Risk Factors for Late Linkage to Care and Delayed Antiretroviral Therapy Initiation Amongst HIV Infected Adults in Sub-saharan Africa: a Systematic Review and Meta-analyses

BACKGROUND
Late treatment initiation threatens the clinical and public health benefits of antiretroviral therapy (ART). Quantitative synthesises of the existing evidence related to this is lacking in sub-Saharan Africa (SSA), which would help ascertain the best evidence-based interventions. This review aimed to systematically synthesise the available literature on factors affecting linkage to care and ART initiation amongst HIV-infected adults in SSA.


METHODS
Systematic searches were undertaken on four databases to identify observational studies investigating factors affecting both HIV care outcomes amongst adults (age ≥19 years) in SSA, and were published between January 1, 2015 and June 1, 2021. RevMan-5 software was used to conduct meta-analyses and Mantel-Haenszel statistics to pool outcomes with 95% confidence interval and <0.05 level of significance.


RESULTS
Forty-six studies were included in the systematic review, of which 18 fulfilled requirements for meta-analysis. In both narrative review and meta-analyses, factors related to health care delivery, individual perception and sociodemographic circumstances were associated with late linkage to care and delays in ART initiation.


CONCLUSION
This review identified a range of risk factors for late linkage to care and delayed ART initiation amongst HIV-infected adults in SSA. We recommend implementation of patient-centred intervention approaches to alleviate these barriers.

intervals. Heterogeneity between studies in effect measures was determined using Chi 2 test and I 2 statistic, and an I 2 value of 75% was considered as high heterogeneity [37]. We used RevMan-5 software [38] to calculate pooled odds ratios by applying Mantel-Haenszel statistics for each outcome and a forest plot to present the results.

Results
The electronic literature search identi ed 2597 articles, of which 451 were duplicates and 2064 were irrelevant to the review question (based on the title and abstract appraisal). An additional 36 articles were removed after the full text review that was based on the eligibility criteria (i.e. studies conducted on ineligible populations, qualitative studies, intervention studies, review articles or articles lacking the desired outcomes: not reporting on linkage to care or ART initiation). Among the remaining 46 studies that were included in the review, 18 met the criteria for meta-analysis. Figure 1 depicts the selection process and number of articles excluded and retrieved at each stage.

Methodological quality
Almost three-quarters (72%) of the studies were assessed as 'moderate' or 'strong' quality in regard to ensuring the representativeness of participants, and 61% of them were scored as 'moderate' regarding the appropriateness of the study design. Most studies (70%) were assessed as having a strong performance in controlling confounders (i.e., controlled at least 80% of relevant confounders). Only ten (22%) studies described the validity and/or reliability of the data collection tools, of which three studies were assessed as 'strong' in this regard. Similarly, nine (20%) studies considered the risk of drop-out and withdrawal, and three of them reported a follow-up rate of more than 80% (a strong performance). This criterion was inapplicable in most (67%) of the studies. Overall, one study was assessed as 'strong' and 24 other studies (52%) were assessed as having a moderately strong methodological quality on the EPHPP tool (see Additional le 5).
Five studies considered care engagement at CD4 count < 200cells/mm 3 as late linkage to care [8,11,48,59,69] and one study at CD4 count ≤ 100 cells/mm 3 [15]. One other study de ned late linkage to care as diagnosis at CD4 count of ≤ 500 cells/mm 3 and/or any of the WHO clinical stages [7].
Interestingly, a study by Maheu-Giroux et al [27] in South Africa determined linkage to care by estimating the length of time between HIV infection and engagement in care.
Twelve of 25 ART initiation studies measured the rate of ART initiation after engagement in care [10,12,13,23,39,40,42,46,51,54,56,65] with two of these de ning delayed ART initiation as commencing ART at CD4 count < 150 cells/mm 3 or at WHO clinical stage IV [10,51]. One study de ned delayed ART initiation as having a CD4 count below or at 200 cells/mm 3 and/or AIDS de ning illness at treatment start [62]. The remaining 12 studies measured the length of time between ART eligibility (based on guidelines available at a particular period of time) and ART initiation [9,14,20,41,43,49,55] or between HIV diagnosis and ART initiation including same day treatment (i.e. initiating treatment on the date of diagnosis) [52,53,63,64,66]. In this review, we used more inclusive de nitions for both outcomes. Accordingly, we de ned late linkage to HIV care as engagement in care at CD4 count < 350cells/mm 3 or at WHO clinical stage III/IV, and delayed ART initiation as starting HIV medication at CD4 count < 350cells/mm 3 or WHO clinical stage III/IV.

Linkage to HIV care
Summary of care linkage results are presented in Table 1. Maheu-Giroux et al [27] identi ed a median time to care linkage after HIV infection of 4.9 years. Among studies that investigated the rate of linkage to care since diagnosis, the rate was within three months of diagnosis [12,49,50,56,58,66,70]. The rate ranges from 24% in Tanzania [12] to 93% in South Africa [50]. Two Tanzanian studies estimated the rate of linkage to care within six months of diagnosis and reported a rate of more than 70% [19,45]. Contradictory results were reported by two studies, in Ethiopia [20] and South Africa [17]; while the former study reported care engagement in 75% of PLWH within one week of diagnosis, only 46% were linked to care within 12months in the latter. However, a more recent study in South Africa reported a rate of 55% within 12-weeks of diagnosis [60]. . Two Ethiopian studies [11,59] reported prevalences of 67% and 60%, when considering baseline CD4 counts of < 200 cells/mm 3 as late presentation for care. Another study in Kenya [15] identi ed a prevalence of 23%, de ning late presentation as engagement in care at CD4 count ≤ 100 cells/mm 3 .
Structural, psychosocial, behavioural and sociodemographic factors were reported to be associated with late linkage to care. Eight studies identi ed healthcare delivery factors [7,12,15,19,45,47,49,61]. More than an hour travel time to reach a clinic [12,15], accessing care at a rural healthcare facility compared to an urban health facility, and having diagnosis through community-based approaches compared to health facility-based approaches were identi ed as risk factors for late linkage to care [12,15,19,45,47,49]. In contrast, Fomundam et al [7] in South Africa identi ed a

ART initiation
Results of studies that investigated ART initiation are presented in Table 2. Two studies in South Africa [53,64]   Six studies determined the rate of ART initiation at various time intervals after treatment eligibility [9,14,20,41,43,55]; within one, two and three months. The rate varied from 41% in Rwanda to 48% in Ethiopia within the rst month [20,41], 75% both in Kenya and South Africa within the second month [43,55], and from 67% in South Africa to 78% in Senegal within the third month [14,65].
Various factors were reported to in uence ART initiation, some of which were akin to those in uencing linkage to care. Eight studies identi ed service delivery factors as barriers to ART initiation [9, 10, 13, 40, 41, 43, 49]. Relative to diagnosis in voluntary counselling and testing (VCT) services, lower odds of ART initiation were reported in PLWH who were diagnosed at health care facilities with a high volume of patients, and in those who enrolled in care through inpatient wards and provider initiated counselling and testing (PICT) services [10,41,43]. While Brown et al [13] found a lower likelihood of initiating ART in PLWH who were diagnosed through community-based approaches compared to health facility-based approaches, Lopez-Varela et al [49] and Rentsch et al [12] identi ed no association between a testing modality and ART initiation. A lower likelihood of ART initiation was reported in PLWH residing more than 2km away from the nearest health care facility, in those who experienced perceived communication barriers with health care providers, and in those with low awareness about HIV care [9,10,40].
Eight studies reported behavioural or psychosocial factors relating to ART initiation [9,10,23,40,52,53,56,65]. A lack of perceived susceptibility to, and understanding of the severity of, the consequences of late treatment, as well as a lack of belief in the health bene ts of early treatment predisposed PLWH to delayed ART initiation [40]. Testing that was undertaken due to symptoms, and patients readiness to commence treatment, were positively associated with ART initiation [10,56], however patient desire for repeated testing was found to predict delayed ART initiation [23].
Nash et al [10] found an association between psychological distress and delayed ART initiation, but Cholera et al [54] reported no association between these variables. While using any substance as a coping mechanism decreased the odds of ART initiation [52], PLWH who reported drinking alcohol were 76% less likely to initiate ART than those who did not [56].
Factors related to social support (such as having a regular partner, living in a two-adult household and the presence of another household member taking ART) were positively associated with ART initiation [53,65]. Perceived social stigma and failure to disclose HIV status predicted delayed ART initiation [56] although this nding has been contradicted by another study [51].

Meta-analyses of factors affecting linkage to HIV care and ART initiation
Eighteen studies involving a combined total of 27,396 people were included in the meta-analyses to assess factors affecting linkage to care and ART initiation. People in younger age groups (< 35 years) were 29% ( Fig. 2a was removed from the analysis of the effect of age on ART initiation because of a high level of heterogeneity. Employed people and people who travelled for more than an hour to reach a clinic were more than 1.3 ( Fig. 3a; OR: 1.32; 95%CI: 1.14-1.52, I 2 = 14%) and 1.2 ( Fig. 3b; OR: 1.27; 95%CI: 1.15-1.39, I 2 = 57%) times more likely to be presented late for care, respectively. A study by Lifson et al [59] was excluded from the analysis of employment and presentation for care due to a high level of heterogeneity.

Discussion
Timely initiation of ART is essential to prevent AIDS and non-AIDS related comorbidities and mortality [2,3], as well as reducing the likelihood of new HIV infections [4,5]. This review demonstrated substantial disparities in the rates of linkage to HIV care and ART initiation across nations in SSA and between settings within a given nation. Overall, care linkage and treatment initiation rates are considerably low in most settings as compared to the second target of UNAIDS 95-95-95 goal, which aims to initiate treatment in 95% of HIV infected individuals [71]. Through the literature synthesis, we identi ed healthcare delivery (structural), psychosocial, behavioural and sociodemographic factors as determinants of late linkage to care and delayed ART initiation amongst HIV infected adults in SSA.

Structural factors
Our meta-analyses identi ed distance to ART sites as the main risk factor for late linkage to care in SSA countries. Similar ndings have been reported by a previous review in which transport costs associated with distant ART clinics was the most cited barrier to care in the region [32].
Although PLWH tend to engage in care more when it is easily accessible [9], many PLWH in SSA may be required to travel long distances, sometimes on foot, to access HIV care due to shortage of transport or associated costs [16,30]. This could be a major concern for low wealth index PLWH households, and partially explain why they are less likely to be linked to HIV care and commence treatment. There have been substantial expansions of ART services in the region in recent years, yet only a few public health care facilities provide the services at a district level [72], underscoring the need for the use of optimal task shifting [73,74] and service integration strategies [75] to reach all people who need treatment.
This systematic review and meta-analyses showed that PLWH initiate ART late when enrolled at clinics with a high volume of patients and diagnosed through community-based counselling and testing approaches. Community-based HIV testing approaches have substantially increased the number of people eligible for ART in SSA [72]. However, insu ciency of appropriately trained staff continues to be a main challenge to initiate treatment in all infected individuals [76]. Studies show that the more PLWH are satis ed with pre-ART care and understand the information given by service providers, the greater the likelihood of ART initiation [10,43,77]. In contrast, when clinic operating hours are not well tailored with PLWH's daily routines, timely clinic visits diminish, which leads to delays in treatment commencement [24]. This may help to explain why employed individuals are more likely to be linked to care late relative to their unemployed counterparts, as demonstrated from our meta-analyses. Enhancing after-hours services and workplace programs may help combat this problem, as would providing ART training for lower level health care staff [78,79].

Psychosocial factors
HIV status disclosure was signi cantly associated with an increased likelihood of linkage to care in the current meta-analyses, which is concordant with prior reviews conducted in SSA [30,32]. Status disclosure enables PLWH to access social support which reduces the negative in uence of social stigma, one of the barriers to accessing care in this review [17,31]. This is evidenced by the nding that married PLWH and those who lived with adults (particularly with those who use ART) are more likely to commence treatment. Conversely, failure to disclose HIV status increases the likelihood of care disengagement during pre-ART period [22].

Perceptions related to clinical conditions
Our review showed that clinical circumstances such as having a higher baseline CD4 count (> 350 cells/mm 3 ) and lower levels of WHO de ned clinical stages are associated with late linkage to care and delayed ART initiation. This is consistent with a narrative review conducted previously in the region, which reported initiation of ART at a very low CD4 count in most PLWH [30]. At asymptomatic stages of HIV infection (i.e., at high CD4 count and low WHO clinical stages), PLWH often feel healthy and may perceive that they do not need treatment [19,31]. During these stages, PLWH may also hesitate to accept a HIV positive diagnosis, thus requiring repeated testing that can lead to delayed linkage to care [10]. However, the current review also demonstrated that HIV-related symptoms alone may not always be su cient to prompt ART initiation, but patient readiness and con dence that the treatment is safe and e cient is also required. Structural factors related to prioritisation of the sickest patients, and low absorptive capacity of health care facilities may also contribute to initiation of ART at low CD4 count in SSA [30]. The rapid expansion of the program in the region may hopefully mitigate these structural barriers [72]

Other behavioural factors
Using positive reframing as a coping strategy was found to be associated with a high rate of linkage to care. This is consistent with previous ndings that showed the positive impact of a desire for good health on care engagement [87]. PLWH with such forethought commence ART hoping that their general health would be improved because opportunistic diseases could be prevented, which could also ultimately minimise social stigma due to HIV-related illnesses [18].

Sociodemographic factors
In this review, males and younger PLWH (below 35 years of age) are more at risk of late linkage to care and delayed ART initiation compared to females and PLWH of older age groups (35 years and above) respectively. The lower rates of linkage to care and ART initiation in males and younger PLWH in the current review support ndings of previous reviews conducted in SSA [30,32]. PLWH of younger age groups are known to have low awareness of their HIV status, and are more likely to experience and adversely react to stigmatisation, as well as engage in substance use [88], which is a signi cant predictor of delayed ART initiation in the current review. Moreover, younger PLWHA struggle with disclosure of their HIV status which may lead to limited access to information and material supports [89,90].
Contextual and cultural norms related to masculinity can play a strong part in hampering health seeking behaviour in males [31,91]. Females tend to be more engaged in health care systems through programs focusing on maternal health. Therefore, adaptation of health services and treatment options to the needs of men and younger people may help close the gaps in linkage to care and ART initiation.
In interpreting the ndings of this review, the following important limitations should be considered. The included studies represent only a few nations of SSA, which restricts the generalisability of the ndings. Representativeness is also restricted due to the wide variability of outcomes across geographical locations. Because most studies used a retrospective cohort or a cross-sectional design, causality between the exposure variables and the outcomes cannot be assured even though important risk factors for late linkage to care and delayed ART initiation have been adequately explored. Rates of linkage to care and ART initiation were measured at varying lengths of time using different reference points, which impacted the development of precise estimation of the outcomes. However, relatively more inclusive measures were taken to embrace a range of results in the analysis. As there has not been a standardised de nition for delayed ART initiation, the included studies de ned the outcome differently, following the available treatment eligibility guidelines within a particular period of time. To minimise this discrepancy, we used the highest and lowest cut points to ensure generalisability of the ndings to all included studies. In addition to substantial heterogeneity between studies in effect measures (with respect to some of the exposure variables), only 54% of the included studies were scored at moderate or above in the overall quality assessment, which may lower the quality of evidence. Although the review used a systematic search strategy, there exists a possibility of missing relevant studies because screening was undertaken by a single reviewer and unpublished data were not explored. Due to time and resource constraints, we included only studies published in English language which may increase the risk of publication bias, and we did not report a funnel plot due to the small number of studies (n < 10) included in the analysis for each exposure variable [92]. Finally, despite efforts in this regard, we were not able to contact authors of primary studies regarding incomplete data, which restricted the analysis of factors for delayed ART initiation.

Conclusions
This systematic review and meta-analyses identi ed a range of risk factors for late linkage to care and delayed ART initiation amongst HIV infected adults in SSA, which included: health service delivery, psychosocial, behavioural and sociodemographic circumstances. We recommend implementation of patient-centred intervention approaches to alleviate barriers and to reinforce best practices and lessons learned from high achieving settings to those with particular challenges. Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
Not applicable.
Authors' contributions TGF developed the search strategy; conducted searching, screening of the articles, data extraction and analysis; drafted the manuscript. ERM participated in the quality assessment of the studies and subsequent revisions of the manuscript; GT contributed to and reviewed the manuscript.

Figure 1
Study ow diagram. Study selection process and reasons for exclusion.

Figure 2
Forest plot of associations between linkage to care and age (a), and ART initiation (b). Lower likelihood of linkage to care and ART initiation in people in younger age groups (<35 years).