A rapid review and synthesis of the effectiveness of programmes initiating community-based antiretroviral therapy in sub-Saharan Africa

Background Community-based antiretroviral therapy initiation (CB-ARTi) has the potential to reduce attrition by increasing access to care, reducing patient costs, decongesting clinics and ensuring improved uptake of ART. There is a paucity of research that identifies successful implementation of CB-ARTi in sub-Saharan Africa (SSA). Objectives The aim of the study was to review and describe the evidence on the effectiveness of CB-ARTi programmes that start ART in communities in comparison with the current standards of care in SSA. Methods A rapid review of grey and published peer-reviewed literature between January 2009 and July 2019, by using PubMed, PDQ-Evidence, Google Scholar, clinical trial databases and major HIV (human immunodeficiency virus) conference websites, was conducted. Search terms used included ‘community-based’, ‘home initiation community models’, ‘antiretroviral therapy’, ‘clinical outcomes’, ‘viral suppression’, ‘retention in care’, ‘loss to follow-up’, ‘HIV’ and ‘sub-Saharan Africa’. Results The search yielded 90 articles and reports following the removal of duplicates. After initial screening and full-text screening, six articles remained and were included in the qualitative narrative synthesis. This included four randomised control trials and two cohort studies of specific interventions comparing CB-ARTi with the standard of care in SSA. There is evidence that CB-ARTi can increase access to HIV-testing services, linkage to ART, retention in care and viral suppression rates and is possibly not inferior to facility-based healthcare. Conclusion CB-ARTi has the potential to increase access to HIV services to people living with HIV in SSA. The results mentioned previously suggest that CB-ARTi models could prove to be equal and possibly not inferior to facility-based ones and warrant further investigation.

Introduction systematic reviews and meta-analyses include communitybased HIV testing, facilitated referrals for ART initiation, education, treatment supporters and active adherence reminder devices such as mobile phone text messages. However, these have shown mixed results, with each intervention also being found not to produce significant effects in some settings. 5 Despite this, interest is increasing in the role of out-of-facility or community-based ART initiation (CB-ARTi) as an essential approach for universal access to HIV care.
Community-based ART initiation has the potential to reduce attrition by increasing access to care, reducing patient costs, decongesting clinics and ensuring improved uptake of ART. 3,6,7 There is a need for the updating of the status of evidence that supports the implementation of CB-ARTi models. Such evidence should include its impact on the clinical and behavioural outcomes such as retention in care and viral suppression amongst patients initiating ART in SSA. The objective of this article is to review and describe the evidence of the effectiveness of CB-ARTi programmes that start ART in communities in comparison with the current standard of care in SSA, namely the initiation of ART in traditional facility-based hospitals and clinics.

Methodology
This rapid review used a streamlined systematic method to capture the evidence from current community-based approaches to the initiation of ART in SSA. There were two research questions. (1) What are the essential elements of evidence-based models of CB-ARTi that inform policy in SSA? (2) How do the reported clinical outcomes, for example, retention in care and viral suppression, amongst patients initiating ART in community-based settings compare with traditional standards of care in SSA?
The rapid review approach streamlines traditional systematic review methods to gain efficiency and accelerate the review process, whilst still aiming to produce valid conclusions. 8 We focussed on a narrow topic, used a limited rather than an exhaustive range of search terms and restricted the analysis and synthesis. We also restricted the search of grey literature (material written for professionals and disseminated outside of peer-reviewed journals) to key websites and only considered studies published in English since January 2009. Furthermore, we performed single (vs. dual) screening of titles and abstracts by using Rayyan, a web-based semi-automated screening software (https://rayyan.qcri.org/welcome).

Search strategy
Studies were identified through bibliographic searches of PubMed and PDQ-Evidence publication databases by using the following terms and variations: 'community based'; 'home initiation'; 'community models'; 'home care services'; 'health facility'; 'clinic'; 'antiretroviral therapy'; 'antiretroviral'; 'clinical outcomes'; 'patient adherence'; 'patient compliance'; 'viral load'; 'viral suppression'; 'retention'; 'retention in care'; 'loss to follow-up'; 'attrition'; and 'HIV'. The search strategy was limited to studies conducted in SSA and published in English from January 2009 through July 2019. The PubMed search strategy is summarised in Box 1.
We also conducted a grey literature search that was limited to abstracts from the following major HIV-related conferences: Conference on Retroviruses and Opportunistic Infections (CROI), the International AIDS Conference, International Conference on AIDS and STDs in Africa (ICASA), Southern African HIV Clinicians' Society Conference and the South African AIDS Conference. In addition, we searched clinical trial databases for planned or ongoing research via the U.S. National Library of Medicine Register of Clinical Trials (clinicaltrials.gov) and the WHO (World Health Organisation) International Clinical Trials Registry Platform (ICTRP). Three technical experts on HIV care and treatment were also contacted by e-mail for comment on the appropriateness of the identified literature and additional articles.

Study inclusion criteria
Studies were included in the review if they: (1) reported on community-based models of ART initiation; (2) measured clinical outcomes (e.g. linkage to care, adherence, reported stigma, retention in care and virologic suppression) of patients initiated on ART in the community compared with patients receiving the current standard of care; and (3) used observational and experimental methods with primary data within randomised controlled trials (RCTs), crosssectional or cohort (prospective and retrospective) designs. We excluded articles that did not meet all three criteria from the review.

Study selection and data extraction
All references captured by the search engine were uploaded to Rayyan and duplicates were identified and removed. The eligibility of the articles was assessed in two steps, the first entailing all titles and abstracts being assessed by the primary reviewer. In the second step, full texts for all the articles deemed relevant to the research questions were read in full and analysed to confirm their eligibility. The primary reviewer extracted data from all eligible articles by using a standard data collection form to collect information about the publication date, study setting, design and interventions,

Quality evidence assessment and risk of bias
The quality of evidence and risk of bias (RoB) of the studies included in the final review was assessed by one reviewer by using the McMaster University's Quality Assessment Tool (https://merst.ca/ephpp-tools/) from the Effective Public Health Practice Project (EPHPP). The secondary reviewer verified the quality of evidence and the RoB assessment.

Ethical consideration
This article followed all ethical standards for research without direct contact with human or animal subjects.

Study selection and characteristics of included studies
The process of study identification and selection concluded with 90 citations being identified after removing duplicates ( Figure 1). After screening the titles and abstracts, six full-text articles were selected for critical review (Table 1). These consisted of four completed studies and two ongoing RCTs ( Table 2). Four studies were included for the qualitative synthesis of patient outcomes, which varied between studies, and included case finding (HIV seropositive yield), linkage to ART initiation, ART adherence, retention in care and viral suppression (Table 3).
We excluded all studies that did not report on ART initiation in the community (out-of-facility). Interventions were implemented in Lesotho, Malawi, Nigeria, South Africa, Tanzania and Uganda. The four completed studies were assessed by using the McMaster University's Quality Assessment Tool, and individual studies ranged in quality from 1 (strong) to 3 (weak), with the overall average being 2.
The two ongoing RCTs included in the model description were not assessed for quality as they had not been completed at the time of reporting.

Models of community-based antiretroviral therapy initiation
Community-based ART initiation modalities include home, mobile and workplace as part of an HIV-testing campaign. The review identified two main models of CB-ARTi, with some variations across countries ( Table 2): (1) on-site ART initiation and community-based ART monitoring and resupply 7,9,10 and (2) a hybrid model with on-site community ART initiation and referral to local clinics for ART monitoring and resupply. 3,4,11 In addition, the review identified key activities addressing seven areas of CB-ARTi service delivery, namely (1) demand creation, (2) patient access for HTC, (3) provider roles, including task shifting, (4) laboratory and clinical evaluation, (5) adherence preparation, (6) place and time of ART initiation and (7) follow-up care. 3,4,9 Examples of these CB-ARTi programme activities, including summaries of information on the purpose of the activity, populations targeted and strategies used, are provided in Table 2.

Effects on clinical outcomes
Two completed RCTs, 3,4 one quasi-experimental prospective cohort study 9 and one retrospective interrupted time series cohort study 11 were included for the qualitative synthesis of patient outcomes. Two of the studies reported on HIV test uptake and case finding, all four on linkage to ART initiation, three on retention in care and one on viral suppression (Table 3).

Human immunodeficiency virus test uptake and case finding
One RCT reported on HIV test uptake, whilst one RCT and one retrospective interrupted time series cohort study reported on HIV case finding as an outcome. In urban slums in Malawi, MacPherson et al. 4 reported that there was no significant difference in the uptake of HIV-self-testing (HIVST) kits between offering optional home initiation of HIV care after self-testing (home group) compared with HIVST followed by facility-based services only (facility group). However, participants in the home group were more likely to report a positive HIVST result (6.0%) compared with the facility group (3.3%). In addition, the median CD4 T-cell count at ART initiation was higher amongst home (219 cells/µL) than facility initiators (154 cells/µL).
Oladele et al. 14 found that introducing two models of community ART delivery services resulted in more HIVpositive individuals being identified per 100 000 population in 14 high-burden local government areas in Nigeria in the 12 months after the models were introduced compared with Records identified through database searching (n = 92) PubMed = 82 PDQ-Evidence = 10 the 12 months before. Model A (immediate on-site initiation) identified 11 374 versus 5352 per 100 000 population, whilst Model B (HIV diagnosis up to baseline evaluation and referral for ART) identified 907 versus 152 per 100 000 population. Furthermore, preliminary data from the delivery optimization for antiretroviral therapy (DO ART) study 13 suggest that 80% (320/398) of persons testing HIV positive in rural Uganda were eligible for same-day ART initiation, with men accounting for more than half the persons eligible (169/320; 53%).

Linkage to antiretroviral therapy initiation
Four studies reported on linkage to ART initiation as an outcome. The study by MacPherson et al. 4 in urban slums in Malawi found that the cumulative incidence of ART initiation was significantly higher in the home (2.2% of residents) compared with the facility group (0.7% of residents). Labhart et al. 3 found that linkage to ART within 90 days after enrolment was higher in the same-day home-based ART initiation group (94/137; 68.6%) than in the facility-based care group (59/137; 43.1%) in rural Lesotho. Tun et al. 11 found that at 6 months, 256/256 (100%) of the community-based intervention group and 181/253 (71.5%) of the facility-based comparison group self-reported as being linked to care and on ART amongst sex workers in Tanzania.

Retention in care and loss to follow-up
Three studies reported on retention in care or loss to followup as an outcome. MacPherson et al. 4 found that at 6 months, 52/181 (28.7%) of the home group and 15/63 (23.8%) of the facility group were lost to follow-up. In addition, the rate of loss to follow-up was higher amongst the home group

Viral suppression
One RCT reported on viral suppression as an outcome. Labhardt et al. 3 reported that ART-naïve participants who were assigned to the same-day home-based ART initiation were more likely to remain in care at 12 months and achieve VL suppression. The authors reported that in the 11 to 14-month window after enrolment, 69/137 (50.4%) of the same-day group and 47/137 (34.3%) of the facilitybased care group achieved documented viral suppression Study nurse conducts a one-on-one structured education and counselling session by using a leaflet (5-10 min).

Place and time of ART initiation
Within the client's home, at the second home visit if client was eligible.
Within the client's home, at the first visit (same day). (VL < 100 copies/mL). In each group, 14/137 (10.2%) had no documented VL, whilst the remaining patients did not attend the health facility within that time frame.

Effects on behaviour
Two studies measured self-reported medication adherence, whilst one measured self-reported internalised stigma and an ongoing one reported on patient acceptability.

Medication adherence
MacPherson et al. 4 found that, based on clients completing an adherence questionnaire (at 2-4 weeks, 3 months and 6 months), 19/164 (11.6%) and 3/60 (5.0%) ART initiators in the home and facility groups, respectively, self-reported missing at least one dose of ART in the past 4 days at any assessment point (p = 0.14). Tun et al. 9 found that medication adherence was not significantly different amongst those with a completed adherence questionnaire, with 37/214 (17.3%) and 25/152 (16.4%) self-reported missing at least one dose of ART in the past 7 days, and 2/214 (0.9%) of the same-day group and 9/159 (5.7%) of the usual care group self-reported stopping taking ART for more than 30 days continuously (p = 0.008).

Internalised stigma and acceptability
Tun et al. 9 used a validated six-item scale to assess participants' feelings of shame and guilt as a result of living with HIV and found that a community-based intervention group was less likely to report high levels of internalised stigma compared with the facility-based group (26.6% vs. 39.9%; p = 0.001). This supports the findings by Wyatt et al. 6 who reported that amongst 50 DO ART study participants in Uganda, home initiation was associated with decreased concerns about disclosure risk at facilities. The authors also reported that other participants perceived home initiation and community follow-up to have many advantages compared with facility-based care, including being convenient, saving time and money otherwise spent on travel to clinics, and being responsive to individual needs. Additional benefits reported include reaching hard-toreach populations, for example, FSWs, 9 men at trading posts and those only available in the evenings and at weekends. 3,6,13

Cost analysis
MacPherson et al. 4 reported that the average cost of the home-based ART services was US$97.11 per patient assessed.

Summary of the evidence
A rapid review method was used to quickly capture the current evidence on the essential elements of evidence-based models of community-based (out-of-facility) ART initiation and the reported outcomes amongst patients initiating ART in community-based settings in SSA. We searched two databases, five conference websites and two registers of clinical trials for intervention studies evaluating CB-ARTi models, with four completed and two ongoing studies being included in this review.
The review identified heterogeneity in interventions, study design, location and definition of outcomes measured as a major obstacle to interpreting and synthesising the data on CB-ARTi. For example, some of the authors report on retention in care and loss to follow-up at 6 months 4,9 versus 12 months, 3 and linkage to care at 3 months 3 versus 6 months. 9 These differences are highlighted in Tables 1 and 2.
However, despite the limitations noted above, the existing data suggest that CB-ARTi could be more effective in increasing the uptake of HIV testing and improving case finding at a population level 4,11 than in facility-based ART initiation. Other advantages include improved linkage to ART initiation, 3,4,9,11 which results in similar rates of viral suppression. 3 One study conducted before the universal test and treat (UTT) era reported a higher rate of loss to followup amongst the home group, 4 and two other studies conducted in the UTT era reported significantly higher retention in care rates in the same-day group. 3,9 In addition, two studies found that there was no difference in selfreported medication adherence, 3,9 and one reported low self-reported internalised stigma. 9 The above results suggest that CB-ARTi models are equal and certainly not inferior to facility-based healthcare.

Strengths and benefits of community-based antiretroviral therapy initiation
Community-based HTS are an essential pillar towards reaching HIV epidemic control. However, it is estimated that almost two-thirds of patients are lost in the process from community-based HIV testing to facility ART initiation without specific interventions, with higher loss to follow up (LTFU) rates amongst African cohorts. 14 Hence, many authors and technical experts believe that CB-ARTi has considerable potential to address the gap between HIV diagnosis and ART initiation. 3,9,10,11 The strengths of the CB-ARTi models noted in the literature include reducing the structural barriers, such as cost of transport to the clinic, time saved otherwise spent on travel to clinics, flexibility of hours and location of service delivery and addressing stigma associated with traditional healthcare (especially for key populations), which in turn appear to result in better access to ART initiation for 'hard-to-reach' populations, such as men, FSWs and other key populations. 3,6,11,14 Concerns and barriers related to community-based antiretroviral therapy initiation There are concerns that CB-ARTi may shift scarce resources and ART initiations from healthcare facilities, which may result in their being unsustainable. 4,9 However, CB-ARTi identified patients with higher median CD4 cell counts, which in turn may influence cost-effectiveness favourably through reduced morbidity and mortality. 4 In addition, studies in Malawi and Nigeria both reported that the rates of facility-based ART initiations remained stable, whilst community-based initiations provided extra numbers. 4,11 Another issue raised in the literature relates to the potential for increased rates of loss to follow-up, 4,11 which means that additional adherence support measures should be put in place as CB-ARTi moves from small pilot studies to programmatic scale-up.
Concerns have also been expressed about the perceived lack of confidentiality, especially in smaller communities. 4 However, participants from Uganda and Tanzania reported CB-ARTi to be an acceptable option that is perceived to have many advantages compared with facility-based care. 6,9,13 Knowledge gaps and future directions The purpose of this rapid review was to synthesise and describe what is currently known on the topic of CB-ARTi. Based on the findings, several considerations for future research and practice in the field of CB-ARTi are evident. Firstly, it is essential to consider the fit of CB-ARTi initiatives within the context of the local epidemic conditions to ensure that they complement existing healthcare systems. This includes addressing existing facility-level barriers, such as long waiting times and poor staff attitudes, to increasing uptake of ART.
Secondly, given the results, which suggest that CB-ARTi models are possibly not inferior to facility-based ones, it is crucial to revisit existing policies on decanting stable virally suppressed patients to community ART distribution models after 12 months. This is especially important for key and other priority populations, such as men, adolescents and young people whose retention in care is often hindered by facility-level barriers.
Thirdly, it is important to consider the needs of the population by developing partnerships with community leaders and community-based organisations to overcome potential barriers related to lack of confidentiality, and widespread stigma and discrimination. In addition, designing CB-ARTi initiatives that are integrated into general health campaigns can also address the potential lack of confidentiality. Community-based ART initiation models should therefore include assessments for noncommunicable diseases, screening for sexually transmitted infections and tuberculosis (TB), HIV pre-exposure prophylaxis (PrEP) and other HIV prevention services, family planning services, as well as alcohol and drug rehabilitation services. Finally, further research is needed regarding the impact and estimated costs, as well as the cost-effectiveness of CB-ARTi models.

Limitations
Our results and conclusions might be susceptible to the bias related to the limits of a rapid review. These include restricting the search to literature published in English language and two electronic databases (although we used the PubMed database that contains by far the most significant number of health and medical journals).
The search was also complemented by grey literature searches from major HIV-related conferences, clinical trial databases and brief technical expert consultations.
A second limitation is that only one author screened and selected the titles and abstracts from the total set of documents retrieved, extracted data from all eligible articles and assessed the quality of evidence and RoB. However, this author is knowledgeable about the content area. The second author checked all the extractions for accuracy and verified the quality of evidence and RoB assessment.
As a result of the heterogeneity in intervention, study design, location and definition of outcomes measured, we were not able to combine the results to estimate overall intervention effect or draw conclusions on the relative effectiveness of community-based versus facility-based ART initiation. Varying definitions of the outcomes of interest also impacted the comparability of the results. This major obstacle has been highlighted by Rosen et al. 15 who have proposed standardised primary and secondary outcomes for research on accelerating ART initiation. Finally, some of the studies assessed are still incomplete, and the final results with higher numbers of participants (higher statistical power) may differ significantly from some of the preliminary findings presented.

Conclusion
This rapid review identified a small but rich set of information on the topic of CB-ARTi. However, after weighing the existing evidence, it appears that there is evidence that CB-ARTi can increase access to HTS, linkage to ART, retention in care and viral suppression rates, and is possibly not inferior to facility-based healthcare. The results reached earlier suggest that CB-ARTi models could prove to be equal and possibly not inferior to facilitybased ones and warrant further investigation. The apparent promise and pitfalls of CB-ARTi and the increasing interest of policymakers in its potential as a strategy to increase linkage to care and ART uptake in the era of UTT indicate that careful monitoring of the evidence base is warranted.