First, a major result of this analysis is that across all countries and modes of testing, high percentages of clients reported having received the services around testing that are considered important in most guidelines. As shown in , across the four countries, 89% reported having a pre-test meeting, and more than 75% reported a complete pre-test index. Consent, confidentiality, and overall satisfaction indices were high, at 86%, 79%, and 81%, respectively. Provision of post-test services was generally lower: though 99% received their results, just over half reported a complete post-test index, and 71% of HIV-positive respondents reported a complete referral index. This suggests that there is still a good deal of room for improvement to provide advice and referral after HIV testing, but that services around testing do not appear to be associated with major breaches of autonomy or privacy. When we compared reports by HIV-positive and HIV-negative respondents, we found them to be similar, and where they differed, HIV-positive respondents reported more favorably on the services received, which does not support the notion that HIV-positive individuals receive worse services due to stigma.
A second major result of this analysis is that there were some significant associations between outcome measures and modes of testing, but differences across modes of testing were not consistent with the notion that one mode of testing performs better than the others. As shown in and , unadjusted analyses indicate somewhat more favorable reports at integrated testing sites on a few outcome measures, while analyses adjusting for covariates and clustering of facilities show slightly higher likelihood of better outcomes for VCT. In general, the results do not support the notion that testing at integrated health facilities functions less adequately than at stand-alone VCT sites.
Taken together, these results suggest that the rights of clients, particularly regarding informed consent and confidentiality, can be protected even as testing is scaled up at health facilities. Some studies have suggested that consent and confidentiality may be jeopardized by inadequate facilities and procedures, poor counseling, or the power differential between providers and patients 
, while other studies have documented high acceptability and improved referral, with no reduction in the utilization of health services when testing is routinized 
. The results of this analysis are consistent with studies showing that routine testing is compatible with ethical practices 
, and this should provide some reassurance about the possibility of scaling up HIV testing through PITC without compromising individual protection.
The association of VCT with more positive outcome measures than integrated or PMTCT testing in Uganda—specifically pre- and post-test counseling, receiving results, and satisfaction—can be attributed to the role of nongovernment organizations, which mobilized early on, when VCT was the only mode of provision, to encourage testing, increase awareness, and fight stigma and discrimination. The less favorable outcomes for pre- and post-test counseling in Burkina Faso may be attributable to the more recent initiation of PITC in the country, which had, for a long time, relied on VCT provided via freestanding facilities and national campaigns. PITC facilities are comparatively less resourced in Burkina Faso, and the guidelines in place at the time of the survey did not recommend all of the post-test items that we included in our indices. Instead, the guidelines in Burkina Faso referred to explaining the meaning of results, providing advice on prevention, and answering questions. When measured against more demanding standards, testing services fell short. These results underscore the need to clarify and harmonize guidelines on testing and counseling and to support their implementation.
The data were collected in 2008–2009, at an important time for the scale-up of HIV testing programs. In the countries where the study was conducted, as well as globally, 2006–2007 was a turning point for the increased availability of treatment, the formulation of guidelines, and the implementation of programs to increase knowledge of HIV status. The results reflect the scale-up of these efforts and are likely indicative of subsequent trends.
The generally favorable outcomes that are documented in this study suggest that it is possible to expand client- and provider-initiated testing while protecting patients' rights. This should not be surprising. The standards defined by guidelines on testing, such as informing clients, obtaining consent, and providing information, are not especially exacting, and can be achieved in most settings. Other studies, such as in the United States 
and in Egypt 
, have reported similarly high levels of service provision ad satisfaction.
It is possible that the high values we found in reported measures of satisfaction result in part from a contrast with clients' general experience with other health services, which tends to be suboptimal, so that, by comparison, the quality HIV testing services appears to be satisfactory. It is also possible that because respondents were interviewed at health facilities, they may have felt that they should not appear ungrateful for the services they receive, and hence may have given more favorable answers. In order to assess the extent of such social desirability bias, we conducted a sensitivity analysis comparing outcomes among respondents who tested at facilities where interviews were conducted and those who reported testing at facilities not included in the MATCH project. We found that outcomes among respondents tested at MATCH facilities tend to be similar to, or less positive than, those who did not test at MATCH facilities, suggesting that the facility environment did not substantially influence respondents' reports on the care they received.
Other studies in Kenya and Malawi have found that counseling may be neglected when testing is scaled up, and that clients may perceive testing as compulsory 
. Qualitative data such as were used in these studies are important complements to surveys. They provide insights into potential gaps in the functioning of services, highlighting research questions regarding counseling and consent for further investigation and suggesting areas for improvement. Analyses are underway on the texts transcribed from responses to open-ended questions in the MATCH survey, and from the semi-structured interviews carried out in parallel with the survey. The texts provide a complementary way to assess clients' views, and the analyses should yield more detailed insights into the reasons for clients' positive or negative evaluations of their testing experiences.
There are limitations to this study, including the fact that the outcome measures may not account for important social, cultural, and health service variations. Our measures of referral and medication use do not include specific details, and hence provide only a rough measure of referral. In addition, our analysis distinguishes modes of testing as discussed in the literature, and compares VCT and PITC (including integrated and PMTCT testing), but the reality of services on the ground is complex, and the boundaries among types of testing facilities are not always strict. It is also possible that the lack of significant associations between testing mode and outcomes at the level of individual countries reflects the smaller numbers of respondents in some of the categories of testing, and that larger samples may have been able to detect more associations.
A further limitation comes from using a sample of convenience, based on systematic rather than random selection, and hence not necessarily representative. This strategy was deemed to be appropriate, since the goal was to compare testing experiences across modes of testing. MATCH respondents are more urban, educated, and wealthy than respondents to national surveys—except in Malawi, where they are similar 
. This is not surprising, since users of health facilities tend to have higher socioeconomic status than the general population. Thus, there are reasons to believe that these results are not gross overestimates and that they reflect the reality of testing in the selected areas.
A strength of this study is that it used the same outcome measures and methodology across sites. Hence, results can be compared across countries to address policy questions about modes of testing. While there are considerable differences, both within and among countries, in the structural factors that define health services and accessibility and the meanings attached to testing, the motivations behind it, and its consequences, there is a degree of consistency in the results that likely reflects the state of services around testing in these different settings.
Despite variations within and across countries in the outcome measures for pre- and post-test counseling, consent, confidentiality, referral, and satisfaction, respondents' reports indicate that the frequencies of positive responses on these measures are generally high, suggesting that overall, people testing for HIV do not appear to experience adverse outcomes such as coercion or breaches of confidentiality, and that the majority of people who are tested are satisfied with the services they receive.
The fact that we found few consistent associations between mode of testing and outcome measures is also encouraging. Provider-initiated modes of testing (integrated medical testing and PMTCT testing) were not more likely to be associated with less favorable outcome measures than VCT. These results should provide some reassurance that the efforts to provide HIV testing in different countries through different approaches can help improve knowledge of HIV status and referral to appropriate care.