Migration and HIV/AIDS in South Africa

Abstract

Although migration and HIV/AIDS have been examined separately in South Africa, researchers are still far from understanding in detail just how and to what extent the two are interconnected. The connections are difficult to unravel because HIV/AIDS arrived in the country at a time when population mobility and systems of labour migration were undergoing rapid transformation. Without a proper understanding of the complexity of the country's new migration regime, it will not be possible to understand either the role of mobility in the spread of the pandemic or the vulnerability to infection of mobile populations. This paper reviews the current state of knowledge on the interconnections between mobility and HIV and argues for more research that will further understanding of migrant vulnerability and the development of appropriate policies and models of intervention and care.

1. Introduction

Research in Africa has long demonstrated that the prevalence and patterns of spread of infectious disease are closely associated with patterns of human mobility (Stock, 1977). With regard, specifically, to sexually transmitted infections (STIs), a seminal study on HIV and migration in Uganda, for example, showed a strong correlation between HIV infection and migration status (Nunn et al., 1995, 1996). People who had moved within the previous five years were three times more likely to be infected with HIV than those who had lived in the same place for more than ten years. People who had moved frequently had more sexual partners, on average, than those who had moved less frequently. In another important study of seasonal migration and HIV in Senegal, the authors concluded that HIV was ‘mainly transmitted first to adult men through sexual contacts with infected women met during their seasonal migration and second to their wives or regular partners once they are back home’ (Pison et al., 1993). This study suggests that the predominant pattern of spread is from returning male migrants to their rural partners and similar findings have been reported from Ghana and Mexico (Decosas, 1996; Santarriaga et al., 1996).

In a study of male factory workers in Zimbabwe, HIV positive men were more likely to live apart from their wives and to have multiple sexual partners than HIV negative men (Bassett et al., 1990). In Ghana, Anarfi has indicated that migration ‘acts to increase the extent of sexual networking’ (Anarfi, 1993). In a rural community in KwaZulu-Natal, people who had recently changed their place of residence were three times as likely to be infected with HIV as those who had not (Abdool Karim et al., 1992). In Kenya, a recent study concluded that ‘migration is a critical factor in high-risk sexual behaviour and that its importance varies by gender and direction of movement’ (Brockerhoff & Biddlecom, 1999). A recent HIV/AIDS risk assessment of Lesotho and Swaziland clearly demonstrates that migration has played a major role in the dissemination of HIV/AIDS. The assessment also highlights various ‘migrant sites’ of high vulnerability (Wilson, 2001).

There is also considerable evidence that migrancy plays a key role in increasing susceptibility to HIV, sexually transmitted infections (STIs) and other opportunistic infections (for example, Abdool Karim et al., 1992; Quinn, 1994; Nunn et al., 1995; Decosas et al., 1995; Colvin et al., 1995, 1998; Brockerhoff & Biddlecom, 1999). Migrant men are more likely to have casual partners and to engage with sex workers, many of whom are themselves migrants (Carael et al., 1995; Hunt, 1989). Research has tended to show that people who migrate in search of work are at risk in their places of work or that people who are highly mobile, such as truck drivers, are at high risk of contracting STIs themselves or that people living close to trucking routes are more likely to be infected with STIs than those who live further away (for example, Carswell et al., 1989).

Decosas & Adrien suggest that the association between migration and HIV is more likely to be a consequence of ‘the conditions and structure of the migration process than the actual dissemination of the virus along corridors of migration’ (Decosas & Adrien, 1997). They argue that a focus on the routes of spread tends to direct attention to the migrants themselves rather than the socio-economic context of migration. This in turn can easily lead to policies aimed at restricting movement and stigmatising vulnerable groups of people. Indeed, ‘the fact that population movements distribute HIV is secondary to the fact that certain types of migration cause HIV epidemics’ (Decosas et al., 1995). Without a proper understanding of the social, behavioural and psychological consequences of particular forms and patterns of migration, it will not be possible to understand the consequences of migration for the spread of HIV and the vulnerability to infection of mobile populations (see Haour-Knipe & Rector, 1996).

To effect this conceptual refocus on ‘the social (and sexual) disruption’ that accompanies various types of migration and forms of mobility, a number of reorientations are required. First, there needs to be a much more detailed and nuanced understanding of migration itself, of the complex and dynamic spatial and temporal patterns of migrancy in its different forms. Second, building upon this analysis, we need to understand the particular vulnerabilities of migrants as migrants (and those with whom they interact) and hence the economic, social, sexual and gender regimes associated with migrancy in its many different manifestations. Third, as generic HIV/AIDS interventions seem to be having so little impact in migrant settings and situations of high mobility, there is a need for models of intervention that are sensitive to the modalities of mobile people. Finally, as attention is increasingly directed towards models of care, there is a need to develop interventions appropriate to the situation of migrants and their divided households.

After reviewing the general evidence on the connections between HIV/AIDS and migration, this paper seeks to move the South African debate from the macro- to the micro-scale. Our basic premise is that none of the above objectives can be adequately reached without attention to the micro-geographies of mobility, social connectivity and sexual behaviour. When the findings of research in three disparate settings are reviewed, the complexity of the connections between migration and HIV/AIDS begin to emerge. But while each local situation has its own distinct characteristics, it does not follow that lots more local studies are necessary before we can say anything ‘representative’ or make sound recommendations for workable prevention strategies and models of care. The case studies presented here represent three instances of a more general set of like situations across the country. The case study areas are spaces of vulnerability, places in which to observe why migrants and those with whom they come into contact are highly susceptible to HIV infection, and hence to develop approaches to decreasing this vulnerability. If workable interventions, based on a sound understanding of local regimes of migration and sexuality, can be developed in disparate case studies such as these, then such best-practice models could have much wider purchase for resisting the ravages of the epidemic.

2. Connecting migration and disease

In South Africa, the precise linkages between migration and the health of migrants (as well as those with whom they come into contact) are complex and difficult to unravel. The main emphasis in migration studies has been on the determinants of migration rather than the health consequences of migration (Crush et al., 1991; McDonald, 2000). Studies linking health and migration have focused mainly on migrants at their place of work, especially on the gold mines, and to a much lesser extent on rural labour-sending areas (Leger, 1992; Packard & Coetzee, 1995). Work on the relationship between vulnerability to infection and the whole process of migration is comparatively new (Campbell 2003; Cohen 2003).

Historically, long-distance migrancy to the mines and towns was largely male. To ensure that a regular and regulated supply of labour was maintained, a regional system of circular migration was developed by the Chamber of Mines of South Africa and enforced by the apartheid state and neighbouring colonial administrations. The Chamber of Mines set up an employment agency, known since 1976 as The Employment Bureau of Africa (TEBA), which recruited men from other countries to work on the mines. TEBA had the right to recruit foreign workers as well as South Africans, but without giving the former rights of permanent residence.

To ensure that migrant miners did not remain in the mining centres of South Africa they were prohibited from bringing their families with them and were given annual contracts which obliged them to return home once a year with no guarantee of continued employment. As a result, the pattern of migration that developed early in the century, and persists to this day in the mines, was predominantly that of circular migration in which men leave their partners in rural areas in search of urban employment, and return home as frequently as money, contracts and distance allow. This form of migration had the added benefit, from the point of view of the mining industry and the state, that the rural areas of Southern Africa provided a form of social security for the men on the mines where the costs fell entirely on rural families.

The size of the mine workforce fluctuated over time, as did the numbers of mine-workers from any one area. However, the proportion of foreign miners on the gold mines never dropped below 40 per cent (Crush et al., 1991). In the early 1970s, it rose as high as 80 per cent and is currently around 55 per cent (Crush et al., 2001). Mozambique, Malawi, Swaziland and Botswana remain the primary foreign suppliers. Malawians had a significant presence on the mines until the late 1980s (Chirwa, 1995, 1998). In 1987, workers in the South African gold mines were tested for HIV and correlations were made between HIV status and country of origin. The prevalence among mine-workers from Malawi was 4 per cent; among those from other countries, including South Africa, it was about 0.03 per cent (Brink & Clausen, 1987). The Chamber of Mines stopped recruiting in Malawi when the Malawian government refused to allow testing of migrants on recruitment (Chirwa, 1995). The number of Malawians employed on the South African mines dwindled from 13,000 in 1988 to no more than a handful by 1990 (Crush et al., 1991).

It once seemed relatively easy for researchers to discern clear causal linkages between this system of oscillating migration and the prevalence and spread of disease. In the late 1940s, Kark suggested that the widespread prevalence of gonorrhoea and syphilis in both urban and rural areas of South Africa was a consequence of the migrant labour system (Kark, 1947). Jochelson has recently affirmed this argument, suggesting that the migrant labour system ‘facilitated the transmission of venereal and endemic syphilis to new regions, and into communities without previous exposure to the disease’ (Jochelson, 2001; see also Setel et al., 1999). Wilson argued more broadly that the system of housing male migrants in hostels away from their wives and families led to massive social and health dysfunction including ‘family break-ups, bigamy, prostitution, alcoholism, violence, corruption, venereal disease, tuberculosis and malnutrition’ (Wilson, 1972). The migrant labour system clearly increased the vulnerability of migrants to infection and greatly facilitated the spread of sexually transmitted, as well as other, infectious diseases.

Throughout the 20th century, the migrant labour system played a central role in the spread of infectious diseases such as tuberculosis and STIs such as syphilis and gonorrhoea. The system also made miners and their dependents particularly vulnerable to infectious and occupational disease. The interconnections between migrancy and disease have been noted in several other studies. Packard (1987), for example, details the way in which the migration of men between urban and rural areas led to the spread of tuberculosis in South Africa during the early part of the 20th century. Other studies have revealed how vulnerable the current generation of migrant miners are to occupational lung disease (Williams et al., 1998). Many have returned home, placing very severe social and economic burdens on individuals, households and the migrant labour-sending communities as a whole. In the Eastern Cape, one study found that 24 per cent of ex-miners were eligible for compensation. The mining industry paid out approximately US$500 000 in unpaid compensation as a direct consequence of this study (Trapido et al., 1998). Legal efforts are now being made to ensure that the remaining one to two million ex-gold mine-workers, from the neighbouring countries as well as South Africa, receive the compensation to which they are entitled under law.

The connections between migrancy and disease are now more difficult to unravel. In part this is because HIV/AIDS arrived in the region at a time when population mobility and systems of migrant labour were undergoing considerable change. Migrancy is, by its very nature, highly dynamic and has changed dramatically in scope, scale and diversity over the last two decades. Because of the complexity and diversity of forms of contemporary mobility, it is much more difficult to ‘map’ the prevalence and spread of disease onto spatial patterns of migration than it was, say, two decades ago (Kalipeni et al., 2004).

Several important migration changes coincided with the advent of HIV/AIDS in the last 20 years. First, the collapse of apartheid brought new opportunities and reasons for migration across borders within the region. Migrants from neighbouring countries and further afield see South Africa as a new place to trade, shop, access essential services, work and seek asylum. The number of people crossing South Africa's borders in both directions has burgeoned since 1990 to over seven million a year (Crush, 1999). So, too, has South Africa's formal trade with the sub-continent, with goods being carried in the main by long-distance truckers. Informal sector cross-border trading has also expanded dramatically since the end of apartheid. Trading is highly gendered, with women playing a major role in the buying and selling of goods across international boundaries throughout the region (Peberdy & Rogerson, 2000).

Second, there has been significant growth in the levels of urbanisation in South Africa. Internal migration, primarily oscillatory in nature, has grown significantly since the ending of influx controls and the pass laws in the mid-1980s. At the time of the 2001 Census, for example, as many as 40 per cent of the population of Johannesburg was made up of migrants from outside Gauteng (Peberdy et al., 2004). Other parts of South Africa have experienced similar growth in internal migration (Bekker, 2001; Kok et al., 2003; Posel & Casale, 2003). One consequence has been the displacement of the rural poor to the towns, as manifested in the ubiquitous informal settlements that dot the urban landscape (May, 2000). What is clear is that many of the new urban residents retain rural linkages and order their lives around the two poles of the ‘stretched household'. The role of oscillating migration in the diffusion of HIV/AIDS and the vulnerability of internal migrants to infection has been suggested but not yet systematically explored (Bekker & Swart 2003; Booysen, 2003; Collinson et al., 2003; Lurie 2003).

The third significant shift has been in the gendering of migrancy. Women are becoming considerably more mobile, migrating for formal and informal work in ever-growing numbers and travelling more frequently for a variety of social and other reasons (Dodson, 2000). A recent cross-sectional survey carried out in a rural area of KwaZulu-Natal found that about one-third of adult women were migrants at the time of the survey but, unlike their male counterparts, female migrants tended to stay closer to home, and were therefore able to return home more frequently (Lurie et al., 1997a, b). In Lesotho, retrenchments on the gold mines have led to an upsurge in migration by women seeking work on South African farms and the new textile factories of Maseru (Ulicki & Crush, 2000).

The greatest degree of continuity with the past is probably in the mining industry, which persists with its regional contract labour system. In the 1990s, the South African gold mining industry employed half a million migrant mine-workers from rural areas in South Africa, Botswana, Mozambique, Lesotho and Swaziland. The number of men living in single-sex hostels has dropped slightly as the mining industry experimented with the provision of married quarters, but in the late 1990s, 89 per cent of miners still lived in single-sex hostels and only 2.1% per cent lived in married quarters. The rest found their own accommodation in the local communities (Crush, 1995). In 1995 the assessors of a judicial commission of enquiry, set up by the government and chaired by Justice Ramon Leon, to investigate the regulation of occupational health and safety in the mining industry, visited three hostels on two mines. Each room was occupied by between 12 and 20 men, giving an average of just over five square metres per person. The assessors were ‘shocked by the conditions in which food was prepared’ and by ablution facilities ‘so squalid as to shock the most hardened’ (Leon et al., 1995). With men living in single-sex hostels, many away from their wives and families for months at a time, they are highly prone to engage in sexual behaviours with women from neighbouring townships that increase the vulnerability of both to STIs.

The spatial perimeter of hostels has become much more permeable than formerly, facilitating higher levels of social contact between migrants miners and people living near the mines. As well as those living in settled urban communities around the mines there are numerous ‘hotspots’-informal settlements where women come to sell alcohol and sex (Campbell, 2003). Developments in the transportation industry (particularly the deregulation of the taxi industry) and the removal of political barriers such as pass laws and influx controls have also had a significant impact on the behaviour patterns of migrant mine-workers, making them much more mobile than in the past. While Mozambican migrants still follow the old pattern and return home once a year, migrants closer to home have effectively become long-distance commuters. In Lesotho, for example, 60 per cent of migrant miners now return home at least once a month (Sechaba Consultants, 1997). On one hand, this could lessen the pressures for infidelity and family breakdown in the future. On the other hand, it means that the rural areas are far less ‘insulated’ than they were from a virus such as HIV.

Accompanying this shift in migrant behaviour, there has been a marked upsurge in the informal movement of impoverished rural women to the mining areas from within South Africa, particularly the Eastern Cape. Accompanying these women are many so-called ‘country wives’ who seek to maintain relationships and access to mine wages by moving to town. As Moodie (1995) points out, ‘if policy makers are worried about the instability of marriages of black mine-workers, the most constructive approach would be to extend the right of mine visits to foreign wives and girlfriends as well’. Migrant women are often forced to the margins of the local economy. Common income-earning strategies include brewing and commercial sex work, often in combination. Others enter longer-term, more stable relationships with migrant miners (Lubkemann, 2000). A recent study of women who self-identified as migrants in Carletonville showed that they were at significantly higher risk of being infected with HIV than women who did not migrate. Indeed, most of the commercial sex workers in one of the ‘hotspots’ in Carletonville were migrants from rural areas in South Africa or neighbouring countries (Zuma et al., 2003).

In spite of the attention paid to the gold mines, the majority of migrant workers are employed in other sectors. Migrants work in manufacturing, agriculture, construction and services and domestic service. While single-sex hostels are discussed almost exclusively in relation to gold mines, they are common outside of the mining industry as well. Within a 12 km radius of Durban, for example, there are at least seven men's hostels with an excess of 43,000 officially registered beds, one women's hostel with over 1,000 beds and one mixed hostel with 11,000 beds.

The particular vulnerabilities to HIV of people (migrant and non-migrant, mobile and relatively immobile) associated with South Africa's changing regime of migrancy are currently poorly understood. The evidence suggests that migrants and their households in town and countryside are particularly at risk. So, too, are the residents of non-migrant communities with whom migrants interact on a daily basis. They offer to migrants social interaction and escape from the drudgery of work (Campbell, 2003). Many of these impoverished non-migrant communities have high rates of unemployment and interaction with migrants represents their only means of access to wage income.

An emerging issue requiring far more attention is the way in which the pandemic is itself generating new forms of migration (Dodson & Crush, 2004). This includes increased rural–urban migration as a result of decline in agricultural production; the movement of healthy migrants to take up job vacancies caused by AIDS deaths; the movement of infected people to be closer to health-care facilities or providers; and the return of sick people to rural homes for support. To date, most work has been carried out (particularly in Lesotho and Malawi) on the migration of children who have been orphaned by the epidemic (Urassa et al., 1997; Young & Ansell, 2003; Ansell & Young 2004; Ansell & Van Blerk, 2004). This phenomenon is predicted to grow substantially over the next decade.

3. Macro-geographies of HIV

One area requiring further exploration is the role of these new forms of mobility in the astonishingly rapid dissemination of HIV in South Africa. As a starting point, this section of the paper documents the macro-geography of the HIV/AIDS epidemic in South Africa in the 1990s.

Data on the extent and patterns of migration in South Africa at a macro-level is generally poor (Swanevelder et al., 1998; Williams et al., 2000a, c). However, certain general insights can be gleaned from the demography of the provinces structured by age, race and gender (SSA, 2000; Udjo & Hirschowitz, 2000). In most provinces the population is overwhelmingly African; only in the Northern and Western Cape is it predominantly coloured. The Northern and Western Cape and also Gauteng have a significant proportion of whites; most Indian people live in KwaZulu-Natal.

Census data suggest that significant numbers of African adults aged between 20 and 50 years are leaving the Eastern Cape and the Northern Province and in both cases there are more men leaving than women. Similarly, in KwaZulu-Natal a deficit of African men compared to women between the ages of 20 and 50 years reflects differential out-migration by men. Many of them are clearly migrating to Gauteng, giving rise to a peak in the age distribution for men between the ages of 20 and 50 years in that province, and to a lesser extent to the Western Cape and the Free State. The legacy of apartheid in the Western Cape has meant that the province has been relatively isolated from the perspective of migration but it has become increasingly integrated with the Eastern Cape.

To fully understand the impact of migration on the spread of HIV, data on the places of origin and destination of the migrants are needed on a much finer geographical scale. Given that HIV/AIDS will kill mainly young adults the impact is likely to be felt most keenly in Gauteng, although if people go home to die their deaths may occur in other provinces.

The prevalence of infection in each of the provinces is shown for 2001 in Figure 1. Trends in the data from antenatal clinics for each province suggest that prevalence will not increase substantially above current levels. The prevalence in the Western Cape is likely to remain significantly below the current level in the Eastern Cape and the prevalence there below that of KwaZulu-Natal (Williams & Gouws, 2001). These trends will become clearer over the next few years. In Figure 2 the provincial prevalence data for 1998 was combined with census data for the population density for the country to show the geographical distribution of infected people. A similar exercise could be attempted now with more recent census data but was not considered necessary to make the general point about intra-province unevenness in prevalence. The intrinsic doubling times at the start of the epidemic varies from 11 months in the Western Cape to 15 months in KwaZulu-Natal. It is possible that the initial epidemic in KwaZulu-Natal spread relatively slowly but once it was established there, migration within South Africa led to more rapid growth within the other provinces even though they will probably peak at lower prevalence rates than KwaZulu-Natal (Coffee et al., 2000; Williams & Gouws, 2001). A recent study confirms these observations and suggests that HIV prevalence has now reached a peak with about 4.69 million infected people (HSRC, 2002; Rehle & Shisana, 2003). The incidence has decreased substantially in the past five years, but simply because a high proportion of those at high risk are already infected. Their model suggests that the annual number of deaths due to AIDS will peak at just less than half a million in 2008.

Figure 1: Prevalence of HIV infection (%)

Figure 2: Distribution of people with HIV, 1998

The HIV data highlight the significance of migration in spreading disease over wide geographical areas but also between urban and rural communities in Southern Africa. By analysing a range of age-specific prevalence data, it is possible to identify four different patterns of infection in South Africa among (a) women attending antenatal clinics; (b) women in the general population; (c) men in the general population; and (d) male migrant workers (Williams et al., 2000a, b). Within each of the four groups represented by the different patterns, the shapes of the age prevalence curves have not changed significantly over time, suggesting that the prevalence is increasing at the same rate for people of all ages and that the shape of the age-prevalence curves gives a reasonable estimate of the variation of the relative age-specific risk of infection.

The difference between the patterns of infection in women attending antenatal clinics and women in the general population is slight for two reasons (Williams et al., 2000a, b). Because young women who attend antenatal clinics must have been sexually active, the antenatal clinic data may overestimate the prevalence of infection among young women. On the other hand, because older women who are at high risk of HIV are also at high risk of contracting syphilis, which may lead to abortions, or gonorrhoea, which may render them sterile, the antenatal clinic data may underestimate the prevalence of infection among older women.

While the prevalence of infection gives a measure of the cumulative risk of infection, the incidence of infection gives the rate at which new infections are acquired and is therefore a more sensitive measure of the growth of the epidemic. It is, however, more difficult to measure incidence than prevalence. Ideally, one would follow a cohort of people for a year or more and determine the number who became infected with HIV during the course of the study. Because of the cost, time and ethical difficulties associated with cohort studies, two alternative methods have been developed for estimating incidence.

The first method estimates incidence by combining age-specific estimates of the current prevalence and of the time trend of overall prevalence for the preceding ten or so years (Williams et al., 2001). The second method uses a so-called de-tuned or sensitive-less sensitive (SLS) enzyme-linked immunosorbent assay (ELISA) test, which relies on the fact that HIV antibody levels build up over several months after infection so that people who have recently been infected with HIV have low antibody concentrations (Janssens et al., 1998).

Both these methods have been used to estimate the incidence of HIV-infection among women in the Hlabisa district of Kwazulu-Natal, as shown in Figure 3 (Gouws et al., 2002). The two independent methods confirm that 25-year-old women in rural Hlabisa, a source of migrant labour, have a 27 per cent risk of being infected with HIV each year. A similarly high incidence of 18 per cent per annum was found in a cohort study of sex workers operating at truck stops in the KwaZulu-Natal Midlands between 1996 and 1999 (Ramjee, 2000).

Figure 3: Incidence if HIV infection, Hlabisa, 1998 (antenatal clinic data)

In order to estimate the likely impact of HIV on the demography of South Africa we need to know the incidence of infection among people according to race, age, gender and province. Unfortunately, there are no accurate data available that make it possible to compare directly the prevalence of infection among people of different races. The blood transfusion services collect data by race (Table 1) but exclude those who it regards as being at high risk, so that the sample is extremely biased. Rates among black women donating blood was only about one-fifth of the rates seen in women attending public antenatal clinics in the same year.

Table 1: Prevalence of HIV infection in people attending blood transfusion services, 1998 (first-time donors only)

	African (%)	Coloured (%)	Indian (%)	White (%)
Male	4.134	0.213	0.203	0.077
Female	6.738	0.588	0.080	0.052
Total	5.397	0.391	0.161	0.065
Ratio (m/f)	0.61	0.36	2.54	1.18

The antenatal clinic surveys include the race of the participants but in the 2001 survey; for example, there were only 74 Indian and 95 white women. The sample is too small for prevalence in these groups to be assessed. Currently the best source of data in this regard is from a study carried out by the Human Sciences Research Council. In this study, the prevalence of infection in black, coloured, white and Indians adults was 19 per cent, 7 per cent, 6 per cent and 2 per cent, respectively (HSRC, 2002).

The impact of HIV/AIDS on mortality is felt among women at younger ages than among men. For black men the impact in the Eastern Cape and KwaZulu-Natal will be mitigated by the fact that so many men between the ages of 20 and 50 have already left those areas to work as migrants in Gauteng and the Free State. More women than men will die and South Africa faces the prospect of a continuing rise in the number of older children and adolescents, with a dramatic decline in the number of parents.

The overall patterns of African population decrease were predicted using a statistical model developed by two of the authors (Williams and Gouws) (see Table 2). In relation to KwaZulu/Natal, for example, Table 4 shows that in 2005 there will be about 150,000 fewer children, 400,000 fewer adult women and 200,000 fewer adult men than would have been the case without AIDS. The provision of anti-retroviral therapy could largely eliminate mother-to-child transmission (Coutsoudis, 2000; Wilkinson et al., 2000; Abdullah et al., 2001). Without AIDS the population would be expected to increase by approximately one million people over the next five years; AIDS will reduce this to 250,000. The biggest impact will be in KwaZulu-Natal, which has the highest overall prevalence, and in Gauteng, which has a high prevalence but also a very high proportion of people between the ages of 20 and 40 years, precisely those most likely to be infected and subsequently to die.

Table 2: Predicted decrease in African population (in thousands)

Province	Children (male or female)	Adult females	Adult males
Eastern Cape	35.6	142.8	63.0
Free State	12.0	107.9	74.8
Gauteng	53.6	291.4	227.2
KwaZulu/Natal	152.8	414.2	206.4
Mpumalanga	27.2	110.6	52.0
Northern Cape	1.6	9.4	3.4
Northern Province	12.0	75.4	33.8
North West	36.4	114.4	77.2
Western Cape	1.0	11.0	7.8
Total	332.2	1 277.1	745.6

A recent study in KwaZulu-Natal has shown that mortality began to increase sharply in the late 1990s and AIDS is now the leading cause of death for people between the ages of 16 and 60 years (Hosegood et al., 2004). A review of the impact of AIDS on adult mortality in over 20 sub-Saharan African countries showed that most had experienced increases in mortality in the 1990s and this was associated with HIV (Blacker, 2004).

The immediate impact of AIDS will be felt through the direct effects of deaths on people, their families and society. The greater long-term impact of AIDS is likely to be felt through changes in the age structure of the population. We see that, in the model, the number of young people in KwaZulu-Natal aged 0 to 20 will increase by 340,000 while the number of adults aged 20 to 50 will decline by 200,000.

For African people the overall patterns of change are broadly similar but the probable difference in impact between the high- and low-prevalence provinces is seen if we compare KwaZulu-Natal and for the Western Cape. The impact of the epidemic is by no means negligible in the Western Cape, but the adult numbers will only be 20,000 less than expected without AIDS while the equivalent figure in KwaZulu-Natal is approximately 600,000. Even allowing for the fact that KwaZulu-Natal has eight times as many black people as does the Western Cape, the model suggests that the effect in KwaZulu-Natal will still be four times greater than in the Western Cape.

The above analysis provides macro-level insights into the geography of HIV prevalence and incidence and, by extension, the geography of vulnerability. As the data improve so too will the reliability of the projections. However, it is clear from this preliminary exercise that spatial patterns of prevalence and vulnerability are linked with the mobility of the population in complex ways. To understand the nature of these linkages, it is necessary to move from the macro- to the micro-scale. The next section of the paper therefore focuses on three contrasting migration situations which have been examined in some depth. Extensive biomedical, behavioural and socio-economic studies have been carried out at two sentinel sites, one in Hlabisa, a rural district of KwaZulu-Natal, the other in Carletonville, an urban gold-mining centre in Gauteng. Data have also been collected at STI, family planning and tuberculosis clinics, as well as referral hospitals and workplaces. Particular risk groups that have been studied include sex workers and truck drivers. These studies provide considerable insights into the connections between migrancy and HIV/AIDS at the local level.

4. Micro-geographies of HIV

4.1 Compounding the epidemic

Carletonville is the biggest gold mining complex in the world. The mines employ approximately 70,000 men, almost all of them migrants from rural areas, living in single-sex hostels without their wives or families. The Mothusimpilo (‘Working-together-for-health’) Project was designed to demonstrate the feasibility of using sustainable, community-based interventions to reduce the transmission of HIV/AIDS (Campbell, 2003). The location of Carletonville is indicated in Figure 2 and Figure 4 provides a more detailed map of the area. The most important urban settlement in the district is the township of Khutsong.

Figure 4: Carletonville District

Working on gold mines is extremely dangerous; miners suffer from high rates of debilitating and potentially fatal diseases such as tuberculosis and silicosis. Accidents, including rock-bursts, are common and men regularly face the prospect of breaking their backs or being killed. Campbell (1997, 2000a) argues that in order to deal with the psychological trauma of life on the gold mines with little or no social and emotional support, men develop a culture of masculinity that involves drinking and engaging in frequent commercial sex which enables them to face the physical and emotional trauma associated with the work that they do. Miners are socially isolated and there are few opportunities for any kind of human contact or intimacy in the hostels in which they live. Consequently, with alcohol and intercourse with sex workers the only means of ‘entertainment', the resulting patterns of HIV infection are inevitable (Ijsselmuiden et al., 1990; Moodie, 1994). Campbell argues further that norms of masculinity that enable miners to cope with and to survive the lonely, harsh and dangerous conditions of work on the mines, also reinforce a ‘macho’ sexuality which increases their likelihood of being infected with HIV (Campbell, 1997, 2000b).

Miners earn an average of about US$250 per month so that the monthly wage bill in the area is almost US$20 million per month. If, for example, 10 per cent of this is spent on alcohol and sex this will give rise to an informal sector industry worth approximately US$2 million per month in a community where only 53 per cent of the men and 24 per cent of the women are employed (Gilgen et al., 2001). Women migrants from rural areas in South Africa, and from neighbouring countries, are attracted to places such as Carletonville. Unsurprisingly, given the limited employment opportunities for women in a mining town, Carletonville has a very substantial informal sector providing alcohol and sex to mine-workers. Much of the sex work takes place in what are referred to locally as ‘hotspots’ (informal camps close to mine compounds). This industry is unregulated, still illegal, and the women have little access to health services.

In August 1998 a confidential survey was conducted on random samples of (a) 1,185 men and women aged 13–59 years stratified by housing type in Khutsong, (b) 899 mine-workers stratified by hostel and (c) 145 sex workers (Williams et al., 2000c). Blood and urine samples were tested for HIV, HSV, syphilis, gonorrhoea and chlamydia, and participants answered a modified version of the UNAIDS four-centre study questionnaire. The questionnaire included sections on demography; social factors including education and income; knowledge of STIs including HIV; attitudes towards HIV and to people infected with HIV; details of sexual practices and networking; and social capital. The survey was repeated in August 1999 but the Khutsong sample was limited to those aged 15–25 years to obtain more precise information on young people.

Gender differences in the age-specific prevalence are striking when comparing young men and women in the Carletonville region (Figure 5). The prevalence of infection is close to zero for both sexes before the age of 15 but increases rapidly thereafter, reaching 39 per cent in 20-year-old women but only 8 per cent in 20-year-old men. The peak prevalence is 58 per cent at 24 years of age among women and 45 per cent at 32 years of age among men. The median age at first sex is close to 16 for both men and women in Carletonville and the prevalence of infection increases rapidly thereafter. Older men and women are probably unlikely to be having high-risk sex and when they were young there was little or no HIV, so the fall in prevalence among older men and women is expected.

Figure 5: Age and gender prevalence of HIV, Carletonville and KwaZulu-Natal

While there are important gender differences between men and women, the shape of the age-prevalence curves have not changed over time and are similar in urban and rural settings. Figure 5 shows the age-specific HIV prevalence for rural men and women in KwaZulu-Natal in 1991 and for urban men and women in Carletonville in 1998. Although the overall prevalence in the later urban survey is approximately ten times higher than in the earlier rural survey, the shapes of the age-prevalence curves for men and women are statistically the same. This suggests that the shape of the age-specific prevalence curves reflects the age-specific risk of infection, and the differences between the shapes of the age-prevalence curves for men and women are particularly interesting, especially the very rapid rate of increase among young women compared to men. Several factors have been advanced to explain these differences, including female genital mutilation (which is common in parts of Africa but not in Southern Africa) (Brady, 1999; Kun, 1999), rape and coercive sex (Miller, 1999), ‘dry sex’ (Ruganga & Kasule, 1995; Sandala et al., 1995; Beksinka et al., 1999) and the increased vulnerability of women at the age of puberty (Garenne & Lydie, 2001). Perhaps surprisingly, none of these factors has been shown to explain significantly the population levels of prevalence, although they clearly increase the risk for individual people, and the overall differences between the age-specific prevalence of infection in men and women can best be explained in terms of the age matching of sexual partners (women have sex with men who are, on average, five years older than they are) and the fact that female-to-male transmission is approximately three times more likely than the reverse (MacPhail et al., 2002).

Comparing the age-specific prevalence of HIV-infection for miners with men in the surrounding community is especially revealing: from 25 to 55 years of age the prevalence is essentially constant but the decline in prevalence among older men in the community is not seen in mine-workers, suggesting that migrant mine-workers have the same risk of infection at the age of 50 as they have at the age of 25 (Williams & Gouws, 2001). Because the mines offer few opportunities for privacy, intimacy or any kind of social support and there are few recreational facilities, it is not surprising that miners, of all ages, turn to alcohol and sex. It is also interesting to note that the overall level of infection among mine-workers is similar to that of men living in the local community. As the miners are almost all migrants from distant rural areas, many of them will choose to return home to die and as the mines commonly repatriate seriously ill men on medical grounds, it is likely that the levels of infection among mine-workers are kept artificially low. Unfortunately, there are no data on the levels of infection among miners who have been retrenched or repatriated.

4.2 Rural risk

About 215,000 predominantly Zulu-speaking people live in the Hlabisa district of northern Kwazulu-Natal. The location of Hlabisa district is indicated in Figure 2. Homesteads are widely scattered and people depend on subsistence farming, migrant labour and pensions. The South African Medical Research Council (MRC) has conducted research in the area since the early 1990s (Wilkinson et al., 1998). Following an STI survey carried out in 1996, a mass media campaign was developed to increase awareness of and treatment-seeking behaviour for STIs, to strengthen STI case management in the public and private sectors, and to design strategies to reduce STIs among migrant workers and their partners (Wilkinson et al., 1998).

Hlabisa has been identified as a potential site for HIV vaccine trials. A Community Advisory Board was elected in 1998 to promote a partnership among researchers, research participants and community members, and to ensure that the needs and concerns of the local people are properly considered, understood and dealt with (Frohlich et al., 2002). Estimates of HIV prevalence and incidence in the general population have been made, and baseline data collected on key demographic and health variables between 1999 and 2001 (Gouws et al., 2000).

One of the striking features of the epidemic of HIV in South Africa is the lack of the obvious differences in the prevalence of infection between urban and rural areas found in East Africa, for example (Gregson et al., 2001). This is clearly illustrated with reference to data from the Hlabisa district of northern Kwazulu-Natal. Although Hlabisa is a rural area, the prevalence of infection among antenatal clinic attendees increased from 4.2 per cent in 1992 to 34 per cent in 1999, almost exactly matching the prevalence in KwaZulu-Natal as a whole, where antenatal HIV prevalence increased from 4.5 per cent to 33 per cent (Gouws et al., 2000). One of the reasons is the extent to which Hlabisa is integrated into local and long-distance systems of migration.

The extent of migration is clearly shown by the demographic profile of men and women in Hlabisa. The number of women in the community falls rapidly after the age of 17 years (Figure 6) and the number of men even more rapidly, leading to the extraordinarily distorted sex ratio (inset in Figure 6). In Hlabisa the proportion of men falls to about 35 per cent at 32 years of age, and then increases again as migrant workers return home, often with severe and debilitating disease (Wilkinson et al., 1998).

Figure 6: Demographic profile of Hlabisa

Of those who migrate for work from Hlabisa, Figure 7 shows the percentage of men and women measured in a cross-sectional survey who seek work in various destinations. Men tend to migrate further than women, with 32 per cent of men but no women going to Johannesburg and 41 per cent of women and only 8 per cent of men going to nearby Nongoma and Matubatuba. About 60 per cent of adult men and 30 per cent of adult women in Hlabisa spend most nights away from the district (Lurie et al., 1997a, b). A study of the impact of migration on HIV transmission was conducted with men from Hlabisa who work on the gold mines in Carletonville (800 km away) and in factories in Richards Bay (80 km away) and with their partners who stay in Hlabisa. For comparison, a group of non-migrant couples in Hlabisa were also recruited (Lurie et al., 1998).

Figure 7: Female and male migrant destinations from Hlabisa (% of migrants)

The study permits closer examination of the relationship between migration and HIV in the rural areas (Lurie, 2001). Migrant men from Hlabisa district were recruited at two frequent migration destinations, Carletonville and Richards Bay. Later, their Hlabisa-based partners were recruited and, for comparison, a group of non-migrant couples based in Hlabisa were also recruited. Biological and social data were collected from all participants. The prevalence of HIV among migrant men was 25.9 per cent compared to 12.7 per cent among non-migrant men, so that migrant men were twice as likely to be HIV-positive as non-migrant men. The study found that migrant men were significantly more likely than non-migrant men to be infected with HIV. Among women, however, having a migrant partner did not confer any additional risk of HIV infection.

The study also shed light on the question of directionality of spread of the epidemic. It has long been argued that, in relation to migration and HIV, it is migrant men who become infected while away, and then return home to infect their rural partners (Lurie, 2000). The assumed directionality of spread has therefore been from returning male migrants to their rural partners. In this study, migrant couples were significantly more likely to be HIV discordant (one partner infected, the other not) than non-migrant couples and, among these discordant couples, the woman was the infected partner in nearly 40 per cent of the cases (Lurie et al., 2000). These findings certainly illustrate the complex interaction between sexual networks, migration and HIV, for these women could only have been infected by men other than their husbands. Frequently the reason for these sexual liaisons is because in deprived areas women are pushed towards selling sex or exchanging it for favours, as often sex is the only commodity that they have to sell in order to survive (Evian, 1993; Wilkinson et al., 1999; Dladla et al., 2000).

4.3 On the road

The vulnerability to HIV of sex workers and their truck-driver clients has been well documented in Africa (see Burayo et al., 1991; Marcus, 1996; McLigeyo, 1997; Marck, 1999; Mukodzoni et al., 1999; FHI, 2000; Bryan et al., 2002). Truck drivers, owing to the migratory nature of their profession, tend to have multiple partners. Since 1992 the MRC has conducted research among an estimated 800 sex workers operating at truck stops on the national highway in the KwaZulu-Natal Midlands between Durban and Johannesburg (Ramjee et al., 1998). Between August 1996 and June 1998, 477 women from five truck stops were screened for HIV as part of a microbicide trial and in 1998 ten sex workers were trained as fieldworkers to collect socio-demographic data from among their clients and to obtain a saliva sample to determine the HIV status of their clients (Ramjee et al., 1998; Ramjee & Gouws, 2002).

A total of 320 truckers was interviewed. Their average age was 37 years and they had been in the profession for an average of 8 years. Seventy per cent reported having wives or regular girlfriends. As many as 66 per cent of the men reported having had an STI in the previous six months and 37 per cent always stopped for sex along the route. Twenty-nine per cent said that they never used condoms with sex workers, while only 13 per cent had ever used condoms with their wives. Among the sex workers at these truck stops 42 per cent said that they had engaged in anal sex with their clients. Only 23 per cent reported ever using condoms during anal sex.

In terms of the geography of trucking, all the truck drivers had travelled to three or more provinces in South Africa and 65 per cent had been to neighbouring countries (including Zimbabwe, Malawi, Mozambique, Zambia, Botswana, Namibia, Swaziland and Angola). Their high-risk sexual practices and highly mobile lifestyle ensure rapid transmission of the disease along major transportation routes. Some 180 of the truckers (56 per cent) were HIV positive. Prevalence varied between truck stops from 52 per cent to 95 per cent (Table 3). Prevalence rates were also high but varied among sex workers at each of the truck stops. The age-prevalence curves were quite different for truckers and sex workers. HIV prevalence among men increased significantly, with age to a high of 69 per cent among men aged 55–59 years. For the sex workers, prevalence peaked at 20–24 years.

Table 3: HIV prevalence rates at KwaZulu-Natal truck stops (% HIV positive)

Truck stop	Truckers	Sex workers
A	57	44
B	52	42
C	50	62
D	52	74
E	95	64

The high HIV prevalence and low condom use among truck drivers and sex workers, as well as the complex web of travel and sexual mixing, creates a milieu that is highly conducive to the spread of HIV and other STIs. The extent to which they contribute to the spread of HIV throughout the region needs to be examined if the epidemic of HIV is to be contained.

5. Conclusion

Much could be done to reduce the impact and the spread of HIV in South Africa. Mother-to-child transmission could be substantially reduced using standard drug regimens. Control of curable STIs would reduce transmission of HIV. The effective promotion of condoms, which are seldom used at present, and a reduction in high-risk sexual behaviour would have an effect in the longer term.

In all these interventions special attention should be given to people at high risk of infection, which includes not only commercial sex workers, but also migrants and the partners of migrants. Tuberculosis prophylaxis could substantially reduce tuberculosis morbidity and mortality among those with HIV and this is particularly important in the context of gold mining, where the high rates of silicosis and HIV lead to a situation in which the incidence of tuberculosis is about 3,000 per 100,000 men per year, and among men with both silicosis and HIV approximately 16,000 per 100,000 men per year, probably the highest rate ever recorded (Corbett et al., 2004). In November 2003, the National Department of Health in South Africa announced an operational plan to provide comprehensive care, management and treatment for HIV/AIDS in the public health sector, including the provision of free anti-retroviral (ARV) therapy (DoH, 2003). While ARV therapy will significantly reduce AIDS-related mortality, the evidence from places such as Brazil is that it will not reduce transmission itself. Ways will still need to be found to dramatically reduce the rate of new infections. Greater effort also needs to go into the development of epidemiological models to understand the current state and likely future course of the epidemic, to provide a context for planning and designing interventions, and to evaluate the effectiveness of such interventions (Bastos et al., 2001).

However, none of these interventions is likely to be effective without a sound understanding of the reasons why Southern Africa is the worst-affected region in the world, why the epidemic has spread in this region more rapidly than in any other, and why there are such great differences in the infection rates in different provinces, between men and women and, critically, between migrants and non-migrants. It is almost self-evident that a major determinant of the spread and the rate of increase of HIV in Southern Africa is migration, which takes a particular form in this region. However, while more is known about both migration and HIV/AIDS in Southern Africa than anywhere else on the continent, few studies have considered both together.

Although both migration and HIV have been examined separately in South Africa, we are still far from understanding in detail just how and to what extent migration affects the spread of HIV. Communicable diseases cannot spread from one area to another unless they are carried by people who move between different places; but we still do not know how much such ‘movement’ is needed to ensure that an epidemic of HIV, which is already established in one area, will become established elsewhere. While it is likely that migrant men carry their infections home with them, to what extent does the absence of men in rural areas put women who remain at home in a situation of high risk? Even if migration is necessary to spread the infection in the early stages and even if the extent and nature of such migration determines the rate of spread of infection, does migration become unimportant (as a determinant of risk) once the infection has established itself over a wide geographical area?

Part of the reason for the fact that we still do not understand the precise role of migration in the spread of HIV is that studies of migration and disease tend to concentrate on the urban, or ‘receiving’ areas with little attention being paid to people living in the rural or ‘sending’ areas. Furthermore, there have been very few well-designed epidemiological studies documenting the relationship between migration and infectious diseases. Even more importantly, at this late stage of the Southern African HIV epidemic, there have been few intervention programmes, even on a small scale, which attempt to reduce transmission among migrants and their rural or urban partners.

This paper has highlighted the current state of knowledge about the linkages between HIV/AIDS and migration and proposed a conceptual basis for greater understanding of that relationship, but it is abundantly clear that there are large gaps in our knowledge of the extent to which migration, and the particular forms of migration that are found in Southern Africa, explain why the levels of infection in this region are so much higher than anywhere else in the world. Applied social and epidemiological research on the migration and HIV nexus is an essential complement to any national and local strategy of AIDS prevention. So, too, can it assist in understanding and coping with the burden of AIDS, which will be felt particularly acutely by migrant communities, urban and rural.

Migration has clearly fuelled the epidemic of HIV in Southern Africa, but infection is now so widespread that it seems likely that migration is no longer driving the epidemic. However, programmes to control the epidemic will certainly be considerably less effective if migrant workers continue to spread infections. Programmes aimed at supporting migrants should be given the highest priority, but much more work is needed to provide an understanding of the social, behavioural and sexual context of the lives of migrants.

Perhaps most importantly, policy issues need to be addressed, including the nature and extent of migration, the rights of migrants, and the kinds of services to which they have access. This must be conducted for those in both the formal and the informal sector, and even illegal migrants must be able to access the health services without fear of exposure. The HIV/AIDS pandemic threatens to devastate much of Southern Africa. Dealing with the epidemic must be given the highest priority and treated with the greatest urgency. However, unless the issues of migration and disease are understood and dealt with effectively, it is unlikely that the greater struggle to control and manage AIDS can be won.

Acknowledgments

This paper was prepared under the auspices of the Southern African Migration Project with support from CIDA and DFID. Mark Lurie also wishes to thank the Wellcome Trust (UK), and the U.S. National Institutes of Health (Grant numbers 1K01 MH069113–01A1, and 5T32 DA13911).

Additional information

Notes on contributors

Jonathan Crush

Jonathan Crush, Southern Africa Research Centre, 152 Albert Street, Queen's University, Kingston, ON K7 3N6, Canada.