Systematic review and meta-analysis of community and facility-based HIV testing to address linkage to care gaps in sub-Saharan Africa

HIV testing and counselling is the first crucial step for linkage to HIV treatment and prevention. However, despite high HIV burden in sub-Saharan Africa, testing coverage is low, particularly among young adults and men. Community-based HIV testing and counselling (testing outside of health facilities) has the potential to reduce coverage gaps, but the relative impact of different modalities is not well assessed. We conducted a systematic review of HIV testing modalities, characterizing community (home, mobile, index, key populations, campaign, workplace and self-testing) and facility approaches by population reached, HIV positivity, CD4 count at diagnosis and linkage. Of 2,520 abstracts screened, 126 met eligibility criteria. Community HIV testing and counselling had high coverage and uptake and identified HIV-positive people at higher CD4 counts than facility testing. Mobile HIV testing reached the highest proportion of men of all modalities examined (50%, 95% confidence interval (CI) = 47–54%) and home with self-testing reached the highest proportion of young adults (66%, 95% CI = 65–67%). Few studies evaluated HIV testing for key populations (commercial sex workers and men who have sex with men), but these interventions yielded high HIV positivity (38%, 95% CI = 19–62%) combined with the highest proportion of first-time testers (78%, 95% CI = 63–88%), indicating service gaps. Community testing with facilitated linkage (for example, counsellor follow-up to support linkage) achieved high linkage to care (95%, 95% CI = 87–98%) and antiretroviral initiation (75%, 95% CI = 68–82%). Expanding home and mobile testing, self-testing and outreach to key populations with facilitated linkage can increase the proportion of men, young adults and high-risk individuals linked to HIV treatment and prevention, and decrease HIV burden.This article has not been written or reviewed by Nature editors. Nature accepts no responsibility for the accuracy of the information provided.


S77
G lobally, there are around 2.3 million new HIV infections annually, 80% of which occur in sub-Saharan Africa 1 . Despite the high burden, only one-third of adults in sub-Saharan Africa have been tested for HIV in the past year and less than 50% of HIV-positive individuals know their status 2,3 . Knowledge of one's serostatus is vital for accessing lifesaving antiretroviral therapy (ART) and linking to HIV prevention. Conventional facility-based HIV testing and counselling (HTC) has not achieved high testing coverage in sub-Saharan Africa and will probably be insufficient to meet UNAIDS ambitious 90-90-90 targets -90% of HIV-positive people knowing their status, 90% of HIV-positive people who are aware of their status on ART, and 90% of people on ART virally suppressed 4,5 . Barriers to facility testing include distance from clinic, long wait times, costs (transportation, lost wages and childcare), confidentiality concerns, low perceived risk and infrequent contact with the health-care system 6 . In addition, patients often present at facilities late in the course of their illness, increasing HIV morbidity, mortality and transmission 7 . Community-based HTC (conducted outside of a health facility) has the potential to overcome these barriers, achieve high coverage, and identify asymptomatic HIV-positive individuals at high CD4 counts 8,9 . In addition, community HTC may reach more men, young adults, and key populations than facility HTC. Community-based strategies also require minimal infrastructure allowing for easier scale up 10-12 . Community HTC modalities include: home, mobile, workplace, index partner/family members (sexual partners or family members of HIV-positive individuals) and as part of a campaign. Uptake and demographics of populations reached can vary widely by modality 9 . A large number of studies on HTC have been conducted in sub-Saharan Africa and a previous systematic review was completed in 2012, but facility testing was not included and uptake in men and young adults was not assessed. In addition, several large-scale interventions have been published since 2012 (refs 11, 13-15). Recently, the World Health Organization released guidelines that strongly recommend implementing community HTC 16 . As most countries have multiple and varying epidemics, UNAIDS recommends creating regional policies tailored to the macroepidemic rather than nationwide approaches 17 . Local policymakers will need to determine the optimal combination of community HTC interventions to increase testing in the context of their country's HIV epidemic.
To provide evidence for decision makers, we summarize the literature on community and facility-based HTC. We characterize each modality by population coverage, since high coverage is beneficial to both HIV-positive and -negative people. HTC can reduce risk behaviour in HIV-negative individuals, while providing a means to link them to primary prevention (including circumcision and pre-exposure prophylaxis (PrEP)) [18][19][20][21] . We evaluate effectiveness in reaching men and young adults (both groups have low HIV testing and poorer clinical Data screening and extraction. M.S., R. Y. and R.V.B. screened abstracts for initial inclusion. Disagreements were adjudicated by reviewing the full text. M.S., R.V.B., R. Y. and G.T. reviewed papers for eligibility and used a standardized extraction form to characterize eligible studies (Supplementary Information 2). Study quality was rated low, moderate or high based on representativeness of underlying population, follow-up (present or absent), assessment of outcomes, and number of outcomes presented. Costs were inflated to 2012 US dollars by converting to local currency units, multiplying by the ratio of each country's gross domestic product deflator (2012 deflator divided by base year deflator) and converting back to US dollars 27 .

Statistical analysis.
Random effects meta-analysis of single proportions with binomial exact confidence intervals (CI) was used to summarize results. Proportions were stabilized using the Freeman-Tukey double arcsine transformation unless the number of events was less than ten, in which case a logit  Coverage is defined as total number of people tested/total number of people in the target population. Uptake is defined as total number of people tested/total number of people offered testing. Bars indicate 95% confidence intervals of random effects meta-analyses. N, sample size; PITC, provider-initiated testing and counselling; VCT, voluntary counselling and testing.

S79
transformation was used because of convergence issues. Heterogeneity was quantified using the I 2 statistic. For modalities with enough data (ten studies or more), trends were examined by year before 2005 (when the HIV rapid diagnostic test was introduced), country and facilitated linkage. Analyses were conducted in R software using the metaprop function in the meta package 28 .

RESULTS
We identified 126 eligible studies out of 2,520 abstracts (Supplementary Figure  S0.a). Overall, 64% of studies were rated moderate or high quality (Supplementary Information 2). Most studies included in our analysis evaluated facility and home HTC. We identified far fewer studies on other types of community HTC: home with self-testing (n = 2), workplace with self-testing (n = 2), index partner/family member (n = 5), key populations (n = 5), campaign (n = 5) and workplace (n = 4). Forest plots of each outcome by modality are provided in the Supplementary Information with pooled estimates presented here. I 2 values of pooled estimates varied from 90% to 100%, reflecting high heterogeneity in study designs and countries included (Supplementary Information). The countries represented varied by outcome with the greatest number of countries having data for home and facility HTC coverage, uptake and tester demographics. Far fewer studies reported CD4 count at diagnosis and linkage to care outcomes; studies containing these data were mainly conducted in South Africa, Kenya and Uganda. All home self-testing studies were conducted in Malawi and the most key population studies were conducted in Nigeria. Overall, the largest number of studies were conducted in South Africa.

DISCUSSION
Across modalities, community HTC successfully reached target groups (men, young adults and first-time testers) with higher coverage than facility HTC (Table 1). High uptake of community HTC reflects acceptability of testing outside of health-care facilities. Community HTC identified HIV-positive individuals with higher CD4 counts who were likely to be earlier in their disease course. Combined with the potential of community HTC with facilitated linkage to achieve high linkage to treatment with similar retention rates as facility HTC, this suggests that scaling up community interventions could reduce the morbidity, mortality and transmission associated with late or non-initiation of ART. Although community interventions test a large number of HIV-negative individuals, HTC can reduce risky sexual behaviour 74 and provide a means to link uninfected persons to primary prevention. This is particularly crucial for young women, who have high HIV incidence and can benefit from PrEP 21 . Preventing HIV infections averts future treatment costs as well as morbidity. A recent modelling study found that ART scale up should be combined with primary prevention such as PrEP to achieve maximum HIV reduction 148 . High coverage of HTC can also reduce stigma around testing.
Each HTC modality reaches distinct subpopulations and a combination of strategies will probably be necessary to achieve high ART coverage. Mobile and campaign HTC had high uptake (97%), as individuals who present at a mobile van or during a campaign are probably seeking out testing, but home HTC also achieved high uptake among people who were offered testing (82%). Home HTC also attained high population coverage, probably because offering testing doorto-door removes substantial barriers, including eliminating the need to actively seek out HIV testing 149 . However, home HTC is less likely to reach men and young adults. A recent home HTC intervention in Botswana reached 85% of women in the target population compared with just 50% of men 150 . This may be because women are more likely to be home at times when the intervention is conducted.
Campaign HTC has the potential to attain high coverage in large catchment areas and identify HIV-positive individuals at high CD4 counts (one-third of newly diagnosed HIV-positive individuals had a CD4 count of 350 cells μl −1 or less compared with two-thirds or more for facility HTC). The multidisease focus of campaigns may reduce stigma of HIV testing interventions. Our results suggest that campaign HTC can be a successful strategy for countries seeking to increase overall testing coverage in a short time frame.
Home HTC with self-testing reached the greatest proportion of young adults of all modalities examined 11 and is a promising strategy with high uptake 151 . Young adults (age 15 to 24 years) represent 39% of new infections in those over 15 years old 23 , but have lower access to HTC and HIV care and poorer clinical outcomes than other age groups 24 . Home HTC with self-testing had slightly lower coverage and reached fewer first-time testers than home HTC administered by counsellors. The World Health Organization recommends HIV self-testing as an option for individuals who are unable or unwilling to receive counsellor-administered HTC. However, supervision improves interpretation of results 151 and a reactive self-test should not be considered a definitive diagnosis, as standard testing is needed to confirm results. More studies evaluating linkage to care following a positive self-test are needed 16 .
Mobile HTC is the most effective strategy for reaching men -a target group in sub-Saharan Africa. Men are more likely to be lost at each step of the HIV treatment cascade; they are less likely to undergo testing, more likely to start ART at an advanced disease stage and more likely to interrupt treatment -all of which leads to increased morbidity and mortality 22 . Qualitative studies highlight men's preference to test outside of facilities 152 , so scale up of community interventions can meet this need. Future studies could investigate HTC at predominantly male workplaces, nightclubs or bars.
Index testing of sexual partners through active contact tracing is an efficient high-yield method that should be scaled up. HIV positivity was 55% in this group and the intervention attained a high coverage (41%). The HIV prevalence we report is similar to that found in the literature -45-50% in cohabitating partners of HIV-positive adults, most of whom are unaware of their status 48 . Interestingly, high coverage of males was achieved only through active contact tracing, whereas passive tracing identified more women (Supplementary Figure S18). Facilitated linkage strategies are a key component of successful community-based HTC. Individuals testing at an HIV facility generally have higher rates of linking to care and initiating ART than those who test outside the health-care system. However, we found that high linkage rates (comparable with, or higher than, facility HTC) can be achieved with community HTC when individuals are followed-up to encourage linkage.
Although scaling up community HTC with facilitated linkage is important, the benefits of improving facility HTC coverage should not be overlooked. Consistent with previous studies, our analysis finds opt-out facility PITC had much greater uptake than referring patients to VCT 56 . However, coverage of PITC in health facilities is low, demonstrating missed opportunities to identify HIV-positive individuals and to link them to care. For example, a Ugandan hospital reported only 50% of inpatients with HIV-related diagnoses were tested for HIV before leaving the hospital 86 . PITC is an underused strategy in sub-Saharan Africa and scaling up testing would provide a safety net for those who do not independently seek HTC 61,112 . Because PITC identifies mainly symptomatic HIV-positive individuals with low CD4 counts as well as those with health-care access, it should be coupled with other modalities to maximize population coverage.
Our review identified gaps where additional evidence is needed. A large proportion of CD4-count and linkage data came from South Africa, with Uganda and Kenya also well represented. South Africa has the lowest percentage of firsttime testers, reflecting the successful scale-up of HTC. There are fewer studies from other parts of sub-Saharan Africa, which may limit how much the pooled estimates can be generalized. Also, few studies followed patients longitudinally and measured linkage to care, ART initiation, retention and viral suppression. In addition, although many studies evaluated home HTC, more data are needed for other community modalities, including campaign and workplace.
Data were also limited for key populations. Despite having an HIV prevalence up to eight times higher than the general population, interventions for key populations are scarce and scale up is urgently needed 115,153 . Key population interventions can reduce the spread of HIV in the general population 154 . Currently, numerous policy barriers exist that restrict the availability and access of HIV-related services for MSM and CSWs, including police harassment and criminal laws 155 . Only three HTC interventions were targeted to MSM and only one was targeted to CSWs and PWIDs. Most key population HTC studies were from Nigeria; data are needed from other parts of sub-Saharan Africa. We report a high HIV positivity combined with a high proportion of first-time testers in MSM and CSW groups, highlighting the need for service expansion. We found a lower HIV prevalence in PWIDs compared with MSM and CSW groups, reflecting sexual transmission as the main mode of HIV spread in sub-Saharan Africa. Successful HTC programmes for key populations are community based (particularly mobile) as many high-risk groups are marginalized and do not have access to conventional health systems 122 . Community-based HTC for MSM and PWIDs have been shown to have higher acceptance and greater HIV yield than clinic referral for HTC 115 . In addition, self-testing is a potential strategy to reach key populations, as it demonstrates high acceptability and is considered convenient and private 156 .
Costs of community-based and facility-based HTC vary by modality, country, scale of intervention, linkage strategy and costs included. Generally, community-based HTC and integrated facility HTC costs were comparable. However, standalone HTC had the highest cost per person tested, indicating that integrated HTC may be more cost-efficient than stand-alone services (Supplementary Table S3).
The limitations of our analysis included the heterogeneity across studies, which may not be accurately reflected in the pooled estimates. Differences in study design, geographical location (country, urban or rural area) and intervention year added to the heterogeneity. To address this, we used random effects meta-analysis and stratified on key variables (year <2005, country and facilitated linkage). In addition, large numbers of HIV-positive individuals were lost to follow-up in studies that reported linkage, so we considered these individuals unlinked in our analyses. If individuals linked at another clinic, our estimates may be conservative 157 . Furthermore, assessment of linkage to care differed by study (self-report or clinic records review), as did time to linkage assessment, which varied from 1 to 12 months after HTC. In addition, CD4 count at diagnosis and ART uptake among those with eligible CD4 counts could only be assessed in community HTC interventions employing point-of-care CD4, as studies that report CD4 only for those visiting a clinic would not provide accurate denominators. Only studies reporting linkage to care among those eligible for ART were included in our main analysis. Also, estimates of coverage vary in their precision because some studies conducted population enumeration and others used census estimates of the catchment area. Finally, proportion of first-time testers, men and young adults tested are crude measures of relative uptake. For example, for home HTC, it is not possible to discern whether the 40% of those tested being

S83
men reflects a lower coverage of men, or a greater coverage of women, or a combination of the two. Future studies reporting the number of men, firsttime testers and young adults offered testing compared with those accepting testing would increase the accuracy of these measures. Our findings on uptake, HIV positivity and CD4 count at diagnosis are similar to a previously published meta-analysis 9 . This analysis characterizes linkage and populations reached by HTC modalities to inform policymakers who are charged with addressing gaps in testing. Facility HTC, although important, is unlikely to be sufficient to curb the HIV epidemic because many people in sub-Saharan Africa do not have regular access to health care. Scaling a combination of community HTC, mobile testing to reach men, self-testing to reach young adults and outreach to high-risk populations, as appropriate to the local epidemic setting, is crucial to achieve high knowledge of serostatus and linkage to HIV treatment and prevention in sub-Saharan Africa.