Adaptation and validation of the Diabetes Management Self-Efficacy Scale to Brazilian Portuguese 1

Objective: to perform the cultural adaptation and validation of the Diabetes Management Self-efficacy Scale for Patients with Type 2 Diabetes Mellitus with a Brazilian population sample. Method: cross-sectional methodological study in which the adaptation and validation process included the stages recommended in the literature. Construct validity and reliability were assessed with 200 adults with type 2 diabetes mellitus. Results: the items indicated by the panel of judges and by the target population were adjusted in the cultural adaptation to improve clarity and understanding. The instrument's four factors remained in the confirmatory factor analysis with factor loadings of items greater than 0.30, except for factor 4; convergent validity, verified by the multitrait-multimethod analysis, presented inter-item correlations from 0.37 to 0.92, while for discriminant validity, 100% of the items presented greater correlation in their own factors. Cronbach's coefficient alpha for the total scale was 0.78, ranging from 0.57 to 0.86 among factors. Conclusion: semantic, cultural, conceptual and idiomatic equivalences were achieved and the instrument's Brazilian version also presented psychometric properties that showed evidence of reliability and validity. Thus, it can be applied both in clinical practice and research. Self-efficacy is useful for planning and assessing educational interventions, as well as predicting behavior modification in self-care.


Introduction
Self-efficacy (SE) is a key concept from Bandura's Social Cognitive Theory and is defined as one's perception of one's own ability to organize and perform actions (1) .
Beliefs regarding one's own ability vary in a given context or situation; i.e., beliefs are not uniform and are a result of personal experiences, others' verbal persuasion, social modeling through observation of other people's performances, and physical and emotional states. Therefore, SE is not a characteristic of one's personality nor does it reflect personal competencies themselves.
Rather, it reflects one's belief or judgment regarding one's competencies (2) .
The SE dimensions that influence human behavior include magnitude, strength and generality. Magnitude refers to one's perceived degree of difficulty to perform a given task; strength refers to one's personal conviction one is able to perform a specific task; and generality is seen as one's ability to extend SE from one situation to another (2) .
In this study, SE is addressed in the context of care provided to individuals with diabetes mellitus (DM), for which a challenging regime is often imposed in which self-care behaviors range from the adoption of a healthy lifestyle, up to the handling of devices and inputs, to selfmonitoring blood glucose levels and the administration of insulin (3) .
Self-care behavior requires personal judgments and decisions that tend to be difficult for individuals. These behaviors require technical and cognitive skills, which, in turn, are frequently associated with one's SE beliefs (4) . SE, therefore, influences an individual's decisions regarding adopting or not adopting a given behavior, whether this individual will persevere in this behavior or not, and how well s/he will respond to obstacles and relapses of undesirable prior behaviors (5) .
When DM education is considered as a strategy to develop self-care behavior, as well as to motivate its implementation and maintenance, SE is highlighted in the planning and assessment of educational interventions (6) and in predicting behavioral changes over time (7)(8) . Therefore, culturally adapted and validated instruments to assess SE need to be available both to be used in clinical practice and in research.
One of the instruments used to assess SE in the performance of behaviors aimed to control type 2 DM (DM2) is the Diabetes Management Self-Efficacy Scale for Patients with Type 2 Diabetes Mellitus (DMSES). Originally developed in the Netherlands (6) , it was later adapted and validated for the Australian (9) , Turkish (10) , Chinese (11) , and Arabic (12) contexts.
In Brazil, there are no valid instruments specifically to assess SE among individuals with DM2. Therefore, this study's objectives included the cultural adaptation and adherence; and exercise); self-observation activities (controlling/observing and recording blood sugar or sugar in the urine, body weight, the condition of feet skin, and general health condition); and self-regulating activities (correction of hypo-and/or hyperglycemia, preparation for vacation, diet changes, or self-regulation when there is weight gain, acute disease or stress) (6) .
The scale's items express not only the behaviors individuals with DM are supposed to perform in each of the activities but also ask whether individuals feel capable of performing such behaviors. After a literature review and expert assessment, the study in which the original instrument was developed established that all items would begin with the expression "I think I'm able to" because this statement reflects "strength", which is considered the main component of SE (6) .
All the items present a response pattern from "definitely can do" up to "cannot do at all", with scores ranging from one to five, respectively. The instrument's global average score determines SE; that is, the score of each item is totaled and then divided by the number of items (20). High scores indicate high SE (6) .
The psychometric properties of the original instrument, which is available in English and Dutch, were assessed with a sample of 94 Dutch adults with DM2 and presented good internal consistency, the Cronbach's alpha of which was 0.81 (6) . The Australian version also showed internal consistency (α=0.91), though the authors recommend further psychometric tests to be applied with larger samples to check for potential redundancy of items (9) .
The Cronbach's alpha of the Turkish version (10) was 0.88 and the exploratory factor analysis revealed three factors instead of the four presented by the original scale (6) .
www.eerp.usp.br/rlae 3 Pace AE, Gomes LC, Bertolin DC, Loureiro HMAML, Van der Bijl JJ, Shortridge-Baggett L. minimum interrater agreement of 80% (15) , which resulted in the consensual Portuguese version 2 (CPV2). This assessment on the part of a panel of judges before the back translation identified potential errors or difficulties in understanding items, so that the instrument can be refined until the version that will be used with the target population is achieved (13)(14) .
The CPV2 was back translated by two bilingual independent translators, not the initial translators, with fluency in both English and Portuguese, in addition to having knowledge concerning Brazilian culture, who were not aware of the study's concepts (16) . This stage was intended to assess whether the Portuguese version reflected the content of the original version in English.
Two back translations were obtained (BT1 and BT2), which were analyzed together with the translators and researchers to reach a consensual version in Afterwards, a semantic analysis using the CPV2 was conducted with 18 people of the target population, who were not included subsequently in the study sample.
The participants' suggestions were incorporated into the instrument and the new version was once more submitted to the authors of the original version. The authors agreed with the changes and the final Brazilian version was obtained (FBV).
A pretest was applied with the FBV to assess the pertinence of items and the answer options. A form containing all the FBV's items, which were divided into subsets of items totaling five subsets with four items each, was applied. The participation of three individuals with DM2 was required for each subset, totaling 15 respondents. This form was based on studies conducted by the DISABKIDS group and addresses the relevance of each item and the difficulty answering each question, as well as whether the answer options were clear. If necessary, the respondent was allowed to rewrite the item in his/her own words (16) .

Stage 2 -Initial assessment of the adapted instrument's psychometric properties
The study's sample was composed of 200 adults with DM2, of both genders, receiving medical treatment with insulin and/or anti-diabetics, presenting no chronic complications in an advanced stage, recruited through their medical files on the day they attended return visits with healthcare professionals. The participants were verbally invited to participate in the study, received clarification regarding the study's objective, and read a free and informed consent form; those who consented signed the form. The sample size met the criterion of at least five and a maximum of ten respondents for each item in the instrument (17) .
The authors explain that cultural characteristics or the connotation of words in the target language may account for such a result.
The Cronbach's alpha found in the study in which the Chinese version was developed and adapted (C-DMSES) (11) was 0.93. The study also showed, through criterion validity, that the C-DMSES could predict self-care activities in a similar fashion to the Arabic version, the Cronbach's alpha of which was 0.91. The Arabic study showed that four out of its five domains predicted self-care behaviors (12) . The first step was to choose the translators, which according to recommendations of experts in methodological studies (13)(14) , must be professionals who have mastered both English and culture and who present distinct profiles, that is, one translator who has knowledge of the concepts investigated by the instrument in order to obtain an equivalent translation from a clinical perspective and another translator who must have no knowledge in regard to these concepts so that the translation reflects the usual language of the target population. The Lagrange multiplier test, which establishes the need to relocate an item to another factor, was also performed to improve the correlation among the items in the same factor. Similar to the Wald test, it shows the extent to which Chi-square statistics will change if an item is relocated to another factor.
Construct validity was then performed by means of convergent and discriminant validity obtained through multitrait-multimethod matrix (MTMM), which describes the magnitude of correlations between items and factors.

Convergent validity verified in initial validation studies is
satisfied when the linear correlation between an item and the factor to which it belongs equals 0.30. For discriminant validity, the linear correlation between an item and its factor is expected to be greater than its correlation to other factors in most correlations considered (17) .
Reliability was estimated by Cronbach's alpha and by the correlation of each factor with the total scale. The level of significance adopted was 5%.

Pretest of the final Brazilian version (FBV)
Assessments concerning the relevance of each item, difficulties answering them, and options for answers were considered satisfactory by those who took part in the pretest, who also agreed with the construction of the items. The individuals in the target population did not manifest the need to make any changes in the redaction or how the items are presented, thus, the instrument was considered easily understood and accepted. Therefore, this stage of adaptation was considered to be complete and we then proceeded to validate the translated instrument.      Table 2).
By estimating factor loading, we verified that most items obtained significant loading in their factors, except factor 4 ( Table 3). The exclusion of items 1, 2 and 3 and the relocation of item 15 to factor 4 was suggested.    (Table 4).

Reliability analysis
The internal consistency of the four domains, calculated by Cronbach's alpha, was 0.78 for the total scale. Note that higher values were found for factors 1 and 3, and moderate values were found for the remaining factors (Table 5).  even though SE is related to one's perception of one's ability at the present moment, people can easily imagine themselves to be able or unable to perform a task when considering a hypothetical situation (1) .
Note that the self-monitoring of blood glucose demands visual acuity, and manual and cognitive skills, in addition to financial resources, which represent a barrier to many people, especially older individuals and those with low levels of education or income (18) . Therefore, lack of understanding of this item may be related to procedural issues of self-monitoring. This result is similar to that found in the case of the Arabic version, in which the respondents had difficulties in regard to items related to the self-monitoring of blood glucose because they do not have the inputs necessary in their homes, and, for this reason, such a practice was not part of their routines (12) .
The suggestion to add the word "celebration" to item 16 was considered pertinent because, in the Brazilian culture, not only parties, but also celebrations of all sorts involve the consumption of food and drinks. Eating is a social activity, the symbolic functions of which permeate interpersonal relationships and reveal the structure of daily life, satisfying not only a biological need but a social function, as well (19) .
In regard to item 18, the term "monitor" was considered more pertinent than "control", especially because, in the context of an individual with a chronic disease, the control of DM depends more on the individual him/herself than on healthcare personnel (17) . The decisions made by an individual with DM to control the disease have a greater impact on their wellbeing than those made by professionals (20) . Healthcare practitioners should remain spectators, monitoring and providing guidance to empower chronic patients (21) .
Some sociodemographic characteristics stand out in the validation process: a higher number of women, adults, low educational levels, and retired individuals, characteristics that are similar to those reported by studies concerning the original version (6) , as well as the Australian (9) , Turkish (10) , and Chinese (11) versions.
Descriptive analysis of the scale revealed that the items with the lowest and highest averages were respectively items 16 and 18. One descriptive study with a qualitative approach conducted with 24 Brazilian adults with DM, which aimed to identify difficulties concerning the treatment of the disease, reports that one of the main difficulties for these individuals involves following their diets, especially during social events, revealing that breaking a diet and the desire for foods are common circumstances in the lives of people with DM (22) .
On the other hand, the high SE verified in this study, regarding being able to attend medical return visits to monitor the disease, is similar to the results reported by descriptive Brazilian studies addressing individuals with DM and/or hypertension (21,23) , which show the extent to which medical care is valued in the Brazilian culture (11) . One cross-sectional study designed to assess the relationships among SE, self-care and the glycemic control of Jordanian adults with DM2, using the Arabic version of DMSES, also reports that the highest average was found for the item concerning medical follow-up (11) .
The results concerning the descriptive analysis of items reinforce the belief that SE can vary among selfcare behaviors (12) , that is, an individual may consider him/herself able to attend medical visits to monitor the disease but not able to follow a diet regime. Each behavior, in turn, may require different skills and types of knowledge, in addition to different levels of motivation and confidence (5,12) .
The distribution of items in the factors according to the CFA of the Brazilian version remained similar to the original version (6) . In the first phase of the CFA, however, three items (1, 2 and 3) that compose factor 4 (blood sugar) were indicated for exclusion, in addition to item 15.
The DMSES proposes the assessment of SE in the performance of self-care behaviors concerning three types of activities, respectively named "activities that are essential for the treatment of diabetes", "self-observation activities" and "self-regulating activities" (6) . Items 1, 2 and 3 of factor 4 refer to one's belief in one's ability to selfmonitor and correct blood sugar and correspond to "selfobservation activities" and "self-regulation" (6) . Therefore, the exclusion of these items in this stage of validation could nullify central aspects of the instrument and hinder global understanding of the SE construct, in addition to the de-characterization of specific characteristics of the studied sample that would possibly interfere in these self-care activities.
Correcting blood sugar by changing dosages of medications is not a common practice among the people with DM2 addressed in this study, due to a lack of www.eerp.usp.br/rlae 9 Pace AE, Gomes LC, Bertolin DC, Loureiro HMAML, Van der Bijl JJ, Shortridge-Baggett L.
knowledge and skills (data not shown). One descriptive study conducted in the same setting as this study with a sample of adults with DM2 showed that, among self-care activities, blood sugar monitoring was performed with a frequency below what is recommended (24) .  3-physical exercise (10) . Variations in the distribution of items may be related to cultural factors or to the sample size or clinical characteristics (10)(11) .
In regard to the construct validity, the results of the and Blood sugar.
The convergent validity of the Australian (9) and Chinese (11) versions of DMSES were verified through linear correlations with the instrument General Self-Efficacy Scale (GSE), developed in Germany and adapted to 28 languages. Correlations of 0.52 (p<0.001) and 0.55 (p<0.01) were found for the Australian and Chinese versions, respectively, and confirmed the validity of these two versions to assess SE. Note that this modality of analysis was not performed in this study because we did not identify any similar instruments validated for our target population (individuals with DM2).
The analysis of reliability for the Brazilian version resulted in a Cronbach's coefficient alpha for the total scale of 0.78. The lowest alpha, 0.57, was found for factor 4 and the highest was 0.86 for factor 3. High internal consistency was verified for factors 1 and 3 and moderate consistency was found for the remaining factors.
A review study addressing the psychometric properties of instruments aiming to assess subjective phenomena reports that values above 0.50 are consider reasonable (25) .
The original version presented a Cronbach's alpha of 0.81 (25) for the total scale and the lowest and highest values were 0.71 and 0.79, respectively. In contrast with this study, the original version's factors 1 and 2 (6) presented the highest internal consistency.

Conclusion
The process of adapting and validating the Escala de autoeficácia no controle do diabetes para pacientes com diabetes mellitus tipo 2 [Self-Efficacy Scale for the of care and assessment of educational interventions.
Therefore, we stress the importance of studies providing culturally adapted and valid instruments to assess selfefficacy and contribute to the advancement of knowledge concerning the care provided to individuals with this form of diabetes.