Study design, setting and participants
This was a cross-sectional study using questionnaires to assess stigma, beliefs about medicines and adherence to medication. The participants were recruited from the outpatient clinics of one public hospital and one private hospital in Jayapura. These are two out of three hospitals providing HIV care in Jayapura. We have selected those two hospitals as they were the first ones to provide HIV care by teams of specialists for 10 years. ART is dispensed free of charge to patients.
Participants were included if they were at least 18 years old, were on ART for more than six months, and signed informed consent. Patients were excluded if more than 50% of data was missing. Participants were recruited and informed about the study while they were collecting their ART in the hospital. The recruiters, one pharmacist and three nurses who were health care professionals from the two HIV study hospitals, explained the aim of the study. They also emphasized that all information was kept confidential and that the decision whether to participate or not did not affect their treatment in any way. After the participants signed the informed consent, an appointment to complete the questionnaires was made for the following month as part of their next visit to the hospital.
The sample size was calculated based on the use of a two tailed test with a level of significance of 0.05. It was assumed that approximately 77% of the population were adherent [16]. Therefore, at least 284 participants were required to conduct the study.
Instruments
We have chosen to assess adherence [18], beliefs about medicines [9] and stigma [19] with widely used instruments [10, 20, 21].
Adherence
We used the self–reported Medication Adherence Rating Scale (MARS) [18]. This consists of 10 questions. This questionnaire assessed the patients’ medication taking behavior in the past week. An example of a statement is: “I take my medication only when I am sick”. The response is binary, yes/no. A total MARS score is calculated as the sum of all items with a maximum score of 10 [18]. A higher total score indicated higher levels of adherence. We dichotomized the response into adherence and non-adherence. Sensitivity analyses were performed with MARS scores of 7, 8 and 9 as the cut-off (Additional file 1: Appendix). The cut-off score of 8 was the most sensitive and was therefore used for the analysis.
HIV stigma
The HIV Stigma-Sowell scale consists of 13 questions. These 13 items represent 3 types of stigma, namely distancing, blaming and discrimination [22]. A question representing distancing is: “I thought other people were uncomfortable being with me”. A question representing blaming is: “I felt I would not get good health care if people knew about my illness”. A question representing discrimination is: “I felt compelled to change my residence because of my illness”. Answers are scored on a 4 point Likert scale (as: 1 = not at all, 2 = rarely, 3 = sometimes and 4 = often.) The distancing and blaming scales were assessed with 4 questions each. The discrimination scale was assessed with 5 questions. The total score ranges from 13 to 52 with a higher total score indicating a higher level of stigma.
Beliefs about medicines questionnaire (BMQ)
The BMQ assesses patient’s concepts about medication use in general and patient’s beliefs about the medication they use [9]. The BMQ-General part of the questionnaire consists of two scales asking about their views on overuse and harm related to medication (four questions each). An example of a statement on overuse is: “doctors use too many medicines”. An example of a statement on harm is: “most medicines are addictive.” The BMQ-Specific part consists of two scales about the necessity and concerns of patients regarding their medicine (5 questions each). An example of a statement on necessity is: “my antiretroviral medication protects me from becoming worse”. An example of a statement on concern is: “I am sometimes worried about becoming too dependent on antiretroviral medication”. All 18 statements were scored on a 5-point Likert scale as 1 = strongly disagree, 2 = disagree, 3 = uncertain, 4 = agree, 5 = strongly agree. The total from every scale of the BMQ was calculated. Higher total scores on necessity indicated patients being positive and seeing the advantages taking their medication. Higher total scores on concern, overuse, and harm indicated concerns about the treatment.
Translation
Since all instruments were developed and validated in English, we translated the instruments forwards into Bahasa Indonesia and backwards into English. The forward translation was done by two Indonesian certified translators. Both versions were assessed by DPA as proofreader and reviewer with a lot of experience in translating questionnaires. A final, reconciled Bahasa version was agreed on. This version was back-translated by an English native speaker who had no information about the original versions of the questionnaires. The backward translation was modified several times because the target language does not recognize verb tenses [23]. Therefore, the final version included words related to time. The final questionnaires were tested in 67 PLHIV who did not participate in the main study. The Cronbach’s alpha coefficient to test the internal consistency of the questionnaires was: 0.94 for the MARS scale, 0.82 for the HIV stigma scale and 0.6 for the BMQ scale, indicating good to moderate consistency. Participants indicated they understood the questionnaires.
Data collection
At the appointment, participants were handed out the three paper-based questionnaires and were also given a questionnaire to collect sociodemographic information (age, gender, marital status, whether or not they had children, education, employment, sexual orientation, religion, ethnicity, receiving support in medicine-taking from social network, and ART regimen). They completed the questionnaires in the waiting area of the hospital. The recruiters were available for questions. Information about the age and types of ART were collected from PLHIV’s medical records. Data were collected between September and November 2016.
Descriptive statistics
The sociodemographic variables were analyzed using descriptive statistics. The categorical variables were presented as frequencies and percentages. The continuous variables were summarized either as mean (standard deviation) or median (interquartile range) depending on the nature of the variables. The outcome variable was binary coded as non-adherent and adherent. Chi-square test for categorical and independent sample t-test or Mann-Whitney U test were used to determine the association between the independent variables and adherence. We included the following independent variables in the analyses: age, gender, marital status, having children, education, employment, sexual orientation, religion, ethnicity, social support in medicine-taking, ART regimen, distancing, blaming and discrimination of the stigma scale and necessity, concern, overuse and harm from the BMQ questionnaire.
Statistical modelling
We included all independent variables with a p value of < 0.20 in the descriptive analysis in a univariate logistic regression to examine the effect of the independent variables on adherence. Then, independent variables with a p value of < 0.20 from the univariate analysis were included in the multivariate logistic regression analysis [24]. In the multivariate logistic regression analysis, we used a backward elimination procedure to select the final model with all independent variables being significant with a p value of ≤0.05. Finally, odds ratios of independent variables and their 95% confidence interval were presented. All analyses were done using two-tailed tests at a significance level of 0.05. The statistical analyses were performed using the Statistical Program for Social Sciences (SPSS) version 24.0 for Windows.