Trends and correlates of HIV testing amongst women: lessons learnt from Kenya

Background A majority of women in Kenya do not know their HIV status and are therefore unable to take preventive measures or medication in order to prolong their lives. Objectives This study investigates the key determinants of HIV testing in Kenya and documents how these changed over the 1998–2008 period. Method This study uses data from the 1998, 2003 and 2008 Kenya Demographic and Health surveys. Principal components analysis was used to compute indices of HIV knowledge, HIV-related stigma, media exposure and decision making. Survey logistic regression analysis was used to determine factors that had a statistically-significant association with ever having been tested for HIV. Results Testing was significantly higher in 2008 compared with the previous surveys. In 1998, 14.7% of the women had tested for HIV. The rate increased to 15.0% in 2003 and then to 59.2% in 2008. In the 1998 and 2003 Kenya Demographic and Health surveys, respondents’ age, region of residence, education, knowledge of someone who had died from HIV-related illness and media exposure were the main determinants of testing. In the 2008 study, HIV-related stigma, occupation and the partner's level of education were found to be associated with HIV testing. Conclusion Despite efforts to scale up voluntary counselling and testing in Kenya over the 1998–2008 period, HIV testing amongst women is still quite low. Prevention and control programmes in Kenya need to focus on reducing HIV-related stigma, increasing access to testing in rural areas and increasing access amongst women with little or no education.


Introduction
The HIV pandemic is a global concern that has had a profound impact on many aspects of modern society. 1 By the end of 2011, there were approximately 34 million people living with HIV globally. 2 The sub-Saharan Africa region was the hardest hit by the pandemic and was home to 69% (23.5 million) of all people living with HIV, accounting for 11%-12% of the global burden. 3 Women accounted for more than half of all people living with HIV worldwide and nearly 60% of all HIV infections in sub-Saharan Africa. 4 In Kenya there were approximately 1.6 million people living with HIV, which corresponded to a national prevalence rate of about 7.1%. 5 Voluntary HIV Counselling and Testing (VCT) has been described as the process by which an individual undergoes confidential counselling in order to enable the individual to make an informed choice about learning his or her HIV status and thereafter take appropriate action. 6 It is also described as the process whereby individuals or couples undergo pre-test counselling, risk assessment, a same-day rapid HIV test, post-test HIV prevention counselling (often not received in traditional testing) and referral for medical and support services by trained counsellors. 7 Recent studies have shown that VCT is a cost-effective intervention for reducing HIV-related risk behaviour, particularly when it serves at-risk couples. 8,9 It plays a pivotal role in the publichealth response to the HIV epidemic and is a vital point of entry to HIV services including primary prevention, prevention of mother-to-child transmission, antiretroviral therapy, management of HIV-related illnesses, tuberculosis control and psychosocial support. 10,11,12 VCT also plays a pivotal role in reducing the stigma and discrimination with regard to people living with HIV and is key in the prevention of mother-to-child transmission. 6,13 In resource-poor settings, including many sub-Saharan countries with generalised epidemics, VCT is becoming increasingly available, but study results conflict with regard to the potential impact of VCT in promoting a reduction in risky behaviours. 14 There is evidence to suggest that more than 80% of people living with HIV in Kenya do not know that they are infected. 13 The Government of Kenya, in partnership with donor organisations, has undertaken an ambitious programme to expand VCT services throughout the country using standalone sites and those integrated into other health facilities. 15 Kenya has had a phenomenal expansion of VCT sites from only three in 2000 to 865 sites in 2007. 16 Despite the rapid scale-up initiative, the use of VCT services still remains very low. 17,18 Several factors have been identified in the literature as determinants of HIV testing. There is, however, no consensus on the main determinants of use. Factors identified include the respondent's age, gender, marital status, education attainment, wealth status, HIV serostatus, area of residence, religion, awareness of a nearby HIV or VCT clinic, belief that everyone should know their HIV status, knowledge of someone with HIV, HIV-related stigma and reported willingness to change sexual behavior if HIV testing was positive, amongst others. 7,16,19 The aim of this study is to identify the main factors that influence HIV testing and the utilisation of VCT amongst women in Kenya. The study also investigated trends of, and patterns in, uptake of VCT over the period 1998 to 2008. We present HIV-testing rates by selected covariates and/or determinants and documents demonstrating how these have evolved over time.

Methodology Data extraction and manipulation
Data on HIV testing and other relevant covariates were extracted from the 1998, 2003 and 2008 Kenya Demographic and Health Surveys (KDHSs). These surveys were designed to provide data that can be used by various stakeholders in order to monitor the population and health situation in Kenya. Data were collected on fertility, family planning and maternal and child health. Other data collected also included HIV prevalence, domestic violence and malaria statistics.
The 1998 KDHS was the third national demographic and health survey conducted in the country. Based on a multistage cluster-sampling approach, a nationally-representative sample of 7881 women aged 15-49 and 3407 men aged 15-54 were interviewed. The 2003 KDHS was implemented using a similar sampling methodology to interview 8195 women aged 15-49 and 3578 men aged 15-54. Finally, the 2008 KDHS was the fifth national demographic and health survey conducted in the country. A total of 9057 households were selected using a multistage cluster-sampling process whereby 8444 women aged 15-49 and 3465 men aged 15-54 were interviewed. The samples in each case provide estimates for Kenya as a whole, for urban and rural areas in each of the eight provinces.

Ethical considerations
This study was based on secondary data with all participant identifiers removed. Survey procedures and instruments were approved by the Scientific

Analysis plan
In this study, we restricted our analysis to women aged 15-49 years and considered the response to the question, 'Have you ever been tested for HIV?' to be our primary response variable. Several other variables identified from the literature were cross-classified with this variable. A Pearson's chisquare test was then used to detect any association between the response variable and the categorical variables identified.
Survey logistic-regression analysis, with stepwise elimination, was carried out using STATA 11.0 under the svy command and statistically-significant covariates were identified. All p-values less than 0.05 were considered to be significant.
Separate statistical analysis was carried out for each of the 1998, 2003 and 2008 data sets.

Covariates
The independent variables entertained included:

HIV knowledge-perception index
The HIV knowledge-perception index was created based on responses to the following questions: 'Can a person reduce risk of getting AIDS (Acquired immune deficiency syndrome) by not having sex at all?', 'Can a person reduce the chances of AIDS by always using condoms during sex?', 'Can a person reduce the chance of AIDS by having one sex partner with no other partner?', 'Can a person get AIDS from mosquito bites?', 'Can a person get AIDS by sharing food with person who has AIDS?' and 'Can a healthy person have AIDS?'. On the basis of their factor scores, respondents were classified as having low-, moderate-or high knowledge with regard to the causes of HIV and the basic issues surrounding the disease.

Stigma index
Stigma has been defined in the literature as an attribute or label that sets a person apart from others and links the labelled person to undesirable characteristics. 21 Specifically, stigma related to AIDS has been defined as 'the prejudice, discounting, discrediting, and discrimination that are directed at people perceived to have AIDS'. 22 A stigma index was created based on responses to the questions: 'Willing to care for relative with AIDS'; 'Person with AIDS allowed to continue teaching'; and 'Would buy vegetables from vendor with AIDS'. Based on factor scores, respondents were classified as having low-, medium-or high HIV-related stigma.

Media-exposure index
A media-exposure index was also computed using principal component analysis based on responses to questions posed on the frequency of watching television, listening to radio and reading newspapers. The respondents were then classified as having low-, medium-or high media exposure.

Decision-making index
The  The results indicate an inverted U-shaped relationship between age and testing, with the probability of testing for HIV peaking at age 25 to 29 years for the 1998 and 2008 data Amongst the behavioural determinants considered, a positive relationship was found between HIV-testing rate and age of first sexual encounter. The nature of the relationship between HIV-testing status and women receiving sex for gifts differed for each year considered. In 1998, the rate of HIV testing was higher amongst women receiving gifts for sex (21.9%) as compared with those who had not (16.1%). In the next time periods, significantly-higher testing rates were found -amongst low-risk women (i.e. those who did not accept gifts for sex).  (2003), we found significantly-lower rates of testing in Central and Rift Valley provinces as compared with Nairobi province, as was also the case in other regions in the country.

Trustworthiness
As detailed in the methods section of this study, the study design and methods used to obtain the data in this study are sound scientifically. For each of the three years, multistage cluster-sampling procedures were used to select the clusters, households and hence the women included in this study. The results presented in this study can therefore be generalised to all women aged 15-49 years living in Kenya during the study period. The scales used in this study have been pretested and used in other studies. The reliability of all the indices developed was assessed using the Cronbach's alpha coefficients and thereafter used to evaluate their impact on VCT use.

Discussion
In this study, we described patterns and trends in HIV testing in Kenya over the period of 1998 to 2008. Our database consisted of national surveys conducted in 1998, 2003, and 2008. It is apparent from this study that the proportion of women being tested for HIV had increased significantly over time. Although only marginal improvement was seen over the 1998 to 2003 period, a marked increase in the testing, from 15.0% in 2003 to 59.2% in 2008, was noted. This could possibly be attributed to government initiatives to expand VCT services throughout the country using stand-alone sites and those integrated into other health facilities. 15 Despite this improvement in access to VCT services, as well as the implementation of wide-ranging interventions by both government and non-government organisations in Kenya, HIV prevalence over the period 1998 to 2008 remained relatively high and unchanging. 23 A large proportion of people living with HIV were unaware of their HIV status, 24 posing a major challenge in preventing HIV transmission and providing effective care to infected persons.
We also set out to interrogate the main sociodemographic, sociocultural, behavioral and HIV knowledge-based determinants of HIV testing and to determine whether the effects of these drivers remained the same over time.
In the first two surveys (KDHS-1998 and KDHS-2003), we found sociodemographic characteristics such as age, region of residence and respondent level of education, alongside socioeconomic status, media exposure and the knowledge of someone who had died of HIV-related illness to be the main determinants in a woman's decisions to test for HIV. In the subsequent survey, a woman's occupation, level of HIV-related stigma and her partner's level of education also played an important role in HIV testing.  Various socioeconomic variables have been shown to be important in explaining HIV testing within the sub-Saharan Africa context. 25,26,27,28,29,30 Studies suggest that the poor and less educated are less likely to test for HIV. Our findings are consistent with these studies. 25,30,31,32 In line with other studies in literature, 31,33 we also found that women employed in highly-skilled jobs were more likely to test for HIV than those who were in lower-skilled jobs or unemployed. Highlyskilled employment possibly alludes to better education and better access to information, thereby granting the women access to sound information on HIV and the importance of VCT. We also found a strong association between HIV testing and the partner's level of education. The probability of testing was found to increase with an increased partner level of education.
We also found higher HIV testing rates in urban areas compared with the rural areas. As suggested elsewhere, 25 these variations could possibly be attributed to lower access to testing centres in the rural areas. Our results suggested improved access to HIV testing services over time, with a larger percentage improvement occurring in urban areas.
We found a significant regional variation in testing rates. Testing rates in Nyanza province were still quite low despite the fact that the region has an HIV prevalence of 14.9% (double the national average) and carries 30% of the national burden. 34 Religion has also been identified as a key determinant of HIV testing in many other settings. 25,30,35 Despite the documented evidence of differences between Muslims and those with no religious affiliation, 36 we found no significant difference in HIV-testing rates between the two groups. These low testing rates by Muslims have been linked to adherence to religious tenets that give protection against sexually-transmitted HIV. 29 Adjusting for other covariates, however, we found no association between religion and HIV testing across all three surveys.
In this study we also interrogated the relationships between HIV testing and three constructs (HIV-related stigma, HIV knowledge and media exposure) that have been suggested in other settings. In line with previous studies within sub-Saharan Africa, 28,29,33,36,37,38,39,40,41,42,43,44,45 our unadjusted results suggested a strong and inverse association between HIVrelated stigma and HIV testing. The impact on HIV testing as a result of knowing someone who had died of an HIV-related illness has been mentioned by various authors. 25,30,46 Results from this study showed that knowing someone who has died of an HIV-related illness increased the chances of testing for HIV by more than 70% for both the 2003 and 2008 surveys.
There are several other studies, however, that have suggested that such knowledge has no impact on the HIV testing. 47 Studies also suggest that depending on the relationship with the individual who has died from HIV-related illness and the potential stigma attached to HIV infection, one may not want to know one's HIV status.

Limitations
In this study, information about HIV-testing status was based on self-reported responses. This could potentially be affected by recall bias if the test was offered a long time ago and could also be affected by the tendency of respondents preferring to give socially-desirable answers. Measurement of stigma, a key variable, may not have been satisfactory due to insufficient questions to cover the all the possible dimensions of HIV-related stigma. It is assumed that participants in each of the surveys were independent of each other and that the responses to the questions on testing were also independent. Our analysis did not account for any spatial heterogeneity or autocorrelation that may have been present in the data.

Conclusion
This study used cross-sectional data from the 1998, 2003 and 2008 Kenya Demographic and Health Surveys to determine trends and patterns in VCT uptake in Kenya. We found a marked improvement in testing rates over the last five years but noted that HIV testing amongst women is still quite low. The study has revealed a socioeconomic differential in HIV testing rates and indicates the need to aim health promotion programmes at women of lower income, education and employment cadres. It is therefore necessary to encourage and support the use of VCT, alongside programmes that encourage behavioural change to reduce the incidence of HIV. HIV prevention and control programmes in Kenya need to focus on reducing HIV-related stigma, increasing access to testing in rural area and on programmes that increase access amongst those with little or no education.