Our results indicate that the majority of older people aged 50-plus years rated their functional ability favourably (median by gender and HIV status from 69 to 88). The lowest scores were observed for the multi-dimensional health state, suggesting a much lower health status than reported using the functional ability or quality of life measures separately. These differences could partly be explained by the underlying methodology in computing the scores for these health measures (WHO tools used for functional ability and quality of life assessment are simple arithmetic additive scores, while the health state score was generated using Item Response Theory and Rasch models). However, in spite of these differences in methodologies, correlates of the WHODAS and HSS scores were very similar for both HIV-infected and HIV-affected people. This may be reflective of the stronger contributions from the health domains describing physical functioning than those that are more subjective to the composite health state score. Our findings highlight and support previous findings that the use of a single health outcome measure may be helpful to describe the overall health status of older people, but may also have limitations [52
]. Using only the composite health score would have underestimated the health status of our study participants. An investigation of the contributions of the specific domains to overall health status needs to be undertaken for a more precise description of the health of older people and to inform the design of interventions [53
Our findings suggest that the effect of being HIV-affected differs between those who are affected via having an HIV-infected adult child and those affected via an HIV-related death of an adult child. Older people who had lost an adult child due to HIV were more likely to be in poor physical and emotional health than those with a living HIV-infected adult child (Table ). The death of an adult child is likely to take its toll on the physical health of older people who have had to care for the adult during the time of sickness [54
] and who may additionally be emotionally affected upon death [31
]. The death of an adult child may furthermore place greater household responsibilities on the older person as there may be loss of household income from the deceased adult [29
] as well as orphaned grandchildren who may require financial and social support. Thus death of an adult child is likely to strain the older person physically, emotionally and financially, which in turn is likely to contribute to their poorer physical and emotional health relative to older people still living with an infected adult.
Overall, combining the two HIV-affected groups into one, we further found that HIV-infected participants had better functional ability, quality of life and overall health state than HIV-affected participants. This may seem counterintuitive in that ill-health may be expected to be more prevalent among HIV-infected people [41
], but this difference may partly be explained by the enhanced health care that this group receives as part of their regular clinic visits for antiretroviral (ART) treatment. These findings are consistent with a study by Louwagie et. al. [16
] who compared health related quality of life (QoL) of patients on Highly Active Antiretroviral Treatment (HAART) to those awaiting HAART, and showed that patients on HAART had on average a higher health-related QoL score than those awaiting HAART. Other studies in South Africa have also demonstrated the beneficial effects of HIV treatment on health and well-being. In a study in the Free-State province of South Africa, where changes over a 12-month period in the physical and emotional quality of life of people on ART were examined [17
], it was shown that at follow-up people on treatment had fewer adverse events than at baseline; adverse events were negatively associated with physical and emotional quality of life. Evidence of ART leading to improved health can also be inferred from previous work in our study community that showed that ART has contributed to declining mortality among adults [56
]. The considerable evidence that HIV treatment is effective in achieving sustained improvement in the health and well-being of HIV-infected people [17
], clearly contributes to the superior health status of HIV-infected people we observe.
Our results do suggest that as age advances irrespective of HIV status, our study population in rural South Africa is increasingly associated with poorer functional ability and overall health state, with major gender differences. Women reported poorer health status than men among both HIV-infected and HIV-affected participants. These results showing a male advantage in self-reported health in later life are consistent with other studies [11
]. A pooled analysis of data collected in four African and four Asian sites, whose study instruments as in the present study were adapted from WHO’s Study on global AGEing and adult health (SAGE), reported that older women had significantly lower health scores than older men at all age groups [10
]. According to findings from a nationally representative study from Thailand [63
], a larger part of women’s remaining life expectancy in old-age is spent in a disabled state. These gender differentials in health are said to be more complicated and nuanced than can be explained by biological or medical factors alone [64
]. Hirve and colleagues [65
] argue that this female disadvantage in health may be accounted for by advancing age, societal norms concerning women, occupation, lower education attainment and lower empowerment. The societal norms and institutionalisation that tend to fuel this sex disparity in health mostly occur around life’s central foci of ‘paid work or unpaid family work’ [64
In South Africa, people in the age range considered for this study come from a generation renowned for migration of men to the mines and cities for paid work while the women remained in their rural homes with the care burden for children and those with disabilities [66
]. This has meant that men and women are exposed to different health-related risks as well as resources across the life-course, and has highly likely contributed to the sex disparities in health we observe. Independent of HIV status, older women are clearly more vulnerable than men to poorer health and functional ability limitations, which are a function of circumstances over the life course.
Being in the highest two household wealth quintiles was strongly related with better quality of life, even after adjusting for other factors in HIV-infected participants. This is consistent with a study among older people aged 50+ in Pune district, India, which found that older people in higher household wealth quintiles were more likely to report better quality of life than those in lower wealth quintiles [65
]. However, our results and context differ from the Pune district study in that in their case there was no ready access to government cash transfers and they did not find a significant association between gender and quality of life. In our study area government cash transfers in the form of old-age pensions are widely available and we find a significant association between gender and quality of life, as well as between having some income source and quality of life. Most of this income, which is linked to the quality of life of older people in rural South Africa, is from non-contributory government cash transfers or grants; therefore rapid increases in the proportion of older people poses serious challenges to their well-being by threatening the sustainability of the cash transfers programme.
Gender, advancing age, education and income were independently strongly associated with the health and well-being of older people in this study. The factors reported here associated with health status were similarly reported on by others using similar study instruments [10
]. In six of these studies all individuals aged 50+ were eligible for inclusion and in two studies a random sample of households containing at least one older person aged 50+ was targeted. For this study, individuals among HIV-infected and HIV-affected clusters of older people within the community were selected. Another main methodological difference to these other studies is that they applied binary logistic regressions to the quintile health scores, where they defined those in the highest two quintiles as healthy and the rest as unhealthy. The decision as to at which quintile the cut-off into healthy and unhealthy should be is highly arbitrary [43
] and different results may be obtained if different cut-off points are used. Ordered logistic regression analyses, which make use of the quintile distributions without an arbitrary cut-off, were used in this study. Despite these methodological differences, the findings confirm that health and wellbeing of older people varies by socio-demographic characteristics such as age group, gender, education attainment and income, but is further strongly linked to whether the older person is HIV-infected or HIV-affected.
There are, however, some limitations to our findings. In addition to our small sample sizes, participants into our study were purposefully sought to be categorised into HIV-infected or HIV-affected groups - this could have biased our findings. Some of the potential sampling bias was corrected by applying sampling weights and making use of survey tools in the statistical software package used in the analyses. Another limitation of the analysis is the possibility for some of the participants categorised as HIV-affected to also be HIV-infected themselves. This is likely to occur if such participants were tested and/or accessing HIV treatment outside of the Hlabisa sub-district or from private practitioners and hence not captured in the Hlabisa HIV Treatment and Care Programme. It is, however, highly unlikely that participants were accessing treatment and care outside the sub-district given the logistical and financial implications of travelling significant distances particularly for older people. Thus, this potential bias was assumed to be very negligible. Although a blood specimen was collected from all participants as per study protocol these specimens were not tested for HIV antibodies. All participants were informed that no HIV testing would be done on the specimens. However, an earlier study from our study population showed that HIV prevalence in the population 50+ in 2008 was 9.5% (95% CI 8.4-10.7), with an incidence rate of 0.5% (95% CI 0.3–1.0) [35
]. Therefore, we do not expect to have had many infected people in groups 3 or 4 to significantly bias our findings.
We urge caution in the interpretation of our results, particularly the association of age with poorer functional ability and health state because of small-numbers, especially in the oldest age group, and limited statistical power. The results may also have been affected by a healthy selection effect into the WOPS - those that participated in the study may be survivors from their cohorts. Furthermore, the study was cross-sectional, thus it is not possible to make causal inferences between the socio-demographic factors considered and health status. We are limited in generalising our findings to the general older population of South Africa since our study participants came from a population under constant surveillance with ready access to a comprehensive HIV care and treatment programme [40
]. In addition we have not controlled for other household factors such as number of HIV-infected persons in the household, living arrangements (living alone, in skip-generation household or multigenerational household) and cash transfers to other household members as that was beyond the scope of this analysis. Our results nonetheless make an important contribution to understanding the correlates of health and well-being of older people in rural South Africa.