on the translated and back-translated versions. Revisions were made until the Luganda language specialist and the original native speakers Adaptation and validation of the KDQOL-SF™ version 1.3 questionnaire in Uganda were all in agreement that the wordings were appropriate to the regional vernacular and linguistically and conceptually equivalent to the English version. See Appendix 1 for the Luganda version of the tool.

Phase II: Focus group discussion with healthcare providers
The focus group discussion was held in the haemodialysis unit at Kiruddu National Referral Hospital and involved nine healthcare workers and two interviewers. It aimed to determine face and content validity and the practicability of employing the tool, the utility of the results obtained and the integrability of the tool into daily work routines. Participants were given copies of the questionnaire and, during the discussion, both the English and Luganda versions were read aloud. Participants were asked about clarity of the questions, relevance of the content to daily patient care, comprehensibility, feasibility for integration within routine care and the practicability of using the questionnaire. The focus group discussion was audio recorded and later transcribed verbatim by a bilingual native Luganda speaker, who then translated the Luganda sections into English. The first author (PB) took field notes while a bilingual research assistant with extensive experience in qualitative interviews moderated the discussion. Participants' responses and reactions were noted. The lead author read and re-read the transcript to familiarise herself with the data to determine thematic areas of interest, face and content validity, comprehensibility, feasibility of integration into routine care and practicality of use. These were then reviewed by the study team and some changes were made to improve the content relevance. Recommendations for integration into routine care were documented as well.

Phase III: Cognitive interviews with individual patients
Interviews took place in the outpatient clinic, the inpatient ward and the haemodialysis unit, and aimed to determine the acceptability, comprehension, comprehensiveness and interpretability of the tool, as described earlier. The first author (PB) and a palliative care nurse with extensive experience of qualitative research methods interviewed five participants aged >18 years who had stage 5 CKD (estimated glomerular filtration rate <15 mL/min/1.73 m 2 ). Questionnaire items were read out to the participants, and answers and reactions recorded. Participants were asked what the items meant to them and whether they were clear or unclear; they were also asked about relevance of the items. Recordings of the interviews were transcribed and translated as in phase II. We used thematic analysis to summarise the concerns regarding acceptability, comprehension, comprehensiveness and meaning.

Phase IV: Field testing the modified questionnaire
The participants for this phase were again adult patients with stage 5 CKD. Consecutive non-probability sampling was used to recruit participants, using a 1:2 ratio of patients on haemodialysis (HD) treatment relative to patients not receiving HD. This was because of the low HD patient numbers. We followed the sample size recommendations of Mundfrom et al. [17], who have suggested a minimum number of participants of 3-20 times the number of variables, with absolute numbers ranging from 100 to over 1,000 participants. Participants were recruited by two trained research assistants, who understood the need to develop rapport and establish trust with participants, to avoid influencing their responses and to accurately interpret the meanings expressed by participants. The questionnaire Adaptation and validation of the KDQOL-SF™ version 1.3 questionnaire in Uganda Correlations of item scores with the general health score, analysis of floor and ceiling effects, and calculation of Cronbach's α coefficient was administered by the two research assistants as our experience in a pilot study had revealed that self-administration of the questionnaire by patients resulted in poor data quality.
The analyses used baseline data which were previously collected [18]. Demographic data were summarised using medians and interquartile ranges or counts and proportions, where appropriate.
Floor and ceiling effects: ceiling effects were taken as being the percentage of respondents with scores of 100 and floor effects were the percentage having a score of 0. Ceiling and floor effects should be less than 20% to ensure that the scale captures the full range of potential responses within the study population, and that change over time can be detected [19].
Reliability: we assessed the internal consistency of reliability using Cronbach's α coefficient. A value of 0.70 and above was adopted as the criterion of adequate internal consistency reliability [20].
Validity: Besides face and content validity, we also assessed discriminant validity. We postulated that people who had greater morbidity and older subjects would have lower QOL scores. We thus examined QOL scores for patients with hypertension, diabetes mellitus and also based on treatment modality. In addition, we explored variation in QOL scores based on education level and work status. The QOL scores were compared using correlation analysis.
Pearson's correlation was applied to the normally distributed variables and Spearman's rho when criteria of normality were violated. The tests were two-tailed and a P value of <0.05 was considered to indicate statistical significance. Data were analysed using Stata version 12.0 (StataCorp LLC, Texas, USA).
Ethical approval for this study was provided by the Makerere University School of Medicine Research and Ethics Committee (#Rec Ref 2018-005) and the Mulago Hospital Research and Ethics Review Committee (MHREC 1543). The Uganda National Council for Science and Technology also reviewed and approved the study (HS 2573). All participants provided informed consent and were fully aware of their right to withdraw from the study at any time.

rESULtS
The numbers and characteristics of the participants involved in the various study phases are summarised in Table 3.

Phases I, II and III
During the process of forward and back translation (phase I) as well as the item-by-item review (phases II and III), we identified items for modification whereas those found to be comprehensible and consistently interpreted by participants were left unchanged. English words, phrases and activities which are not common in Uganda, such as somewhat, groceries, flight of stairs and vacuum cleaning, were replaced with more frequently used words and activities such as slight, items bought from the market, climbing ten steps and mopping (Table 4). Modifications were made based on suggestions by the participants and the research team members.
The focus group discussion lasted 168 minutes, whereas the cognitive interviews variously lasted from 42 to 87 minutes. Regarding feasibility and practicability of employing the tool, participants found it to be feasible but identified its length as a limitation. Both health workers and patients identified the questions and results as useful and felt that it was justified to integrate the tool into daily work routines. In the individual cognitive interviews, participants noted that the questions were relevant and helped them talk about their illness in more detail than they had ever done before.
"… it also gives us an opportunity to talk about things that we have been wanting to talk about, like sexual issues." ICI P2 They also commended the flow of questions, especially for topics they considered sensitive. "… for us in dialysis, you have the four hours so it's not too much, unless someone wants to be sleeping, but there is time."

ICI P3
Following these suggestions, culturally relevant adaptations were made to the tool, as shown in Table 4.

Phase IV (table 6)
There were 364 participants recruited for field testing the questionnaire, an adequate sample size according to the recommendations of Mundfrom et al. [17]. There were approximately 4.5 respondents per item, and more than 19 per domain. Most respondents were male (60%), married (62%), aged less than 50 years (62%) and had some form of employment (92%). The highest response rate was noted for the emotional well-being subdomain (363/364) and the lowest rates were for the staff encouragement (135/364) and patient satisfaction (148/364) subdomains. Most subdomains (16/19) had a response rate of over 90%.
The highest mean QOL score was noted for staff encouragement (97) and the lowest for burden of kidney disease (24).
Regarding the distribution of responses to each subdomain, physical functioning (33%), work status (24%) and burden of kidney disease (15%) had the highest proportion of floor effects with the other subdomains having less than 10% floor effects. Social support (44%), patient satisfaction (28%) and pain (26%) had the highest proportion of ceiling effects with the other subdomains having less than 10% ceiling effects. See Table 5. The presence of comorbidities, especially stroke and diabetes, was associated with lower QOL scores (Table 6).
Having diabetes was associated with the kidney disease composite score (P = 0.008) and the overall score (P = 0.007), whereas a history of stroke was associated with the physical composite score (P = 0.029) and the overall QOL score (P = 0.039). Of note, we found no difference in QOL scores based on sex, age, treatment modality and socioeconomic factors such as educational level and employment.

DISCUSSION
The KDQOL-SFTM version 1.3 questionnaire has successfully been adapted culturally and is fit for purpose as well as valid and reliable in Ugandan patients with end-stage kidney disease. Validation of a tool is important if it is being introduced into a different setting [21] and, to the best of our knowledge, this is the first validation study of this tool in East Africa.
We encountered differences in the terms used in daily language. Terms and activities such as playing golf, bowling and walking a block are not common in our setting and were substituted by phrases such as sweeping the compound or walking a mile. Similar substitutions have been made in other cultural settings such as in Brazil [22] and Turkey [23] and this increases comprehension and the usability of the tool.
The KDQOL-SF questionnaire was designed to be selfadministered but, in our setting of a low literacy rate, we found that response rates were better when the questionnaire was administered by a research assistant. This may affect answers provided because some questions are sensitive to the data collection method [24,25]. This could have been the reason for low response rates seen in the staff encouragement and patient satisfaction subdomains. The training of research assistants in this setting is very important to reduce biases such as social desirability bias, interviewer bias and acquiescence bias [25].
The floor and ceiling effects were used to assess the distribution of responses to each item during field testing. This suggested a fair distribution of responses, with most items having less than 10% floor and ceiling effects. Physical functioning (32%) had the highest proportion of floor, followed by work status (24%) and burden of kidney disease (15%); these proportions were lower than has been seen in other settings in Mexico [26], South India [27] and Egypt [28]. On the other hand, social support (43%), patient satisfaction (27%) and pain (26%) had the highest proportion of ceiling effects in our study, similar to several reports from other settings [26][27][28]. Careful attention to clarification of the meaning of items is important, especially for items which might overlap (and therefore might not be specifically distinguished) such as burden of kidney disease and symptom burden.
Internal consistency reliability scores were below 0.7, the standard required, for eight out of eighteen items, with emotional well-being (0.41) scoring lowest, followed by staff encouragement (0.50) and energy/fatigue (0.51). In an Egyptian study [28], the lowest scores were for quality of social interaction (0.23) and work status (0.28) and in a Danish study, the lowest scores were for quality of social interaction (0.43) and social support (0.67) [29]. These low scores may be due to the varying interpretations of items which might appear similar and the scores of these items therefore need to be interpreted with caution.
Regarding discriminant validity, the scores on our tool could distinguish between subgroups of patients based on their comorbidities but not on age, sex or socioeconomic factors, as has been found in other studies [22,26,28]. Differences in scores among study populations may be due to variations in the composition of their demographics; for example, our patients were relatively young compared to the more elderly participants in other settings. Also, in a public hospital setting such as ours, the socioeconomic status of patients is uniformly low, leading to the possibility that the questionnaire is unable to discriminate different subgroups due to the absence of variation.
Future studies involving the questionnaire could explore why employment and social support did not lead to differences in quality of life, as has been seen in other settings [16]. Moreover, a confirmatory factor analysis could be completed to evaluate whether a shorter version of the scale can be used with the same measurement validity in our setting to reduce the length burden and therefore complete the questionnaire more quickly.

CONCLUSIONS
The Ugandan version of the KDQOL-SF version 1.3 is reliable and valid. It may be used in research or in routine clinical care to measure HRQOL and changes in response to interventions.
Users of this tool in resource-limited settings should be aware of poor response rates when self-administered, as well as for items such as staff encouragement and patient satisfaction when the tool is administered by healthcare workers, and the need for care in the administration and interpretation of items which may seem similar.

Funding
PB is a PhD scholar supported by a DELTAS Africa Initiative grant #DEL-15-011 to THRiVE-2. The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences' Alliance for Accelerating Excellence in Science in Africa (AESA) and is supported by the New Partnership for Africa's Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust (grant #107742/Z/15/Z) and the UK government. The views expressed in this publication are those of the authors and not necessarily those of the funders. The funding body approved the study but did not contribute to the design, ethical review or execution of the study. Okunoonyereza kuno kugenderedwa kumanya endowooza z'abantu ku bulamu bwabwe nga balina obulwadde bwensigo. Ebinaava mu kunoonyereza kuno bijja kutuyamba okugoberera nga bw'owulira ate nga bw'osobola okukola emirimu gyo egya bulijjo mu bulwadde buno.