Gebrekristos, Hirut T. MPH*; Resch, Stephen C. MPH*; Zuma, Khangelani PhD†; Lurie, Mark N. PhD‡
THE MIGRANT LABOR SYSTEM in South Africa involves circular patterns of migration between rural homes and work areas. The mines offer single-sex hostels as the most common type of accommodation available to labor migrants from rural areas of South Africa and other neighboring countries.1 The literature highlights 2 distinct but related reasons for why the migrant labor system encourages risky sexual behavior. The first is the separation of familial and stable sexual relationships. The second is the presence of commercial sex in mining towns.2–4 The circular nature of the migrant labor system in South Africa has also developed circular forms of sexual networks between rural areas and labor centers.
HIV prevalence rates are higher among migrant men compared with nonmigrant men in South Africa. In a study comparing the prevalence of HIV among men in 2 broad migration arrangements and their respective partners against nonmigrating men and their partners, Lurie et al. found a 27.7% HIV prevalence among migrant men who returned home 2 to 4 times a year and a 15.2% HIV prevalence among their partners.5,6 HIV prevalence was 22.4% among migrant men who returned home once a month and 20% among their partners.5,6 Nonmigrating couples have the lowest HIV prevalence, where HIV prevalence among men is 17.8% and 12.9% for their partners.5,6 The data indicate that the lowest HIV rates are found among couples that remain together year round. This suggests that establishing family housing where partners live together may dramatically decrease the incidence of HIV among migrants and their partners.
We present the potential impact of family housing on annual risk of HIV infection among migrating men and their rural partners. Using a binomial model, we estimate the annual risk of HIV infection given current HIV prevalence data, migration patterns, and serodiscordance data among couples. We then compare this estimate with a predicted annual risk of HIV infection resulting from establishing family-style housing.
To quantify the impact of establishing family-style housing, a Bernoulli model of HIV transmission is used. Described in detail elsewhere,7,8 the model treats HIV transmission as a random event with an occurrence probability that is a function of population HIV prevalence and the probability of HIV transmission per sex act within a partnership. Estimates for the annual risk of HIV infection among migrants and their partners are considered separately to highlight potential differences that family housing may have for each member of the couple. The risk of HIV transmission estimated for family housing is calculated for short- and long-stay migrants* and their respective partners. The proportion of HIV serodiscordance is central to estimating risk of HIV infection associated with family housing among couples. The benefit of the intervention to any individual couple is dependent on the couple’s joint HIV status. Couples in which both partners are HIV-positive cannot benefit from a prevention intervention. Each member of a serodiscordant couple will benefit differently from the intervention. If the migrant is negative and his partner is positive, he may be better off drawing at random from the commercial sex worker (CSW) pool than having sex with his HIV-positive partner. If the migrant is positive, he gets no benefit from family style housing, and his partner’s risk is increased. Both members of a concordant negative couple, however, stand to benefit from family housing.
The equations used to estimate the impact of family housing among migrants and their partners is also described in Table 1. For rural partners in discordant relationships, in which the miner is HIV-infected, the annual risk of HIV infection is calculated by the following equation: 1-(1- RYX)N*(1-Hi), where RYX is the male to female transmission rate, N is the number of sex acts for men per year, and Hi is the duration of time migrants spend in the work areas (where i is 1 for short-stay, 2 for long-stay). Among HIV-negative concordant couples, the annual risk of infection for the rural partners is driven by the probability that her male partner is infected that year by a CSW. We assume that male partners who become infected by CSWs, become so mid-year, and therefore only half of the contacts between the man and his rural partner in that year are potentially infectious. The rural partner’s ARI is represented by the equation 1-(((1-Rxy*W)N*Hi)+(1-(1-Rxy*W)N*Hi)*(1-Ryx)N*(1-Hi)/2) where Rxy captures the female to male transmission rate, W represents the HIV prevalence among CSWs, and (1-Hi)/2 represents our assumption that male partners who become infected mid-year on average. The annual risk of HIV infection among migrants in HIV discordant relationships is estimated by 1-((1- RXY)N*(1-Hi))*((1- RXY*W)N*Hi). The annual risk of HIV infection among miners in HIV-negative concordant relationships is estimated by 1-(1-RXY*W)N*Hi.
Based on data from Lurie et al, couples were classified by their serodiscordance status.5 The change in risk for each group was calculated and these changes were weighted by the proportion of couples in each category resulting in a predicted impact of family housing. This analysis was performed for both short- and long-stay migrants. Because there is limited HIV serodiscordance data available for the population of interest, 1-way sensitivity analysis was conducted for the proportion of serodiscordance among migrants and their partners. In addition, a sensitivity analysis was also conducted for HIV prevalence among CSWs, proportion of sex acts that migrants have with CSWs after the interventions, and per sex act probability of HIV transmission (RXY and RYX). The parameter values used in the model are reported in Table 2.
HIV Prevalence Among Sex Workers.
Published estimates of the HIV prevalence rates found among CSWs in South Africa range from 0.503 to 0.69.9,10 HIV prevalence among CSWs (W) used in the model is taken from HIV prevalence measures among CSWs in the mining communities.10
Infectivity Rates (RYX and RXY).
Studies in the United States and Europe have estimated the probability of HIV transmission per sexual act of vaginal intercourse to vary from 0.0001 to 0.0014.11–13 Estimates for vaginal HIV infectivity rate per sex act for men who engage in sexual contact with sex workers vary from 0.056 to 0.100 in studies in Thailand and Kenya.14,15 The base-case infectivity rates (RYX and RXY) used in this model were from the Rakai study in Uganda.16 For the base-case, a male to female rate of infectivity is 0.0009 and female to male infectivity rate used is 0.0013. However, because the infectivity rates are heavily dependent on the presence of other sexually transmitted infections, particular sex act engaged (penile–vaginal, penile–anal intercourse), the stage of disease progression of the infected person, and the HIV subtype, there is significant variation associated with RYX and RXY in published literature.11,14,15,17–19
Number of Sex Acts per Week.
The sexual transmission of HIV links infectivity within sex acts; the model assumes 2 sex acts per week (N = 104 sex acts per year). This assumption is in keeping with the infectivity rates used in the model based on data reported by Gray et al.16 The base-case in this model was 2 acts per week for men. The number of sex acts per week for women was determined by where the men spent their time. For example, if men spent only 25% of their time at home, and engaged in sex acts at a rate of 2 per week, their rural partners would average 1 sex act every 2 weeks.
Percent of Sex Acts with Commercial Sex Workers.
Short-stay migrants spend approximately 75% (H1) of their time at the work area. Long-stay migrants spend 90% (H2) of their time at the work areas, returning home 2 to 4 times per year for stays of 1 to 2 weeks. The percent of time migrants spend away from home was used to estimate the percent of sex acts migrants have with CSWs.5 The probability that a miner is at home (1-H1 or 1-H2) during a given year is also calculated using the duration of time migrants spend living in migrant communities. For example, short-stay migrants spend approximately 25% of the year at home. Because rural partners’ exposure is dependent on the proportion of time migrants spend at home, rural partners will not have a full year of exposure to HIV. The model assumes an average midyear estimate for the partners to be exposed and acquire HIV by including (1-Hi)/2. This captures the reality that migrants may become infected during sex acts with CSWs at different points during the year and in the process possibly become a risk for their rural partners on their return home.
This model also assumes that migrants engage in sex with CSWs while at the work areas and with their rural partner when at home. The potential benefit of a family housing intervention relies heavily on the change in the number of sex acts between migrant workers and CSWs. The extent to which migrants who live with their partners in family housing will seek out sex workers or engage in casual sex with other women is unclear. Given family housing, we assumed in the baseline calculations that migrants spend 10% of their sex acts engaging with sex workers. This is supported by data from several African countries where the percent of sex with CSWs ranges from 4% to 20%.
The intervention’s main objective is to increase the proportion of sex acts in which the migrant’s sex partner is their rural partner instead of a CSW. Therefore, the intervention’s impact will depend on the serodiscordance profile of the population. For HIV-negative migrants with HIV-negative rural partners, this change in mixing pattern lowered their annual risk of HIV infection. However, family housing would increase the annual risk of HIV infection for the HIV-negative partner in discordant couples. Table 3A and 3B report the results. When weighting these results by the proportion of couples in each category based on Lurie et al serodiscordance data, 0.0254 is the estimated net reduction in the annual risk of infection for short-stay migrants and their partners. A net reduction of 0.0305 is estimated for long-stay migrants and their partners.
In the absence of family housing, the annual risk of HIV infection among migrants in concordant HIV-negative partnerships is 0.068. For HIV discordant couples, in which the rural partner is infected with HIV, migrants have a 0.099 risk of HIV infection. Using baseline parameter values, the family housing intervention reduces annual risk of infection in short-stay migrants by 0.058 and by 0.0004 for their rural partners. HIV-negative rural partners of HIV-negative short-stay migrants would face a slight decrease in their annual risk of HIV infection with the implementation of family housing. HIV-negative rural partners of HIV-positive migrants will do far worse under family housing. For HIV-positive short-stay migrants, their HIV-negative partners’ risk increases by 0.058. In the same way, couples in which the short-stay migrant is HIV-negative and rural partners are HIV-positive, the annual risk of HIV infection for the short-stay migrant would increase by 0.024.
Within the current migration arrangement, an annual risk of 0.081 is estimated for concordant HIV-negative long-stay couples. In the context of HIV discordance, in which the rural partner is HIV-infected, the long-stay migrant has a risk of 0.093 in the absence of family housing. Establishing family housing among current long-stay migrant couples reduces the annual risk of HIV infection among migrants by 0.071 and there is an increase in risk for their partners (0.000001). Because the rural partners will have increased their sex acts with long-stay migrants, there is a slight increase in the annual risk of HIV infection among HIV-negative rural partners in HIV-negative concordant long-stay couples. The annual risk of HIV infection increases by 0.072 for rural partners of HIV-positive long-stay migrants. Similarly, the annual risk of HIV infection increases by 0.030 for HIV-negative long-stay migrants with HIV-positive rural partners under family housing.
A 1-way sensitivity analysis was conducted for 4 parameters critical in this model: the proportion of serodiscordance, HIV prevalence among CSWs, the percent of sex migrants have with CSWs after the family housing intervention, and the per sex act probability of HIV transmission (RXY and RYX). In Figure 1, the 1-way sensitivity analysis illustrates that reduction in the annual risk of HIV infection associated with family housing increases as the proportion of concordance among both short- and long-stay couples increases. The impact of various values for HIV prevalence among CSWs was assessed for concordant negative short- and long-stay migrants and their respective partners (Fig. 2). As the HIV prevalence among CSWs approaches that found among rural partners, the difference in the annual risk of HIV infection between the single-sex hostel and family housing arrangement decreases for both short- and long-stay migrants and their partners. Similarly, in Figure 3, higher rates of sexual contact with CSWs after establishing family housing increases the annual risk of HIV infection for short- and long-stay migrants and their respective rural partners. Family housing is estimated to have the highest impact in the context in which migrants’ sexual contact with CSWs is 0 for both short- and long-stay couples. The influence of the per sex act probability of HIV transmission on family housing was examined in Figure 4. The benefits of family housing increases as the probability of RXY increases, whereas increasing the probability of RYX reduces the impact of family housing.
In most plausible scenarios, the intervention results in a net reduction in annual risk of HIV infection for concordant HIV negative couples. If other interventions that are currently ongoing such as sexually transmitted infection treatment and prevention or condom promotion were also included, we may see greater reductions in annual risk of HIV infection. However, migrants and their rural partners do not benefit mutually from the intervention, where family housing increased the risk of HIV by a small margin (0.00001) for partners of long stay migrants (Table 3). For serodiscordant couples, the risk of HIV for the negative partner will increase under family housing because they will have more sex acts with their infected partners. This highlights the need for appropriate HIV counseling and testing activity concurrent with the implementation of family housing.
The potential benefit of family housing intervention rests on the assumption that it will reduce the absolute number of sex acts between seronegative migrant workers and seropositive CSWs. The model assumes that whenever migrants have sex at the mines, they do so with CSWs. This extreme scenario in which CSWs, a population associated with a high prevalence of HIV, is likely to overestimate the annual risk of HIV infection at the mining centers in the model. However, the sensitivity analysis indicates that the benefit of family housing holds with a wide range of HIV prevalence for CSWs, in which lower HIV prevalence for CSWs serves as a surrogate marker for sex with other non-CSW women who are likely to have lower levels of HIV infection. Although this article does not examine the cost-effectiveness of this intervention, it is possible that less costly interventions could be considered in lowering the number of sex acts migrants have with an identifiable group of CSWs. The gains estimated under a range of HIV prevalence values for CSWs in this model suggests that there may not be a discrete and identifiable group of women in the mining communities to target for interventions that may reduce the number of sex acts with CSWs.
The binomial model only approximates observed data. Studies have shown that over the short-term, binomial models underestimate and over the long-term, they overestimate the per-partnership probability of transmission.21–24 Empiric findings suggest that over the long-term, as many as two thirds of serodiscordant couples will remain serodiscordant despite hundreds of sex acts.23 Speculation that this is the result of host factors or changes in infectivity over the course of infection has been discussed.21,25,26
The analysis provided in this article highlights some general principles for considering the distributional aspects of the effect of HIV prevention. When HIV prevalence is heterogenous between subgroups of a population, interventions that simply shift contacts from 1 subgroup to another can have the characteristics of a zero-sum game. In addition, timing is a critical element for a successful intervention. As the HIV prevalence among the subgroups in the model approaches a level where there is “0” difference, the benefits associated with family housing will diminish. However, other social familial benefits outside the scope of this article may continue to support this intervention, regardless of the HIV-related benefits.
We modeled the impact of family housing on the probability of HIV infection among different subgroups while holding prevalence constant. The probability of becoming infected increases with the number of sex acts in the model. Over a long period of time, the probability of being infected approaches 1 in any scenario where the initial HIV prevalence is greater than zero. A dynamic model would continually update prevalence level (and therefore risk levels) in the population and thereby give more accurate long-term projections. Such a model could capture the indirect benefits of reduced transmission as well as the short-term gains, but would require considerable amounts of additional data and assumptions. The 1-year period on which this analysis focused is appropriate from a policy perspective and is sufficient for short-term projections.
Heterosexual vaginal sexual intercourse is the only type of sex considered in the model. However, heterosexual intercourse may include anal intercourse, which is associated with higher risk of HIV transmission. The exclusion of anal intercourse in the model may present conservative estimates for the annual risk of HIV infection. Similarly, the only mode of HIV transmission considered is that associated with heterosexual sex. The magnitude of other modes of transmission present in the population may change the estimated benefits of establishing family housing. The model also does not account for concurrent partnerships, because it assumes that migrants have only 1 rural partner and that each sex act with a CSW is drawn at random from a pool of CSWs. However, these are simplifying assumptions that may result in higher estimated risk of infection than would be observed. Furthermore, some migrants’ rural partners may face HIV infection risk from other men, yet these risks were assumed to be zero. The exclusion of this estimate is likely to produce a slight underestimation of annual risk of infection for rural partners and migrants.
The model does not consider the influence of other sexually transmitted infections in facilitating HIV transmission. Ramjee et al indicate in a study of CSW in KwaZulu-Natal that 42% of sex workers have syphilis, 14.3% have gonorrhea, and 12.7% have genital ulcers.9 Data from the migration project indicate that 10.5% of migrants have syphilis and 21.1% have gonorrhea. Considering that sex workers and migrants in South Africa have high sexually transmitted infection prevalence and that sexually transmitted infections facilitate transmission of HIV,19 the estimates provided in this article may be conservative.
The HIV prevalence data used for the 2 broad categories of migrants included men who live in single-sex hostels at the mines (long-stay migrants) and men who live in other forms of accommodation but who are generally not miners (short-stay migrants). Understanding the impact of establishing family housing on HIV transmission for miners using the 2 broad categories of migrants may present a limitation. There may be unique elements of living in a single-sex hostel environment and being a miner that supports risky sexual behavior compared with other forms of accommodation and other types of migrants. However, modeling the impact of family housing on HIV transmission among both groups of migrants provides a broader range of scenarios for understanding how changing single-sex hostels into family housing may impact mining communities.
Sweat and Denison note that the most dramatic reductions in mortality and morbidity have historically resulted from addressing structural and environment factors.27 Although family housing may be a useful structural intervention for HIV prevention, qualitative research conducted by 1 of the authors (HG) suggests the potential for unintended negative consequences. We may find that some migrants and their partners do not necessarily want to live with one another near the work areas. Extended families are supported by the remittances from each miner. If the immediate family of the miner is no longer back “home,” we may unintentionally remove a social support network. Also, we learned that some women come to join husbands and find that their husbands are with someone else. In which case, they are left in an unfamiliar environment without much familial or financial support. Some of these women stay in the mine communities and turn to commercial sex for income, which would shatter our initial objectives of reducing HIV transmission. In addition, we may find that the migrants live with their “home” partner and continue to simultaneously engage in sex with CSWs. In this scenario, family housing would not lead to any gains for miners or their partners.
The possible challenges involved in establishing family housing, however, does not detract from the positive benefits associated with this intervention for HIV-negative concordant couples. The actual or hypothetical impact of converting single-sex hostels into family housing on HIV transmission has not been examined. Although this model does not provide all the answers, it does provide a framework with which we may understand the value this structural intervention has on HIV transmission.
1. Crush J, James W. Crossing Boundaries: Mine Migrancy in Democratic South Africa. Albion: Institute for Democratic Alternatives in South Africa, 1995.
2. Hunt CW. Migrant labor and sexually transmitted diseases: AIDS in Africa. J Health Soc Behav 1989; 30:353–373.
3. Jochelson K, Mothibeli M, Leger JP. Human immunodeficiency virus and migrant labor in South Africa. Int J Health Serv 1991; 21:157–173.
4. Parker RG, Easton D, Klein CH. Structural barriers and facilitators in HIV Prevention: A review of international research. AIDS 2000; 14:S22–S32.
5. Lurie MN, Williams BG, Zuma K, et al. Who infects whom? HIV-1 concordance and discordance among migrants and non-migrant couples in South Africa. AIDS 2003; 17:2245–2252.
6. Lurie MN, Williams BG, Zuma K, et al. The impact of migration on HIV-1 transmission in South Africa: A study of migrant and nonmigrant men and their partners. Sex Transm Dis 2003; 30:149–156.
7. Allard R. A family of mathematical models to describe the risk of infection by a sexually transmitted agent. Epidemiology 1990; 1:30–33.
8. Pinkerton SD, Abramson PR. The Bernoulli-Process model of HIV transmission: Applications and implications. In: Holtgrave D, ed. Handbook of Economic Evaluation of HIV Prevention Programs. New York: Plenum Press, 1998.
9. Ramjee G, Abdool Karim SS, Sturm W. Sexually transmitted infections among sex workers in KwaZulu-Natal, South Africa. Sex Transm Dis 1998; 7:346–349.
10. Williams BG, MacPhail C, Campbell C, et al. The Carletonville-Mothusimpilo Project: Limiting transmission of HIV through community-based interventions. S Afr J Sci 2000; 96:351–359.
11. Royce RA, Sena A, Cates W, et al. Sexual transmission of HIV. N Engl J Med 1997; 336:1072–1078.
12. Padian N, Shiboski SC, Jewell NP. Female-to-male transmission of human immunodeficiency virus. JAMA 1991; 266:1664–1667.
13. De Vincenzi I, for the European Study Group in Heterosexual Transmission of HIV. A longitudinal study of human immunodeficiency virus transmission by heterosexual couples. N Engl J Med 1994; 331:341–346.
14. Mastro TD, Satten GA, Nopkesorn T, et al. Probability of female-to-male transmission of HIV-1 in Thailand. Lancet 1994; 434:204–207.
15. Cameron DW, Simonsen JN, D’Costa LJ. Female to male transmission of human immunodeficiency virus type 1: Risk factors for seroconversion in men. Lancet 1989; 2:403–407.
16. Gray RH, Wawer MJ, Brookmeyer R, et al. Probability of HIV-1 transmission per coital act in monogamous, heterosexual, HIV-1 discordant couples in Rakai, Uganda. Lancet 2001; 357:1149–1153.
17. Nicolosi A, Correa Leite ML, Musicco M, et al. The efficiency of male-to-female and female-to-male sexual transmission of the human immunodeficiency virus: A study of 730 stable couples. Italian Study Group on HIV Heterosexual Transmission. Epidemiology 1994; 5:570–575.
18. Fleming DT, Wasserheit JN. From epidemiology synergy to public health policy and practice: The contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect 1999; 75:3–17.
19. Rottingen JA, Cameron DW, Garnett GP. A systematic review of the epidemiologic interactions between classic sexually transmitted diseases and HIV. Sex Transm Dis 2001; 28:579–597.
20. Wasserheit JN. Epidemiological synergy. Interrelationships between human immunodeficiency virus infection and other sexually transmitted diseases. Sex Transm Dis 1992; 19:61–77.
21. Rottingen JA, Garnett GP. The epidemiology and control implications of HIV transmission probabilities within partnerships. Sex Transm Dis 2002; 29:818–827.
22. Downs AM, De Vicenzi I. Probability of heterosexual transmission of HIV: relationship to the number of unprotected sexual contacts. European Study Group in Heterosexual Transmission of HIV. J Acquir Immun Defic Syndr Hum Retrovirol 1996; 11:388–395.
23. Jewell NP, Shiboski SC. Statistical analysis of HIV infectivity based on partner studies. Biometrics 1990; 46:1133–1150.
25. Kaplan E. Modeling HIV infectivity: Must sex acts be counted? J Acquir Immun Defic Syndr 1990; 3:55–61.
26. Shiboski SC, Padian NS. Epidemiologic evidence for time variation in HIV infectivity. J Acquir Immun Defic Syndr Hum Retrovirol 1998; 19:527–535.
27. Leynaert B, Downs AM, De Vicenzi I. Heterosexual transmission of human immunodeficiency virus: variability of infectivity throughout the course of infection. European Study Group on Heterosexual Transmission of HIV. Am J Epidemiol 1997; 146:350–357.
28. Sweat MD, Denison J. Changing public policy to prevent HIV Transmission: The role of structural and environmental interventions. In: Holtgrave D, ed. Handbook of Economic Evaluation of HIV Prevention Programs. New York: Plenum Press, 1998.