Towards Universal Voluntary HIV Testing and Counselling: A Systematic Review and Meta-Analysis of Community-Based Approaches

In a systematic review and meta-analysis, Amitabh Suthar and colleagues describe the evidence base for different HIV testing and counseling services provided outside of health facilities. Please see later in the article for the Editors' Summary


Introduction
HIV is a leading cause of death and hinders economic development globally [1]. In 2010 there were 2.7 million new infections and 1.8 million deaths among the 34 million people with HIV worldwide [2]. Antiretroviral therapy (ART) prevents HIV-related mortality and has transformed HIV infection into a manageable chronic condition [3]. Moreover, ART can virtually eliminate vertical and heterosexual HIV transmission [4,5]. To harness ART's population effects on life expectancy and transmission, people must be diagnosed and enrolled into care early in the course of HIV infection [3]. Learning HIV status can also decrease unprotected sex and provide an opportunity for males without HIV to consider circumcision to reduce susceptibility to HIV [6,7].
Since 2007 WHO has recommended provider-initiated HIV testing and counselling (HTC) to all persons seen in all health facilities in generalised (i.e. antenatal HIV prevalence ≥ 1%) epidemics and in specific services in concentrated epidemics [8]. WHO also recommends couples testing and counselling with support for mutual disclosure in all health facilities and in other settings [9].
Since rapid HIV tests are highly sensitive and specific, they should be used where feasible [ The reasons for limited uptake of HTC are complex and include service-level barriers (shortages of trained health workers, restricted and limited availability of free tests, frequent stock-outs of tests, and a limited number of testing sites) and patient-level barriers (long distances to clinics, limited funds for transportation, lack of HIV awareness, and fear of a positive result due to enduring stigma) [14]. Access to and uptake of HTC also varies by gender and population group.
For example, in generalised epidemics women have higher rates of testing than men, largely because of their more frequent contact with health services. Moreover, adolescents are also poorly serviced by many programmes which primarily rely on facility-based HTC. Although provider-initiated HTC in antenatal care clinics has often been successfully implemented in concentrated epidemics, access and uptake of HTC by key and vulnerable populations remains poor with current implementation approaches. This has highlighted the need to consider community-based HTC approaches to support increased equitable access and earlier diagnosis.
Over the past 20 years a number of countries have offered HTC outside of health facilities through stand-alone voluntary counselling and testing sites. However, this approach is often costly, favours those with higher income and literacy rates, does not reach all people at risk of HIV infection, and is sometimes associated with stigma [14]. Recently, other community-based approaches have been developed to support increased and potentially more equitable access to HIV testing at the district-level [15,16]. Previous research has systematically reviewed the impact of house-to-house HTC in developing countries [17]; however, substantial evidence has accumulated since this review was published. Moreover, the evidence regarding the impact of other community-based HTC approaches has not undergone formal systematic review or synthesis. The objective of this study is to systematically review all community-based approaches of HIV testing and counselling.

Conduct of systematic review
This systematic review will be conducted in accordance with PRISMA guidelines [ be searched for future and on-going studies. Experts in the field will be contacted to identify unpublished research and on-going studies.

Search strategy and selection criteria
The search strategies (Table 1) were designed with a librarian to identify studies including community-based HIV testing and counselling. Per recommendations from the PRISMA Group, eligibility criteria were based on key study characteristics: population, intervention, comparator, outcome, and design [18]. Specifically, studies were included when (1) the study population included people in generalised or concentrated HIV epidemics, (2) the intervention was community and facility-based HTC, (3) the comparator was facility-based HTC alone, (4) the outcomes were CD4 value at diagnosis; testing uptake (i.e. the proportion of the study population accepting HTC); community HIV incidence; linkage to care rates; proportion of people receiving their first HIV test; or cost effectiveness, and (5) the study design was a randomised trial or observational cohort study. Community-based HTC approaches were defined as those outside of health facilities (Table 1). Given the difficulties and expense of conducting large-scale studies for community-based HTC, studies without a comparator arm will also be included. Three of the investigators, ABS, NF, and WA, will independently screen abstracts of all identified articles and then match the full texts of all articles selected during screening against the inclusion criteria. Articles failing to meet these inclusion criteria, or featuring fixed stand-alone voluntary testing and counselling sites, will be excluded from this review.

Data extraction
ABS will complete the data extraction using a standardised extraction form comprising five tables. The first table will summarise the characteristics of study participants. The second table will include information on the community-based testing approaches, including: pre-test demand creation, multi-disease components, study design, linkage to care, provision of incentives, and required number of visits. The third table will summarise the reported outcomes.
Subsequent tables will include studies with a comparator arm. The fourth table will summarise study methods (analytical model used and variables included in the model, information on the comparator arm, and information on outcome ascertainment). The final table will focus on quality assessment.

Statistical analyses
For statistical analyses, studies will be stratified based on the community-based testing approach.
A funnel plot with the effect measures on the x-axis and standard error of the log for the effect measures on the y-axis will be created to assess publication bias and the Egger and Begg tests will be used to test the funnel plot's symmetry. If studies are similar enough to combine, metaanalyses will be performed and statistical heterogeneity will be assessed. Effect measures will be entered as the natural log of the effect measure and standard error as the natural log of (95% upper limit ÷ 95% lower limit) ÷ 3.92 [19]. Fixed-effect models assume that the magnitude and direction of an interventions' effects are identical across studies and that observed differences among study results are due solely to chance [20]. Random-effects models assume that the magnitude and direction of an interventions' effects are not identical but follow a distribution [20]. Since it is possible that the magnitude and direction of community HTC's impact could differ for reasons other than chance, random-effects models will be used for all analyses. An Isquared statistic will be used to measure heterogeneity [21]. I-squared statistics near 25% indicate low heterogeneity, values near 50% indicate moderate heterogeneity, and those above 75% indicate high heterogeneity [22]. If there is moderate to significant heterogeneity in estimates, potential causes, including pre-test community sensitisation; study design; linkage to care; provision of incentives; and required number of visits, will be explored using sensitivity analyses.
STATA version 10.0 will be used for all analyses.

Quality assessment
For the quality assessment, studies will be stratified based on study design (i.e. randomised controlled trial or observational study). Per recommendations from the Cochrane Collaboration, the Collaboration's 'Risk of bias' tool will be used to assess bias in randomised trials with a comparator arm [20]. This tool rates studies based on six criteria in four sources of bias. The presence of random sequence generation for allocation into intervention and comparator arms, and attempts to conceal this allocation, will be used to gauge selection bias. Blinding of study participants, personnel, and outcome assessment during the conduct and analysis of the studies will be used to gauge performance and detection bias. Incomplete outcome data, through review of participants excluded from outcome analyses or lost to follow-up, will be used to gauge attrition bias. Selective reporting of outcomes, time-points, subgroups, or analyses, will be used to gauge reporting bias. A criterion for other forms of bias will also be used. Based on these criteria, studies will be scored out of 100%.
Per recommendations from the Cochrane Collaboration [20], the Newcastle-Ottawa Quality Assessment Scale will be used to assess bias in observational studies with a comparator arm [23].
This scale rates studies based on eight criteria in three sources of bias. Each criterion is worth one point except confounding, which is worth two points. Selection bias will be assessed using four criteria: (1) representativeness of the cohort in the intervention arm the average person in the community from which study participants were drawn, (2) representativeness of the cohort in the comparator arm to the intervention arm, (3) ascertainment of HTC, and (4) demonstration that the outcome was not present at the start of follow up. Adjustment for a patient-level barrier (distance to testing site, income level, or education level) will be used to judge whether appropriate methods were used to address confounding. Measurement bias will be assessed using three criteria: (1) assessment of outcome, (2) adequate follow-up to detect the outcome, and (3) ≤ 30% of participants lost to follow up during the study. Based on these criteria, studies will be scored out of 100%.
For randomised trials and observational cohort studies, studies scoring ≥ 67% will be considered to have a low risk of bias, those scoring 34-66% will be considered to have an unclear risk of bias, and those ≤ 33% will be considered to have a high risk of bias.
The quality of evidence will be assessed using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) system to guide programme managers and other policy makers on national community-based HTC strategies (Appendix, [24]).