The comparison of the two PLACE-surveys conducted in two townships in Livingstone in 2005 and 2010 indicated clear improvements over the five year period in condom availability and outreach of peer education activities in venues where people meet new sexual partners, particularly in the intervention community. In addition, interviews with people socializing in the venues indicated marked changes in their sexual behaviour. There were decreases in the reported number of sexual partners and the proportion reporting engaging in transactional sex and increases in reported condom use in both communities. However, the increase in reported condom use with the previous partner met in the venue was particularly sharp in the intervention community where most of the respondents reported obtaining the condom in the same venue. It thus seems likely that the increase in reported condom use in high risk places in the intervention community was partially due to the condom distribution and peer education intervention in these places.
Based on previous research there is mixed evidence to back up the hypothesis that targeting HIV preventive activities at high risk places has an impact on the behaviour of people socializing there. A randomised controlled trial in Nicaragua, found that providing condoms in motels lead to increased condom utilisation among guests, but the presence of leaflets and posters promoting condoms did not [5
]. PLACE-surveys in East London, South Africa, found increased reporting of condom use and a reduction in multiple partnerships among guests socializing in venues where people met new sexual partners over a three year period, and this could possibly be attributed to a behaviour-change intervention targeting these venues [13
]. However, a randomised controlled trial conducted in Kingston, Jamaica, where venues where people met new sexual partners were randomised to a site-based intervention (including on-site HIV testing, condom promotion, and peer education), found no significant differences in reporting of number of partners or consistent condom use between guests in intervention and control venues. However, there were several factors that possibly could explain the lack of impact: implementation difficulties (condoms and educational materials not always being available in the intervention venues), spill-over effects due to patrons visiting both intervention and control venues, national HIV prevention campaigns, time-gap between the intervention and the post-intervention survey, and other interventions being run in some of the control venues [14
The overall objective of this targeted condom distribution and peer education intervention was to reduce the incidence of HIV and other STIs, but we did not measure any biological outcomes of the intervention. Measuring effects on incidence of HIV in a population requires a much bigger sample size and investment. Very few other intervention studies targeting high risk places have attempted to do so. However, reported STI cases indicate a decreasing prevalence of STIs overall in both communities during the intervention period, and the relative decline in reported cases was sharper in the intervention than the control community (38% vs. 16% from 2008-2010) (clinic registries of the public clinics serving the intervention and control communities). The sharper decline in Maramba may be partially due to our intervention, but any attribution must be done with caution. A study in Zimbabwe with a much stronger design, a cluster-randomised trial, which included a peer education and condom distribution component, did not find any impact on HIV and STI incidence [15
Condom distribution to high risk places may obviously be organized in different ways, and using peer educators to do this is possibly not the cheapest (at least if the peer educators are provided financial compensation) and quickest way. Nonetheless, using peer educators provides an additional opportunity to engage people in discussions about HIV prevention. Studies indicate that peers of the same sex are an important source of information about sex-related issues among young people [16
]. It is likely that a higher number of individuals socializing in the venues would have been reached with HIV-related information in this study if the peer educators had been available when the venues were busier. However, it is not possible to distinguish whether the increased reporting of condom use among guests was a result of the combination of condom distribution, condom demonstrations and behaviour change discussions conducted by peer educators or of improved condom availability alone. People who had been in contact with a youth peer educator were not more likely to report using condoms. On the other hand it seems likely that the peer educators would have had a bigger impact if persons of different ages had been recruited although the evidence for peer educator effectiveness from other studies again is mixed. A randomised controlled peer education intervention study among male beer hall patrons in Zimbabwe, which included condom information and demonstrations and recruitment of men of different ages, did not find any impact of the intervention on unprotected sex with non-marital partners [17
], and a review of youth peer education intervention studies conducted between 1998 and 2005 found no impact on condom use [18
]. Nonetheless, a review of studies on youth peer education interventions for HIV prevention in low- and middle-income countries conducted between 1994 and 2008 found that such programs often resulted in increased HIV-related knowledge and increased reporting of condom use, but that there was less evidence for an effect on sexual abstinence and number of partners [19
It is likely that lack of knowledge among new staff explained why not all representatives interviewed in the intervention venues in the follow-up survey reported that condom distribution had taken place there. At the same time, the external monitor revealed that continuous availability of condoms was not fully achieved although this was one of the most important objectives of the intervention. It was expected that the peer educators would need some time to sort out the demand for condoms in the venues, and this may explain why as many as a quarter of the venues lacked condoms in the first monitoring round. Receiving feedback from the monitor probably motivated the peer educators to ensure that they distributed sufficient condoms after this. However, some peer educators seemed to continue to underestimate the demand for condoms, possibly because they did not visit the venue frequently enough, for example in relation to busy weekends.
The improved condom availability in the control community could indicate a trend towards improved condom distribution, i.e. that it was easier for venue owners to obtain free or subsidised condoms that could be given or sold to customers. It is also possible that the increased availability was partly a spill-over effect from the intervention as some of the peer educators lived in the control community and reported that they had distributed condoms in bars and night clubs in their own neighbourhood too since they had not been aware that it would serve as a comparison during the evaluation. Since we found that there was an increase in the proportion who believed that condoms were effective as HIV prevention, it is also possible that venue staff and owners may have experienced an increased demand for condoms from guests, and this may have motivated them to make efforts to offer condoms. The reduction in high risk behaviours reported both among respondents in the intervention and control communities may indicate a general trend. In addition to influence from national campaigns, local prevention efforts carried out by different NGOs in partnership with the DHMT may have had an impact. These have included drama, training of youths, improved VCT and PMTCT services, free provision of condoms and STI treatment services for female sex workers, and promotion of subsidized condoms in high risk places (personal communication with former District Director of Health, Dr. Chinyonga). A behaviour change in the general population would explain the decline in HIV and syphilis prevalence observed among young pregnant women in Livingstone during the period 2002-2008 [4
] and also be in line with behaviour changes reported in other studies in the region [21
The assignment of the intervention was not randomised. Thus we cannot rule out that there were other important differences between venues and respondents in the intervention and the control community, which were not related to the intervention, but which could explain some of the observed changes. Since there was a rather long period of five years between the baseline and follow-up surveys, it is possible that other programs and changes had taken place in both the intervention and the control community. Fortunately, the adjusted logistic regression analyses indicated that differences in types of venues included in the surveys did not influence the main findings. The low number of refusals in the baseline and the lack of refusals in the follow-up surveys may be an underestimation. Since only two interviewers conducted interviews in the same venue, it would be easy for people to shun them if they were reluctant to be interviewed. Thus the respondents interviewed may not have been representative of all the guests in the venues. If people who were willing to be interviewed were more likely to be consistent condom users, we may have overestimated condom use. In addition, we cannot rule out that the intervention itself made some people change what venues they preferred socializing in. People who knew they were taking high risks might have wanted to avoid places where peer educators talked about HIV and prevention.