Interventions to improve or facilitate linkage to or retention in pre-ART (HIV) care and initiation of ART in low- and middle-income settings – a systematic review

Introduction Several approaches have been taken to reduce pre-antiretroviral therapy (ART) losses between HIV testing and ART initiation in low- and middle-income countries, but a systematic assessment of the evidence has not yet been undertaken. The aim of this systematic review is to assess the potential for interventions to improve or facilitate linkage to or retention in pre-ART care and initiation of ART in low- and middle-income settings. Methods An electronic search was conducted on Medline, Embase, Global Health, Web of Science and conference databases to identify studies describing interventions aimed at improving linkage to or retention in pre-ART care or initiation of ART. Additional searches were conducted to identify on-going trials on this topic, and experts in the field were contacted. An assessment of the risk of bias was conducted. Interventions were categorized according to key domains in the existing literature. Results A total of 11,129 potentially relevant citations were identified, of which 24 were eligible for inclusion, with the majority (n=21) from sub-Saharan Africa. In addition, 15 on-going trials were identified. The most common interventions described under key domains included: health system interventions (i.e. integration in the setting of antenatal care); patient convenience and accessibility (i.e. point-of-care CD4 count (POC) testing with immediate results, home-based ART initiation); behaviour interventions and peer support (i.e. improved communication, patient referral and education) and incentives (i.e. food support). Several interventions showed favourable outcomes: integration of care and peer supporters increased enrolment into HIV care, medical incentives increased pre-ART retention, POC CD4 testing and food incentives increased completion of ART eligibility screening and ART initiation. Most studies focused on the general adult patient population or pregnant women. The majority of published studies were observational cohort studies, subject to an unclear risk of bias. Conclusions Findings suggest that streamlining services to minimize patient visits, providing adequate medical and peer support, and providing incentives may decrease attrition, but the quality of the current evidence base is low. Few studies have investigated combined interventions, or assessed the impact of interventions across the HIV cascade. Results from on-going trials investigating POC CD4 count testing, patient navigation, rapid ART initiation and mobile phone technology may fill the quality of evidence gap. Further high-quality studies on key population groups are required, with interventions informed by previously reported barriers to care.


Introduction
The rapid roll-out of antiretroviral therapy (ART) programmes over the past decade in low-and middle-income countries has resulted in an estimated 9.7 million HIV-positive individuals receiving ART by the end of 2012, compared to just 300,000 in 2000 [1]. HIV testing and counselling, the entry point to care, has also been brought to scale at an impressive pace [1]. However, overall programmatic success has not reached its full potential due in part to the high attrition occurring in the period between HIV testing and ART initiation. Mathematical modelling studies have shown that achieving the maximum population-level prevention benefits from the proposed universal test and treat strategies is dependent on successful programmatic outcomes, including optimum linkage to care rates [2Á4]. It is estimated that only 65% of those eligible for treatment as of the end of 2012 were currently receiving ART [1,5,6], and by raising the ART initiation threshold to CD4 B500 cells/mL as per World Health Organization (WHO) guidelines released in July 2013, coverage is estimated to be considerably lower [1]. If major progress towards universal access to ART is to be attained by 2015 then there is an urgent need for HIV programmes to strengthen the existing HIV continuum of care pathway and ensure that increased access to HIV testing is accompanied by improved linkage to care.
To date, attempts have been made to quantify the losses occurring along the HIV cascade and describe the barriers to care, although data from low-and middle-income countries outside sub-Saharan Africa are scarce. Three recent systematic reviews have consistently shown major attrition occurring along each step of the HIV cascade in sub-Saharan Africa for adults. Less than half (45Á46%) of individuals not yet eligible for ART were retained in pre-ART care and only two-thirds (65Á68%) of ART-eligible individuals initiated ART [7Á9]. A recent HIV cascade analysis conducted in India found that approximately 70% of HIV-positive patients linked to care within three months post-diagnosis, with only 65% of pre-ART patients retained in early care and 67% of ARTeligible patients who initiated ART within three months [10]. A systematic review investigating patient and programme level factors associated with retention in care during the pre-ART period and linkage to ART care in sub-Saharan Africa found that the commonly cited barriers included psychosocial (stigma and fear of disclosure), economic (inability to afford transport costs, distance to healthcare facility) and health system (long waiting times, shortage of health care workers) factors [11]. Moreover, poor linkage to care has been reported among groups such as pregnant women [12] and children [13]. Among ART-eligible pregnant women in low-and middle-income countries, 38Á88% failed to initiate ART, though point of attrition assessed by the different studies differ, with financial constraints and fear of stigma identified as the main obstacles to ART care [12]. High rates of pre-ART mortality and loss to follow-up prior to starting ART among children have also been reported in studies from sub-Saharan Africa [13]. According to the new ART eligibility criteria recommended by the WHO guidelines, the number of HIV-positive individuals deemed eligible for ART will increase considerably; however, it will decrease the time period from entry to care and ART eligibility and likely reduce attrition in the pre-ART period [1,14].
Patients in the period between HIV diagnosis and ART initiation can be divided into three groups: those that enter into HIV care post-diagnosis, those that remain in pre-ART care until ART-eligible, and those that initiate ART once ART eligibility is established. Rosen et al. describe these three periods as stage 1, 2 and 3, respectively [12]. Several interventions to improve ART adherence and retention have been rigorously evaluated [15], including treatment supporters [16], nutritional support [17Á19], mobile-text messages [20]. However, research on interventions to increase linkage to and retention in pre-ART care and initiation of ART has only recently emerged. Interventions that have been tested include point-of-care (POC) CD4 count testing [21Á23], referral slips [24], patient navigators [25], home-based HIV testing and ART initiation [26,27]. The aim of this review is to assess the effect of interventions to improve or facilitate linkage to or retention in pre-ART care and initiation of ART in low-and middleincome settings.

Methods
Randomized trials, comparative non-randomized observational studies, and comparative before-and-after studies were eligible for this review. Cohort studies were only included if the control (comparator) group was contemporary (delivered at the same time and in the same country). Studies were eligible if they included HIV-positive patients in low-and middle-income countries (as defined by the World Bank [28,29]) at any point before initiating treatment and described an intervention aimed at improving linkage or retention in pre-ART care or initiation of ART. Primary outcomes were defined as proportion of HIV-positive patients 1) retained in pre-ART care, 2) linked to care and 3) initiated on ART as defined by study author and 4) time to linkage or initiation of ART. Secondary outcome assessed were: 1) patient satisfaction with care, as defined by the study authors 2) cost to the provider and 3) cost to the patient and family.
The literature was searched for available systematic and narrative reviews investigating linkage to pre-ART, HIV and ART care and retention in pre-ART care. Five systematic reviews were identified [7Á10,30] and the reference lists were hand searched. Additional searches were conducted in four electronic databases for primary studies: Medline, Embase, Global Health and Web of Science from the 1st of January 2004 (start of ART delivery at scale in low-and middle-income countries) to the 10th of February 2013. The detailed compound search strategies are described in Supplementary Tables 1Á3. Three electronic databases were searched for on-going trials: Cochrane Central Register of Controlled Trials (CENTRAL), WHO International Clinical Trial Registry Platform and ClinicalTrials.gov. Furthermore, a total of 117 experts (Supplementary Tables 4 and 5) were contacted to identify additional studies either completed or on-going; these experts were authors of studies included in this and previous reviews [7Á10] and these experts were asked to provide additional names of people who were subsequently contacted in a second round. Finally, reference lists of all studies identified by the above methods were screened for potentially eligible studies. Conference abstract databases of the following conferences were searched: Conferences on HIV Pathogenesis and Treatment of the International AIDS Society (IAS, 2001Á2011), Conferences on Retroviruses and Opportunistic Infections (CROI, 2006Á 2013) and the International AIDS Conferences (AIDS, 2001Á 2012). Additional searches were undertaken to identify subsequent publications of studies presented at conferences. These searches were conducted in Ovid Medline using the first and the last authors' names combined with ''HIV'' and the country the study was conducted in. Authors of publications with potentially relevant data were contacted in case clarification was needed.
All references identified by the compound search strategy were imported into EndNote. After duplicates were removed, titles and abstracts were examined independently by the two authors (DG, KK). The full text of potentially relevant studies was obtained and the inclusion criteria were applied independently using a standardized eligibility form. Data extraction was performed by KK using a standardized data extraction form. Information was extracted regarding study (citation, start and end dates, location, study design and details), participants (eligibility criteria, age range, gender, population size, specific high-risk groups, HIV disease progression), intervention (detailed description of the intervention and standard of care in the control groups) and outcomes (proportion of individuals with a positive outcome and time to accessing HIV or ART care). Drawing on knowledge from previous systematic reviews of risk factors and barriers to care, interventions were categorized as follows: health system interventions, patient convenience and accessibility, behaviour interventions and peer support, and incentives. In addition, the stage the intervention was targeted at in the HIV cascade was also highlighted. The risk of bias was assessed by the Cochrane Effective Practice and Organisation of Care (EPOC) Group criteria for randomized-controlled and non-randomized-controlled trials [31] and the Newcastle-Ottawa Scale for observational studies [32].

Study characteristics
A total of 11,129 potentially relevant citations were identified, of which 24 were eligible for inclusion ( Figure 1). In addition, 15 on-going trials investigating interventions to improve linkage and retention in pre-ART care were identified. Four publications with potentially relevant data could not be included as authors were unable to provide relevant data [18,24,33,34]. Characteristics of the included studies are summarized in Tables 1 and 2. Twenty studies were published in peer reviewed journals, one was an unpublished study and three were conference abstracts. There were 21 studies from sub-Saharan Africa (nine countries), two studies from India [35,36] and one study from Cambodia [37]. The following patient subgroups were included: patients co-infected with tuberculosis (TB) (n 01) [37], children (n01) [35], adolescents (n01) [38], inpatients (n01) [39], injecting drug users (n 01) [36] and pregnant women (n 09) [40Á48]. No studies were identified that investigated interventions in any other key population group (men who have sex with men, sex workers, the elderly, prisoners and  38], service integration, mainly in the setting of antenatal care (n 06) [40,42,45Á48], and wider health systems interventions including improved communication, referral and teaching (n 04) [37,41,44,49] and incentives (n02) [35,36]. No study provided data on patient satisfaction and costs. Most studies assessed interventions for stage 3 (ART initiation) (n014) and stage 2 (completion of ART eligibility screening) (n09) of the HIV cascade, with only six studies assessing the impact of interventions for stage 1 (linkage to HIV care).

Impact of interventions
The possible impact of interventions was assessed by comparing proportions of individuals completing staging, enrolling in care, accessing follow-up care or initiating ART in the intervention and control group or by comparing the median time to a positive outcome in individuals receiving the standard of care or the intervention. The definitions of successful outcomes were heterogeneous across studies (Tables 1 and 2) and thus effects of the similar or different interventions were difficult to compare. Studies often reported multiple outcomes along the different steps of the pre-ART period.

Integration
Integration of ART and antenatal care increased enrolment in HIV care and ART initiation in five out of six observational studies (Tables 1, 2 and 3) [40,42,46Á48], with the largest effect observed in a before/after study from Zambia, with pregnant women in the intervention group twice as likely to enrol in HIV care [adjusted odd ratio (aOR) 02.06 (1.27Á3.34)] and initiate ART [aOR 02.0 (1.37Á2.95)] within 60 days post-diagnosis [47]. An observational study from South Africa examining the impact of integrated ART services within an antenatal care setting showed that pregnant women in the intervention group were less likely to initiate ART compared to the control group [hazard ratio (HR)00.62 (0.37Á1.04)] [45].

Comprehensive HIV care
Three before/after studies investigating packages of health systems interventions aimed at improving referrals, communication and teaching showed an increase in ART initiation by 15% in pregnant women in South Africa [44], by 27% in eight weeks in TB patients in Cambodia [adjusted HR (aHR) for ART initiation within eight weeks post-TB diagnosis 02.60 (1.87Á 3.62)] [37] and by 15% in unselected adults in Swaziland [49].

Task shifting
A post hoc analysis of a cluster randomized trial primarily investigating the effect of task-shifting to nurses on mortality found no difference in attrition during pre-ARTcare six months post-enrolment in South Africa [relative risk (RR) 0 1.02 (0.92Á1.18)] [51].

Patient convenience and accessibility
Immediate CD4 count testing POC CD4 count testing at the time of HIV diagnosis increased the proportion of patients completing an ART eligibility assessment (44% vs. 79%) and reduced time to ART initiation (48 to 20 days pB0.0001) [21]

Home-based ART initiation
A cluster randomized trial from Malawi showed that optional availability of home-based eligibility assessment following HIV self-testing and ART initiation resulted in a threefold increase in the risk of population-based ART initiation [risk ratio02.94 (2.10Á4.12)] [26].

Inpatient testing and counselling, home visits
Similarly, a higher proportion of patients randomized to receive HIV testing as inpatients linked to HIV services (21%) compared to patients who were referred for outpatient HIV testing following discharge from hospital (16%) [39]; however, the result did not reach statistical significance (p00.174).

Behaviour interventions and peer support
Extended counselling and peer support A randomized controlled trial conducted in Uganda showed that intensified post-test counselling and monthly visits by a peer support worker almost doubled linkage to HIV care five months post-diagnosis [HR01.8, (1.42Á2.1)] [52]. A peer support intervention among pregnant women in a cluster randomized trial conducted in the Democratic Republic of Congo led to a 39% increase in linkage to HIV care [HR01.39 (1.01Á1.91)] [43], but these results have yet to be formally published.

Peer navigator
Visits by peer navigators also increased the likelihood of seeking care following a community-based HIV testing campaign in a observational study conducted in Kenya at 10 months for men (aHR 01. 35  Control: HIV self-testing kits were available for all cluster residents through community counsellors. All individuals who requested self-testing could do so in the privacy of their home. They were asked to return the used kit, but were told that they were not required to disclose the result, but if they wish to do so they would receive post-test counselling. All selftesting participants received a self-referral card to access HIV care (and ART if eligible) at local health facilities Intervention: Self-testing and facility-based HIV care was available for all community members as in the control clusters. Individuals who disclosed a positive result to the community counsellor could opt to receive home-based initiation of HIV care. If requested a study nurse would undertake a home visit within 3 days and perform confirmatory testing, ART eligibility assessment (CD4 count, WHO stage), undertake TB screening and provide INH and the first of the two treatment preparedness sessions. At a second visit individuals who met national eligibility criteria received 2 weeks of ART before transfer into the routine clinic system.

Food
Two studies from India investigated the effect of food incentives on retention in pre-ART care and ART initiation. In the first study, the introduction of monthly food supplements increased the proportion of children regularly visiting the clinic over a one year period from 81 to 93% [HR02.89, (1.09Á7.63)]; however, this study was a before/after design and these findings are subject to survival bias as more children were reported dead and lost to follow-up in the control arm (9%) than the intervention arm (4%) [35]. The second study, a randomized controlled trial in injecting drug users, showed that the median time to ART initiation was seven days in individuals receiving food vouchers compared to 58 days in individuals not receiving food vouchers [36].

Medical
A third study, from Kenya, found that scheduled and regular visits during the pre-ART period for the purpose of cotrimoxazole refill increased pre-ART retention at 12 months post-enrolment by 20% [50].
On-going trials Fifteen on-going or planned trials were identified ( Table 4). The majority (n 013) of trials are underway in sub-Saharan Africa, six of them in Kenya, and seek to assess interventions for injecting drug users (n 03) and pregnant women (n05). Proposed interventions include POC CD4 count testing, rapid (same day) ART initiation, integration, peer educators, intensive counselling, assisted partner notification, homebased male partner testing and short text messages. The majority of trials will be completed by January 2015.

Quality of studies
Most studies included in this review were observational studies; either times series (before/after studies) (n 010) or contemporaneous comparative studies (n07). There were three cluster randomized controlled trials [26,43,51] and four individual randomized trials [36,39,52,22]. Age, sex and CD4 count distributions in control and intervention groups, outcome verification and ascertainment, time delineation of the outcome and controlling for confounding and clustering are described in Table 5. Among the 20 peer-reviewed studies assessed for quality of the evidence, the majority (n08) were rated as having an unclear risk of bias; four studies were classified as being at high risk of bias.
Of the four peer-reviewed randomized trials assessed for study quality, the majority were categorized as having low risk of bias (n 03) with key quality concerns related to the random sequence generation, allocation concealment and reporting. Of the 16 remaining observational studies, most had an unclear risk of bias (n07) and four studies were rated as having a high risk of bias, with the main quality issues related to uneven distribution of baseline characteristics, no adjustment for confounders, and lack of data related to patient follow-up.

Discussion
This review identified 24 published studies and 15 on-going trials assessing interventions to improve or facilitate linkage        Observational study with control group 0-0-0 1-0 0-1-0 High risk Pfeiffer [48] Observational study with control group 1-0-0 0-0 0-1-0 Unclear risk Tsague [40] Observational study with control group 1-1-1 0-0 0-1-0 Unclear risk Larson [23] Observational study with control group 1-1-1 1-0 0-1-0 Unclear risk Stinson [45] Observational study with control group 1-1-1 1-0 0-1-0 Unclear risk Topp [53] Observational study with control group 1-1-1 1-1 0-1-0 Low risk Choun [37] Before/after study 1-1-1 1-0 0-1-1 Low risk Kundu [35] Before/after study 1-0-1 0-0 0-1-1 High risk Kohler [50] Before/after study 1-1-1 1-0 0-1-1 Low risk to or retention in pre-ART care and initiation of ART. The majority of these studies were conducted in sub-Saharan Africa and most were observational in design. Patient centred strategies to minimize clinic visits, integration of ART care into antenatal services, education and support for those enrolled in pre-ART care, and incentivized clinic attendance all showed promising results. However, there is a paucity of studies focusing on specific key populations and few studies have investigated a combination of interventions and the impact of these interventions at multiple steps along the pre-ART period. This review suggests that whilst some interventions are specific to certain points along the cascade (e.g. POC CD4 testing for the improvement of linkage to pre-ART care), there are interventions that could potentially support linkage and retention along the entire HIV cascade both in pre-ART and ART care (e.g. integration of HIV services, food incentives, patient navigators) and thus warrants further investigation. Attending pre-ART clinic visits is a major challenge among patients in low-and middle-income countries due to high transport costs, distance to health care facilities, inability to take time off work, and long clinic waiting times [10] as well as the lack of services (perceived or actual) provided during the pre-ART period [9,10]. Interventions such as POC CD4 testing, home-based ART initiation and integrated service delivery, aimed at decreasing the number of pre-ART visits prior to ART initiation, have proven successful, particularly for those patients immediately eligible for ART. The use of POC CD4 technology or immediate CD4 count testing in South Africa and Mozambique have shown higher completion of ART eligibility assessment, shorter time to staging and ART initiation, and an increase in ART initiations [ 21Á 23,38]. Home-based ART initiation is a viable strategy to effectively improve population-level ART initiation as indicated in a trial offering HIV self-testing from Malawi [26]. ''Fast tracking'' of patients through the provision of counselling and ART eligibility assessment in TB and ANC settings facilitated prompt ART initiation with decreased numbers of clinic visits [37,41]. Results from a randomized controlled trial investigating same day ART initiation on the day of HIV diagnosis, currently underway in South Africa, may provide further evidence on the effect of rapid ART initiation on patient outcomes [63]. However, the long-term effect of reducing the number of clinic visits prior to ART initiation remains unclear. There has been no study to date, except for the Malawian study [26], that has followed-up patients beyond the pre-ART period, and it is plausible that the improvement in linkage and ART initiation is offset by increased default rates once on ART. Given the high HIV burden in this resourceconstrained setting, there is a clear need for prioritizing patients based on their baseline CD4 count and urgency to link to care, with more support (e.g. patient navigation, fasttracking, home-based ART initiation) provided to patients with lower CD4 counts compared to those with higher CD4 counts.
Apart from reducing clinic visit frequency, inflexible clinic hours have been identified as another barrier to accessing care, especially among the working population [10,69,70]. However, none of the included studies investigated the effect of providing services outside traditional clinic operating Before/after study Burtle [49] Before/after study

High risk
Observational cohort studies: Selection: studies could score a maximum of 3, the first score for representativeness of the exposed cohort, the second score for the selection of the non-exposed cohort and the third score for ascertainment of exposure. Comparability: studies could score a maximum of 2, the first score for controlling for important factors, the second score for controlling for any additional factors.
Outcome: studies could score a maximum of 5, the first 2 scores for outcome assessment (a. independent blind assessment, b. record linkage or verification), the second score for time lineated and clear definition of outcome, the last two scores for follow-up (a. complete follow-up, b. subjects with missing outcome assessment unlikely to introduce bias).
hours (e.g. evenings and weekends). In addition, interventions targeted at reducing waiting times through increased medical and administrative staffing levels, or appointmentbased systems, are yet to be evaluated. A qualitative study from Tanzania investigating referral systems between testing and treatment found that the provision of a transport allowance was a key facilitator to linkage to ART care [24]; however, no study has investigated the impact of transport provision, subsidies, or vouchers on linkage to pre-ART or ART care. Several studies have investigated the effect of bringing HIV care services closer to where patients reside through decentralization and service integration to overcome issues of distance and difficulties to pay for transport in the context of pre-ART care. A recent systematic review assessing the influence of various models of decentralized HIV treatment and care on initiating and maintaining ART concluded that partial decentralization can decrease attrition of patients on ART [71]. Currently, there is recommendation from recent WHO guidelines to decentralize HIV services from secondary care (e.g. district hospitals) to primary health care clinics [14]. This often goes along with task shifting ART initiation and patient management from doctors to nurses. This is supported by a cluster randomized trial of nurse-led ART initiation in South Africa which showed no detrimental impact on retention in pre-ART care, or immunological and virological outcomes at six months, suggesting that task shifting is a viable approach in some settings [51]. The provision of ART initiation services within or near to ANC clinics as part of a ''package'' of PMTCT interventions was shown in two studies to increase the proportion of eligible women on ART at delivery [41,44]. In contrast, a positive correlation was seen between distance from the clinic and enrolment for HIV care in an ANC study, suggesting that more distant services were preferred, perhaps due to issues of confidentiality and stigma [25]. These differences are likely to be context specific and therefore this area merits further investigation in different context. Of note, integration of services (i.e. ANC, post-natal and HIV) without clear delineation of the roles, responsibilities and authorities between services could disrupt the continuum of care for women on ART during pregnancy and thus programme planners should consider these requirements prior to implementation.
This review identifies that peer support and appropriate counselling may facilitate linkage to care. Stigma has been identified as a barrier to care at all steps of the HIV cascade, including HIV testing [72], linkage to care [10,73], and ART adherence [74,75]. Fear of disclosure of one's HIV-positive status is also a reported barrier to care [10], but if achieved, disclosure can support linkage to care [10,76]. Appropriate counselling has been shown to promote disclosure of HIV status to family members, improve adherence to treatment, and promote psychological wellbeing [77Á83]. A randomized control trial conducted in Uganda showed that combined post-test counselling by trained staff with home visits by community support agents for on-going counselling also improved linkage to ART services [52]. The costs and feasibility of this labour-intensive approach on a large scale need assessment and further investigation of the effect of different models of post-test counselling on linkage to care is therefore warranted. The use of community-based peer support provided by health care workers and expert patients has been associated with higher rates of ART initiation in the general adult population and ANC settings [43,44]. Several on-going studies will provide evidence into the effect of support systems on ART initiation in high-risk groups such as injecting drug users [54,66,68], pregnant women, and their partners [56]. Mobile phone technologies have been used to trace and support patients once on ART [19,78], and two on-going trials will provide clarity around the effectiveness of the use of mobile technology in pre-ART retention [57,60].
There have been few studies on the use of incentives to improve retention in pre-ART care, including nutritional support, free medications, and conditional cash transfers. Food insecurity was highlighted as a barrier to pre-ART care in a recent systemic review [10] and studies assessing the use of food incentives have shown improved linkage to care in children and injecting drug users [35,36]. Given the potential of improved nutrition to improve treatment outcomes for HIV-positive patients, this area deserves further attention, although costs and feasibility need careful assessment. The provision of free co-trimoxazole was shown to increase retention in pre-ART care for ART-ineligible patients from 63 to 84% in a Kenyan study [50], and further investigation into the cost-effectiveness of medical incentives on retention in pre-ART care are required. Conditional cash transfers have previously been used to promote HIV testing [84Á86], and to encourage patients to maintain their HIV-negative status [87,88], but their effect on retention in pre-ART care has not yet been assessed.
This review showed a lack of studies testing interventions aimed at key populations. Men have less access to and coverage of HIV testing and ART in most settings and are at higher risk of attrition from ART and pre-ART services. No study investigating interventions specifically targeted at men were identified [10,89Á97]. In addition, despite the high rates of pre-ART attrition amongst children [12] and poor treatment outcomes in adolescents [98Á100], only one study was focused on each of these respective groups [35,38]. Ways to support linkage to care for other vulnerable groups such as migrants, prisoners and those in closed settings, sex workers, men who have sex with men, people who inject drugs, and those with HIV-TB co-infection should also be assessed.
Overall, most studies investigated the impact of individual interventions on one stage of the cascade, mainly stages 2 and 3 and few studies have investigated interventions for stage 1, linkage to HIV care.
A single intervention is unlikely to solve attrition across all steps in the pre-ART period and to address the different reasons for attrition. Although, certain interventions such a POC CD4 counting are effective in improving time to linkage to HIV care and completion of ART eligibility screening, it may be a less suitable intervention for improving pre-ART retention, in contrast to food and medical incentives. The effect of interventions targeted at a single point in the pathway might be limited when assessed across the entire pre-ART period. Loss to follow-up prevented earlier on in the pathway might not guarantee retention later on. However, most of the planned or on-going trials investigate single interventions with limited follow-up time. Therefore, a combination of interventions to address the main points at which patient drop-out of care during the pre-ART period or while on ART are most likely to be successful. This package should include interventions aimed at increasing patient convenience and accessibility (e.g. POC CD4 count testing, inpatient testing, home visits and possibly home-based ART initiation) as this was found to be effective in increasing ART eligibility screening and initiation rates; health system interventions such as integration of ART and ANC care as this appears to increase HIV care and ART enrolment; behavioural interventions and peer support (e.g. intensified post-test counselling and peer support) as these were found to increase linkage to care. HIV programmes should first determine the patient groups at risk for drop-out in their cohorts and reasons for dropout (e.g. too ill to travel to the clinic, disclosure and stigma issues) in order to inform the selection of an effective package of interventions. Programme planners should be mindful of several key factors when selecting interventions i.e.: the operational needs of specific interventions (e.g. staff duties, supplies, data collection systems), challenges in implementing and maintaining interventions (i.e. POC CD4 count testing, integration of care), interventions studied (e.g. incentives) could be context-specific and may be need to be tailored to the target population in their settings. Furthermore, programmes adopting the WHO guidelines supporting earlier ART initiation will likely observe a significant reduction in the number of people with HIV required to remain in the ''pre-ART'' period, however these programmes will need to implement interventions to improve immediate linkage to ART care as majority of people will require ART care.
The main strengths of this review are the use of a broad search strategy that included an extensive expert consultation, allowing identification of both published and unpublished studies as well as information on on-going trials, a rigorous assessment of study risk of bias and quality. This assessment found that the evidence base and methodology of this current review was subject to several limitations. Firstly, the quality of studies was limited with the majority of studies being observational, using time series rather than concurrent control groups, and subject to unclear risk of bias. Outcome ascertainment was rarely adequately described and time delineation was often lacking, making it difficult to compare outcomes and calculate summary estimates. Assessment of cost-effectiveness and patient acceptability were absent from all studies. Publication bias is likely present as studies with unfavourable results are less likely to be published. In addition, HIV programme evaluation reports from resource-limited settings might not necessarily be published and thus might have been missed by this review. While the literature review was extensively conducted in four international databases including primarily studies published in peer reviewed journals, regional databases such as the African Healthline and LILAC were not searched. The review was limited to low-and middle-income countries. Thus some potential interventions from high income countries which might be applicable to low-and middle-income countries were not included in the review. Most of the studies included were conducted in sub-Saharan Africa, limiting the generalizability of our findings beyond this region and even within the region there is considerable heterogeneity of current service delivery models and approaches. Most studies assessed interventions within the ANC and PMTCT setting and thus findings may not be generalizable to general HIV care programmes. Finally, the majority of studies looked at the effect of a single intervention on a single point in the HIV continuum.

Conclusions
The overall findings from this review suggest that streamlining services to minimize patient facility visits, providing adequate counselling, medical and peer support, and providing incentives may decrease attrition between HIV testing and ART initiation. Further implementation research, focused programme evaluation and studies with rigorous study designs investigating the impact of individual interventions and a combination of interventions across the HIV cascade, including for the various groups at high risk of attrition, is warranted in order to inform HIV policy and programmes. There is a specific need to also evaluate interventions for key populations as currently they are not only disproportionately affected by HIV but have lower access to and coverage of HIV testing and ART.