In a nationally representative sample of Kenyans aged 15–64, 1 in 10 married or cohabitating couples was affected by HIV. Regional variations were pronounced, with the highest proportion of couples in Nyanza province being HIV-affected and the lowest in North Eastern province. Nyanza province is located on the shore of Lake Victoria in the tropical part of Kenya with the highest rates for HIV and other infectious diseases in the country while North Eastern province is mostly arid and less populated. Younger age in women, an increasing number of lifetime partners in women, HSV-2 infection in one or both partners, and lack of circumcision in the male partner were associated with HIV infection in discordant couples compared with uninfected couples in multivariable analyses. Our findings suggest a scenario where young women may enter a marriage or cohabitation when they are already HIV-infected or HIV/HSV-2 –co-infected, especially if women have multiple sexual partners early after sexual debut. Young women may also be HIV-uninfected but start the relationship with a man who is already HIV-infected or HIV/HSV-2 –co-infected.
Despite considerably less power to show significant associations, HSV-2 infection in both partners and lack of circumcision in the male partner were associated with being in a concordantly infected couple compared to a discordant couple, indicating the need to address these risk factors in prevention programs for discordant couples. Our findings were consistent with the literature 
and therefore confirmed previously reported associations at a national population level in Kenya.
Married or cohabitating couples are a population at high risk for HIV transmission and acquisition in Kenya. Discordant couples represent a particular high risk group. Also, partners in the acute phase of a new infection pose a high risk for onward transmission within the couple or if they have unprotected sex outside of the couple. Without intervention, 8–12% of HIV-infected adults living in couples will transmit HIV to their partners annually 
. If all the 338,000 uninfected partners became infected HIV prevalence in Kenya could increase by up to 2% (based on the KAIS 2007 prevalence of 7.1% 
). Currently, prevention messages often ignore couples, focusing on casual partnerships despite the frequent lack of knowledge within HIV-affected couples of the risk of transmission within their partnership or through perinatal transmission. Lack of knowledge may be increased by a lack of risk awareness within a stable relationship that results in low condom use, as was confirmed in our study. The desire in female partners to have a child in the next 2 years did not remain significant as a risk factor for discordant couple status in the multivariable analysis; however, getting pregnant in a natural way requires unprotected sex, which raises the question of adequate risk counseling and family planning.
The importance of prevention interventions targeted to couples is increasingly recognized 
. Interventions need to encourage couples to be tested together early in the relationship and mutually know their HIV status by learning their test result at the same time or disclosing their status to each other. Partner testing for HIV-infected persons in care and treatment programs should be standard practice as recommended by the World Health Organization 
. Our results indicated a considerable need for family planning counseling, which should be an integral part of all testing and counseling services. Understanding of HIV discordance, even among health providers, is limited, and common misconceptions may undermine risk reduction efforts 
. Both lack of male circumcision and HSV-2 infection have been shown to be associated with increased HIV transmission 
. Our findings confirm that male circumcision will benefit couples by increasing the male partner's likelihood of remaining HIV uninfected. HIV testing and counseling offer an opportunity to refer uncircumcised, HIV-uninfected men for voluntary medical male circumcision. Strong associations between HSV-2 infection and both HIV discordance and concordance reinforce the need to further advance our knowledge on the role of HSV-2 as a biological cofactor in HIV acquisition and transmission 
and develop policies and program guidelines for HSV-2 as a risk factor for HIV infection in couples. HSV-2 suppression with twice daily acyclovir assessed in clinical trials did not prevent HIV acquisition 
or transmission 
. An antiretroviral vaginal gel has recently shown promising results in reducing HSV-2 acquisition 
. In addition, a HSV-2 vaccine, if it is approved, should be considered to reduce the risk of HSV-2 infection in HIV uninfected and discordant couples 
Similar proportions of couples as found in our study may be affected by HIV in other countries in Sub-Saharan Africa with a generalized HIV epidemic similar to Kenya's. The vast majority of these couples are unaware of their HIV infection and many perceive themselves to be at low risk with no need for testing. Opportunities for expanding couples testing and counseling exist in Kenya, including integration of partner testing into HIV care and treatment programs, enhanced partner testing in prevention of mother to child transmission and tuberculosis programs, and home-based HIV couples testing. Couples testing and counseling should be a central component of Kenya's national HIV prevention strategy.
Our analysis had some limitations. Cross-sectional surveys do not allow for determination of the sequence of cause and effect which complicates interpretation of associations. Because of the cross-sectional design, we were unable to include some factors associated with HIV acquisition in our models (e.g. correct knowledge of status). The way the household questionnaire was constructed only allowed including couples in which the head of household was the male partner and co-wives in polygamous partnerships could not be analyzed. Some data were not available from the study, e.g. whether respondents had tested for HIV as a couple. Finally, Kenya's population structure with over 40 ethnic tribes of considerable cultural differences may have resulted in some differences in self-reporting; however, the direction and magnitude of potential reporting bias is unknown.