KAIS showed that only one-quarter of ever sexually active men and less than half of ever sexually active women aged 15–64 years in Kenya in 2007 had been tested for HIV. Furthermore, more than 80% of HIV-infected adults aged 15–64 years were unaware of their HIV infection, and among those, one-third reported being uninfected but had a laboratory-confirmed positive HIV test. Time since last HIV test and higher CD4 counts suggests that many of these had sero-converted since their last HIV test.
Considering that in 2003, only 15% of all Kenyans aged 15–49 years had ever been tested (compared to 37% of that age group in KAIS), Kenya made substantial progress in expanding testing 
. However, overall testing coverage in 2007 remained far below Kenya's national goal of testing 80% of all adolescents and adults 
, and gender difference in testing rates demonstrated a need for special efforts to bring HIV testing to men. Higher testing rates in women of reproductive age and the finding that half of these women reported that they their last HIV test was during antenatal care confirmed the importance of ANC services for HIV testing in women but also raised the question of sufficient access to HIV testing for women who do not get pregnant and older women who are less likely to get HIV tested during ANC. In addition, the large proportion of people with HIV infection reporting a previously negative HIV test indicates ongoing incident infection and the need for more frequent HIV testing than once in a lifetime.
Testing rates for both men and women were higher in urban areas (highest in the capital Nairobi), in better-educated and wealthier persons, and in persons who had contact with health facilities. In multivariable analyses, variables that were independently associated with ever HIV testing for both men and women included age, province, education and contact to health services (outpatient visits in men, hospitalizations in women). In men only, wealth index and condom use at last sex were independently associated with ever HIV testing. In women only, marital status, ever having been pregnant, and perceived HIV risk were independently associated with ever HIV testing. The considerably lower odds for ever testing in women aged 35 years and older compared to men of that age confirmed that women are testing at younger age, most likely driven by ANC services during pregnancy. All provinces with significant AOR were negatively associated with ever HIV testing when compared to Nairobi. This included North-Eastern, the province with limited need for HIV testing, given a HIV prevalence of <1%, and Nyanza with a 40% lower odds for ever HIV testing in women yet an HIV prevalence in women of 17% compared to 10% in Nairobi. Nyanza is a mostly rural province on the shore of Lake Victoria with the highest rates for HIV in the country, at least partially due to a low prevalence of male circumcision, while North Eastern province is less populated and ethnic groups in this area traditionally circumcise. Independent associations with ever HIV testing of both education and wealth (men only) in our study highlight linkages between HIV infection, access to services, and socio-economic status 
. The main reason for not testing for HIV among those never previously tested was low perception of risk, which has been also reported from Uganda 
We found that considerable opportunities for testing were missed during general or pregnancy-specific contacts with health facilities. Our study suggests that more than 90% of all persons in Kenya with undiagnosed HIV infection who had never been tested could potentially be identified through a combination of provider-initiated testing and door-to-door testing in high prevalence provinces. However, coverage of door-to-door testing can decrease when family members cannot be reached at home 
The 2008 National Guidelines for HIV Testing and Counselling in Kenya promote a diversified approach to reduce the number of missed opportunities for providing HTC including client-initiated, provider-initiated, self-testing, home-based testing and mass HIV testing campaigns 
. The guidelines call for integration of HTC into other health services to allow for early detection and better health care for persons living with HIV 
. Our findings show that targeting sexually active men in general, sexually active non-pregnant and older women (e.g. ≥35 years), and rural and disadvantaged populations should be a priority for prevention efforts, as well as increasing general knowledge about HIV risks in a country with a prevalence of 7%. Standardized quality-control measures for HIV testing, partner testing and mutual disclosure of testing results are additional programmatic implications.
In times of increasing restrictions of funding, national strategies need to consider the most cost-efficient interventions. Menzies and colleagues estimated costs and effectiveness of four HTC strategies in Uganda in 2003–2005 
. Door-to-door HCT had the lowest cost per client tested ($8.29) and per client who tested for the first time ($9.21) compared to costs of $11.68 and $14.73, respectively, during hospital-based HCT. However, cost per HIV-positive individual identified was considerably higher for door-to-door HCT ($163.93) than for hospital-based HCT ($43.10). Door-to-door HCT was able to reach more clients as couples (21.6%) than hospital-based HCT (3.2%). Although these results may not be entirely transferable to Kenya, this study confirmed that a mixture of different types of HTC facilities will allow contributing to Kenya's national targets of achieving 80% of the sexually active population to know their HIV status 
, maximize preventive effects that have been shown to be strongest among HIV-positive clients and discordant couples 
, and identify as many persons as possible living with HIV but do not know their status. Further operational research is needed to determine the ideal mixture of services for a country like Kenya and the frequency of repeat testing needed for specific risk populations and in high HIV prevalence areas to identify persons with recent infection early and enroll them into care and treatment programmes. Home-based testing was acceptable to over 80% of persons aged 15–64 years in KAIS 2007 and may help achieve the national testing goal.
Program and survey data in Kenya suggest that testing coverage has continued to increase since KAIS. In 2008–9 
, 40% of men and 57% of women aged 15–49 years in Kenya had been tested for HIV and had received results at least once in their lifetime (up from 26% and 45% for age 15–49 years, respectively, in KAIS 2007). Nevertheless, given persistent incidence 
, on-going provision of testing will be critical for Kenya's HIV prevention, care and treatment efforts. While testing rates continue to increase in the country and this may place Kenya more towards the higher end of testing rates in sub-Saharan Africa 
, consolidated efforts are needed to reach Kenya's national goal and allow all HIV-infected persons access to life saving treatment. Similar proportions of persons unaware of their HIV status as in our study may be found in other countries in Sub-Saharan Africa with a generalized HIV epidemic similar to Kenya's, indicating the need to include both laboratory testing results for HIV and interview data on prior testing history and current HIV status in population-based HIV surveys.
Our study had several limitations. Cross-sectional surveys do not allow for determination of the sequence of events in time. KAIS was not designed to assess HIV testing among high-risk populations, such as sex workers, men who have sex with men, or intravenous drug users (Kenya has increased surveillance efforts for these populations since 2010). Some data were not available from the study, e.g., whether respondents had not tested for HIV because no transport was available to reach a testing site or they were not able to pay for the transport, or whether they had tested as a couple. There is no generally accepted definition of a high HIV prevalence area; therefore, missed opportunities were reported across Kenya without excluding provinces with relatively low prevalence such as North-Eastern. High rates of undiagnosed infection suggest limited coverage of testing services and relatively high incidence; however, they may also partially reflect reporting bias due to misunderstanding of prior results, denial, misreporting, or false-negative test results. Finally, Kenya's population structure with over 40 ethnic groups of considerable cultural differences may have resulted in some differences in self-reporting. The example of reporting bias of HIV results was discussed above. In general, the direction and magnitude of any potential reporting bias is unknown.
In spite of these limitations, by including for the first time questions on HIV status and CD4 count testing among HIV-infected persons, Kenya's nationally representative HIV survey helped inform HIV program planning with unprecedented detail. Our findings illuminated both the high rates of undiagnosed infection throughout Kenya and the clear opportunities for expanding testing coverage to meet the national and 2008 United Nations goals of universal access to HIV prevention, treatment, care and support 
. Knowledge of HIV status could help protect millions of people from transmitting HIV unknowingly, from suffering unnecessarily from opportunistic infections, and from dying prematurely with no access to treatment. Three decades into an epidemic, which has already claimed more than an estimated 15 million lives in Africa alone 
, the urgency of universal testing access could not be clearer.