For the selection of the population-based sample, we used a stratified two-stage probability sample design with the assistance of the Central Statistics Office at the Ministry of Finance and Development Planning in Botswana. The first stage involved the selection of enumeration areas with probability proportional to measures of size, where the measure of size is the number of households in the enumeration area as defined by the 2001 Population and Housing Census. We randomly selected 89 enumeration areas in total, 41 from large urban settlements, 28 from small urban villages, and 20 from rural villages, agricultural lands, and cattle posts. Field researchers then systematically selected households within each enumeration area under the guidance of field supervisors. With a target sample of 1,200 households and an anticipated 85% response rate, 1,433 households were randomly selected. We used random number tables within each household to select one adult member that met our inclusion criteria, and up to two repeat visits were made. No replacements were made if participants could not be contacted after two visits.

Heavy alcohol consumption was a strong and consistent correlate of all sexual risk behaviors examined for both men and women, including unprotected sexual intercourse with a nonmonogamous partner, having multiple partners, and paying for or exchanging sex for money or other resources. Parallel to findings from our multivariate regression analyses, 45% of participants identified alcohol use as the single most important HIV risk factor in semistructured questions. Our results from this large probability sample of adults in Botswana reinforce the findings from a number of smaller studies that alcohol use is strongly associated with a number of risky sexual behaviors in sub-Saharan Africa [2–4]. For example, in a study among 324 male beerhall patrons in Zimbabwe, Fritz et al. demonstrated that the number of days of alcohol consumption was correlated with the number of episodes of unprotected sex with a casual partner and with episodes of paying for sex. In another study by Simbayi et al. among 149 men and 72 women receiving sexually transmitted disease services in Capetown, South Africa, participants with more problematic drinking had significantly more sexual partners over the previous month, were more likely to have received money for sex, and were more likely to have a history of sexually transmitted diseases. These results were not stratified by gender. Dunkle et al. showed that problem drinking was associated with significantly higher odds of sex exchange in women attending antenatal clinics in Soweto, South Africa [17]. Our data extend this previous research by demonstrating that these associations hold in a population-based study, that the relationships between alcohol use and risky sexual practices are similar in both men and women, and that there is a dose-response relationship between alcohol use and unprotected sex as well as other risky sexual practices among both genders. Our findings, in conjunction with those of others, strongly argue for the need to target alcohol use and abuse in HIV prevention programs. While causality can not be determined from a cross-sectional study, the consistency of results across many studies (including the reported associations between alcohol use and incident HIV infection [6]), the dose-response relationship between alcohol and risky sex, the strength of the associations, and the biologic plausibility all suggest that alcohol use is in fact a cause rather than a consequence of risky sexual behavior.

To date few policies have been implemented in Botswana and elsewhere in Africa to address the strong overlap between alcohol use and HIV. A study by Fritz et al. has shown that it is methodologically feasible and culturally appropriate to carry out HIV interventions in Zimbabwe beer halls [2]. Additional risk reduction strategies that could be considered include educational campaigns targeting alcohol and HIV in schools and other social venues, interventions that limit alcohol licenses or increase taxes on alcohol, and the bolstering of programs for the prevention, treatment, and rehabilitation of alcohol abuse. Alcohol use in Botswana and elsewhere in Africa has deep-seated cultural and social meanings related to social status, gender identity, and family and communal structures [27] that must be taken into account in the design of effective alcohol reduction strategies.

The most significant correlates of risky sexual behavior were similar for men and women. In addition to alcohol use, intergenerational sex was strongly and consistently associated with all risky sex variables for both men and women, and was also strongly associated with heavy drinking. Previous qualitative studies have shown that intergenerational and transactional sexual relationships are often initiated in drinking establishments [13]. Intergenerational sex can be a major contributor to propagating and sustaining HIV in sub-Saharan Africa, as it can account for substantial amounts of HIV transmission between different age groups [12]. In view of the strong overlap between risky alcohol use, intergenerational sex, and sexual risk-taking, our results attest to the need for multidimensional approaches in HIV prevention programs that simultaneously target a number of high-risk practices. For example, reinforcing or raising the drinking age may also be effective at decreasing intergenerational sex, and programs to promote economic independence for women may help overcome the financial dependency that promotes intergenerational sex and sex exchange [28].

We found that lack of control in sexual relationships was associated with having multiple partners for both men and women, and with sex exchange for women. Not surprisingly, women were significantly more likely than men to report lack of control in sexual relationships, and were also more likely to consider lack of control in sexual relationships and a partner's refusal to use condoms as key barriers to condom use. These findings are consistent with studies showing that lower relationship control and forced sex for women are associated with both inconsistent condom use [29] and higher HIV seroprevalence [30]. Higher negotiating power within sexual relationships and economic independence were also found to be positively associated with condom use in a small cross-sectional study among 71 women in Gaborone, Botswana [22]. HIV prevention programs are more likely to be effective if they address the pervasive gender discrimination that helps to perpetuate the spread of HIV in Botswana and elsewhere in Africa [30]. Interventions should also target the complex inter-relationships between alcohol use and gender economic imbalances.

A few important gender differences were notable in correlates of sexual risk-taking. For example, older age was associated with lower odds of unprotected sex for women, and the reverse trend was apparent for men. This is consistent with statistics on HIV prevalence in Botswana (with evidence that HIV prevalence is three times as high among younger women than younger men [31]) as well as with the aforementioned finding that intergenerational sex is associated with a greater likelihood of unprotected sex since these relationships are most common between younger women and older men. Men but not women with more frequent contact with health providers were more likely to have multiple partners, and men who reported poor health were more likely to pay for sex. It is more likely that frequent contact with health providers and poor health are a consequence and not a cause of these risky sexual practices. HIV testing was associated with a lower likelihood of having multiple partners for men but not women. This may be because men are in a better position to change their sexual behavior after HIV testing compared to women, who may lack control over sexual negotiation, as discussed above. Greater symptoms of depression were associated with having multiple partners for women and paying for sex for men. This is consistent with studies in developed countries that have shown a strong link between depression and high-risk sexual behavior [32,33], but no studies to our knowledge have previously examined this association in an African context. In view of the high prevalence of depressive symptoms reported in this study (nearly 30% of participants screened positive using the Hopkins Symptom Checklist), it is important to further explore depression as both a possible cause and consequence of the high-level of HIV transmission in sub-Saharan Africa.

In addition to the cross-sectional design, a number of limitations affect interpretation of our results. Self-reporting may introduce bias, because it can be influenced by social desirability. To minimize self-reporting bias, we did not ask about HIV status, assured confidentiality and privacy in all interviews, and carefully trained interviewers on asking sensitive questions in a nonjudgmental fashion. While risky sexual practices may have been underreported, this would not necessarily affect the associations we found between risky sex and alcohol use as well as other covariates. Another limitation was that our measures on alcohol use did not address either alcohol dependence or impairment in social functioning. Findings would have been strengthened if we used the internationally validated WHO Alcohol Use Disorders Identification Test (AUDIT) as a screening tool for alcohol use. Future validation studies of Western standardized alcohol abuse screening tools in African settings should pay careful attention to the fact that consumption of home-brewed drinks may be more difficult to quantify. Finally, Botswana has a number of unique features that may limit generalizability to neighboring African countries, such as its relatively high per capita income, comparatively extensive health care infrastructure, strong donor involvement, and strong government commitment to combating HIV. Nonetheless, similar results from studies in other African countries [2,4,10,13] strongly support these findings and their applicability elsewhere in Africa.