The research protocol was approved by the Research Ethics Committee of the Hospital of Universidade Federal de Juiz de Fora, MG, Brazil (CAAE 77831517.0.2002.5133) and all participants gave their informed written consent before being included in the study. Translation, cross-cultural adaptation, and validation of DATE-Q was performed with prior permission from the original authors and consisted of two phases: (1) translation and cross-cultural adaptation and (2) establishing psychometric properties.

This study followed the five steps proposed by Beaton et al:27 (1) Translation into Brazilian Portuguese - two translations from English into Brazilian Portuguese (T1 and T2) were generated independently by two native speakers of Brazilian Portuguese; (2) Synthesis of the translations – the two initial translators and a physical therapist summarized the two translations into one version (T1.2), disagreements were solved by consensus; (3) Back translation – two native English speakers not familiar with the original version of the questionnaire translated the T1.2 version independently back into English (BT1 and BT2); (4) Expert Committee - a committee of experts which included T1, BT1, a physical therapist, a dietitian, an expert physician in diabetes, and an associate professor with expertise in methodology consolidated all previous versions of the questionnaire and developed a pre-final version of the questionnaire in Brazilian Portuguese to be used in a pilot test;27 (5) Pilot testing of the pre-final version – 30 patients with diabetes were selected by convenience from a secondary health service in a Brazilian city according to the following eligibility criteria: diagnosis of type 1 (T1D) or type 2 diabetes (T2D), age 18 years or older, and able to physically exercise. All of them were invited to self-complete the pre-final version of the questionnaire.

The understanding of each item was assessed by individual interviews based on the think-aloud method.28 The participants were interviewed to answer what each item meant to them and to explain the significance of words and expressions.29 To accurately identify possible uncertainties, the interviewer recorded the participants' full answers to each item, and the percentage of uncertainties for each item was computed. The clarity of each item was assessed by a Likert Scale ranging from 1 (I do not understand completely) to 5 (I do understand completely).

After completing the fifth step, the last adaptations were made to obtain the version to be psychometrically tested in the next phase of this study.

Two-hundred adults with diabetes were enrolled in this phase. The sample size calculation was based on Hair and Anderson’s recommendation of a minimum sample size of 10 individuals per item, and/or at least 100 participants.30 These individuals were recruited from primary and secondary health services in a Brazilian city and invited to the study according to the same eligibility criteria described previously, in addition of not having participated in the pilot test.

All recruited individuals were asked to answer a six-item test31 to screen for cognitive status, and only potential participants who achieved a score of 4 or greater were invited to sign the consent form and were included in the study. All participants completed a form to report clinical and sociodemographic information. Participants were randomized to one of four subsamples (Fig. 1). The generation of the randomized allocation sequence was performed by the principal investigator (PI) using an online tool, and the allocation sequence was retained in a password-protected file. The randomization information was provided from the PI to the assessor only when the assessments were about to be started to ensure allocation concealment. Participants who did not complete all steps of this study phase, according to the subsample to which they was randomized, were excluded.

Figure 1.

Flow chart of participants recruitment, inclusion, and randomization. SIS, Six-Item Screening tests; A1C, Glycated hemoglobin; ISWT, incremental shuttle walking test.

(0.13MB).

The time for completion of the DATE-Q was recorded. Participants completed the questionnaire with neutral monitoring by the assessor.

Subsample 1 — participants randomized to this subsample shared with the research team the glycated hemoglobin (A1C) values from laboratory tests completed as part of their routine clinical care because it is used as a gold standard to estimate the average blood glucose control.6 From the results of previous studies,32,33 it was hypothesized that A1C values and DATE-Q total scores would be negatively associated, and these variables were considered for the assessment of the following psychometric property: construct validity.

Subsample 2 — participants randomized to this subsample performed the incremental shuttle walking test (ISWT) to measure functional capacity.34 From the results of previous studies,35,36 it was hypothesized that ISWT distance in meters and DATE-Q total scores would be positively associated, and these variables were considered for the assessment of the following psychometric property: construct validity.

Subsample 3 — participants randomized to this subsample completed the DKN-A to objectively evaluate disease-related knowledge.25 The DKN-A is a 15-item self-administered multiple-choice questionnaire validated in Brazil, and its total score was considered for the assessment of the following psychometric property: construct validity.

Subsample 4 — participants randomized to this subsample completed the instrument twice, with an interval of seven to 21 days between completions. This process was used to assess the following psychometric property: test-retest reliability. Both tests were completed in the same location by the same assessor.

The IBM Statistical Package for Social Sciences (SPSS) version 22.0 software was used for storage and data analysis. The normal distribution of data was verified using the Shapiro-Wilk test. Variables with normal distribution were expressed as mean and standard deviation, while those with non-normal distribution were expressed as median and interquartile range. The DATE-Q data were compared as total scores as well as scores in each domain of the original questionnaire (self-management, long-term complications, being active, healthy eating, and psychosocial well-being) according to age, diabetes type, insulin therapy, time elapsed since diagnosis, educational level, and household income using unpaired t-tests. Comparisons for more than two-samples were performed with one-way ANOVA or Kruskal–Wallis test when numerical variables were tested and with Chi-Square Test when categorical variables were tested. The power of the sample was calculated based on the size of Cohen’s effect of these comparisons and values >0.8 were considered significant.37 A significance level of 5% was adopted for all statistical tests. The psychometric properties of the questionnaire were determined based on randomized subsamples.38

The internal consistency was assessed by Cronbach's alpha and considered satisfactory when α > 0.6.37–39 Construct validity was assessed by calculation of Spearman’s correlation coefficient between DATE-Q total scores and A1C values, ISWT distance (meters), and DKN-A total scores. Correlation coefficients <0.19 were considered very weak, from 0.20 to 0.39 weak, from 0.40 to 0.59 moderate, from 0.60 to 0.79 strong, and >0.80 very strong.30,38 Reliability was assessed by intraclass correlation coefficient (ICC), considering the values <0.40 as low reliability, from 0.40 to 0.75 moderate, from 0.75 to 0.90 substantial, and >0.90 excellent.30,37,38 The presence of ceiling and floor effects for the 20-item DATE-Q was determined based on the proportion of occurrence of the highest and lowest scores for the total sample, with values below 15% considered acceptable.38

According to the Expert Committee, there was no divergence between the translated and back-translated versions compared to the original questionnaire. Cross-cultural adaptations were made based on the Expert Committee's recommendations and results from pilot testing, which resulted in 10 items being culturally adapted. All versions of the questionnaire (i.e. from the original to the final version) are available at https://data.mendeley.com/datasets/wfcwb3k79h/1.

Thirty patients with diabetes (median (interquartile range) age of 58 (49–64) years old; 21 females; 87% with T2D; 60% insulin-treated; fasting glucose 115 (92–151) mg/dL; glycated hemoglobin 8.0 (6.5–8.8)%; educational level from illiterate to postgraduate; household income from less than 1 to more than 9–12 times the Brazilian minimum wages received monthly; six-item screening test scores of 5.4 ± 0.8) participated in the pilot test.

The most common value assigned to all DATE-Q items on the Likert scale was 5, and the percentage of uncertainties was lower than 20% in all items, with the exception of items 3 and 6, as many participants did not know what A1C was.

For the second phase, 242 patients with diabetes were approached, of which 231 (95%) met the inclusion criteria and were enrolled in the study. Of these, 200 (83%) concluded the study (Fig. 1). Sociodemographic and clinical characteristics of these participants and DATE-Q total scores, as well as the comparison between the four subsamples are presented in Table 1. The proportion of participants of subsample 1 on insulin therapy and oral antidiabetics use were significantly higher and lower, respectively, compared to those from other subsamples. Participants’ mean time to complete the DATE-Q was 5 min and 51 s, and the completion rate was 100% for all items.

The Brazilian Portuguese version of DATE-Q item descriptions and mean score are described in Table 2.

The internal consistency was satisfactory (Cronbach’s alpha = 0.6) and reliability was moderate (ICC = 0.5). The DATE-Q total scores were strongly correlated to DKN-A total scores (ρ = 0.7; P < 0.001; Fig. 2), and weakly correlated to A1C values (ρ = 0.2; P =  0.1) and ISWT distance (ρ = 0.3; P =  0.01). Ceiling and floor effects were not found as only 2 participants scored the maximum total and no participants scored the minimum total.

Figure 2.

Correlation between DATE-Q total scores and DKN-A total scores (n = 50). ρ = 0.7; P < 0.001.

(0.03MB).

Table 3 shows the results from the DATE-Q answers compared by age, diabetes type, insulin therapy, time elapsed since diagnosis, educational level, and household income. Participants under 65 years old, who live with T1D, who have a higher educational level or household income, achieved significantly higher DATE-Q “hits” rates and DATE-Q domains’ scores than their counterparts. In addition, insulin-treated participants demonstrated significantly higher DATE-Q “hits” rates and scores in self-management and long-term complications domains than those who were not treated with insulin. Conversely, participants living with diabetes for less than 10 years demonstrated significantly lower scores in self-management and long-term complications domains compared to their counterparts.