Cette page existe uniquement en anglais
Home > Activities > Committees >AIDS (1995-99) > Conference Report
Durban, South Africa, 3-6
organized by the IUSSP Committee on AIDS
Some 15 years after the symptoms of AIDS were first identified, HIV continues to spread. Nowhere has the impact on the general population been greater than in Sub-Saharan Africa. Despite our growing understanding of the virus and the factors that determine its transmission, we are still far from being able to predict with any accuracy the final shape of the epidemic and the way it will affect either the social or the demographic structures of the continent. Active investigation of the impact of HIV on fertility, mortality and social organisation prompted the IUSSPs committee on AIDS to convene its first conference on the socio-demographic impact of AIDS in Africa.
Co-sponsored by the university of Natal and held in Durban from February 3 - 6, 1997, the conference attracted policy-makers and development workers specialising in HIV/AIDS issues as well as academics. The juxtaposition made for interesting debate, although (or perhaps because) their specific areas of interest were rather diverse. Many policy-makers and development workers were from South and southern Africa and were interested in how the disease might affect the highly mobile and largely urbanised societies on which they focus. The overwhelming majority of research presented was, on the other hand, centred on a small number of rural cohort studies in East Africa.
From the opening address given by Lieve Fransen, head of the European Commissions task force on HIV/AIDS, policy makers made it clear that what they wanted from demographers was clearly presented data to inform their decisions about how best to use limited resources to minimise the impact of the disease. Demographers, for their part, suggested throughout the conference that they would welcome more collaboration with the people who use the data they produce, but remained circumspect about their ability to fill the laundry list of data requirements presented by the policy makers.
The four-day conference was organised into sessions on fertility, household and family structure, mortality, measurement issues, orphanhood and policy implications. The final session was dedicated to group sessions discussing what demographers could bring to the government planning process and to the private sector, donor and NGO efforts in the fight against AIDS.
The relationship between HIV infection and childbearing is clearly important, not least because much of what we know about levels and trends of infection in different areas comes from testing pregnant women at ante natal clinics. The more we know about how HIV infection relates to pregnancy, the better idea we have of how our sentinel surveillance data relate to women in the general population.
In a presentation that sought to clarify the interaction between HIV and fertility, Simon Gregson considered the possibility of a direct physical effect of the virus on the probability of conception. He also posited that changes in behaviour that result from a growing awareness of HIV and a desire to avoid infection or vertical transmission might change childbearing patterns.
Behaviour change might work in both directions. Where post-partum abstinence and long periods of breast feeding are common, women may feel the practices push their husbands into multiple partnerships and therefore increase their risk of infection. Curtailing abstinence and breast feeding may increase fertility. Deliberately seeking to increase fertility in response to the high-mortality environment created by HIV would also increase the average number of children per woman.
On the other hand, where the most common response to the fear of infection is decreased sexual activity or increased condom use, behaviour change is likely to limit childbearing. In high-HIV prevalence situations, death and disabling illness is also likely to cut down the time spent in sexual unions. Overall, Gregson and his colleagues from the Blair Institute in Harare felt that the forces reducing fertility were likely to outweigh those promoting fertility.
The two papers that followed examined in greater detail the physical relationship between seropositivity and fertility. Lucy Carpenter, in the first of several papers at the conference based on a Medical Research Council (MRC) cohort study in a rural ugandan population, looked at fertility in the seropositive over a six year period. After adjusting for age, she and her colleagues found that HIV infected women were 23 percent less likely to bear children than the seronegative. They found no significant relation between fertility and markers of previous syphilis infection.
The effect of HIV infection on the likelihood of being pregnant was even more dramatically demonstrated in data from a second longitudinal cohort study in rural uganda. David Serwadda and his collaborators found that HIV positive women were half as likely to be pregnant as uninfected women, and the drop in fertility appeared to become more acute as infection progressed. By the time women showed symptoms of AIDS, they were nearly 80 percent less likely to be pregnant than HIV negative women, even taking age differences into account.
The study showed that active syphilis cut pregnancy by just under a third independently of HIV, but that other STDs had little effect. And HIV infection appeared to suppress fertility regardless of behavioural differences such as number of partners.
It may be that the reduction in fertility differs at different stages of the epidemic, and with various mixes of behavioural and physiological factors. These papers showed that fertility is diminished even when controlling for behavioural differences. But they reminded participants that behavioural variables such as age at first sex, rates of partner exchange and coital frequency, some of which may alter in response to perception of risk of infection, can be important factors in reducing fertility in their own right.
Household and family structure
A study of households covered by the MRCs general population cohort of 10,000 individuals in 15 villages in south-western uganda attempted to examine changes in household structure over six annual survey rounds and to relate them to the progress of HIV infection. The proportion of households headed by women remained relatively constant over the six years studied but there was a significant increase between rounds one and six in the skipped generation structure, in which households tend to be headed by grandparents and include grandchildren but few adults in the economically active age brackets. These skipped generation structures have the highest 'dependency ratio' - that is, ratio of elderly and young people to those of economically active age. Participants noted that in a heavily HIV-affected area, classic measures of dependency which look just at age may underestimate the burden on household resources of sick and unproductive adults.
Experience of HIV in the household is likely to affect peoples perception of risk, so it was interesting to see quantified the difference between people affected and households affected. In a study of rural villages in ugandas Rakai district, Joseph Konde-Lule showed that where HIV prevalence was 11 percent, over 20 percent of households were affected by HIV. Prevalence was higher among household heads than the general population - in trading centres where 35 percent of all adults were seropositive, infection among heads of household ran at 43 percent.
A descriptive survey of widowhood by James Ntozi painted a sorry picture of the lot of ugandan women who had lost their husbands. Widow-inheritance is clearly likely to spread HIV unless it can be dissociated from sexual relations. But attempts to educate people about the dangers of widow inheritance is likely further to stigmatise widows, and may also contribute to their being denied access to family resources. It was not clear how these problems might be resolved.
Back to the MRC cohort of rural ugandans for an assessment of HIV-related mortality over six annual survey rounds. Taking the indicator closest to the heart of most policy-makers - life expectancy - Jimmy Whitworth showed a dramatic reduction of 16 years in life expectancy at birth among the HIV positive, compared with those not infected. It was noted that the life expectancy among the seronegative (60.5 for women and 56.5 for men) was remarkably low for a rural population in East Africa. It is possible that those who contract HIV are prone to behaviours that would put them at higher risk of ill health and death even in the absence of HIV.
Whitworth and colleagues estimated the HIV attributable mortality fraction for adults at 41 percent. In women between 20 and 24 it rose to 70 percent.
In a rural cohort of 20,000 in north-western Tanzania, Ties Boerma and colleagues found that although HIV prevalence among adults between 15 and 44 was under seven percent, over a third of all adult deaths were classified as HIV/AIDS related. The next most prominent killer, diarrhoea, trailed far behind, causing just over six percent of deaths. At these levels of HIV prevalence, Boerma and colleagues calculated the probability of dying between the ages of 15 and 60 at 42 percent for men and 37 percent for women - about five times the likelihood commonly recorded in industrialised countries.
In the early days of the epidemic before sero-surveillance systems were instituted, the course of the epidemic was even more difficult to discern than it is now. Looking at death registries in Côte dIvoire, Michel Garenne and colleagues noted a departure from stable or declining trends in age-specific death rates from 1986. Attributing to AIDS the deaths in excess of those that would have been expected had the established trends continued, Garenne and colleagues attempt to reconstruct the dynamics of HIV infection by fitting mathematical equations to the shape of the mortality curves. The best fit gave a peak in infection in men between 1987 and 1988. Women showed two peaks - one in 1985 and one in 1991. Garenne posited that the first female peak represents the infection of sex workers. Those sex workers infected their clients, accounting for the second peak, who in turn passed the virus on to women in the general population, producing a second peak of infection in women.
This and the following paper presented by Immaculate Nakanaabi illustrated the difficulties of classifying deaths in an age of AIDS. Retrospective verbal autopsies are unlikely to lead to correct classification of deaths that occurred several years previously, particularly when the disease of interest was at the time little known and has since become heavily stigmatised.
Simply ascertaining levels of adult mortality with any degree of accuracy has long proven a challenge for demographers in several Sub-Saharan African countries. Ian Timaeus presented an evaluation of the orphanhood method in the context of stable levels of HIV in a rural population. He found that with fairly simple adjustments made to compensate for the likelihood that seropositive mothers will transmit HIV to any future children and that the child would therefore not survive to report orphanhood, the method is fairly robust.
Although curiously scheduled in a session on orphanhood, George Bicegos paper on the use of national survey data to estimate mortality dealt squarely with measurement issues. The sisterhood method was examined as an indirect method of estimating adult mortality. Although it has been found wanting as a method of assessing maternal mortality, Bicego argued for questions on sibling mortality to be retained in Demographic and Health Surveys as a more general tool for estimating adult mortality. DHS sibling survival data gathered in uganda was compared with data from longitudinal studies. Although actual levels of mortality differed, the same gradient in mortality from high to low prevalence areas was visible in survey as in longitudinal data.
Since HIV is incurable, its prevalence is determined by its incidence and the survival time between infection and death. John Blacker and Basia Zaba began from this premise to try and model the likelihood that an individual would die from HIV-related causes, given a certain level of prevalence. using data from Kenya, they estimated that where prevalence is under eight percent (which implies that one person in 13 is seropositive) one person in three will die of HIV-related causes. Participants discussed the difficulties of using single sex closed population models to investigate HIV, which is transmitted between the sexes and which has various direct and indirect impacts on fertility as well as mortality and therefore contributes to a dynamic population structure.
The interaction of HIV with other diseases further complicates the mortality picture. A case control study conducted in Malawi and presented by Judith Glynn showed that when under three percent of the adult population was infected with HIV, some 17 percent of TB was associated with the virus. But the proportion of TB associated with HIV rises dramatically with HIV prevalence. As HIV prevalence rose towards 11 percent, nearly four TB cases in 10 were related to the virus. The impact of HIV on TB is greater still, since a rise in the number of TB infections (and a lower likelihood that cases will be cured) increases the spread of the disease in the population at large, regardless of HIV infection.
A better understanding of how HIV relates to TB, malaria and other communicable diseases is essential if policy-makers and service providers are to use precious resources to best effect. Conference participants were concerned not just with measurement of these and other demographic effects but with their diffusion. In a presentation focusing on the information needs of civil society as well as government planners, Melody Manyasha called for the collection of more socio-cultural data that would lend context to epidemiological information. Better behavioural and contextual data would at once help us to interpret epidemiological trends and suggest interventions appropriate to a particular social milieu.
One of the social consequences of HIV has been a rise in orphanhood. Several papers attempted to quantify a rise in orphanhood, examine the social coping mechanisms in caring for children without parents and identify the breaking points in those coping mechanisms.
In several rural societies child fostering was shown to be common regardless of orphanhood. Mark urassa presented data from rural Tanzania showing that over a third of children whose parents were alive did not live with both their biological parents, and over one in ten lived with neither surviving parent. Two households in five were home to children who were neither indigenous to the household nor orphans, nearly three times as many as those housing orphans.
In the population studied, orphans and foster children seemed to fare as well in terms of health and schooling as biological children, at least into the teen years when children not living with their birth parents were more likely to drop out of school.
Looking at the sex of the head of household, Helen Aspaas found that discrimination in favour of indigenous children did surface in male-headed households in uganda. Female household heads were more likely to treat orphans and foster children the same as their own children.
Long-established systems of care and fostering centred on the extended family are certainly able to absorb a certain amount of the shock of the increase in orphanhood due to HIV, but falling fertility, urbanisation and the migration of labour, often across borders, is eating into the extended family structures that provide the cushion.
Working at the point where traditional extended family mechanisms become saturated, Geoff Foster documented the appearance of child headed households in Zimbabwe. As noted, the problem is likely to be more acute in more urbanised societies and those where migration is common. And it will become more acute still in the next generation, because the orphans of people who lost their own parents in the early years of the HIV epidemic will have no grandparent generation to care for them. When older children are forced to drop out of school or sell assets or sex to support younger children, they may be increasing their own vulnerability to HIV infection. Institutionalisation is not the answer, but policy-makers should encourage social support for and beyond the extended family system.
Working again from data gathered over five years in a rural ugandan cohort, Alex Kamali calculated that 40 percent of orphans had lost their parents to HIV. Orphaned children were far more likely than non-orphans to be HIV positive. Among the HIV negative, mortality was 30 percent higher for children who had lost a parent than for those with two surviving parents, although because numbers were small the results were not significant.
A review of data available to demographers trying to project the epidemic was presented by Simon Gregson and Basia Zaba. The quality of inputs into models clearly determines in large part demographers ability to produce projections likely to be of practical use to policy makers. And yet it is frequently policy-makers who determine, directly or indirectly, the extent and reliability of the data collected. Sero-surveys at a national, community-based level are unknown in south and eastern Africa. Even national sentinel systems based around ante-natal clinics or blood donors are rare. With reliable estimates of HIV prevalence, demographic models can produce fairly solid projections of AIDS deaths and demographic impact up to around 10 years, because the bulk of deaths over that period will be of people already infected or being infected at current rates of incidence. But as behaviour changes so will incidence; medium and long-term projections depend largely on future trends in behaviour, and it is for policy makers, not demographers, to try to influence what those will be.
As someone with experience of modelling for insurance companies in the private sector, Peter Doyle warned against believing that models were reality. Good modelling ought to provide policy-makers with indications of possible interventions that, if applied, should then alter the outcome of the situation being modelled.
However good a model, policy-makers may choose not to act on it for a number of reasons. Alan Whiteside argued that the more policy-makers are involved in the modelling process - in deciding what questions most need to be answered, in defining realistic inputs etc. - the more likely they will be to act on its outcome. And the more clearly we can define potential points of intervention and demonstrate the effect of acting at that point, the more likely it is that action will be taken.
Following the presentations, a panel discussion resumed the main themes of the conference. Although data is still imperfect; we can work with traditional tools such as indirect estimation to make the best of it. Clear presentation to policy makers of what we do know and what it implies is a fundamental responsibility of a researcher. Researchers need to think more about the end users of their data; by working more closely with other disciplines to help interpret prevalence data epidemiologists can contribute more to the search for interventions.
During the final afternoon of the conference, participants split into two groups, one dedicated to PUBLIC policy, the other focusing on private sector, development and donor needs. It was here that the specific needs of South Africa became most apparent. Both fora concluded that more information on the course of the epidemic in highly mobile urban populations was needed. Both felt that dissemination of research and information would help pressurise governments into recognising the scale of the HIV epidemic and the social and behavioural forces that shape it. There is a desperate need to think ahead about the capacity of current social structures to cope with problems of care, of service provision, of funding, of orphanhood. Demographers can and must contribute to the process of planning both to prevent and to cope with the continuing spread of HIV.