|Home | About | Journals | Submit | Contact Us | Français|
ICMJE criteria for authorship read and met: DH OM TH BM BC TM CB SG. Agree with the manuscript's results and conclusions: DH OM TH BM BC TM CB SG. Designed the experiments/the study: DH TH TM SG. Analyzed the data: DH OM TH BM TM SG. Collected data/did experiments for the study: DH TH TM BM SG. Wrote the first draft of the paper: DH. Contributed to the writing of the paper: DH OM TH BM BC TM CB SG.
While dramatic gains in the availability of antiretroviral medications in developing countries have been achieved , there is growing consensus that, unless prevention efforts can be made more effective, there will ultimately be no victory in the fight against HIV/AIDS –. Maintaining tens of millions of people on treatment throughout their lifetimes will not be sustainable or affordable, particularly as drug resistance may increasingly result in the need for much more expensive second and third line medications. Although there have been promising breakthroughs in a few other areas, notably male circumcision and prevention of mother-to-child transmission (PMTCT) ,,, it is widely recognized that behavior change must remain the core of prevention efforts –.
While the often cited prevention success stories of Thailand  and Uganda , are inspiring and informative, some of the specific socio-cultural, historical, and other factors in the southern African region—now the global epicenter of the HIV pandemic—are distinctive. In these “hyper-endemic” settings, where adult HIV prevalence ranges from 12% to 26% , HIV transmission is highly generalized, whereas Thailand's epidemic was much more concentrated. There, HIV transmission was driven mainly by brothel-based sex work—enabling the aggressive “100 percent condom” programs to be feasible, enforceable, and effective ,. The unprecedented HIV decline and associated behavior change in Uganda, mainly involving large reductions in multiple sexual partnerships ,–, occurred some 20 years ago and under rather different contextual and programmatic circumstances.
More recent examples of HIV prevalence reduction are emerging, including from Kenya, Haiti, the Dominican Republic, Malawi, and Ethiopia –,–. Given the severe HIV epidemics that continue to plague parts of sub-Saharan Africa, there is an urgent need for studies identifying the proximate as well as underlying causes for these encouraging trends. In this paper, we review and summarize the principal findings of our comprehensive interdisciplinary analysis (commissioned by two United Nations agencies, the United Nations Populations Fund [UNFPA] and United Nations HIV-AIDS Program [UNAIDS]) of the causes behind the considerable HIV decline in Zimbabwe, including evidence for changes in patterns of sexual behavior and the contextual and possible programmatic reasons for these changes, which we have published in other peer-reviewed journals –. Here we also consider some policy implications of these findings.
HIV prevalence data from national antenatal clinic surveillance and the household-based 2005/6 Demographic and Health Survey (DHS) were used to fit a mathematical model to estimate trends in HIV incidence and AIDS deaths in Zimbabwe (Figure 1A) . Data from the DHS and other longitudinal surveys ,,– were used to examine the possible contributions of changes in sexual behavior to reductions in HIV infection. Published data from focus group discussions with 90 adult men and 110 women in diverse urban and rural sites and several dozen in-depth key informant interviews as well as an extensive historical mapping of prevention programs , were examined in assessing the contributions of different contextual and programmatic factors to observed changes in behavior. Finally, DHS data on various potential proximal and contextual determinants of behavior change for Zimbabwe were compared with similar data for seven other southern African countries to identify distinctive patterns that might help to explain the earlier and faster HIV decline observed in Zimbabwe (Figures 2, S1).
Interpretation of data on the causes of HIV declines in other countries (e.g., Uganda) has proved to be contentious and problematic when drawing conclusions for policy. Therefore, to establish a consensus among key stakeholders on the roles of different potential causes of HIV decline in Zimbabwe, a national stakeholders meeting was held in Harare in May 2008 to examine the evidence assembled during this study. Stakeholders from a broad range of backgrounds attended the meeting, including high-level representatives of civil society and international organizations as well as senior non-political appointees within the Ministry of Health and Child Welfare and the National AIDS Council (participants and agenda listed in Text S2). At the meeting, the proximate and underlying contextual and programmatic factors that could have contributed to Zimbabwe's declining HIV epidemic were ranked systematically according to whether they were considered likely, plausible, or unlikely to be major contributors to the HIV decline, based upon triangulation of the data described above and from other relevant studies – (see Table 1 and Text S1). Four tests were applied in determining this ranking: 1) whether the factor could, based on the epidemiological literature, have been effective in reducing HIV risk at the population level; 2) for underlying factors, the existence of a clear causal pathway to those proximate factor(s) considered to be effective in reducing risk; 3) the extent or coverage of the factor; and 4) the consistency of timing between the primary change or amplification of the factor and the period of most significant risk reduction as estimated by the epidemiological model (about 1999–2003) . Those factors considered effective that had already reached high levels of coverage before the period of most rapid risk reduction and continued to maintain high levels throughout this period were considered possible contributors to the HIV decline, as theoretically they could have helped to curb transmission (without an immediate impact on prevalence) so that later behavioral changes had a greater impact .
HIV prevalence in Zimbabwe increased rapidly in the early to mid-1990s, before reaching a plateau in the late 1990s (peaking at an estimated 29% adult prevalence in 1997 ), and declining after 2000 (down to 16% estimated prevalence in 2007). Mathematical modeling fitted to surveillance data  (Figure 1A) estimates that HIV incidence peaked around 1991 and (as in many other African countries ) declined gradually thereafter, mainly as part of the natural course of the epidemic, primarily due to saturation of infection in high-risk populations ,,. Between about 1999 and 2003, the pace of incidence decline accelerated considerably, which empirical data , and modeling  suggest corresponded to reduced levels of risky sexual behavior.
As illustrated in Table 1 (and see Text S1), the unanimous conclusion from the stakeholders meeting (Text S2) held to assess, triangulate, and interpret the evidence assembled in the review – was that a reduction in multiple sexual partnerships was the most likely proximate cause for the recent decline in HIV risk. Although the DHS surveys indicate there was little change in age of sexual debut between 1999 and 2005/6, an approximately 30% reduction in the proportion of men reporting extra-marital partners occurred (Figure 1B). Similar or larger reductions in multiple partnerships among adults have been reported in other national surveys ,. Since presumably nearly all married people would have sex at least occasionally with their spouses (and given that most adults in Zimbabwe are married), this implies a sizeable reduction in the level of concurrent partnerships, which is a key epidemiological factor . At a rural eastern Zimbabwean research site, where HIV prevalence fell substantially between 1998–2000 and 2001–2003, the fraction of men (and women; data not shown here) reporting concurrent partnerships declined by about 40% (Figure 1B) . In addition, there were considerable reductions reported around this time in the number of Zimbabwean men paying for sex (Figure 1B) . Participants in focus groups and key informant interviews conducted in various rural and urban areas similarly reported that major changes in norms of sexual behavior had occurred, especially after the late 1990s . For example, many informants recounted that, whereas in earlier years it was commonplace for men gathering at locales such as beer halls to be surrounded by women/sex workers, by the late 1990s this norm had changed and it was now typical for men to gather strictly among themselves at such places. Moreover, while earlier it had been considered a proof of masculinity to acquire a sexually transmitted infection (STI), more recently getting an STI (and visiting sex workers) is typically said to be embarrassing or even shameful for Zimbabwean men .
In fact, the prevalence of other STIs was greatly reduced during the early 1990s, mainly due to widespread syndromic management services . Although STI control remains an important public health measure, the data from clinical trials regarding the population-level impact on HIV incidence are increasingly unconvincing (Text S1) ,. However, it has been hypothesized that STI treatment during the early phases of an HIV epidemic may help to reduce transmission (although this is unconfirmed by observational evidence; e.g., given the absence of HIV declines in several other African countries that had also implemented early and aggressive STI control programs). Reported condom use increased steadily during the 1990s (reaching 59% among men for last non-marital sexual encounter in the 1994 DHS), but did not increase further between 1999 and 2005/6 (Figure 1B), and remained very low for regular partnerships . However, there is some evidence for modest improvement in the consistency of condom use among women in casual partnerships ,, a more important measure for reducing infection risk than reported use at last sex .
In assessing the underlying factors for the national prevalence decline (Table 2, Figure S2), high AIDS mortality appears to have been the dominant factor for stimulating behavior change. Both empirical and modeling-derived estimates indicate that AIDS deaths increased dramatically during the mid-to-late 1990s, before stabilizing after 2000 ,. Moreover, men and women in focus groups and interviews repeatedly and consistently reported personal exposure to AIDS mortality and the resulting fear of contracting the virus to be the primary motivation for changes in sexual behavior, particularly reductions in casual sex and other multiple sexual partnerships .
The severe economic decline, taking hold in the late 1990s/early 2000s, appears to have played a considerable secondary role in amplifying patterns of behavior change, particularly partner reduction. Gross domestic product in Zimbabwe began to slump in the late 1990s, declining by about 40% between 1999 and 2005, with average real earnings plummeting by 90% during the same period. And many men in focus groups and interviews reported that having less disposable income has increasingly led to reduced ability to purchase sex or maintain multiple sexual relationships . However, the most severe financial declines occurred after 2002, i.e., after the bulk of HIV incidence decline had evidently already occurred , (Figures 1A, S3). The severe economic and political instability in the country also led to extensive international migration from Zimbabwe around this time. To have contributed substantially to the HIV decline, migration also would have needed to have been highly concentrated among individuals with HIV . In fact, the available data suggest that the opposite was probably the case; e.g., HIV prevalence among pregnant Zimbabwean women living in the United Kingdom peaked at 12%, less than half the equivalent figure seen among pregnant women living in Zimbabwe itself . Therefore, although the political and economic crises in Zimbabwe have been (especially in more recent years) extremely turbulent and of grave concern for humanitarian and other reasons, these factors do not appear to be the predominant ones for explaining the HIV decline that occurred.
In considering the potential impact of prevention programs on the HIV decline (Table 2, Figure S2), condom distribution and promotion efforts commencing in the early 1990s may have contributed through helping build high levels of condom use for commercial and casual sex ,. (And it appears that condoms were usually not promoted in the often highly “sexy” manner as occurred in some neighboring countries such as Botswana, but generally as a strictly “protective” public health intervention.) Voluntary counseling and testing (VCT) and PMTCT programs were, however, unlikely to have contributed significantly to HIV incidence decline as they were scaled-up only after 2002 . Furthermore, the epidemiological evidence for individual- and population-level impact of VCT remains uncertain or weak ,,. During the 1990s, a wide range of prevention and information programs were implemented utilizing the national media along with school-, workplace-, and church-based activities, peer education, and other inter-personal communication interventions . Community-based activities were intensified following establishment of the National AIDS Council in the late 1990s. This range of broader HIV education and prevention programming could have had impact. Focus group and interview participants mentioned a number of prevention programs and awareness/education efforts and many reported that the “B” part of “ABC” was promoted by churches in particular and was “heard” by many community members , yet no specific intervention was cited consistently.
One question arising from this review is why similarly high AIDS mortality and extensive coverage of HIV prevention programs (resulting in similarly high levels of reported condom use, early and large reductions in STI incidence, etc.) in several other countries in the region have not yet led to substantial declines in HIV prevalence (or multiple sexual partnerships) ,,,. Our comparative analysis of eight southern African countries revealed few patterns of association. The HIV epidemic in Zimbabwe is somewhat older than in some other countries in the region, yet HIV prevalence has been declining markedly for over a decade now, which has not occurred to nearly the same extent, for example, in Malawi and Zambia (where HIV arrived even earlier). In addition to the severe economic decline, where Zimbabwe does stand out is in having high levels of both secondary education and marriage, especially in urban men, among whom the greatest level of behavior change evidently has occurred ,, (Figures 2, S1). It appears that this unique combination helped facilitate: 1) a clearer understanding and acceptance of how HIV is sexually transmitted (once such information became widely available through various AIDS education and prevention programs commencing in the early 1990s ), as some studies of schooling levels and HIV determinants have suggested  and 2) a greater ability to act upon “be faithful” messages, given the stronger marriage pattern – in Zimbabwe than that in neighboring countries also having relatively well-educated populations, such as Botswana and South Africa.
In addition, national survey data suggest that between the mid-1990s and the early 2000s, Zimbabweans increasingly received information about AIDS from their friends, churches, and other inter-personal (as compared to official media) sources (Figure S4) ,. A similar pattern has been linked to behavior change in Uganda ,. Furthermore, the Zimbabwean government's early adoption of a home-based care policy  may inadvertently have accelerated the process of behavior change. It has been hypothesized that, when people die at home, this direct confrontation with AIDS mortality is more likely to result in a tangible fear of death among family and friends than when patients are primarily cared for in clinical facilities, such as in Botswana .
It appears that the motivation for behavior change largely arose endogenously from within the population, and may have been partly due to events specific to Zimbabwe, such as the drastic economic decline in recent years. Nevertheless, it is unlikely that significant changes in behavior in response to the increasing levels of mortality could have occurred unless prevention programs had provided effective information and education about the link between risky sexual behavior and AIDS. We had hoped that our review would identify some particularly effective approaches, which could then be strengthened in Zimbabwe and inform prevention programs in other countries. Perhaps one reason that most respondents failed to identify specific effective programs is because it was the cumulative exposure to many programs that helped create a “tipping point” leading to changes in behavioral norms. We also note that government and civil society did promote faithfulness (mainly in the context of a generic “ABC” message), although not as early or as vigorously as Uganda's “zero grazing” campaign during the late 1980s ,,. Furthermore, findings from the qualitative research suggested the considerable impact of popular culture that occurred precisely around the key period of behavioral change of the late 1990s and early 2000s. For example, a (donor-sponsored) documentary Todii (“What shall we do?”) and a related widely popular song released by the famous performer Oliver Mutukudzi addressed the behavioral risks and social consequences of HIV infection .
The behavior changes associated with the HIV decline in Zimbabwe appear to have resulted primarily from increased interpersonal communication about HIV and its association with risky sexual behavior, due to high personal exposure to AIDS mortality and correct understanding of sexual HIV transmission (due to relatively high education levels along with information provided by HIV communication programs), as well as the deteriorating economic situation. However, the substantial shift in social norms that appears to have occurred, such as STI infection having become a cause for shame, suggests that the economic decline was probably more a co-facilitating factor rather than the major reason for behavior change; e.g., reduced income may prevent men from frequenting bars, but wouldn't change their attitude about having an STI.
One lesson emerging from this review is that in Zimbabwe, as elsewhere ,,–, partner reduction appears to have played a crucial role in reversing the HIV epidemic. Public and private sector programs in Zimbabwe are now building upon this knowledge by more assertively warning against multiple and concurrent partners and promoting sexual fidelity, in addition to consistent condom use and other effective approaches such as male circumcision . Similar efforts have begun appearing elsewhere in the region, such as a bold Swaziland campaign highlighting the danger of having “secret lovers” ,,.
A clear consensus was established regarding the conclusions presented in this article at the stakeholders' meeting, following extensive and open debate. Nevertheless, some uncertainty may remain regarding the conclusions reached. We hope that the detailed documentation of the data and criteria used in attributing causality provided here—and in the supporting publications, references, and supplementary text—will allow others to judge for themselves whether they agree with these conclusions.
HIV prevalence has declined in Zimbabwe by approximately 50%. This decline is almost unique in sub-Saharan Africa and it is hoped that the findings presented here may provide important insights for HIV control within the region. Additional investigations, similarly involving rigorous triangulation of data from multiple sources, should be commissioned in other countries where HIV prevalence has also declined substantially.
Levels of marriage and secondary education in eight southern African countries. Same as for Figure 2, except these data also include for (A) urban men; (B) urban women; (C) rural men; and (D) rural women.
(0.02 MB PDF)
Relationships between proximal and distal factors for behavior change and HIV decline in Zimbabwe. This chart illustrates the need to consider different levels of analysis, and suggests that at each level of analysis (including prevalence decline, incidence decline, behavior change, program activities, and the underlying socio-economic/cultural factors) a combination of a few key factors appears to best explain the observed changes.
(0.06 MB PDF)
(0.01 MB PDF)
Sources of information on HIV-AIDS among young men in Zimbabwe. Sources: 1994 ZDHS; 2001/2 Zimbabwe Young Adult Survey . Note that the categories included in the two surveys are not exactly identical. In the 2001/02 YAS “print media” and “hospital” were used (as per chart), but the 1994 DHS had asked for “newspaper” and “health worker,” respectively.
(256 KB PDF)
(0.17 MB PDF)
List of meeting participants and agenda from the May 2008 Stakeholders meeting, Harare.
(0.03 MB PDF)
This study was sponsored by the United Nations HIV-AIDS Program (UNAIDS), the United Nations Populations Fund (UNFPA) and the Zimbabwean Ministry for Health and Child Welfare. TBH and SG also thank the Wellcome Trust for funding support. We thank Exnevia Gomo, Reko Mate, Allison Herling Ruark, Timothy Mah, Doug Kirby, David Wilson, Helen Epstein, Martina Morris, and Ann Swidler for their valuable assistance and comments.
The authors have declared that no competing interests exist.
Some of the studies upon which this paper is based were funded by the United Nations Population Fund (UNFPA), which provided some logistical support as well as helping with coordination between the studies. Two of the authors of this paper (Benedikt and Campbell) are employed by UNFPA and helped edit the manuscript. The United Nations HIV-AIDS Program (UNAIDS) and the Zimbabwean Ministry for Health and Child Welfare also sponsored this study. TBH and SG thank the Wellcome Trust for funding support. The funders had no other role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Provenance: Not commissioned; externally peer reviewed.