Translation, cultural adaptation and validation of the Diabetes Attitudes Scale - third version into Brazilian Portuguese 1

ABSTRACT Objective: to perform the translation, adaptation and validation of the Diabetes Attitudes Scale - third version instrument into Brazilian Portuguese. Methods: methodological study carried out in six stages: initial translation, synthesis of the initial translation, back-translation, evaluation of the translated version by the Committee of Judges (27 Linguists and 29 health professionals), pre-test and validation. The pre-test and validation (test-retest) steps included 22 and 120 health professionals, respectively. The Content Validity Index, the analyses of internal consistency and reproducibility were performed using the R statistical program. Results: in the content validation, the instrument presented good acceptance among the Judges with a mean Content Validity Index of 0.94. The scale presented acceptable internal consistency (Cronbach’s alpha = 0.60), while the correlation of the total score at the test and retest moments was considered high (Polychoric Correlation Coefficient = 0.86). The Intra-class Correlation Coefficient, for the total score, presented a value of 0.65. Conclusion: the Brazilian version of the instrument (Escala de Atitudes dos Profissionais em relação ao Diabetes Mellitus) was considered valid and reliable for application by health professionals in Brazil.

The DAS-3 consists of 33 questions divided in five related subscales: 1) need for special training to conduct educational interventions; 2) seriousness of Type 2 Diabetes; 3) value of strict glucose control for diabetes care; 4) psychosocial impact of diabetes on the lives of people and 5) autonomy of the person with diabetes (9) . It should be noted that the DAS-3 went through an evaluation process with 1,430 health professionals, proving to be valid and reliable, and has been translated and adapted to other countries, with the ability to maintain the original characteristics to measure the construct analyzed (9,(11)(12) .
In order to provide an instrument for use in the Brazilian context, this study aimed to carry out the translation, adaptation and validation of the Diabetes Attitudes Scale -third version .

Method
This methodological study followed the recommendations established in the literature (13) . In the analysis of the conceptual equivalence and items, concepts related to diabetes and to the attitudes construct were explored in order to verify whether the dimensions of the instrument are relevant to the Brazilian cultural context. Considering the viability and relevance of using DAS-3 in Brazil, the following steps were performed.
The translation was carried out independently by two translators, generating the T1 and T2 versions in Brazilian Portuguese. The translated versions were then compared by the same two translators and a third translator, which gave rise to a consensus version (T1-2). Next the instrument was backtranslated to its original language, independently, by two other translators, in order to verify the concordance between the original version and the consensus version (T1-2) (13) .
After these steps, 30 health professionals and 30 from the field of Applied Linguistics were invited to participate as the Committee of Judges (13) . This was a convenience sample. The invitation was sent by e-mail and a link provided for access to the instrument previously uploaded to the web e-Surv platform. The judges were divided into three groups so that each group evaluated 11 statements, since the review of all 33 questions would take longer than 45 minutes. All the participants evaluated the instructions of the instrument and response options so that there was no impairment in the understanding and evaluation of the translated version. The aim was to evaluate the semantic, idiomatic, conceptual and experiential equivalences.
When comparing the original and the translated version, the judges evaluated the instrument according Vieira GLC, Pagano AS, Reis IA, Rodrigues JSN, Torres HC. to the need for retranslation (1 = requires complete retranslation; 2 = requires partial retranslation with many changes; 3 = requires partial retranslation with a few changes; 4 = does not require retranslation) and the relevance of the reduction of the response options (from five options to four options).
After obtaining the responses of the judges, the Content Validity Index (CVI) was calculated, defined by the sum of the relative frequencies of the "3" and "4" responses (14) . The assumption that the higher the CVI, the lower the number of changes needed to improve the text was considered.
A total of 22 health professionals that provided care to people with diabetes mellitus participated in the pre-test stage. In this stage, the questionnaire was sent electronically, and the link to access the instrument was provided. The professionals were asked to respond to the 33 statements of the instrument, to evaluate each statement for ease of understanding and clarity of the information and to present suggestions for improvement of the text (13)(14) .
Finally, in order to verify its validity and reliability, the instrument was applied, through the web e-Surv platform, with health professionals on two occasions with an interval of 15 days between the test and retest (14) .
To calculate the sample size, a psychometric property was chosen that involves both the moment of the test and of the retest, the temporal reproducibility, and an alternative to its measure, the linear correlation. Thus, a significance level of 5%, test power of 80%, standard deviation equal in the test and retest scores and a correlation coefficient of 0.30 (minimum value to be detected in the evaluation of reliability) were considered. The minimum sample size required was 82 professionals. When considering a 20% losses, the final sample size required was 100 health professionals.
The selection of the professionals was performed by convenience from the database of the project entitled "Measurement instruments for educational practices in chronic disease: interdisciplinarity and innovation". Each professional that agreed to take part in the study was asked to indicate other professionals that worked with people who have diabetes. The application of the instrument was conducted in March and April 2016.
The descriptive analysis of the categorical variables was performed by calculating the absolute and relative frequencies and, for the quantitative variables, the means, standard deviation, and percentiles were calculated. The evaluation of the internal consistency was made from the calculation of Cronbach's alpha (15) .
In the analysis of the reliability of the instrument, the Polychoric Correlation Coefficient was used, as the response scale is of the categorical ordinal type (16) . As with the Pearson's linear correlation coefficient, the polychoric correlation The Intra-class Correlation Coefficient (ICC) was also used as a measure of concordance between the total score obtained in the two applications of the instrument, while the Wilcoxon test was used to verify whether there was a statistical difference between the median score of the first and second application of the instrument (11) . Data analysis was carried out using the R † statistical program. The significance level considered for the statistical tests was 5%. answer choices among people who would respond to the instrument; no significant difference within the Brazilian cultural context between the options, "disagree" and "totally disagree".
In order to preserve the comparison between the scores obtained with the original instrument and the instrument translated and adapted in Brazil, it was decided to maintain the score of response options with the range between 1 and 5 points. Thus, the following points were awarded to the statements with scores in direct order: disagree -1 point, no opinion -3 points, partially agree -4 points, agree -5 points. Regarding the statements that have reversed scores (2, 3, 7, 11, 13, 15, 16, 23, 26 and 28), the points were distributed as follows: agree -1 point, partially agree -2 points, no opinion -3 points and disagree -5 points. It is important to note that the "no opinion" option is scored the same in direct and reverse order.
The main changes made in the translated version after the suggestions given by the judges and in the pretest phase were: (1) replacing the term "patient", "user" and "diabetic" with "person with diabetes"; (2) inclusion of physiotherapy, pharmacy, physical education and psychology professionals; (3) changing the expression "self-care plan" to "care plan" and (4) replacing the word "disease" with "chronic condition". After these steps, the final version of the Escala de Atitudes dos Profissionais em relação ao Diabetes Mellitus (EAP-DM) was obtained, as presented in Figure 1.   Vieira GLC, Pagano AS, Reis IA, Rodrigues JSN, Torres HC.
(the Figure 1 continue in the next page...)

Original version Final version
Title Diabetes Attitudes Scale -third version

Instructions
Below are some statements about diabetes. Each numbered statement finishes the sentence "In general, I believe that…" You may believe that a statement is true for one person but not for another person or may be true one time but not be true another time.
Place a check mark in the box below the word or phrase that is closest to your opinion about each statement. Note: The term "health care professionals" in this survey refers to doctors, nurses, and dietitians.

Response options
Mark the answer that you believe is true most of the time or is true for most people. 10…it is important for the nurses and dietitians who teach people with diabetes to learn counseling skills.
11… people whose diabetes is treated by just a diet do not have to worry about getting many long-term complications.
12…almost everyone with diabetes should do whatever it takes to keep their blood sugar close to normal.   Table 3 shows the presence of moderate to high correlations between the items at the test and retest moments.

Region of the country
The reliability analysis of the instrument was supported by calculating the Intra-class Correlation Coefficient, which indicated moderate concordance in all subscales and in the general scale, as presented in Table 4. 23. …people who have type 2 diabetes will probably not get much payoff from tight control of their blood sugar.
24. …people with diabetes should learn a lot about the disease so that they can be in charge of their own diabetes care.
27. …what the patient does has more effect on the outcome of diabetes care than anything a health professional does.
28. …tight control of blood sugar makes sense only for people with type 1 diabetes.
29…it is frustrating for people with diabetes to take care of their disease. 29... ter que cuidar de si é frustrante para as pessoas que têm diabetes.
30…people with diabetes have a right to decide how hard they will work to control their blood sugar.
31…people who take diabetes pills should be as concerned about their blood sugar as people who take insulin. 31...as pessoas que tomam medicamentos orais para controlar o diabetes devem se preocupar com a glicemia tanto quanto as que aplicam insulina.
32…people with diabetes have the right not to take good care of their diabetes.
33…support from family and friends is important in dealing with diabetes. 33...é importante ter o apoio da família e dos amigos para lidar com o diabetes.

Discussion
Opting to culturally adapt an instrument is due to the various advantages already mentioned by the literature, such as savings time and the possibility of comparing the results with studies carried out in other countries (13) .
The studies that translated and adapted the DAS-3 used methodology similar to that presented in this study, differing only in the composition of the specialists that composed the Committee of Judges. Despite methodological differences related to the performance of the Committee of Judges, the DAS-3 has proved to be a valid, reliable, and easy to understand instrument, for use by professionals in different countries (9,(11)(12) .
The main changes in the items of the translated version were related to the change of terms used to describe people who have diabetes and the reduction of the response options. The term "diabetic" is no longer Rev. Latino-Am. Enfermagem 2017;25:e2875.
used, due to the current principles that consider the importance of the autonomy of people living with the condition of diabetes in the process of choices in their care plan. The term "diabetic", used as a noun, labels people who have diabetes from a negative perspective and also implies that all people living with this condition are equal, resulting in the establishment of standardized behaviors that do not consider the life story and the individual needs of these people (17) .
The reduction of the response options should also be highlighted, which was considered relevant by the majority of the specialists. The justifications of the judges for the reduction of response options were related to the discussions presented in the international literature, which demonstrate the existence of differences in response patterns for Likert type scales among people with different education and cultures (18) .
The results of the evaluation of the psychometric properties indicated adequate internal consistency.
Other studies found the presence of variation in the alpha values, which is justified by the instrument being applied in populations with different characteristics.
Nevertheless, the versions translated and validated in other countries have also obtained internal consistency considered adequate (9,(11)(12) .
The median score of the retest can be considered equal to the median score of the test for the majority of the subscales. It should be noted that the differences in medians found for the overall score and the "psychosocial impact of diabetes" subscale, although significant, can be considered small (0.04 and 0.14 points respectively).
The scores for each subscale were found to be similar to the results of a study conducted in Spain (11) .
A moderate to high discrimination capability was observed for the items, verified by the Polychoric Correlation Coefficients ranging from 0.443 to 0.813.
It was not possible to compare these coefficients with studies performed in other countries, since these studies did not use the Polychoric Correlation Coefficient.
In the analysis of the reliability through the stability, an ICC of 0.65 was obtained for the entire scale, demonstrating the temporal stability of the instrument (11) .
It is worth considering that evidence of validity should be accumulated to strengthen confidence in the use of scales. Therefore, it is suggested that this scale be applied with representative and more heterogeneous samples of health professionals, considering the different occupational categories and regions of the country.

Conclusion
It was concluded that the Brazilian version of Diabetes