Cross-Cultural Adaptation and Validation of the Exercise Adherence Rating Scale to Nepali language: A Methodological Study

Background The Exercise Adherence Rating Scale (EARS) is a commonly used outcome tool, which helps to identify the adherence rate of exercises and reasons for adherence and non-adherence. There is no evidence of availability of any measurement tools to assess exercise adherence in Nepalese context and cultural background. Therefore, we conducted a cross-cultural adaptation of the EARS in to Nepali language and investigated its reliability and validity. Methods Cross-cultural adaptation of the EARS was done based on Beaton guidelines. Psychometric properties were evaluated among 18 participants aged 18 years or older with pre-diabetes or conrmed diagnosis of any disease who were prescribed with home exercises by physiotherapists. Any disease that limited participants from doing exercise and individuals unwilling to participate were excluded. Reliability was evaluated through internal consistency, using the Cronbach’s alpha. Exploratory Factor Analysis (EFA) was performed to explore construct validity and conrm its unidimensionality. Receiver Operating Characteristic (ROC) curve was analyzed to identify cut-off score, sensitivity and specicity of the tool. Results The Cronbach’s alpha was 0.94 for EARS-adherence behavior. The EFA of 6-items adherence behavior revealed the presence of one factor with an eigenvalue exceeding one. The scree-plot suggested for extraction of only one factor with strong loading (75.84%). The Area Under the Curve was 0.91 with 95% condence interval 0.77 to 1.00 at p = 0.004. The cutoff score was found 17.5 with 89% sensitivity and 78% specicity.


Background
Exercise adherence is the extent to which person's behavior corresponds with agreed recommendations from health care providers [1]. The bene ts of exercise can only be obtained when a person is adhering to the prescribed exercises. Multiple factors are associated with exercise adherence such as sociocultural factors, knowledge towards exercise, self-e cacy, ethnicity, and economic status of an individual [2]. Within physiotherapy services, the concept of exercise adherence is associated with the performance of the prescribed exercise appropriately following the advice given by the physiotherapists [3]. There is no any gold standard outcome tool to measure exercise adherence rate in Nepalese cultural context and background. People commonly use self-reported diaries to re ect exercise adherence; however, they lack standardization, accuracy and possess self-presentation bias that limit their validity [4]. The Exercise Adherence Rating Scale (EARS) is one of the commonly used outcome tools, which helps to identify the adherence rate of exercises and reasons for adherence and non-adherence [5].
The original English version of the EARS is a 16-item, self-reported questionnaire, which assess adherence of prescribed exercises [6]. The EARS consist of 3 sections. Section 'A' is about prescribed exercise questionnaire. This section consists of 5 questions, which are related to way of doing activities and exercise that people often do to improve their physical quality of life. Section 'B' is about exercise adherence behavior, so called exercise adherence rating scale. This section consists of 6 items, which is an actual measure to identify exercise adherence. This evaluates whether individuals do their exercise as per recommendation or not. Section 'C' is about reasons for adherence/non-adherence of exercises. This section consists of 10 items, which assesses factors that hinders and facilitates the exercises [5]. The internal consistency (0.81), test-retest reliability (0.94), construct validity (70%) and face validity of the original version have been established [5,6]. Acknowledged with good validity and reliability, the EARS scale has been established as an appropriate and feasible tool to assess exercise adherence.
The "cross-cultural adaptation" is a process that looks at both language (translation) and cultural adaptation issues for a questionnaire to use in another setting [7]. The cross-cultural adaptation is important when an instrument has to be used in a different language, setting and time because of diversi ed context of geography, ethnicity, economic status, culture and diseases [2,8]. When there is no any tool available to assess exercise adherence in Nepal, a tool that is valid and reliable in measuring exercise adherence of Nepali-speaking individuals was required. Therefore, the aim of this study was to conduct the cross-cultural adaptation of the EARS to Nepali language and investigate its reliability and validity.

Methods And Materials
Cross-cultural adaptation Beaton Guidelines is one of the commonly used guidelines for the translation and cross-cultural adaptation of measurement tools [7]. As per the suggestion from the developer of the tool, Dr. Emma L Godfrey, we considered Beaton guidelines and an evidence of cross-cultural adaptation process followed in a study by Takasaki et al., in 2017 [7, 9] to cross-culturally adapt the EARS into Nepali language. The ve steps of cross-cultural adaptation were; forward translation, synthesis, back translation, expert committee review and pre-testing, which are described in Fig. 1.
Two independent non-medical translators, who were bilingual in English and Nepali, translated the original English EARS into Nepali language and developed two forward translated versions (FT1 and FT2). Reconciliation meeting was held among two translators and investigators of present study to reach to a consensus without compromising one's feeling and opinion. All minor issues encountered were addressed and resolved, as there were no any major issues. Through reconciliation, a common forward translation (FT12) was synthesized. Two Physiotherapists who were bilingual in English and Nepali then back translated the FT12 version into English. The purpose of the back translation was for validity checking to make sure that the adapted version was re ecting the same item content as the original version [10]. The back-translated versions were reviewed and a consensus version was developed [7,9].
Pretesting was done on 10 individuals with pre-diabetic conditions to explore clarity, understandability, comprehensibility and feasibility of the adapted version of the EARS using a visual analogue scale; ranging from 0 (not clear at all and di cult to understand) to 10 (clear and easy to understand). The average score of 8.10 indicated that the adapted version was clear, comprehensible and understandable. No ambiguity of meaning on any item was reported. Thus, the pre-testing version was considered as a nal version without any modi cation in the original English version. This was similar with the ndings from De Lara et al., in which no di culty was faced, or no any suggestions were given during crosscultural adaptation of the EARS into Brazilian version [11]. Thus, a Nepali version of EARS (N-EARS) was cross-culturally adapted (see Additional le 1).
Validation of N-EARS: Individuals at pre-diabetic stage as well as patients with various conditions were screened. Patients were eligible if they satis ed the following inclusion criteria: (i) individuals with prediabetes (HbA1c level between 5.7-6.4%) [12] or any patients with con rmed diagnosis of any disease who were prescribed with home exercises by physiotherapists, (ii) aged 18 years or older, and (iii) those who gave informed consent. Any disease that limited participants from doing exercise like recent surgery and individuals unwilling to participate were excluded from the study.

Statistical analysis
The descriptive statistics were used to analyze demographic and clinical data. The Cronbach's alpha (α) was calculated to determine internal consistency. Exploratory Factor Analysis (EFA) was performed in the study by Newman-Beinart et al., to determine factors in the original English tool [5]. With an aim to compare ndings with the study, we performed EFA to explore construct validity. Kaiser-Meyer-Olkin (KMO) test and Bartlett's test were used to check for sampling adequacy and sphericity, respectively. The minimum recommended value of 0.60 was considered for sampling adequacy [13]. The Varimax rotation was used during analysis. Eigenvalues were calculated to select number of components in EFA [5].
Receiver Operating Characteristic (ROC) curve was analyzed to identify cutoff score, sensitivity and speci city of N-EARS. Data were analyzed using SPSS (version 21.00). The signi cant level was considered at p < 0.05.

Results
Total of 18 individuals participated in the study. The mean age of the participants was 38 years with standard deviation (SD) of 11.88). Two-third of the participants (n = 12, 66.6%) were females. Similarly, 12 (66.6%) participants were pre-diabetic. Out of remaining 6 (33.3%) participants, one was with anterior cruciate ligament injury of knee (at 3 months of surgical repair), two were with low back ache (one at 18 days and another at 1.5 months), one was with stroke (at 1.5 months who was able to do activities independently), one was with Bell's palsy (at 15 days of disease onset) and one was with cardiac disease (at 3 months after open heart surgery). As shown in Table 1, the mean score for 6-item adherence behavior and 10-item reasons for adherence/non-adherence ranged from 2.17 to 2.83 and 0.83 to 3.39  Test of validity: Construct validity was explored using an EFA. The KMO value for 6-items adherence behavior was 0.73, exceeding the recommended minimum value of 0.60 which veri ed sampling adequacy for the analysis. Bartlett's test for sphericity indicated that correlations between items were su ciently large (Chi square: 110.19, p < 0.001) for factor analysis. Thus, the criteria for sampling adequacy and sphericity for 6-items adherence behavior scale was achieved. As depicted in Table 3, the EFA of 6-items adherence behavior revealed the presence of one factor with an Eigen value exceeding one. The scree-plot suggested for extraction of only one factor with strong loading (75.84%). Since KMO value of 10-items of reasons for adherence/non-adherence was < 0.60, it was not suitable for EFA.  Fig. 2, demonstrated that Area Under the Curve (AUC) was 0.91 with 95% con dence interval 0.77 to 1.00 at p = 0.004. The cutoff score was found 17.5 with 89% sensitivity and 78% speci city.
Correlation: As a means of validating 6-item adherence behavior, the correlation analysis was done between 6-item adherence behavior with 10-item reasons for adherence/non-adherence scale, which

Discussion
The highlights of the study are as follows: the EARS has been cross-culturally adapted to Nepali language. The adapted N-EARS has been validated. The N-EARS showed excellent internal consistency. The EFA indicated good construct validity. The 6-items adherence behavior scale revealed presence of only one factor with strong loading. The cutoff score was 17.5 with sensitivity of 89% and speci city of 78%. The 6-item adherence behavior and 10-item reasons for adherence/non-adherence scale were highly correlated.
Heterogeneous participants with respect to age, gender and diagnosis, were involved in the study. The study site had an easy access to the participants from urban, sub-urban as well as rural areas of Nepal. So, the participants comprised of diverse ethnicity, geographical regions and education level.

Cross-cultural adaptation
The EARS has been cross-culturally adapted to Nepali language based on Beaton guidelines [7]. The forward and back translation as well as adaptation procedure revealed no content or language related issues. Through pre-testing, good clarity and understandability of the N-EARS were demonstrated. In contrast to the ndings of a study by Meade et al., where re-framing for some items was required [6], there was no need of re ning or rede ning any item or words while adapting to Nepali language. The N-EARS has been formatted in such a way so that it is concise, short, easy to administer and looks attractive. In section 'A' of the tool, participants did not have any issues in understanding the questions. However, in agreement with the ndings from the study by Meade et al., they had di culties to complete the answers of the questions when exercises were not prescribed in appropriate dosage or prescribed dosage was not understood properly by the participants [6].

Reliability of N-EARS
The internal consistency was assessed to evaluate the degree of the interrelatedness among the items [14]. The internal consistency of N-EARS was excellent (α = 0.94) for 6-item adherence behavior [14,15]. The internal consistency of the original English versions was 0.8 and that of Brazilian version was 0.88 [5,11]. Present study demonstrated higher internal consistency (α = 0.94) of N-EARS than both English and Brazilian versions. An α value of 0.70 to 0.95 were considered acceptable values [16]. Therefore, the internal consistency of N-EARS was comparable with the values of English as well as Brazilian versions and it was in acceptable range.
Since the recommendation was against adding up of items to calculate a nal score in 10-items for reasons of adherence/non-adherence, we did not determine internal consistence of the 10-items [5]. This was not established even in Brazilian version by De Lira et al., in their study [11].

Validity of N-EARS
The EFA demonstrated adequate construct validity of 6-item adherence behavior scale of N-EARS. The 6item adherence scale revealed one factor solution with a strong loading (75.84%) to exercise adherence. The factor loading was higher than that of the original version which demonstrated 71% factor loading [5] and other self-reported outcome measures [17]. We could not perform EFA on 10-item reasons for adherence/non-adherence as it could not ful ll the criteria of sampling adequacy (KMO < 0.60) [13] which was in contrast with the Brazilian versions (KMO = 0.64) (11). the 10-item would be useful in exploring reasons why participants adhere or do not adhere to prescribed exercises via. single-item question as described by Newman-Beinart et al., in their study [5].
The ROC curve was used to analyze the predictive effect of the 6-item adherence scale [18]. The AUC of the total score of 6-item adherence behavior scale was 0.91 which was statistically signi cant and suggested a predictive validity which is in line with literature evidence [18,19]. The cutoff score of the tool was 17.5 with a sensitivity of 89% and speci city of 78% that discriminates adherent and non-adherent participants with respect to exercises. De Lira et al., in Brazilian version, demonstrated a cutoff score of with the ndings from the Brazilian study. We compared the ndings with the study of Wang et al., in which similar scale for exercise adherence was used and sensitivity (87.20%) as well as speci city (76.34%) they found were in line with the ndings of present study [18]. The cutoff score of 17.5 indicated that any individual obtaining score > 17.5 out of 24 on 6-item adherence scale is said to be adherent to the prescribed exercises. However, the cutoff score has to be cautiously used during interpretation because without knowing the level of exercise that is necessary for treatment to be effective, a cutoff score in assessing exercise adherence may not be useful [5,19]. The cutoff score, sensitivity and speci city re ected a preliminary predictive validity, which was not established even in the original version of the EARS and was a limitation [5]. On the other hand, fully relying on the established guidelines with the back translation re ecting the same item content as the original version supported good face validity of the N-EARS [7,10,20].
The correlation between total score of 6-item adherence behavior and 10-item reasons for adherence/non-adherence demonstrated a validity of the N-EARS. The strength of correlation has been used in describing validity in patient-reported outcome measures [6,21]. The 6-item adherence scale demonstrated strong correlation (0.55 to 0.84) with items 1, 2, 4, 6, 7, and 10 in present study. Therefore, these 6 items are important to consider while nding non-adherence to exercises. The reasons for nonadherence in the participants of a study by Newan-Beinart et al., were item numbers 1, 2, 3, 4, 7, and 9 [5]. Thus, the 10-items adherence/non-adherence gives clear information on reasons for non-adherence to exercise on one-to-one analysis, which may vary from one participant to another.

Strengths and Limitations
The strengths of this study include: (i) The method of cross-cultural adaptation by fully relying on the established guidelines giving a methodological strength; (ii) The reliability and validity were established on pre-diabetic who were healthy during recruitment and on patients with various other health conditions as well. We could evaluate the feasibility of the N-EARS on healthy individuals who were recommended for exercises to prevent disease or remain t and on patients who were prescribed exercises to treat their impairments or activity limitations. Therefore, the reliability and validity were demonstrated in heterogeneous group of participants; and (iii) The N-EARS yielded identical psychometric properties as original EARS. On the other hand, convenient sampling, small sample size and possibility of recall bias were main limitations of this study.

Conclusions
The EARS has been cross-culturally adapted to Nepali language. This study provided excellent internal consistency and adequate face, construct as well as predictive validity of the N-EARS. The N-EARS yielded identical psychometric properties as the original EARS. A cutoff score of 17.5 was found with good sensitivity and speci city. The ndings of present study provided evidence to use N-EARS in research and clinical practice that might facilitate the evaluation of exercise related interventions. Further studies are recommended to investigate other psychometric properties of the N-EARS. Receiver Operating Characteristic (ROC) curve of 6-items of adherence behavior scale