2.2. Study Selection

Each title was evaluated by a single individual and then all abstracts were evaluated by two independent individuals. Differences of opinion at the abstract level were brought to a third individual to consider and make a final decision. Each full text article was reviewed by two independent reviewers using the following standardized inclusion criteria: (1) focused on evaluating an public health intervention that improves HIV linkage to care, ARV adherence, or retention in care; (2) clearly identified as a qualitative study defined by both qualitative research methodologies (e.g., in-depth interviews, focus groups) and qualitative data analysis (e.g., framework analysis, thematic analysis); (3) presented original research data.

2.3. Methodological Rigor Assessment

The methodological rigor of each qualitative research study was investigated using a set of seven items adapted from the CASP (Critical Appraisal Skills Program) tool (Critical Appraisal Skills Programme, 2006), including the following: (1) qualitative research; (2) study context; (3) researcher reflexivity; (4) sampling methodology; (5) data collection; (6) analysis; (7) sufficient evidence to support claims. This assessment has been used in other qualitative evidence reviews (Munro et al., 2007, Carlsen et al., 2007). Methodological rigor was rated as having minor, moderate, or major concerns based on evaluation of two independent reviewers. Differences of opinion on the methodological rigor were brought to a third independent individual to consider and make a final decision. Methodological rigor was not used to screen studies for exclusion.

2.4. Data Extraction

Data were extracted using a standardized data extraction tool that included the following elements: (1) primary source data (e.g., quotes from individuals living with HIV); (2) secondary source data (e.g., interpretation from authors); (3) characteristics of the studies such as location of trial (country, city), type of intervention, key findings, study size, mean age of participants, analytical methodology, study limitations and implications, and inclusions of subpopulation or key populations. The following potential subgroup analyses were specified a priori: men, key populations (sex workers, people who inject drugs, transgender individuals, men who have sex with men, prisoners), pregnant women and breastfeeding women, children, adolescents, and individuals initiating antiretroviral therapy (ART) with a CD4 count greater than 500 cells/mm³. All data was extracted by two people and differences were resolved by discussing with a third individual. More details about the methodology can be found in the original systematic review papers (Tso et al., 2016, Ma et al., 2016, Hall et al., n.d).

2.5. Qualitative Synthesis

We used a framework thematic synthesis approach based on the theory of conceptual saturation (Booth et al., 2012, Thomas and Harden, 2008). Thematic synthesis is one of the approaches recommended by the Cochrane Qualitative Review Methods Group (Noyes, 2011). The thematic synthesis had three steps. First, the two independent individuals who undertook the respective searches identified cross-cutting themes that emerged independent of intervention type. Then, these 26 initial cross-cutting themes were compared across the three types of interventions to identify themes that were present in interventions at more than one point in the care continuum (e.g., adherence and linkage to care). Finally, once tentative cross-cutting themes were identified, primary (i.e., participant quotes), secondary (i.e., author interpretation), and tertiary (e.g., study team interpretations) data were explored with respect to each theme.

Qualitative review findings were assessed based on the CERQual approach (GRADE-CERQUAL, 2015). This is analogous to the GRADE tool for quantitative systematic reviews, providing a transparent method for evaluating confidence in the review finding based on standardized criteria (GRADE-CERQUAL, 2015). The CERQual approach evaluates specific review findings according to the following four components: (1) methodological rigor of the individual studies contributing to the review finding; (2) adequacy of the data contributing to a review finding; (3) coherence of the review finding; and (4) relevance of the review question. Methodological rigor was evaluated using the adapted CASP tool as described above. Adequacy of the data was evaluated by the number of contributing studies, the richness of individual study data, geographic regions included, and World Bank classification of country where the individual study was undertaken. Coherence was assessed by examining how the review finding formed a clear pattern and accounted for variations as appropriate. Relevance examined the extent to which the review finding was applicable in terms of population, setting, and phenomena of interest. Each of these four components was assessed and determined to have minor, moderate, or major concerns. Each of these four components was assessed to achieve an overall confidence level for each review finding as high, moderate, low, or very low. A draft version of the review was presented at the World Health Organization's ARV Operational Guideline Development Group for feedback. Subgroup analysis was not attempted because of insufficient data.

3.1. Participant Gender (High Quality of Evidence) (MacPherson, 2013, Shroufi et al., 2013, Kroeger et al., 2011, Naik et al., 2013)

Two low-income country studies and two middle-income country studies from Africa suggested that the individual's gender influenced the uptake of HIV interventions. Men living with HIV had decreased engagement in HIV care compared to women and also influenced the willingness of their partners to engage HIV services. Three studies found that men were comparatively disinterested in care engagement compared to women (MacPherson, 2013, Shroufi et al., 2013, Kroeger et al., 2011). This disinterest was attributed to gender power norms (MacPherson, 2013, Shroufi et al., 2013), lack of clinic responsiveness to men's perceived needs (MacPherson, 2013), and hesitancy to take women's advice (Shroufi et al., 2013). Two studies found sex-specific delays in intervention uptake (MacPherson, 2013, Shroufi et al., 2013), with particular reluctance from men to seek care until signs of HIV affected sexual function or capacity to work (MacPherson, 2013). Three studies found female participants had concerns about how engaging in HIV services could elicit anger, distrust, or violence from their husbands or male partners (MacPherson, 2013, Shroufi et al., 2013, Kroeger et al., 2011). One mother mentoring intervention promoted partner disclosure which helped women cope with gender differences (Shroufi et al., 2013).

3.2. Participant Socioeconomic Status (High Quality of Evidence) (Bezabhe et al., 2014, Busza et al., 2014a, Busza et al., 2014b, Agbonyitor, 2009, Ferguson et al., 2014, Lazarus et al., 2012, Miller et al., 2010, Sarna et al., 2014, Smillie et al., 2014, Naik et al., 2013, Garland et al., 2011)

Eleven studies from eight countries noted that participant socioeconomic status influenced public health HIV interventions. Most of these studies were from low- and middle-income countries in Africa. Individual socioeconomic status mediated intervention effectiveness. People with HIV living in poverty faced greater challenges engaging in interventions. Some individuals did not disclose their HIV status to family and friends, further limiting their access to resources. Individuals with HIV noted that several projects focused on alleviating poverty were feasible and acceptable in the context of larger HIV interventions. Unemployed people with HIV living in poverty faced greater challenges participating in interventions (Agbonyitor, 2009). Employed people with HIV living in poverty dropped out of care because of competing work demands and the financial cost of missed work time (Busza et al., 2014b, Lazarus et al., 2012, Miller et al., 2010, Sarna et al., 2014, Smillie et al., 2014, Naik et al., 2013). Three studies (Ferguson et al., 2014, Naik et al., 2013, Garland et al., 2011) discussed lack of finances as a barrier to linkage interventions. Six studies (Bezabhe et al., 2014, Busza et al., 2014a, Busza et al., 2014b, Agbonyitor, 2009, Miller et al., 2010, Smillie et al., 2014) noted that people living in poverty faced greater challenges with retention interventions. One study showed how migration to rural areas with better employment prospects disrupted clinic attendance and ultimately adherence (Bezabhe et al., 2014). Individuals living with HIV perceived that entering into care would necessitate all of the financial obligations that comprehensive care entailed and would be a great burden (Garland et al., 2011). Individuals living with HIV noted that the following components focused on alleviating poverty were feasible and acceptable in the context of larger HIV interventions: providing clinical care without cost to the HIV-infected individual (Agbonyitor, 2009, Miller et al., 2010, Smillie et al., 2014), providing ARVs without cost to the HIV-infected individual (Miller et al., 2010, Smillie et al., 2014), providing non-ARV medicines without cost to the HIV-infected individual (Bezabhe et al., 2014), small grants to support groups (Agbonyitor, 2009), paying school fees (Agbonyitor, 2009), and directly giving money to HIV-infected individuals (Agbonyitor, 2009).

3.3. Housing for Homeless and Marginally Housed Individuals (Moderate Quality of Evidence) (Lazarus et al., 2012, Cameron et al., 2009, Johnson et al., 2003, Nunn et al., 2010, Prentice et al., 2011, Anaya et al., 2015)

Six studies from three countries in Europe and North America evaluated housing for the homeless or marginally housed individuals. Housing support delivered as part of the intervention facilitated linkage and retention in care. Housing support in this subset of individuals living with HIV provided a more stable environment, enhancing the overall intervention. Five linkage studies reported on the importance of securing housing and maintaining stable housing for individuals living with HIV (Lazarus et al., 2012, Cameron et al., 2009, Johnson et al., 2003, Nunn et al., 2010, Prentice et al., 2011). One study mentioned how housing support integrated within an HIV intervention created a new local environment that made it easier for them to stay away from alcohol and drugs (Prentice et al., 2011). The achievement of housing support was instrumental to helping patients feel stable to proceed onto linkage of care medical services (Cameron et al., 2009, Anaya et al., 2015). Individuals living with HIV who were homeless and participated in an integrated intervention reported that after they had stable housing, their patient-physician relationship improved and they were more engaged in care (Cameron et al., 2009). Acceptable housing interventions for individuals living with HIV included the following: temporary accommodation for HIV-infected homeless and marginally housed individuals (Cameron et al., 2009), housing for homeless youth (Johnson et al., 2003), housing for sex workers with adherence problems (Lazarus et al., 2012), and shelter-based HIV interventions (Anaya et al., 2015).

3.4. Food Insecurity (Moderate Quality of Evidence) (Bezabhe et al., 2014, Busza et al., 2014a, Agbonyitor, 2009, Lazarus et al., 2012, Cameron et al., 2009, Johnson et al., 2003, Nunn et al., 2010, Prentice et al., 2011)

Eight studies from seven countries focused on food insecurity within public health HIV interventions. Lack of food limited the extent to which individuals living with HIV engaged in interventions and were retained in care. Some individuals living with HIV were unable to have sufficient food to accompany their ARVs, influencing their ARV adherence. Food assistance was a critical determinant of several interventions across the HIV continuum of care. One study found that widows living with HIV were not able to find enough food for themselves and their children (Agbonyitor, 2009). Even among farmers and food sellers living with HIV, they were unable to purchase meat and have a balanced diet (Agbonyitor, 2009). The lack of food prevented individuals from taking their ARV and sometimes affected retention in care (Bezabhe et al., 2014). One study mentioned how free meals within a health center were an incentive to link and be retained in care (Prentice et al., 2011). HIV-infected individuals noted that the following interventions to improve food insecurity were acceptable and feasible: referral to meal assistance (Cameron et al., 2009), reimbursing HIV-infected participants for food purchased for themselves and their children (Agbonyitor, 2009), donated food for HIV-infected individuals (Bezabhe et al., 2014), free meals and snacks (Lazarus et al., 2012, Prentice et al., 2011), dietician advice (Prentice et al., 2011), and donated fertilizer for crops (Bezabhe et al., 2014).

3.5. Transportation Assistance (High Quality of Evidence) (Bezabhe et al., 2014, Busza et al., 2014b, Nsigaye et al., 2009, Agbonyitor, 2009, Jama and Tshotsho, 2013, Kroeger et al., 2011, Lazarus et al., 2012, Miller et al., 2010, Sarna et al., 2014, Naik et al., 2013, Johnson et al., 2003, Magidson et al., 2014, Nunn et al., 2010)

A total of 13 studies, most within low and middle-income countries in Africa, provided evidence on the relationship between transportation and public health HIV interventions. Lack of regular transport to clinic appointments and to pick up ARVs were difficult for parents living with HIV (Busza et al., 2014b, Lazarus et al., 2012, Miller et al., 2010, Naik et al., 2013) and those caring for children with HIV (Busza et al., 2014b). Transportation assistance to the clinic and to obtain ARVs increased the feasibility and acceptability of HIV interventions. Long distances between home and the clinic exacerbated transportation problems (Jama and Tshotsho, 2013, Lazarus et al., 2012). Transportation assistance was particularly useful among poor HIV-infected individuals (Busza et al., 2014b, Agbonyitor, 2009) and those who lived in more remote locations (Agbonyitor, 2009). Transportation assistance included referrals (Cameron et al., 2009, Nunn et al., 2010), money for travel (Bezabhe et al., 2014, Agbonyitor, 2009, Magidson et al., 2014), vouchers (Kroeger et al., 2011), escorts to the clinic (Kroeger et al., 2011), transferring to a proximate clinic (Miller et al., 2010), and home-based services (Agbonyitor, 2009, Kroeger et al., 2011) were all feasible and acceptable to individuals living with HIV.

3.6. Clinical Navigation and Service Orientation (Moderate Quality of Evidence) (Busza et al., 2014b, Agbonyitor, 2009, Lazarus et al., 2012, Miller et al., 2010, Smillie et al., 2014, Cameron et al., 2009, Johnson et al., 2003, Prentice et al., 2011)

Eight studies from eight countries, half in Africa, examined clinical navigation and service orientation. Some individuals living with HIV reported that once they reached the HIV clinic, they were disoriented and encountered challenges accessing clinical care or ARVs. Several public health HIV interventions that aimed to improve clinical navigation were feasible and acceptable to individuals living with HIV. There was confusion about where to be evaluated within medical centers that were often large and intimidating (Agbonyitor, 2009, Lazarus et al., 2012). Long delays in seeing a doctor contributed to a sense of futility among HIV-infected individuals (Busza et al., 2014b, Agbonyitor, 2009, Lazarus et al., 2012). Delays in obtaining ARVs from the pharmacy and stockouts were also common (Busza et al., 2014b, Agbonyitor, 2009, Lazarus et al., 2012, Miller et al., 2010). One retention study found that some HIV-infected individuals referred to missing paperwork (“clinic cards, transfer papers, and proof of travel”) as the reason for not returning for follow-up (Miller et al., 2010). HIV-infected individuals noted that the following interventions to improve clinical navigation were feasible and acceptable: SMS text messaging between HIV-infected individuals and nurses (Smillie et al., 2014), peer support (Lazarus et al., 2012), non-peer support (Agbonyitor, 2009, Cameron et al., 2009), and a one-stop model for care provision (Johnson et al., 2003, Prentice et al., 2011). The one-stop model integrated several types of services within a single site to improve adherence.

This research was commissioned by the World Health Organization HIV Department. Additional support was obtained from the US National Institutes of Health (NIAID 1R01AI114310-02, FIC 1D43TW009532-03). We would like to thank the members of the World Health Organization's HIV Operational Guideline Development Group for helpful feedback on this research. We would also like to thank Fengyu Hu, Weiping Cai, Bin Yang, Wei Ma, Kathryn Muessig, and Chongyi Wei for helpful comments.