The effectiveness of VCT in preventing new HIV infections is an evolving field of research, and subject to some debate. An early meta‐analysis of largely cohort data (27 studies performed before 1994, of which only six were African studies) suggested that VCT was mainly a secondary HIV prevention strategy, as existing published data suggested that effectiveness focused on discordant couples and HIV‐positive men.² Subsequently, two large randomised controlled trials,^3,4 one based in developing countries including Kenya, showed that quality controlled VCT approaches led to significant reductions in risk behaviours, new sexually transmitted infections (STIs), or both. Three recent African studies have conflicting outcomes. In the Côte d'Ivoire, condom use increased only in male VCT clients with tuberculosis.⁵ In Mozambique, condom use increased more significantly in VCT clients, but casual sex also increased.⁶ A study of population‐based VCT found no difference in sexual behaviour,⁷ although these were not clients self‐referring for VCT. Rapid testing has been shown to be as effective as counselling around standard HIV testing.⁸

Different counselling and testing service models (such as DCT, stand‐alone sites, or VCT health centre‐based sites) might vary in behaviour change outcomes. This study was conducted in government primary healthcare centre VCT services, nested in a feasibility and costing study. Primary health centres have been shown to be a cost effective⁹ and feasible¹⁰ base for VCT services in resource‐poor settings, and countries are using this approach.^11,12,13 It is important to consider primary healthcare settings for VCT services in addition to both provider‐initiated DCT at primary healthcare centres and VCT stand‐alone sites to enhance VCT coverage (primary healthcare centres reach very rural populations), and integration of VCT with other activities—for example, STI services, family planning. Limiting primary healthcare services to DCT has opportunity costs if VCT can aid prevention.

It has been unclear whether VCT is associated with increased negative life events,^{14,15,16,17,18} hampered as we are by a lack of information on background rates for comparison. Low disclosure rates may affect VCT impacts on behaviour change and stigma.

Here we present the findings of a before–after study of behaviour change and life events for health centre‐based VCT services. We report on risk behaviour, disclosure rates and life‐event outcomes of VCT. This provides insights into the potential prevention benefits and negative outcomes of VCT provision in health centres. There is considerable interest in enhancing primary care HIV and STI services in developed countries, and this study can inform new approaches in the North.

Three health centres were selected systematically after a review of all health centres in Thika and Nairobi districts. Clinics were selected on catchment area (serving large, poor populations) and facility type (standard, with average number and types of staff) to represent government facilities serving distinct poor populations. This study took place before antiretroviral therapy and DCT activities were in place in Kenya, and when VCT existed predominantly in a few urban private clinics, non‐governmental organisations and district hospitals.¹⁹ As rural VCT activities were under‐developed, the study focused on two rural sites in Thika District (clinics A and B) and one urban Nairobi site (clinic C). Formal permission and consent were sought locally. The new free counselling and simple rapid on‐site testing service used existing infrastructure to mirror future government services. Each had one project‐salaried counsellor, with testing performed by nurses and counsellors. Condom demonstrations and free condoms were provided. Further study details are provided elsewhere.¹⁰ Counselling sessions (guideline^20,21,22 based) used a client‐centred²³ approach involving personalised, interactive counselling, and included risk assessment and coping strategies.

Consecutive VCT clients were approached for recruitment. Transport costs and small gifts were given at follow‐up. Enrolment criteria were: age at least 18 years or married, never tested for HIV, regular partner never tested. Fully informed verbal consent was ascertained. Using a McNemar test for the difference between correlated proportions, with estimates from the multi‐centre VCT efficacy trial data,⁴ a sample size was estimated at 586 patients (assuming 30% attrition and power of 80).

Basic demographic data were collected (including socioeconomic score using household amenities at recruitment), before pretest counselling, by field interviewers (not counsellors), trained in all aspects of data collection. The structured face‐to‐face behaviour survey, adapted from a fully validated Kenyan questionnaire,⁴ was completed by the field interviewer after pretest counselling while clients awaited results. At a 6‐month follow‐up visit, a second two‐part behaviour survey was completed by the field interviewer and counsellor, and repeat HIV testing was offered by counsellors.

Self‐reported recall of STI symptoms and treatment over the preceding 6 months was recorded at baseline and for the follow‐up period (~6 months). Numbers of sexual partners were recorded at baseline (over the past 2, 6 and 12 months) and at follow‐up (for the preceding 2 months and follow‐up period). Partner‐specific behaviour (up to five partners for 2 months) for each time point was recorded. Clients were asked to assess their chance of testing HIV positive. Specific life events were recorded at both time points, and personal HIV‐prevention plans and actions were ascertained. Field interviewers were blinded to HIV status.

With prior consent, up to three tracing attempts were made for follow‐up. Disclosure and life‐event data that might reveal HIV status were not collected in field interviews.

Data were protected and double‐entered using FoxPro and validated. All analyses were performed using SPSS for Windows V9 (SPSS, Chicago, Illinois, USA) and Stata V6.0 (Stata Corporation, College Station, Texas, USA). Bivariate comparisons of categorical variables used χ², and means were compared using the two‐sample Wilcoxon rank‐sum (Mann‐Whitney) test, Kruskal‐Wallis tests and paired t test. A two‐tailed α level of 0.05 was used as criterion for significance throughout. The generalised estimating equations technique investigates the effect of baseline explanatory factors (clinic site, gender, age, socioeconomic score, education, HIV status and clinic site) on outcome variables, measured before and after VCT. Odds ratios refer to after VCT, and are adjusted for explanatory variables (clinic, age, sex, marital status and HIV status; these had no effect on any of the outcomes).

As part of a larger operational research study evaluating new strategies for VCT provision, the design has limitations, as it could not include a control group or biological measures of STI. Social desirability bias may have led to under‐reporting of risky behaviours. In addition, risky behaviours and events may prelude a visit for VCT and may not be typical of a person's behaviour and therefore may “regress” with time regardless of any intervention effect. This is another limitation of an uncontrolled intervention study where people self‐select for the intervention. However, in this study, 58% of clients stated no specific trigger for attending for VCT (just wanting to know status), and only 17% said that they had attended because of risky behaviour. Only 19% were internal referrals from within the clinic (eg, STI attenders). High rates of recruitment and follow‐up meant that the study population appeared to broadly represent those eligible, although some under‐recruitment and increased loss to follow‐up meant that the study population does under‐represent HIV‐positive clients. In addition, the participants may not have been representative of VCT attenders in Kenya generally. These data were collected in the pre‐antiretroviral era; any recommendations must be made cautiously, as behaviour may have changed with antiretroviral availability. However, this new era has also moved the focus towards diagnostic testing, and it remains important to consider missed opportunities for prevention.

Many VCT clients failed to perceive their higher HIV risk compared with the general public (reported elsewhere²⁴). There was also poor correlation between individual risk perception and HIV result, as noted elsewhere,²⁵ making counselling and testing a vital tool in this population.

In addition to a reduction in the number of sexual partners, decrease in STI symptoms, and a small increase in condom use, there was also a trend for change in partner type, with a reduction in sexual intercourse with non‐primary partners after VCT. These changes did not vary by gender, age, marital status or HIV test outcome.

The concurrence of a reduction in numbers of sexual partners with reduction in reported STI symptoms suggests a real change, but without a control group it is not possible to conclude that this is due to the intervention. However, the magnitude of the change and the clients' own perceptions of their HIV prevention behaviour before and after the intervention support this interpretation. Behaviour change was mainly via partner reduction not condom use, which may go against social desirability bias.

Behaviour change has been described for other VCT services.^{4,25,26,27,28} It is encouraging that health centre VCT may mirror these successes. It is disappointing that this study, like others²⁶ but unlike the multicentre efficacy trial,⁴ did not find more impact on condom use despite good knowledge and access to condoms. This may be related to low baseline rates of condom use and prevalent negative attitudes towards condoms.

Disclosure rates for HIV‐positive clients were low, although similar to other studies.^17,29 This highlights the importance of seeking improved strategies for disclosure for HIV‐positive clients—for example, using couple counselling.^18,27 High rates of HIV‐negative disclosure suggest low stigma for HIV testing itself.

In comparison with earlier reports,¹⁷ this study concurs with the multicentre trial in recording no significant harm caused by VCT.¹⁸ Our findings need to be interpreted with care in view of the small numbers of HIV‐positive clients involved and low disclosure rates. Importantly, we found high background rates of violence; this has also been noted in studies in other countries.³⁰