Introduction Testing for HIV is the first step in the cascade of care for HIV-positive individuals who need treatment. Knowledge of HIV status is also an important part of HIV prevention, for both HIV-negative and HIV-positive individuals, and developing innovative means to increase uptake of testing has recently been identified as an international policy priority –. Despite some progress, knowledge of HIV status remains low in sub-Saharan Africa (SSA), where HIV prevalence is highest . National population surveys in six sub-Saharan African countries found that amongst participants living with HIV there was a wide range in the proportion of respondents aware of their status (from 31% in Congo to 69% in Kenya) . Men have consistently been shown to be less likely to have been tested for HIV than women ,. Out-of-facility approaches to offering testing in the community , and the workplace , are means of bringing access to testing closer to clients. By removing distance as a barrier, these methods could be complementary means to scale up HIV testing . Home-based voluntary counselling and testing (HBT) has been suggested as an effective out-of-facility approach for identifying HIV-infected people at an earlier stage of their disease and for enrolling them into care and treatment in a timely manner –. The World Health Organization has recently published a handbook to guide service providers and policy makers in delivering HBT . HBT may reach individuals that community venue-based testing and workplace testing do not because it does not require clients to come forward . In HBT it is the test provider who approaches the client, regardless of his/her perceived risk of having HIV . However, there is uncertainty about HBT and concern that it may be poorly accepted or even harmful, partly owing to the enduring climate of stigma and discrimination around HIV/AIDS in many settings ,. We carried out a systematic review and meta-analysis of the available evidence regarding the acceptability of HBT in SSA, and assessed a number of potential determinants of uptake and programme success. Methods We conducted this systematic review and meta-analysis based on a pre-defined search protocol (Text S2) that conformed to the criteria set out by the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) group  and was in accordance with the PRISMA statement (Text S1). The specific objectives of the study were to summarise the following proportions: “accepted” (or uptake), defined as the proportion of all individuals offered HBT who accepted and had an HIV test performed at home; “received”, defined as the proportion of all individuals who accepted a home-based HIV test who subsequently obtained the result of the test; and “overall”, defined as the proportion of individuals who received a test result among all those offered HBT (including refusals). We also planned sub-group analyses as outlined below. Search Strategy We aimed to summarise studies that described uptake of HIV testing provided at home in SSA. We screened studies published between 1 January 2000 (the onset of programmes providing antiretroviral therapy in SSA) and 24 September 2012. The following study designs were permitted: randomised controlled trials, observational cohort studies, cross-sectional surveys, and programme evaluations. Any study that described an intervention to provide HIV testing at home and reported proportions accepting HIV testing out of all individuals offered a home-based HIV test was included. Where acceptance of testing was reported, it was assumed that testing was performed unless stated otherwise. To avoid duplication we excluded reports that pooled data from previously published studies, and where there was substantial overlap of study participants, we included the study with the most complete information. No language or age restriction was applied to the literature search. To identify studies for this systematic review, we searched the online databases PubMed, Embase, and Global Health (Ovid) and manually searched the bibliographies of relevant articles. We included only peer-reviewed journal articles; conference abstracts were excluded. Anticipating overlap between studies reporting HBT and other community-based strategies, we developed a broad compound search strategy that combined terms for “HIV”, “voluntary counselling and testing”, “home based”, “mobile”, “community”, “workplace”, “couples”, and “self”. We then combined these terms with the individual names of countries in SSA (Text S2). Finally, we excluded all studies that did not report home-based delivery of HIV testing. Eligibility of abstracts and journal articles was determined by one investigator (K. S.) and verified by a second researcher (R. V. d. B.). Two investigators (K. S. and R. V. d. B.) then independently extracted data on study characteristics and outcomes using a standardised form. Any disagreements regarding eligibility or outcome data were resolved by a third investigator (N. F.). The rigour of study processes and research methods was examined using pre-defined criteria, but studies were not excluded for quality reasons. Data Synthesis and Analysis We calculated the proportion of people who accepted HIV testing at home, and the proportion who received their test result out of those (i) who were offered and (ii) who accepted testing. The variance of raw proportions was stabilised using a Freeman–Tukey arcsine square-root transformation, and proportions were then pooled using a DerSimonian–Laird random-effects model –. Pooled odds ratios were calculated for proportions stratified by gender, also using a random-effects model. We report the I 2 statistic to assess the proportion of variability due to between-study heterogeneity, but as this estimate is known to increase as the number of participants contributing to the meta-analysis increases , we also report τ2 as a measure of between-study variance (reported on the arcsine square-root scale). We explored potential sources of heterogeneity through univariate sub-group analysis to determine the potential influence of the following covariates: HIV prevalence ( 13 y and eligible childrenb Counsellors trained for purpose Yes No FP RDTs (Determine and Bioline) 26%c Kranzer, 2008 Malawi, rural district 2005–2006 2,047 Factors associated with HBT refusal 11.4% 18–59 y Trained local VCT counsellors No No Venous blood sampling for ELISA and particle agglutination testing in laboratory 36% Lugada, 2010 Uganda, five rural districts 2005–2007 4,798 Uptake of HBT versus clinic-based testing in household members of HIV-positive index patient 5.6% Any Trained lay field workers No No FP RDTs (Determine screening, Unigold confirmation) Not specified Maheswaran, 2012 South Africa, rural district 2009 1,726 Uptake of HBT and community mobile HIV testing and factors associated with HBT versus mobile testing 22% ≥15 y HIV Counsellors No No Not specifiedd 40% Matovu, 2002 Uganda, rural district 1999–2000 11,709 Uptake of HBT and effects on sexual risk behaviour and HIV acquisition 5.6% 15–49 y Counsellors No No Venous blood sampling for ELISA (×2) testing in laboratory 55% Menzies, 2009 Uganda, setting not specified 2003–2005 49,470 Comparison of four testing approaches: door-to-door HBT, household member (of index HIV patient) targeted HBT, stand-alone, hospital-based VCT 5.6% Any Not specified Yes No FP RDTs (screening test followed by confirmation if HIV-positive; tests not specified) 10% Michelo, 2006 Zambia, one rural, one urban district 2003 5,445 HIV prevalence survey 20.4% 15–59 y Not specified No No Bionor saliva test and “serum test” for saliva-positive or second saliva test Not specified Molesworth, 2010 Malawi, rural district 2007–2008 16,894 To assess the performance of HIV RDTs in a HIV prevalence survey 11.6% ≥15 y Non-laboratory basic health personnel Yes No Venous blood sampling for RDTs (Determine and Unigold in parallel pre-May 2008, serially post-May 2008) Not specified Negin, 2009 Kenya, rural province 2008 2,033 Feasibility, acceptability, and cost of HBT 7.8% 15–49 y Lay counsellors Yes No FP RDTs (Determine and Bioline) Not specified Sekandie, 2011 Uganda, urban district 2009 588 Uptake of HBT and factors associated with HBT 6.5% ≥15 y Trained nurse counsellors No No FP RDTs (Determine screening, Statpak confirmation) 61% Shisana, 2004 South Africa, nationwide 2002 9,963 HIV prevalence survey 26.5% ≥2 y Nurses No Yes—money provided to head of household FP onto filter paper; ELISA (×2) testing in laboratory Not specified Tumwesigye, 2010 Uganda, rural district 2004–2007 282,857 Acceptability and uptake of HBT 5.4% >14 y and eligible children >18 mof Counsellor and laboratory assistant teams Yes Yes—HIV-positive provided with condoms, insecticide-treated bednets, and home water treatment equipment FP RDTs (Determine screening, Statpak confirmation) 9% Welz (1), 2007 South Africa, rural district 2003–2004 19,867 HIV prevalence survey (residents) 27.9% Women 15–49 y; men 15–54 y Trained fieldworkers No No FP onto filter paper; ELISA (×2) testing in laboratory Not specified Welz (2), 2007 As above As above 916 HIV prevalence survey (subset of migrants in the community) As above As above As above No No As above Not specified Were, 2003 Uganda, rural district Not specified 2,373 Uptake of VCT and HBT 4.1% Any Not specified No No Venous sampling, tests not specified Not specified Were, 2006 Uganda, two rural districts 2003–2004 3,338 HIV prevalence and acceptability of HBT among household members of HIV-positive index patient 4.1% Any Counsellors No No FP onto filter paper; ELISA (×2) testing in laboratory; for children 200 cells/mm3) ,. 10.1371/journal.pmed.1001351.t002 Table 2 Assessment of study rigour. First Author, Publication Year Study Process Quality Indicators Research Method Quality Indicators Pre-Test Counselling Donea Consent Provided Test Offered with the Intention of Giving Results to Clientsb Confirmatory Laboratory Testing Done Discordant Results Addresseda Repeat Sampling if Discordant Repeat Visits if Absenteeism Specific Advice if HIV Result Negative Linkage to Care for HIV-Infected Sampling Strategy Described Selective Outcome Reporting Angotti, 2009 Yes Yes Yes No Not specified No No Yes—retest in 3 mo time Yes Yes No Choko, 2011 Yes Yes No No Yes Yes No No Yes Yes No Helleringer, 2009 Yes Yes No No Not specified Yes Yes No No Yes No Kimaiyo, 2010 Yes Yes Yes No Yes Yes Yes Yes—behaviour change and “ABCs” of HIV prevention Yes No No Kranzer, 2008 Yes Yes Yes Yes Yes Yes Yes No No Yes No Lugada, 2010 Yes Yes Yes No Yes Yes Yes No Yes Yes No Maheswaran, 2012 Yes Yes Yes No Not specified No No No Yes No No Matovu, 2002 Yes Yes No Yes Yes Yes Yes No No No No Menzies, 2009 Yes Yes Yes Yes Not specified Yes No No Yes No No Michelo, 2006 Yes Yes Yes Yes Yes Yes No No No Yes No Molesworth, 2010 Yes Yes Yes Yes Yes Yes No No Yes Yes No Negin, 2009 Yes Yes Yes No Yes Yes No No Yes No No Sekandi, 2011 Yes Yes Yes No Yes Yes No Yes—HIV prevention counselling Yes Yes No Shisana, 2004 Not specified Yes No Yes Not specified Yes Yes No No Yes No Tumwesigye, 2010 Yes Yes Yes Yes Yes Yes No No Yes No No Welz, 2007 Not specified Yes Yes Yes Not specified Yes Yes No No Yes No Were, 2003 Yes Yes No No Not specified No Yes No No No No Were, 2006 Not specified Yes Yes Yes Yes Yes No No No No No Wolff, 2005 Not specified Yes No Yes Yes Yes No No No Yes No a Where no information is available “not specified” is indicated for these variables, as we considered it possible that these activities were done but not reported in the paper. b Some studies offered testing but results were not promised, e.g., results available only if client sought the result separately; some studies were entirely blinded, e.g., where testing was done for anonymous population HIV prevalence estimation. Proportion of Individuals Accepting Testing and Receiving Results A total of 524,867 people were offered HBT across the 21 studies, which ranged in size from 216  to 282,857  people. Twelve studies disaggregated data on offer of HBT by gender, with 180,942 men and 198,042 women offered testing overall –,–,. The proportion of those offered testing who were men (in the studies that reported on gender) ranged from 22% to 49%, with an overall proportion of 47%. Across all 21 studies the proportion of people who accepted HBT ranged from 58.1% (95% CI: 57.5%–58.8%) to 99.7% (95% CI: 99.7%–99.8%), with a pooled proportion of 83.3% (95% CI: 80.4%–86.1%) accepting to be tested (n = 474,377) (Figure 2). Heterogeneity was high (τ2 = 0.11). In studies that reported on acceptance of HBT by gender (eight studies) ,,–,,, men were as likely as women to accept testing (78.5% [95% CI: 71.1%–86.0%] versus 81.5% [95% CI: 72.9%–90.1%]). The pooled odds ratio of men accepting HBT was 0.84 (95% CI: 0.56–1.26) compared to women (τ2 = 0.33). Studies that offered targeted HBT to household members of index HIV-positive individuals , achieved higher proportions of uptake than the other studies: 94.0% (95% CI: 82.4%–100%) versus 80.6% (95% CI: 77.2%–84.0%) (p 350 cells/mm3 (and 68% >200 cells/mm3) . Similarly, Menzies et al. found that 69% of HIV-positive individuals identified through HBT had CD4 count >200 cells/mm3. In this latter study, which compared approaches of HIV testing, the proportions of HIV-positive individuals identified with a CD4 count <50 cells/mm3 through stand-alone voluntary counselling and testing and hospital-based testing were 20% and 24%, respectively, while the corresponding proportion was 12% for HBT targeted to household members of known HIV-positive individuals, and 6% for untargeted HBT . This is consistent with other findings that suggest HBT is a useful approach for earlier detection of HIV, initiation of treatment, and better prognosis ,, as well as for higher impact with treatment as prevention ,. A recent pilot study in South Africa found a reduction in mean community viral load 6 mo after the introduction of a HBT campaign . While women are disproportionately affected by HIV in SSA , men have long been known to under-utilise HIV services and to present later for care than women, and consequently they have worse outcomes on treatment ,,. In the studies reviewed here, an overall proportion of 47% of those offered testing were men. This compares favourably with facility-based testing, where testing of males attending the clinic may be as low as 9% . In our meta-analysis of HBT, an almost equivalent proportion of men were offered a test as women, and they were as likely to accept testing, an outcome that gives promise of greatly improving awareness of HIV status for both sexes. Studies that provided results at a distant site even if testing was conducted at home were associated with lower proportions of people receiving results out of those who accepted testing. While this emphasises the benefits of HBT including immediate provision of results in raising awareness of HIV status, it may be of less concern given that rapid diagnostic tests with immediate results are now the norm for voluntary HIV testing globally. Examination of trends by country suggest lower uptake of testing in South Africa, where three out of five studies reported uptake of ≤70% (note that two of these studies were in the same setting in KwaZulu-Natal) (Figure 1) ,. However, the most recent study from South Africa found very high uptake of HBT (91.8%) . Based on the paucity of countries and the number of studies per country available for this review, it would be unwise to draw conclusions about country differences and acceptability of HBT. While the results of sub-group analyses need to be interpreted with caution, they suggest that the running of pre-test sensitisation campaigns may be of little benefit in terms of uptake of HBT. However, these are essentially “ecological” comparisons, which may be confounded by many other differences between the study populations examined. Also, the number of studies where incentives were given was very small (Table 1), and strong conclusions cannot be drawn. Nevertheless, the fact that most of the studies demonstrated similar proportions of uptake of HBT perhaps argues against a strong effect. The finding that studies with a lower proportion of individuals previously tested for HIV (<30%) had a higher frequency of test uptake points to the value of HBT as an effective approach to engage those not previously aware of their HIV status in testing. It could suggest that HBT is effective in achieving initial diagnosis but less so for repeat testing. Targeted HBT of index HIV-positive clients' household members may be an effective way to achieve higher acceptance in settings where more general HBT is not feasible because of resource limitations. Uptake of HBT may be influenced by availability of treatment, as indicated by the fact that the study with the lowest overall success (only 25% of people offered a test received their result) was done at a time when antiretroviral treatment was not available in the communities studied  (although overall there was no effect of “study period”). However, there may be other confounding factors involved, and this study was based on a small sample size; in sensitivity analysis, excluding it from the analysis did not change the pooled estimate of uptake of HBT (data not shown). Three other studies were notable for having <70% receipt of results amongst those who accepted HBT (Figure S1). Two of these studies offered the option of receiving results at a later date ,, while the third  offered the option of receiving results on the same day. Human rights protections should be an integral part of any testing campaign, and every effort should be made to avoid physical, social, and psychological harm to individuals ,. However, the high level of uptake we have found overall seems to indicate acceptability of HBT in the communities studied. There are several strengths and limitations to this review. We used a broad search strategy that allowed us to capture 21 studies (published in 19 articles), resulting in a large overall sample size and giving increased confidence in the pooled estimates. There was high statistical heterogeneity, as expected for pooled proportions in observational studies. We limited our search to studies conducted in SSA over the last decade in order to improve comparability, and used a random-effects model to pool data. We undertook a number of sensitivity and sub-group analyses to explore potential sources of heterogeneity. The non-uniformity of the studies, which were nonetheless looking at uptake of a “uniform” activity (the offer of an HIV test at home), could be considered both a strength and limitation of our review. While it may be a limitation for pooling results, it could be considered a strength that even in a range of study contexts, HBT consistently achieved higher uptake than is seen in facility-based testing. Another limitation was that, as a trade-off to using a broad search strategy, our search was limited to just three databases and to published articles in peer-reviewed journals. We therefore cannot rule out the possibility that we may have missed some studies, or the possibility of publication bias leading to the non-publication of studies with lower uptake. The limited number of studies that provided data on the health status of those identified as HIV-positive by HBT is a further shortcoming that this review was unable to address. Our findings do, however, indicate a number of directions for future research. In particular, key areas for research include linkage to care following HBT, retention in care of those identified HIV-positive through HBT (who are more likely to be clinically well when diagnosed), as well as repeated HBT for ongoing knowledge of HIV status. The option of self-testing with support from HBT staff is an area of research that is highly topical given recent developments in self-testing . The suggestion from this review that the conduct of sensitisation campaigns has little or no impact on uptake of HBT and receipt of results has important implications for programme cost and efficiency and deserves further evaluation. More data are needed on the effectiveness of HBT in detecting previously undiagnosed HIV infection. Sustainability and cost considerations (short- and long-term) are important to help guide policy, and further work on cost-effectiveness is required. Further research on individual-level factors associated with participation in HBT, such as that recently published by Cherutich et al. , would inform implementers on individuals who require further engagement to encourage uptake. A key finding of our review is that HBT is able to reach wide sections of communities in a diverse range of contexts and settings. HBT provides the opportunity to acquire knowledge of HIV status at the doorstep for those who may not otherwise have sought testing, and may be pivotal in providing an effective tool for governments and health service providers to increase access to HIV treatment and prevention, by increasing uptake of testing. We conclude that HBT has the potential to dramatically increase awareness of HIV status in previously undiagnosed men and women in SSA. HBT is a gateway to accessing care early, and the benefits for individual and public health, both for treatment and prevention, make it an invaluable tool in the fight against HIV. Supporting Information Figure S1 Proportion receiving result of HBT. (TIFF) Click here for additional data file. Table S1 Studies reporting individual-level predictors of uptake of HBT. (DOCX) Click here for additional data file. Text S1 PRISMA statement. (DOCX) Click here for additional data file. Text S2 Search protocol. (DOCX) Click here for additional data file.