Article Text

Sociodemographic context of the AIDS epidemic in a rural area in Tanzania with a focus on people's mobility and marriage
1. J T Boerma1,
2. M Urassa2,
3. S Nnko2,
4. J Ng'weshemi2,
5. R Isingo2,
6. B Zaba3,
7. G Mwaluko2
1. 1Department of Epidemiology, School of Public Health, and Carolina Population Center, University of North Carolina, USA
2. 2Tanzania–Netherlands Project to Support AIDS control in Mwanza Region
3. 3London School of Hygiene and Tropical Medicine, Keppel St, London WC1, UK
1. Correspondence to:  Dr J Ties Boerma, Department of Epidemiology, School of Public Health, and Carolina Population Center, University of North Carolina, Chapel Hill, NC 27516-3997, USA;  ties_boerma{at}unc.edu

## Abstract

This analysis focuses on how sociocultural and economic characteristics of a poor semi-urban and rural population (Kisesa ward) in north west Tanzania may directly and indirectly affect the epidemiology of HIV and other sexually transmitted infections (STI). Poverty and sociocultural changes may contribute to the observed high levels of marital instability and high levels of short and long term migration in Kisesa, especially among younger adults. Marriage and migration patterns are important underlying factors affecting the spread of HIV. The most cost-effective intervention strategy may be to focus on the trading centre in which mobility is higher, bars were more common, and HIV prevalence and incidence were considerably higher than in the nearby rural villages. If resources suffice, additional work can be undertaken in the rural villages, although it is not clear to what extent the rural epidemic would be self sustaining if the interventions in the trading centre were effective.

• AIDS
• sexually transmitted disease
• demography
• Tanzania

## Statistics from Altmetric.com

With an estimated 71% of the estimated 34.3 million adults and children living with HIV, the AIDS epidemic in sub-Saharan Africa is much more severe than in any other continent in the world.1 A range of sociocultural, political, and economic factors is thought to favour HIV transmission in many African societies to a much greater extent than elsewhere in the world.2, 3 In addition, within sub-Saharan Africa large differentials in HIV prevalence have been observed between countries, between regions of the same country, between urban and rural areas in the same region, and within rural areas.4–6 Many factors, ranging from socioeconomic and demographic features to sexual behaviour patterns, male circumcision practices, and the prevalence of incurable sexually transmitted infections (STI) have been held responsible for the uneven spread of HIV within the continent.7

Studies in western societies have shown the key role of core groups in the epidemiology of sexually transmitted infections.8 Recent studies in Asia have indicated the importance of bridge populations, of which members have sex with both members of core groups and of the general population.9, 10 There are also studies, mainly among sex workers, in sub-Saharan Africa that have pointed to the importance of core groups in STI epidemics.11–13 On the other hand, the HIV prevalence in several sub-Saharan African countries has reached levels that suggest widespread occurrence of risk behaviours in the general adult population, and this is corroborated by findings from sexual behaviour surveys.14 With the exception of some cities, commercial sex is less prominent than in, for example, Thailand, and core groups appear to be much larger and less clearly identifiable.

Wasserheit and Aral15 have proposed a dynamic typology of STI epidemics which emphasises the distinction between spread networks (characterised by higher rates of concurrent partnerships, by large numbers of sexual linkages throughout the subpopulation, and by some sexual contact with other subpopulations) and maintenance networks (located in subpopulations with relatively lower rates of sexual mixing). The STI epidemic is divided into four phases, starting with an early growth phase, followed by hyperendemic, decline, and endemic phases. Intervention programmes should adapt to the role of the networks in the different phases. In early phases prevention efforts should include efforts raise general public awareness of the “new” STI and improved counselling, detection, and treatment services, with concerted and additional efforts for spread networks. In the subsequent stages prevention strategies should focus more on outreach and community level behavioural interventions in hard to reach populations and less on general population interventions.

Hitherto, research and interventions strategies in the AIDS epidemic in sub-Saharan Africa have primarily focused on raising knowledge and awareness in the general population, on condom promotion to the general public, and on the control of sexually transmitted diseases in the general population or in specific high risk groups. The targeted high risk groups—as representatives of the spread network—are most commonly female commercial sex workers, truck drivers, and sometimes long distance migrant labourers.2, 16 Currently, the case is made for more emphasis on interventions focused on high risk subpopulations, in part motivated by epidemiological considerations but largely driven by resource limitations in proportion to the huge epidemic.17 In order to effectively plan, implement, and evaluate such interventions there is a need for a better description of the composition and dynamics of spread networks.

In this paper we describe the context of the AIDS epidemic in a small rural area in Tanzania, where a population of about 20 000 people has been followed since 1994. This analysis includes the period 1994–98 and focuses on population characteristics rather than on individual risk factors. Special attention is given to population mobility and marriage patterns and how these relate to sexual behaviour and HIV infection at the community level. It is shown that mobility and marriage are critical contextual factors in this population, and sexual mixing is dynamic and diffuse. An area based strategy, focusing on places with high new partner acquisition rates,18 appears an appropriate alternative to conventional sex worker focused approaches, and can help focus interventions in the hyperendemic and subsequent phases of the epidemic.

## DATA SOURCES

Kisesa ward is located in Mwanza Region in northwest Tanzania, about 20 km east of the regional capital Mwanza, along the main road to Kenya (fig 1). It includes six villages with a trading centre along the main road, which have been grouped into trading centre, peritrading centre, and agricultural rural villages for the purpose of this study.6

Figure 1

Map of Kisesa.

A demographic surveillance system was established in 1994 and collects basic demographic information through household visits every five months, and by late 1998 10 rounds had been completed.6, 19 All households are visited each round, and information is collected on residence and survival status of all household members, on pregnancy of women of reproductive ages, and on new arrivals (migrants, newborns). A new person was only listed as a household member if the household respondent had indicated that this person was intending to stay in the household. People who had left the household by the next round were not considered household members. For each resident it was asked whether or not that person had slept in the household the night preceding the visit.

Epidemiological and behavioural surveys of all adults aged 15 to 44 years were carried out during 1994–95 and again two years later. In all, 5820 and 6413 respondents participated in the first and second survey, respectively (response rates 78% and 80%). The surveys included a structured interview on background characteristics, AIDS knowledge and attitude, sexual behaviour, STI treatment, and so on, and collection of a blood sample for HIV and syphilis testing in the first survey and HIV only in the second survey.6

The second survey included a sexual mixing module, which obtained information on all marital partnerships and on the last five non-marital sexual partnerships in the last year. This module included information on the age, marital status, and place of residence of the sexual partner. The 3684 respondents of the sexual network module (1651 men and 2033 women) reported 2439 non-marital partnerships in the last year. Reports were obtained from a total of 1130 male and 803 female spouses, and 554 male and 1990 female non-marital partners with whom respondents had sexual relations in the last year.20

A travellers survey was conducted in 1997. The field workers counted vehicles and interviewed travellers during a single week on all primary and secondary roads in the area. Qualitative methods were used to collect data on mobility, characteristics of bars, and commercial sex. Local informants listed all bars, including traditional brew selling points called pombe shops. Field workers and local informants listed all bar and pombe shop workers and women who frequent such places and are willing to have sex for a small payment or gift. Data on health service utilisation by STI patients were derived from routine records of all health facilities in the study area. All traditional healers in the study area were also interviewed about STI treatment and other conditions.21

## STUDY SETTING

The total population of Kisesa ward was 19 458 in 1994 and grew 2.5% a year to 21 774 by the 10th demographic round in late 1998. The latter included 12 073 people living in the rural villages, 4085 in the peri-trading centre area, and 5616 in the trading centre. The population grew more rapidly in the peri-trading centre area and trading centre (3.7 and 3.4% a year, respectively) than in the rural villages (1.7%). Nearly half of the population is under 15 years (46%) and large cohorts of young people will be moving into the reproductive age span in the coming years. For example, while 10.2% of the population are aged 15 to 19, 13.2% are aged 10 to 14 years.

## HIV/STI EPIDEMIOLOGY

### HIV by age and sex

HIV prevalence among men and women aged 15 to 44 was 5.8% in 1994/95 and 6.6% in 1996/96, while HIV incidence in the intersurvey period was 0.7 and 0.8 per 100 person years among men and women respectively.6 Figure 4 presents the observed HIV prevalence in both surveys and the “expected HIV prevalence,” derived from HIV incidence rates, by age for men and women separately (in two year and, at older ages, three year age groups, with the sample size exceeding 125 in all age groups). The “expected HIV prevalence” represents hypothetical prevalence if a 15 year old were exposed to the current incidence rates until age x and was calculated from the HIV incidence rates.*

Figure 4

(A) Observed and “expected” HIV prevalence among men, Kisesa 1994–97. (B) Observed and “expected” HIV prevalence among women, Kisesa 1994–97. The expected HIV prevalence is based on the incidence rates in the period between the two surveys.

For both sexes the shapes of prevalence curves are similar, with a small increase in the 1996–97 survey. The expected HIV prevalence initially lies close to the prevalence curves. Observed HIV prevalence and expected prevalence curves diverge from age 26 to 27 for women and age 30 to 32 for men, which is likely to be associated both with increased HIV associated mortality and possibly a discrepancy between current and past incidence among older cohorts. For women the curves show an almost linear increase during the first 10–15 years after the initiation of sexual intercourse (15–17 years). The increase among women has a steeper slope than among men. HIV prevalence among women exceeds 5% at about 20 and 10% at 25 years of age. Male HIV prevalence reaches these levels at an age five to six years older.

### HIV by residence and mobility

The surveys revealed striking differences in HIV prevalence and incidence within the small geographical area. HIV prevalence in the trading centre was twice that in the area surrounding the trading centre (within 2 km), and three to four times higher than in the rural villages (within 8 km of the trading centre).6 Analysis of individual risk factors of HIV infection showed that the large impact of the community factors remained after controlling for multiple individual demographic, socioeconomic, biological, and behavioural variables.38 The main community characteristics that affected the risk of HIV included level of economic and social activity, numbers of female bar workers in the community, mobility of the population, and proximity to town. There were also some differences in sexual behaviour between communities, but these were fairly modest and did not explain the effect of community on the risk of HIV.

Other studies have shown an association between HIV prevalence and individual mobility.34 Also, in Kisesa those who moved into the ward had a higher prevalence than those who had lived in the ward all their lives, although the differences were fairly small and became smaller when other variables were controlled for.38 The lower participation rates in a survey of more mobile individuals is, however, an important bias for the individual level analysis. The main reason for non-participation in the survey was travel, short term or long term.6 Current marital status (being divorced or separated) and a history of divorce were strongly associated with the risk of HIV, in analyses of both prevalence and incidence.

### Other sexually transmitted infections

Serological data on other STIs were only available from the first survey in 1994–95, when whole blood was collected and Treponema pallidum haemagglutination assay (TPHA) and Venereal Diseases Research Laboratories (VDRL) tests were done in the laboratory. Overall, 15.5% of 2455 men had a positive TPHA test, including 11.3% who also had a positive VDRL test, which is taken as evidence of recent or current syphilis. Among 2641 women the corresponding figures were 20.5% with a positive TPHA test and 15.8% with a positive TPHA and VDRL test. Positive reactions were least common among men and women aged 15 to 19, and there was little variation from that age onward.

Self reported data were collected for genital discharge and genital ulcer in the 12 months preceding the survey. In 1996–97, 10.2% of men and 6.3% of women reported a genital discharge in the last year, while 10.8% of men and 4.6% of women reported a genital ulcer in the last year. Among men and women with a self reported genital discharge or ulcer, 46.1% and 38.9%, respectively, had visited a health facility for treatment. Traditional healers were the second most popular source of treatment (used by 23.1% and 22.8% of men and women, respectively). During the interviews with traditional healers, however, only a few said they treated large numbers of patients, which may indicate that the traditional healers play a relatively limited role in the treatment of STIs in this area.21

The three dispensaries in Kisesa ward started to provide STI services using the syndromic approach in 1994 (one dispensary in the trading centre) and in 1996 (two rural dispensaries). These clinics saw 393 STI patients in 1996 and 380 in 1997; 59% were women. The leading diagnoses for 1994–97 were genital discharge syndrome (36.4% of all 1141 diagnoses), genital ulcer syndrome (24.7%), and pelvic inflammatory disease (21.8%). As there were approximately 8500 adults aged 15 to 44 living in Kisesa in 1996, the clinic data suggest an incidence less than 2% for genital discharge and genital ulcers. As such, there is a considerable discrepancy between the incidence of STIs based on self reports and on clinic data; this may reflect the non-utilisation of modern health services, the use of services outside the study area, or poor quality of self reported data.

There are no reasons to assume that the pattern of STIs in Kisesa is very different from other settlements in Mwanza Region. Studies in similar populations have shown that Herpes simplex virus (HSV-2) is common, with 20% of men and 50% of women aged 15 to 29 having antibodies39), although HSV-2 was responsible for less than 10% of genital ulcers in clinical studies.23 Similarly high rates of serosyphilis have also been observed, while gonorrhoea and chlamydia infection (mostly asymptomatic) was found in 2–3% of the adult male population40). In a population based survey of men, 2.2% had gonorrhoea and 0.7% chlamydia infection, often asymptomatic.41 In antenatal clinics Trichomonas vaginalis infection was most common (27% of 964 women), followed by active syphilis (10%), Chlamydia trachomatis (6.6%), and gonorrhoea (2.1%).42

## DISCUSSION

In this analysis we have focused on population characteristics of a poor semiurban and rural population in northwest Tanzania, and how these may directly and indirectly affect the epidemiology of HIV and other STIs. The overwhelming majority of the households are poor, which may be an important factor contributing to the high levels of short term mobility and migration within and outside the study area. High rates of annual migration and high proportions of household members not spending the night in the household were observed in all age groups, and especially among women under 25. Male villagers, however, appear to be involved in the bulk of short distance trading, mainly on bicycles, and from rural villages to trading centre and from trading centre to town.

It also appears that socioeconomic and sociocultural changes are affecting marriage systems. Marital instability was high. One in 10 women was divorced or separated at the time of the survey, large proportions of men and women had a divorce in their marital history, and divorce was a common reason for changing residence. The traditional system of marriage support may have weakened, as evidenced by the large proportion of less formal marriages and incomplete wealth transfer in association with marriage. More research on marriage patterns and how they affect mobility and vulnerability to infection in young women is urgently needed.

Sexual behaviour data indicate that premarital sex and multiple partnerships are common, while condom use was low. Most boys and girls were sexually active by the age of 16 or 17. The sexual partners of teenage girls were on average five (non-marital) to seven (marital partner) years older, and a significant proportion of teenage girls had a sexual partner aged over 30.

There are recurrent periods of different levels of risk related to marriage, and short and long term mobility is high. Extensive sexual mixing occurs by place and age and across social boundaries. The high levels of migration, marital instability, adolescent sexual activity, extensive sexual mixing, and sexually transmitted infections in the Kisesa population are all factors that could lead to high levels of HIV incidence. Yet, HIV prevalence and incidence data for 1994–97 suggest that, even though the epidemic has not reached its peak, prevalence levels of over 10% in the whole Kisesa ward adult population are not likely, as the overall incidence was below 1/1000 person-years. HIV prevalence among young women (and young men) was relatively low compared with several other rural and urban areas studyed in eastern and southern Africa,43–47 and mortality among HIV infected persons was close to incidence.19 On the other hand, our study provides little evidence that adolescent sexual behaviour in Kisesa differs from other places with much higher incidence under the age of 20, either in terms of onset of sexual intercourse or in terms of mixing with older age groups. However, data are limited.

The spatial analysis of sexual mixing patterns of non-marital partnerships showed that there is a limited level of mixing between the rural population and the trading centre (and beyond). Nearly eight of 10 partnerships in the trading centre and the rural villages are within the same location. Less than 10% of the partnerships are between the rural villages and trading centre. It is difficult to assess whether this level of spatial mixing is sufficient to enhance the spread of HIV from the trading centre to the rural areas and level the difference. In part, it depends on further mixing of partners within the rural villages. If those who have partnerships with the higher HIV prevalence trading centre (or the regional capital) have multiple partnerships within the study area, rapid spread of HIV is possible. There was only limited evidence that this may be the case. The difference in HIV incidence between the trading centre and rural villages suggests that HIV prevalence differences may become somewhat smaller than was the case in 1996–97, but a significant difference is likely to remain.

The size of the epidemic and consequent human suffering are unprecedented in recent African history, and one's first impression is that a full blown multi-intervention strategy aimed at all population groups in the society seems justified, with relatively more emphasis on the most cost-effective interventions. However, the reality is that in most countries with generalised epidemics it is only possible to follow a much more modest agenda than is desirable. The phase specific intervention strategy based on a dynamic typology of STI epidemics15 provides an epidemiological rationale for focusing interventions according to the phase of the epidemic. The epidemic in Kisesa is in a hyperendemic stage, in which it is most effective to focus interventions on spread networks and single out those with the riskiest behaviour, while maintaining general population interventions. The lack of resources in countries like Tanzania, however, prohibits a full scale intervention and in such situations focusing on spread networks should be a priority. Our analysis has shown that the spread networks are very extensive and fairly diffuse, with members often moving in and out of the network. Whether or not men and women are members of the spread network is influenced by their sociodemographic status, with marital status and mobility playing major roles. Poverty and lack of employment may be important underlying economic reasons. Such extensive networks may have a reproductive rate of infection just above unity and thus contribute to the spread of infection at a fairly slow rate, as opposed to more traditional core groups, such as sex workers, with higher reproductive rates of infection.

In a study in the industrialised world it was shown that focusing interventions on geographic “core” areas rather than on core groups was an effective way of reaching spread networks.48 In the context of Kisesa, which is likely to be similar in much of rural and semi-urban sub-Saharan Africa, this may also be the most effective and the most feasible approach. Such an approach should focus on places where new partner acquisition rates are high, following a methodology to identify high transmission areas developed by Weir et al.18 The “place focus” rather than “people focus” also reduces the risk of stigmatising core groups through the interventions. In Kisesa, focusing initially on the trading centre, with higher mobility, more bars, and higher HIV prevalence and incidence than in the nearby rural villages would appear to be the most cost-effective strategy.6, 38 If resources suffice, additional work can be undertaken in the rural villages, although it is not clear whether the rural epidemic could be self sustaining if the interventions in Kisesa were in fact highly effective. The interventions should include social marketing of condoms, especially in the bars, traditional brew shops, and guest houses, and general AIDS education in places where people meet sexual partners, supported by improved STI services and adolescent sexual health programmes in schools. This high transmission area intervention strategy needs to be complemented with lower intensity basic general population interventions, including increased condom access, and would need to be continued as long as new partnership formation rates are high and until the risk of HIV and other STI are virtually eliminated.

## Acknowledgments

Sponsorship The TANESA project is financed by the Minister for Development Cooperation, the Netherlands. The analyses were supported by the USAID sponsored MEASURE evaluation project.

View Abstract

## Footnotes

• * The HIV incidence rate can be treated analogously to the central death rate, nMx, in a life table, as it is a ratio of events to persons years at risk. In life table terms, the age pattern of stable prevalence associated with a given regime of incidence rates is given by the complement of the proportionate person years function ((1−nLx)/n). Obtaining the person years nLx function from the central rate, nMx, is a standard straightforward calculation, provided we can make some simplifying assumptions about the linearity of change with age in the proportion of susceptibles, lx, in the population—within the narrow two and three year age intervals considered here, such an assumption is justified.

## Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.