The study was carried out in a rural area of Masaka district, southwestern Uganda, and peri-urban Wakiso district, located in and around the town of Entebbe near Kampala, the capital city of Uganda on the shores of Lake Victoria. Since, 1989, the Medical Research Council/Uganda Virus Research Institute on AIDS (MRC/UVRI) has been conducting HIV/AIDS-related epidemiological and clinical studies in these populations. In 1989, an open population cohort study was established in a rural population in Masaka district, which was expanded in the year 2000 to cover a population of about 18,000 people. In this community cohort, annual house-to-house surveys are conducted during which consenting participants are asked to provide demographic and behavioural information and a blood specimen for HIV testing. Birth and deaths are registered. Free access to health care is provided through a clinic which is supported by MRC/UVRI, and HIV testing and counseling services are offered at five outlets within the study area. Further details of the cohort study have been described elsewhere [6, 19, 20].
The MRC/UVRI in Entebbe, in collaboration with The AIDS Support Organisation (TASO), established and followed open cohort study of HIV infected participants from 1994 to 2009 to evaluate a variety of new interventions aiming to reduce HIV associated morbidity and mortality (the Entebbe HIV cohort) [21–23]. ART was introduced in both areas in 2004 and is available free of charge. ART coverage rates are thought to be very high, as HIV testing coverage is high in the rural population and treatment access is good. Local research studies have reported median CD4 cell counts of 100-150 per mm3 at initiation of ARV therapy [23, 24].
The study sample consisted of people aged 50 years and older. Five groups were selected to assess the direct and indirect effects of HIV/AIDS on the health of older people, from own infection to family situation. For each group, 100 respondents including 50 rural and 50 peri-urban residents, were selected from databases from the other studies described above. The study groups consisted of older people who:
Are HIV infected and have been on ART for at least 1 year; and,
Are HIV infected and not yet eligible for ART.
Had an adult child who died of AIDS-related illness;
Have an adult child living with HIV and on ART;
Have no child with HIV/AIDS and are not themselves infected with HIV (comparison group);
The criteria for initiation of ART were determined by the Ministry of Health, based on 2006 WHO guidelines. At the time, eligibility for ART was determined by CD4+ cell count with a cut-off of 200 cells per mm3 and by clinical criteria (stage 3 or 4).
Rural respondents for the first three groups were randomly selected from the General Population Cohort database of MRC/UVRI which covers the whole population in the area. For groups 1 and 2, all available 33 older people from the General Population Cohort were recruited, and an additional 67 rural respondents were randomly selected from the clinical databases of TASO and two other HIV care providers in the district.
The peri-urban study respondents for groups 1 and 2 were randomly selected from the MRC/UVRI cohort and the registers of the HIV/AIDS clinic run by TASO in Entebbe. Respondents for groups 3 and 4 were recruited from registers of the families of the Entebbe HIV cohort participants. The respondents for group 5 were randomly selected from the listing of self-support groups organized by local non-government organizations, unrelated to HIV/AIDS, and from the outpatient clinic of Entebbe hospital.
The study participants were interviewed at home by trained interviewers, using the study questionnaire after obtaining informed consent, and measured on blood pressure, weight, height, vision, grip strength, walking speed and cognition. A blood sample was also collected through a finger prick and filter paper (data from blood tests not available for this analysis). If the health status required further attention, respondents were referred to the MRC/UVRI clinic in the vicinity.
The duration of an interview and examination often exceeded 2 h. Therefore, interviews were spread out over two visits. Less than 1% of the selected participants refused to participate. The study was conducted between June 2009 and April 2010.
Special attention was paid to age reporting. Training and data collection involved using a historical calendar, checking of (grand)children's and parent's ages and relating the age of the respondent to others of a known age. The MRC/UVRI research study databases provided an independent age report, and major inconsistencies were addressed based on further examination of records or further information provided by interviewers or respondents.
The structured questionnaire and observed performance tests were adapted from existing survey instruments of the WHO multi-country Study on Global Ageing and Adult Health (SAGE) which have been used in multiple settings [25, 26]. The tools were translated into Luganda by a translator at MRC and back-translated, pre-tested in 22 respondents, and finalized after reviewing the pilot results. The questionnaire included four sections addressing the following areas: household information; family support networks and transfers; assessment of health and well-being; and, caregiving burden.
For the analysis, we used multiple health measures to assess whether people living with HIV and people indirectly affected by HIV were different from other older people not affected by HIV/AIDS: overall self-reported health and functional status, prevalence of chronic conditions, and a set of biological and clinical markers. Each of these measures was taken from validated composite indices comprised of a series of questions.
Overall health state was measured using a self-reported measure derived from eight health domains, including affect, cognition, interpersonal relationships, mobility, pain, self-care, sleep/energy and vision. Two questions were asked in each health domain, which measured the difficulties faced by the respondents in performing activities, each using a five-point Likert-type response scale. Item response theory with a partial-credit model was used to generate a composite health state score . Following each item calibration, using chi-squared fit statistics to evaluate its contribution to the composite health score, the raw composite score was transformed through Rasch modeling into a continuous cardinal scale, with 0 representing worst health and 100 representing best health . The psychometric properties of the health score have been evaluated elsewhere .
The WHO Disability Assessment Schedule (WHODAS) 12-item instrument was used to assess problems in functioning and disability . The series of questions assessed any difficulties faced by respondents in performing daily life activities. A five-point Likert-type response scale was used, with different weights assigned to different questions. The total score was then inverted to transform it to an index between 0 and 100 (termed WHODASi), with 0 representing extreme problems or complete disability and 100 representing a total absence of disability. Two of the 12 questions were omitted after the pilot test (difficulties in making friendships and dealing with strangers) because of poor endorsement rates and problems with translation equivalence. Respondents either said they did not understand the question, and those who answered only gave affirmative answers.
Given the limitations of self-reported morbidity, symptom questions and a related diagnostic algorithm were used to ascertain possible presence of chronic conditions such as angina, arthritis, asthma and depression. These have been shown to provide a better estimation of disease prevalence. A validated set of symptom questions is not available for diabetes prevalence, which was therefore based on a self-report of the condition. Details of the methods have been described elsewhere . This analysis focused on the presence of at least one of the five chronic conditions.
A hand grip strength test is considered a good indicator of frailty and strong predictor of mortality [32, 33] and was assessed using a Smedley's hand dynamometer. Testing was done with the respondent in a seated position and the elbow flexed at 90°, with the upper arm close to their body. Two measurements were taken for each hand and the best score was used for the analysis. The analysis focused on mean grip strength. Measured height and weight were used to calculate the Body Mass Index (BMI), computed as body weight (in kg) over the squared value of height (in m). Both mean and underweight (BMI below 18.5) were used in the analyses.
Systolic and diastolic blood pressure were taken three times in a sitting position using a Boso Medistar-S wrist blood pressure monitor. The median value was used in the analysis. High blood pressure was defined as having a systole of at least 140 mm Hg or a diastole of at least 90 mm Hg. Distance vision was tested using tumbling-E logMAR chart at 4 m, with poor vision defined as a score of 0.3 units or more. Walking speed is another good indicator of frailty, and predictor of a variety of health outcomes [34, 35]: a timed normal and rapid walk over a 4 m distance was used. Assistive devices such as eyeglasses or a walking stick were allowed, if the respondent typically used them.
The association between study group and background variables was first examined for men and women while adjusting for age. A multivariate regression analysis was conducted for each of the health and functional status indicators to assess the effects of being HIV-positive, with or without ART. The associations were examined in a model including both sexes, with and without controlling for demographic and socio-economic variables, including place of residence, marital status, and education (wealth quintile was dropped because of high multi-collinearity with education). The three study groups with older adults with no HIV-infection were combined into one group, as differences among them were small. (We also analysed separately using the original five study groups with similar results). A multivariate linear regression model was used for all health and functional status measures, except logistic regression was applied in the models for the prevalence of a chronic condition and hypertension. All analyses were conducted using STATA statistical software version 10.0 .
Ethical clearance for the study was given by the Uganda Virus Research Institute Science and Ethics Committee and the Uganda National Council for Science and Technology.