Lack of a comprehensive understanding of national HIV epidemics remains a major challenge for targeting effective HIV programs. Antenatal clinic (ANC) sentinel surveillance has been useful to document trends of generalized epidemics but lacks adequate sexual-behavioral and other associated risk factors for HIV infection, and is not generalizable to men and non-pregnant women [1]. Studies conducted in a number of sub-Saharan African countries showed that ANC data over-estimated prevalence in 15-19 year olds and underestimated prevalence in older ANC clinic attendees [2, 3]. Hospital-based studies have also been conducted in low prevalence areas to understand risk factors [4], but the findings are not easily generalizable to populations living in countries with high HIV prevalence.

Concerns about the representativeness and accuracy of national HIV estimates derived from ANC surveillance have led to an increased demand for surveys and more data on the prevalence and distribution of HIV in the whole population [5]. In sub-Saharan Africa, more than 38 countries have conducted national population-based HIV surveys [6]. These surveys most often were restricted to reproductive ages, collected dried blood spots for biological tests [7], and avoided questions about sexual behavior and HIV status which are considered sensitive. Thus, laboratory and risk factor analyses have been limited. Programmatic data from voluntary counseling and testing (VCT) centers as well as HIV test results from prevention of mother-to-child transmission (PMTCT) programs have been used in some cases to estimate HIV prevalence and investigate associations with demographic, behavioral, and biological factors [8-10]. However, these are limited due to their lack of representativeness of the populations that seek VCT and PMTCT services [11]. For example, PMTCT services are only offered to pregnant women. Additionally, risk behavior, perception of risk behavior, or attitudes towards HIV/AIDS have been shown to influence uptake of VCT services [12].

Sexual transmission is the major route for HIV infection in sub-Saharan Africa [13,14]. In Kenya, sexual transmission contributes over 90% of all HIV infections [13]. The World Health Organization (WHO) categorizes Kenya as having a generalized HIV epidemic, defined as having an HIV prevalence of more than 1% in the adult general population and more than 5% in vulnerable adults such as sexually transmitted infection (STI) clinic patients. Kenya has wide variability in the distribution of HIV infection by geographic, demographic, sexual-behavioral, and biological factors. The Kenya Demographic and Health Survey (KDHS) conducted in 2003 included HIV serology but was limited in the scope and depth of sexual-behavioral and biological indicators collected, and included consenting adults aged 15-49 years [15]. In this era of expansion of prevention, care and treatment programs for HIV, additional variables of interest are needed to facilitate accurate interpretation of HIV prevalence data and associated risk factors [16].

We conducted the Kenya AIDS Indicator Survey (KAIS) in 2007 to provide nationally-representative and comprehensive data on demographic, behavioral, and biologic indicators of HIV/AIDS, beyond that of previous national HIV surveys in Kenya. In addition to providing national prevalence estimates for HIV and sexually transmitted infections, these data provided the opportunity to link HIV status with key demographic, behavioral, and biologic information to identify significant correlates associated with HIV infection in Kenya. We examined factors independently associated with the risk of HIV infection among individuals who were sexually active during the 12 months prior to the survey.

We collected information from 9,691 households. A total of 19,840 individuals were eligible, of which 17,970 (90%) consented to be interviewed, including 10,239 women (93%) and 7,731 men (87%) (Fig. 1). Among the eligible respondents, 9,049 (83%) women and 6,804 (77%) men consented for and provided a blood sample for HIV, HSV-2, and syphilis testing, a response rate of 88% of those interviewed, with similar uptakes rates among men and women. Among those consenting to blood draw 6,447 women (71%) and 5,112 men (75%) were sexually active in the last year. Overall, 57% of KAIS participants were female, the median age was 30 years, 74% resided in rural areas, 70% were currently employed, 61% were married, and <1% of women and 6.6% of men drank alcohol before their last sexual encounter (Table 1). Among the laboratory-confirmed HIV positive respondents, 56% had never tested for HIV infection, 28% self-reported HIV negative and 16% self-reported HIV positive status. We found ethnicity to be co-linear with province and circumcision and choose to present differential risks with a focus on circumcision status and province rather than with a focus on ethnicity.

In 2007, an estimated 7.4% of sexually active Kenyan adults aged 15-64 years were infected with HIV. Correlates of HIV infection among women and men were age, number of lifetime sex partners, residence in Nyanza province, HSV-2 infection, consistent condom use with the last sex partner and lack of circumcision among men. The strongest independent predictors for HIV infection for both women and men were HSV-2 co-infection and higher number of lifetime sex partners.

HIV prevalence was highest in Nyanza Province, where 16.9% of the sexually active adults were HIV-infected. In a sub-analysis for Nyanza Province, we found that age, HSV-2 infection, multiple lifetime sex partners, consistent condom use with the last sex partner and lack of male circumcision were independently associated with HIV. Many of these factors are similar to predictors of HIV infection found at the national level and are consistent with findings from other studies [1,8,16,19].

In Kenya, the adjusted odds of having HIV among sexually active persons with HSV-2 infection were 5-6-fold higher than those uninfected with HSV-2. However, awareness of HSV-2 is very low, even among health care providers, despite the high prevalence of HSV-2 and the potential role of HSV-2 in driving the HIV epidemic [20,21]. HIV-induced immune compromise can result in frequent and persistent HSV disease, while poorly managed HSV replication may influence HIV transmission [22]. Researchers estimate that in settings with high HSV-2 prevalence, such as Nyanza province, HSV-2 infection could contribute to the risk of HIV-infection in more than one in four new cases of HIV [19,21]. Unfortunately, randomized controlled trials that have examined daily acyclovir treatment of HSV-2 among persons with HIV co-infection, or acyclovir prophylaxis among persons without HIV have not demonstrated a protective effect [23].

The results from KAIS show a non-statistically significant increase in overall HIV prevalence from KDHS 2003 (7.4% vs 6.7%) among those aged 15-49. The potential increase in HIV prevalence may be in part due to the survival effects of antiretroviral therapy. It may also indicate increasing incidence and a need to improve and expand HIV prevention programs throughout Kenya, and particularly in Nyanza Province. Appropriate messages on delaying sexual debut, knowledge of HIV status, male circumcision, consistent and correct use of condom with partner of unknown HIV status or known discordant HIV status, and reduction of number of sex partners should be reinforced [8, 24, 25]. Our study showed that widowhood and divorce were significantly associated with higher HIV prevalence. This corroborates findings from other sub-Saharan African countries which show that women who encounter marital disruption through divorce or widowhood were more likely to be HIV infected [26]. On the other hand, divorce is more common among HIV-infected women, particularly those in HIV discordant unions [27].

Ethnicity and province influence both the distribution of circumcision practice and HIV prevalence. Several studies have showed that male circumcision reduces the risk of HIV acquisition among men [28]. High prevalence in Nyanza province could be attributed to low male circumcision rates. Among the general population, 85% of men were circumcised nationally while 48.2% were circumcised in Nyanza Province. In addition, cultural practices such as widow inheritance practiced among the Luo community (the predominant ethnic group in Nyanza) may be a factor [28, 29]. Widow inheritance is a traditional practice in which a designated man takes social and economic responsibility over a woman following the death of her husband [30]. A study by Agot et al. [30] in Nyanza province showed that inherited widows were more likely to have HIV infection compared to those not inherited. Luo women are believed to acquire contagious cultural impurity following the death of a husband. Often, a professional “cleanser” is hired who performs sexual rituals to cleanse the widow. If the spouse of the deceased is HIV-infected, the cleanser acts as a bridge for HIV transmission to other widows hence putting widowed women at a high risk [31]. The Kenya Ministries of Health published The National Guidance for Voluntary Medical Male Circumcision (VMMC) in January 2008 and in November 2008 launched the VMMC program focusing on Nyanza province and other traditionally non-circumcising communities with the aim of reducing new HIV infections and other STIs. The guidance provides a broad policy framework for the integration of VMMC into existing HIV prevention programs.

Respondents reporting ever having used condoms were more likely to have HIV infection. Condom use reduces risk of HIV acquisition and transmission [32] and HIV-infected persons who are aware of their HIV infection are more likely to use condoms [33, 34]. KAIS showed a 4-fold increase in condom use among those who knew their HIV-positive status. Though the association between condom use and HIV infection was not expected, the finding may reflect the success of positive prevention interventions and condom promotion efforts to increase condom use by people living with HIV. These data highlight the need for future HIV surveys to collect more detailed partner-specific information on condom use and knowledge of HIV status of participants and their partners which could help interpret the complex associations between condom use, sexual behavior, and HIV. The application of a laboratory assay that can accurately distinguish recent from established infection could also suggest the temporality of any associations between HIV infection and condom use, and more accurately highlight areas for programmatic focus.

Our study was limited by several factors. About 20% of eligible residents were either not present or declined to participate in the interview and blood draw. Although we do not expect that there is significant participation bias, we were not able to conduct these analyses; however, appropriate weighting was applied to adjust for non-response. These results cannot be generalized to all Kenyans, but only to those that reported recent sexual activity. Additionally, key sexual behavior indicators were based on self reported data. Though KAIS interviewers were trained on asking sensitive questions around sexual behavior and ensuring respondent confidentiality, there is a possibility that these questions were not accurately answered. We did not ask how long after circumcision the men engaged in sexual intercourse. The cross-sectional design of the study limited our interpretation of the temporality of association between the factors examined and HIV infection. The survey also did not ask questions on men having sex with men or injecting drug use activities that are practiced in Kenya and may contribute to new HIV infections [13]. We did not include children due to the relatively low HIV prevalence among this group.

HIV remains a major public health challenge in Kenya. Although various prevention, care and treatment programs have been initiated and expanded in Kenya, evidence based prevention efforts that target known behavioral and biologic factors such as reduction of sex partners, condom use, delayed sexual debut and male circumcision should be enhanced. The wide regional variation in HIV prevalence reinforces the need for targeted prevention interventions focusing on provinces with high infection rates, while at the same time addressing the key behavioral factors that are associated with the risk of HIV infection nationally.