Data source
In this study, older people are categorised as younger old (50–64), young old (65–74 years), old old (75–84 years), and the oldest old (85 years and above) [20]. Data utilised in this study were acquired from the WHO SAGE Well-Being of Older People Study (WOPS). These were population-based HIV surveys conducted in South Africa between 2010 (Wave 1) and 2013 (Wave 2) in collaboration with the Africa Centre Demographic Information System (ACDIS) [21]. The SAGE WOPS study gathers comparable longitudinal data on a variety of health, demographic, and social markers that are relevant to the health and functional status of older persons who are HIV-positive or have HIV/AIDS in their family [20]. In addition, the survey looked at the respondents’ nutritional status, and HIV treatment. Concerning the sampling method, the survey's sample was divided into five groups [20]. At the onset of Wave 1 of the project in 2010, the sample for Group 1 consisted of adults who had been receiving HIV therapy for at least a year. Aged individuals in Group 2 of Wave 1's 2010 cohort who were not receiving HIV therapy or who had only had it for three months or less. The third group of HIV-positive people in Wave 1 of 2010 were those who lived with adult (14–49-year-old) children. Group 4 was made up of elderly people who had experienced an HIV-related death of an adult household member in 2010. The aged who were not receiving HIV therapy or had only received it for three months or fewer in 2013 during Wave 2 were included in Group 5 [20]. The sampling methodology is described in detail elsewhere [22, 23].
Measures
Outcome variable
The outcome variable is based on the question “Have you ever been diagnosed with hypertension”. The response option was "Yes" or "No", which has coded into a binary outcome with Yes = 1 and No = 0.
Independent variables
The following factors were identified and selected as explanatory variables based on literature review [15,16,17], and their availability in the dataset: age, sex, education, employment, body mass index (BMI), marital status, and household wealth index. Age was recoded as (0 = 50–59, 1 = 60–69, 2 = 70–79, 3 = 80 +), sex (coded 1 = male, 2 = female), level of education (recoded 0 = no formal education, 1 = basic, 2 = secondary +), employment (0 = not working, 1 = working), marital status (recoded 0 = married, 1 = divorced/separated, 2 = never married, 3 = widowed). Body mass index of respondents was calculated based on weight and height using standardised computation (0 = underweight, 1 = normal, 2 = overweight, 3 = obese), wealth index (0 = poorest, 1 = poorer, 2 = middle, 3 = richer, 4 = richest). Wealth index variable was computed from respondents’ source of water, toilet facility, cooking fuel, electricity, household assets, and having domestic animals using principal component analysis (PCA). PCA post estimation test was done with Kaiser–Meyer–Olkin of 0.7 indicating a good measure of sampling adequacy. Wealth index was then divided into five quintiles (1 = poorest, 2 = poorer, 3 = middle, 4 = richer, 5 = richest). The comorbidity variables were derived from the questions on whether a respondent has ever been diagnosed of the following health conditions: diabetes (0 = No, 1 = Yes), stroke (0 = No, 1 = Yes), arthritis (0 = No, 1 = Yes), asthma (0 = No, 1 = Yes), heart disease (0 = No, 1 = Yes), cancer (0 = No, 1 = Yes) and depression (0 = No, 1 = Yes).We also derived some lifestyle behaviour variables from the following questions: ‘how many servings of fruits, and vegetables do you eat on a typical day? And ‘Have you ever smoked tobacco or used smokeless tobacco? (recoded 0 = No, 1 = Yes), and Have you ever consumed a drink that contains alcohol? (recoded 0 = No, 1 = Yes). Health-seeking behaviour characterised by the number of clinical visits (recoded 0 = not at all, 1 = once/twice, 2 = three to six times, 3 = more than six times) was also included as an independent variable.
Data analysis
We used STATA Version 14 as the tool for data analyses. Descriptive statistics were used to summarise hypertension status and its correlates. Chi-square test were used to test for differences between categorical variables. Binary logistic regression analysis was used to examine variables associated with hypertension. In all, four Models were fitted in the study. Model I introduced only socio-demographic factors (age, sex, education, employment, wealth status and body mass index). Model 2 adjusted for comorbidities (depression, heart disease, arthritis, asthma, diabetes, cancer and stroke). Model 3 varies from Model 1 & 2 based on the inclusion of lifestyle behaviour (tobacco and alcohol consumption, and fruit and vegetable consumption), and the complete model includes health-seeking (times visited the clinic in the last 12 months) in addition to all variables in preceding models (I-IV).
Ethical approval
This study followed the Declaration of Helsinki. The Ethics Review Committee of the World Health Organization, Geneva, Switzerland, approved the South Africa-SAGE Well-Being of Older People Study (WOPS) Wave 2. All participants signed a written informed consent form. The authors of this paper were not directly involved in the data collection operations. All methods were performed in accordance with the relevant guidelines and regulations. We requested access to the data at: http://www.who.int/healthinfo/sage/cohorts/en/.