Study design
A cross-sectional study design was employed to examine the psychometric properties of the Afaan Oromoo version of the DQOL measure.
Participants
A convenience sampling technique was used to recruit people living with T2D attending the diabetes center of Nekemte Specialized Hospital in Western Ethiopia for their monthly medical check-up between June and August 2020. Included among the people living with T2D were those who 1) had been treated at the diabetes center in the hospital for six months or more; 2) were in a stable medical condition; 3) were aged 18 or over; 4) were cognitively intact, and 5) were able to speak and understand Afaan Oromoo. People living with Type 2 diabetes were excluded if they had a hearing problem.
Ethical approval for the study was obtained from The Hong Kong Polytechnic University. Permission to collect data was obtained from the hospital administrator before the start of the study and informed written consent was obtained from the participants. The confidentiality of the data was ensured through coding.
Translation of the DQOL
Permission to translate and adapt the 46-item DQOL scale was obtained from the scale developers. The DQOL was translated and culturally adapted into Afaan Oromoo according to the six-stage recommendation of cross-cultural adaptation developed by the Institute for Work and Health in 2007 [28]. In stage 1, two versions of the forward translation of the original version of the DQOL were prepared by two translators (a health professional and a naïve translator who is a Ph.D. holder in Afaan Oromoo). In stage 2, a synthesis of the translations obtained in stage 1 was made by the principal researcher, and a reconciled translation of the scale was developed after agreement on any discrepancies was reached. In stage 3, two separate versions of the back-translation of the scale were prepared by another two excellent translators, who were Ph.D. holders in English and native speakers of Afaan Oromoo. In stage 4, an expert panel consisting of seven professionals (one public health expert, one nurse, one Afaan Oromoo language expert, two forward and two backward translators) was formed and they evaluated conceptual, semantic, and idiomatic equivalences of the translated versions of the scale using five-point Likert scale items to calculate the content validity index (CVI). Discrepancies were resolved through discussion until consensus was reached. The CVI of the Afaan Oromoo version of the DQOL tool was ≥ 0.95. In stage 5, 30 people living with T2D were asked to assess the applicability, readability, and clarity of the item content of the expert-evaluated version of the scale [29]. The cultural adaptation was made using locally spoken and acceptable words. The people living with diabetes were requested to suggest the appropriate terms, and amendments were done to the local culture. In stage 6, an amendment was made based on feedback from the participants, using appropriate words and restructuring some sentences in a culturally appropriate way, and the final version of the scale in Afaan Oromoo, the Diabetes Quality of Life-Afaan Oromoo (DQOL-AO), was developed and subjected to psychometric testing.
Sample size calculation
For psychometric testing, the required sample size was determined based on exploratory factor analysis (EFA) using the case-to-variable (rule of thumb) of 10:1 ratio. A minimum sample size of 460 was included.
Instrument
The 46-item DQOL has four major domains: satisfaction (15 items), impact (20 items), social/vocational worry (7 items), and diabetes-related worry (4 items). Items in the satisfaction domain are scored on a five-point scale ranging from 1 (very satisfied) to 5 (very dissatisfied), and items in the impact and the two worry domains are scored on a five-point scale, ranging from 1 (no impact and never worried) to 5 (always impacted and always worried). If an item is not relevant to the respondent, the ‘Does not apply’ option is provided for the social/vocational worry and diabetes-related worry subscales and will not be scored. A lower score in DQOL indicates a better QOL [18].
Sociodemographic variables, namely gender, marital status, ethnicity, religion, educational level, family member usually providing support, and employment status, and patient-related factors such as the diabetes-specific complication (s) and year of first disease diagnosis were collected.
Data collection procedure
Eight data collectors who have experience in data collection were trained in a one-day workshop to ensure they were familiar with and understood the items in the scale and the techniques of conducting interviews for the study. People living with diabetes were approached when they were waiting to see the doctor in the diabetes center of the hospital. After explaining the purpose and study procedure, the data collectors screened the people living with diabetes for their eligibility. Having obtained their informed written consent, the data collectors then administered the questionnaire via face-to-face interviews.
Statistical analysis
All data analyses were conducted using SPSS statistics version 25. The factor structure of the DQOL-AO was examined in two steps. In step 1, item reduction based on the item-total correlation was performed. Any item with an item-total correlation coefficient below 0.3 was removed [30, 31]. In step 2, EFAs were conducted on the items remaining after step 1. The Kaiser-Meyer-Olkin (KMO) and Barlett’s tests checked for the appropriateness of conducting EFA. The factor retention was based on four criteria: (i) eigenvalues > 1; (ii) scree plot; (iii) interpretability of the retained factors; and (iv) factor loadings > 0.4. For items cross-loaded on factors, the retention of the item to the factor was determined by two criteria: 1) a higher loading effect of the item onto the factor and 2) the interpretability of the item. The Cronbach’s alpha value was then calculated to assess the reliability of the subscales and the overall scale of the DQOL-AO. Ceiling and floor effect analysis for subscales and the overall scale were performed to distinguish the proportion of respondents with the highest and lowest QOL scores, respectively [32]. Ceiling or floor effects were judged if more than 15% of subjects reached the highest or lowest score, respectively [33].
The construct validity of the DQOL-AO was assessed by the known group and correlation analysis – Pearson’s correlation was used for continuous demographic variables and an independent t-test or ANOVA was used for categorical variables. In all the analyses, a P-value <0.05 was considered statistically significant.