Psychometric performance of the Brazilian version the “Insulin Management Diabetes Self-Efficacy Scale” for patient with Type 2 Diabetes Mellitus

Model of study: Validation study. Objective: To evaluate the psychometric performance of the Brazilian version of the Insulin Management Diabetes Self-Efficacy Scale IMDSES. Methodology: Reliability (internal consistency Cronbach’s α), convergent and criterion validity were assessed. Data were gathered from 127 patients with type 2 diabetes mellitus (T2DM) in use of insulin through the application of sociodemographic and clinical instruments and the Brazilian versions of the Summary of Diabetes Self-Care Activities Measures SDSCA and the IMDSES. Results: It was observed “ceiling effect” in Insulin domain. Reliability was confirmed in all IMDSES domains and Total Score (α=0.87). The convergent construct validity was supported by significant negative moderate to weak magnitude correlations between the scores of IMDSES domains and the scores of SDSCA domains. Criterion validity was partially supported by the weak magnitude correlation between the General Managements domain and the Total Score of IMDSES and serum levels of glycated hemoglobin (r=0.22; p=0.02). Conclusion: The Brazilian version of IMDSES presented satisfactory psychometric measures and may be useful to evaluate the effectiveness of nursing interventions aimed at optimizing the self-efficacy of the T2DM patient in the management of his treatment.


INTRODUCTION
As a chronic condition, Type 2 Diabetes mellitus (T2DM) requires the individual to adopt complex practices to control its progression and prevent its complications. 1 One of the strategies used to adopt healthy practices is to improve pa-tient confidence in their capacity/ability to perform the behavior, that is, the improvement of their selfefficacy (SE). 2 SE, a central construct of Social Cognitive Theory, can be defined as the individual´s perception of their ability to organize and perform certain actions to achieve a target outcome. 2 A high SE is needed in order to obtain a personal wellness effect aimed at adopting positive behaviors, considering that the higher the individual conviction of the ability to produce desired results, the more intense and lasting will be the attempt to deal with the challenges inherent behavioral changes. 3 The relationship between SE and adoption of health behaviors has been investigated in numerous studies which suggest that the strengthening of SE relates to the adoption of health practices in different clinical conditions 4,5 and among those with Diabetes mellitus (DM). [6][7][8][9][10] Regarding T2DM, SE has been considered an important outcome measure of self-manage-ment interventions. 11 In this sense, studies of pa- 1 2 2 tients with T2DM demonstrated that greater SE predicted better nutrition and medication management, more frequent self-monitoring of blood glucose and enhanced physical activity. While higher levels of SE have been considered indicator of better diabetes control, lower levels have been seen as an indicator of worse diabetes control. 12 Aimed at evaluating the relationship between SE and the best performance by people with T2DM in adopting healthy behaviors, differ-ent instruments that measure how capable the individual perceives himself to play their care plan were created. 13,14,15 Among them, the Insulin Man-agement Diabetes Self-efficacy Scale (IMDSES), 16 developed in the United States to measure SE of adults with diabetes (type 1 and 2) specific for those treated with insulin. Thus, the aim of this study was to expand the evaluation of the psy-chometric properties of the Brazilian version of IMDSES through its use in patients with type 2 diabetes mellitus.

Study design and Setting
This methodological study was conducted in a Primary care Unit and in a specialized outpa-tient clinic of DM of a city university hospital in the Campinas, state of São Paulo, Brazil. Data collection was carried out between May and Octo-ber of 2012.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration; as well as its later amendments or comparable ethical standards. The study was approved by the local research ethics committee (No.1180/2011).

Sample selection and data collection
All patients (age > 18 years) with T2DM insulin applicants were included in outpatient monitoring. The exclusion criteria were: patients us-ing insulin for at least one year, who had chronic complications in the advanced stage as those with chronic renal failure diagnosis on hemodialysis, blindness, heart failure class III and IV and/or motor sequel of stroke were excluded.
The interviews were carried out individually by means of a patient approach in a private setting and consultation with available data in the registries. The sociodemographic and clinical data were obtained and Brazilian versions of the Summary of Diabetes Self-Care Activities Measures were applied -SDSCA and IMDSES 28 through interviews. This alteration of data gathering technique has been performed elsewhere with no harms to participants or the instrument psychometric performance. 17 To estimate the sample size, data from a pilot study among 20 subjects were used. Then sample size was calculated based on the Pearson correlation coefficient, considering as the null hypothesis, correlation coefficient = 0 (zero), α = 0.05 and β = 88%. The correlation coefficients were considered among the IMDSES Total Score and the domains of Specific Food and Glucose Moni-toring Summary of SDSCA, estimating the sample size of 127 subjects.

Measurement
Insulin Management Diabetes Self-Efficacy Scale (IMDSES): the original American version consists of 28 items that measure self-effi-cacy for seven behaviors: General Management, Diet, Exercise, Feet care, blood glucose monitoring, insulin administration and Prevention, Detection and Treatment of Hypoglycemia/Hyperglycemia. The answers are rated on a Likert-type scale of six points with the possibility of a "not apply" answer. 16,18 The cultural adaptation of IMDSES 28-items for the Brazilian context 17 was conducted in subjects with type 1 diabetes mellitus, dependent on insulin, and it has been modified from the original scale. The Likert-type response scale of six points has been changed to four-point and maintained the possibility of a fifth alternative "do not apply". Thus, the Brazil-ian version of IMDSES consists of 20 items distributed in the domains -General Management, Diet and Insulin. In this study we chose to use the Brazilian version of IDMSES with 28 items, without excluding the eight items proposed by Gastal, Pinheiro, and Vazquez. 17 The calculation of the total score corresponds to the average of the scores, with a possible variation of 28 up to 112. The higher the score, the lower the self-efficacy. Item 13 (drafted in the negative) needs reverse scoring for calculating the total score. Tobacco. Also, additional items were included re-lated to self-care recommendations (expanded version) that are not accounted for score calcu-lating. 19 The adaptation of expanded SDSCA into the Brazilian context was also conducted in pa-tients with T2DM. 20 It was chosen in the Brazilian version by the inclusion of items in the expanded version, as an item of Feet Care, items on Medi-cation and Tobacco. Thus, the Brazilian version of SDCSA consists of 15 self-care assessment items in six domains: General Diet, Specific Diet, Physi-cal Activity, Glucose Monitoring, Feet Care and medication; plus three additional items for Tobacco assessment. The score is obtained for each dimen-sion by the average number of days (scale 0-7), zero being the least desirable and seven to the more favorable situation.

Data Analysis
The statistical software program SAS ("Statistical Analysis System") version 9.2 was used for the following analyses: • Descriptive analysis of sociodemographic and clinical variables and the scores of the domains and the total score of the Brazilian version of the IMDSES and SDSCA; • Ceiling and floor effects: floor effect of IMDSES 28-item was evaluated by the percentage of patients who scored the 10% worse possible results of the scale). Ceiling effect was evaluated by the percentage of patients who rated the 10% best possible results of the scale; 21 • Practicability: assessed by time spent in the in-strument application; • Acceptability: evaluated by the percentage of patients who responded all items; • Reliability regarding internal consistency: estimated using Cronbach's alpha coefficient. Satisfactory internal consistency α > 0.70 was considered; 22 • Convergent Construct Validity: assessed through the use of the Spearman correlation coefficient to test the scores of the Brazilian version of the IMDSES and of the SDSCA. In order to test convergent validity, moderate to strong magnitude correlations were hypothesized between simi-lar constructs of IMDSES and SDCDA domains and no correlation or weak magnitude correla-tions were hypothesized between dissimilar constructs; • For the interpretation of the magnitude of the correlation coefficients, the criteria adopted by Fishbein and Ajzen 23 was used, which considers correlations ≤ 0.30 of low magnitude, from 0.31 up to 0.50 of moderate magnitude and > 0.5 of strong magnitude. A significance level of 5% was adopted.

RESULTS
It was found a predominance of women (59.8%) with a mean age of 58.3 (SD = 9.3) years old, with an average schooling of 6.1 (SD = 4.6) years. The average length of diagnosis of DM was of 15.1 (SD = 9.3) years and the mean duration of treatment with insulin was 7.7 (SD = 6.6) years. Among the associated clinical conditions, hyper-tension (85.2%), metabolic syndrome (83.5%), obesity (80%) and dyslipidemia (71.8%) were highlighted. The mean of the HbA1c was 9.0% (SD = 2.1).

Acceptability
Regarding acceptability, all patients responded to all of the items.

Ceiling and floor effects
In the analysis of ceiling and floor effects, assessing the proportion of 10% of the best and worst possible score of the scale, there was no floor effect for domains and Total Score. A ceiling effect was observed in the Insulin domain of the IMDSES, that is, 18.9% of participants scored in the best possible scale score in this domain.

Descriptive analysis
It has been found for the three domains of the Brazilian version of IMDSES as well as for the Total Score; scores above the possible averages, which indicates the self-efficacy level of commitment of the interviewees. The General Management domain had less involvement of self-efficacy level with an average of 2.3 (SD = 0.3) while the Insulin domain had the greater average commitment of 3.0 (SD = 0.5). The analysis of data on SDSCA shows score below the average expected for all domains, especially for Specific diet (3.7), showing a low level of self-care (see Table 1). Table 1 Descriptive analysis of domains and total score of the Brazilian version of the Insulin Management Diabetes Self-Efficacy Scale -IMDSES 28-items and the Summary of Diabetes Self-Care Activities Measures -SDSCA (n = 127).

Reliability Analysis
The reliability regarding internal consistency of the IMDSES was measured through Cronbach's alpha coefficient (α) estimation. Considering satisfactory internal consistency as α > 0.70 Cronbach's alpha 22 was satisfactory for all the IMDSES domains and for the Total Score. The analysis of item-total correlation and Cronbach's alpha if the item is deleted showed that in the Diet domain, coefficient lift to the exclusion of items 1 and 10. In the General Management domain, it was found higher coefficients with the exclusion of items 12, Item 13 and 27 (see Table 2).

Convergent Construct Validity
The convergent construct validity was as-  Table 3).
Medicina (Ribeirão Preto, Online.) 2018;51(2):121-30 Table 2 Correlation item-total, Cronbach's alpha coefficients, and Cronbach's alpha if the item is deleted from the Brazilian version of Insulin Management Diabetes Self-Efficacy Scale -IMDSES 28-items (n = 127).  Table 3 Spearman's correlation coefficients between the domains and total score of the Brazilian version of the Insulin Management Diabetes Self-Efficacy Scale -IMDSES and domains of the Summary of Diabetes Self-Care Activities

Criterion Validity
Criterion validity was tested by correlational evidence between scores of the domains and the Total score of IMDSES and serum levels of HbA1c.
Weak positive magnitude correlations were found between the score of the General Management domain (r=0.22; p=0.02) and the Total score (r=0.22; p=0.02) of IMDSES and the average value of HbA1c.

DISCUSSION
The aim of this study was to deepen the assessment of the psychometric properties of the DM, in which 54% had up to 12 years of study, the short-form IMDSES was easily filled, implying a low cost to the respondent, although it has not been described the average time of application. In other studies using the instrument, the practicability has not been evaluated.
As for acceptability, the results show good acceptance of the instrument since all the patients responded to all of the items. However, a high percentage of responses "Does not apply/not applied was recorded to items 14 and 15, which belongs to the domain of General Management of the Brazilian version of the IMDSES. This find-ing may indicate the need for greater investment in the education of T2DM patients regarding their treatment, since 47.7% of patients enrolled had a diagnosis of peripheral neuropathy.
It has been found a ceiling effect on insulin domain, i.e., 18.9% of participants scored the best possible results of the scale in this domain. The same finding was observed previously 24  Like previous history studies, 16,18 the correlations between the constructs of self-efficacy and self-care by applying the respective scales were of weak to moderate magnitude, confirming the construct validity of the Brazilian version of IMDSES. Although the humble findings given the size of correlations, it is relevant to highlight the capacity of this instrument to correctly measure such a complex construct as self-efficacy; which can be much more defying among a low literacy sample such as ours -this evidence can not be mistreated, since this sample probably represents most of Brazilian population. This relationship supports the underlying assumption that confidence of the individual in their abilities to perform a given behavior determines what behavior it can engage and how intense and lasting will be the engagement to deal with the challenges to achieving the expected results.
Criterion validity was assessed by correlation with HbA1c, which reflects the average blood glucose control in the three months prior to its acquisition, was supported in this study by weak magnitude of correlations between the domain General Management and Total Score of the Brazilian version of IMDSES and HbA1c. Gastal, Pinheiro, and Vazquez 17 found a significant association between self-efficacy level in Insulin domain of the Brazilian version of IMDSES 20 items and the level of HbA1c (ANOVA) and a linear trend that the higher self-efficacy, the lower level of HbA1c. In the study by Hurley & Harvey   16 found an association between all domains of original IMDSES and the level of HbA1c. Despite the small correlations, it is noticeable that the questionnaire was able to demonstrate correlations with such a strong variable such as HbA1c, since relations with objective measures are considered as the most strong relations an instrument may demon-strate. We must highlight that HbA1c is one of the best choices to assess self-efficacy behavior for insulin use, since it is the best serum variable to evaluate glucose levels -which reflect patient's adherence to T2DM as a whole -including medi-cation (as insulin), diet and physical activity. HbA1c has been used in guidelines worldwide as an ex-cellent marker to adherence.
In the literature, there is no consensus on the magnitude of the relationship between the level of self-efficacy and HbA1c. In a recent study, HbA1c levels were significantly correlated with the degree of self-efficacy among type 1 diabetics, but not among individuals with type 2 diabetes, suggesting that the factors that contribute to selfefficacy in diabetes patients with type 1 and type 2 are probably different. 26 As for the limitations, we must recognize that the alteration in the way the instrument was delivered (self-applied to interview) may modify the way participants respond to the questionnaire -although this alteration is supported elsewhere. 17 Besides, limitations regarding the use of self-reported measures must be taken into account, as bias of social desirability (in which the respondent answers what he believes the interviewer expects to hear) and memory.
In summary, the results of this study point to the reliability and partially support convergent construct and criterion validity of the Brazilian version of IMDSES when applied to T2DM patients.
It is expected that the findings of this research contribute to the psychometric refinement of the Brazilian version of IMDSES 28-items with a view to their future use in assessing the impact of nursing interventions on patient´s care diabetic insu-lin applicant types 1 and 2. It is recommended to carry out further studies with sample expanded to assess the validity of known groups (such as good glucose control x poor glucose control, us-ing HbA1c) and the responsiveness of the Brazil-ian version of the IMDSES 28-items in T2DM. It is also suggested the replication of this study among subjects from different socioeconomic status to confirm its practicability and acceptability.

CONCLUSION
The results of this study indicate that the Brazilian 28-item IMDSES requires a moderate time of application and have good acceptability when applied to T2DM patients with low literacy. It seems that the instrument has a potential to measure worsens and improves in self-efficacy, since no relevant ceiling or floor effects were found. The findings point to satisfactory reliability regarding internal consistency and convergent construct validity. Predictive criterion validity tested by the relationship with serum levels of HbA1c was partially supported. The availability of a reliable and valid instrument is an essential tool for evaluating the effectiveness of nursing interventions aimed at optimizing self-efficacy of T2DM outpatients in the management of their treatment.