Introduction

South Africa has the world’s largest HIV epidemic, which has been met with an ever expanding and increasingly robust response since 2004, enabling the development of the single biggest antiretroviral therapy (ART) program globally. There are now over 3 million people on treatment in South Africa, which represents roughly half of the people living with HIV (PLWH) in the country [1,2]. The expansion in treatment availability, first ushered in by the US President’s Emergency Plan for AIDS Relief (PEPFAR) and the Global Fund and its partners [3,4] and now predominantly run through a governmental response, has increased the availability of ART for healthier PLWH [5,6].

South Africa has undergone multiple expansions in ART eligibility, with an increasing immunological threshold for ART initiation, from a CD4⁺ threshold of ≤200 cells/μl at the start of the treatment program to providing ART for all PLWH as of September 2016. The expansion of ART eligibility has resulted in a larger number of individuals being screened for treatment. Simultaneously, funding has shifted from programs primarily funded through external donors, such as PEPFAR and the Global Fund, to clinics run by the South African Department of Health (DOH). This has resulted in transitions from centralized, physician-managed programs to more decentralized, nurse-managed clinics [7–10].

While guidelines have shifted to expand earlier access to ART, it is unknown whether the expansion of treatment eligibility and availability has had an impact on patients’ engagement in care and early retention on treatment. This is particularly critical in the context of a new era of “test and treat” that has shifted the care cascade to be more focused on earlier ART initiation and durable retention [11]. Prior research in the earliest phases of treatment availability in South Africa showed that risk of loss to follow-up and virological failure both increased over successive calendar periods as the number of patients per clinic provider increased markedly over time [5]. An individual site in Durban undergoing a rapid transfer of care from a robust PEPFAR-funded program to community-based clinics reported loss of up to 20% of patients, who may have experienced a treatment interruption [12]. Similarly, a study of over 5,000 patients transitioning care from a physician-managed clinic to a nurse-managed program showed a significantly higher rate of loss to follow-up among patients who were down-referred [13].

We performed a retrospective cohort analysis of data from a large, urban community HIV treatment site in Cape Town to assess the association of South Africa’s HIV treatment eligibility guidelines with pre-ART attrition and early loss from care (<16 weeks). We hypothesized that increasing the CD4⁺ threshold to access ART would increase pre-treatment and early losses. The rationale for this hypothesis was that there could be a healthy cohort effect and/or that the programmatic shift to expanded treatment eligibility could result in a higher patient to nurse ratio, resulting in a crowding effect.

Methods

Results

In all, 4,025 ART-eligible PLWH who were referred to the treatment clinic between 1 January 2009 and 31 December 2013 were included in our sample. The median age in our population was 34 years (IQR 28–41 years) (see Table 1). Nearly 62% were female, and the median CD4⁺ count was 173 cells/μl (IQR 92–254 cells/μl). Overall, individuals in the earlier cohort had significantly more advanced disease than those in the later cohort, with lower CD4⁺ counts at the time of ART initiation (146 cells/μl versus 214 cells/μl, respectively, p < 0.001), and a larger percentage were classified as having a higher WHO stage (61.1% with stage 3/4 versus 42.8%, respectively, p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of all participants by qualification period.

https://doi.org/10.1371/journal.pmed.1002434.t001

Ninety percent (n = 3,640) of the population initiated ART within 16 weeks of first entering the clinic, per pharmacy records (see Table 2). There was no significant difference in the percentage of the population who initiated ART within 16 weeks between the earlier and later cohorts (90.6% versus 90.2%, p = 0.7). Of the 9.6% (n = 385) who did not start ART, 25.7% died prior to starting treatment. The rate of pre-ART death in the earlier cohort was twice that of the later cohort (34.2% versus 16.7%, p < 0.001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Early loss from care and mortality in Cape Town, South Africa.

https://doi.org/10.1371/journal.pmed.1002434.t002

Among the cohort who initiated treatment, 17.7% stopped accessing treatment within 16 weeks of ART initiation, and 1.6% died within the first 16 weeks. ART-eligible individuals in the later cohort were significantly more likely to discontinue care <16 weeks into treatment compared to those initiating treatment prior to 31 August 2011 (19.8% versus 15.8%, odds ratio [OR] = 1.32, p = 0.002). After controlling for baseline CD4⁺, WHO stage, and age, this effect remained significant (adjusted OR [aOR] = 1.30, 95% CI 1.09–1.55). When the analysis was restricted to only the individuals who would have qualified for ART in either cohort (CD4⁺ ≤ 200 cells/μl), the difference in early ART discontinuation remained significant (aOR = 1.34, 95% CI 1.06–1.67).

Fig 1 shows attrition across the cascade from the pre-ART period through to 16 weeks on ART in both the early and later cohorts. Across the full cohort, 157 (3.9%) ART-eligible PLWH died, and 930 (23.1%) were lost from care prior to ART initiation or within the first 16 weeks of starting treatment. This resulted in a total combined early loss of 1,087 (27.0%) ART-eligible PLWH. Over the 5 years of the study period, the total number of people entering care increased over 2-fold, from 776 entering care in 2009 to 1,506 entering care in 2013. During this time, the Gugulethu clinic transitioned from 5 doctors and 5 nurses in 2009 to 3 doctors and 7 nurses in 2013. The standard of care during this period required 8 clinic visits for patients initiating treatment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Attrition in the care cascade.

PLWH, people living with HIV.

https://doi.org/10.1371/journal.pmed.1002434.g001

Discussion

In this cohort study, we assessed early losses from care over a 5-year period as South Africa was expanding its ART eligibility, and shifting from a centralized HIV treatment program with a high number of medical doctors, funded through PEPFAR, the Global Fund, and other partners, to a more decentralized, nurse-led system, supported largely through the South African government. Overall, we found that over one-quarter of this well-established ART-eligible cohort never achieved the long-term benefits of treatment and viral load suppression due to early mortality, failure to start ART, or ART discontinuation <16 weeks from the time of initiation. Patients who entered care in the later cohort were significantly more likely to discontinue treatment early.

Estimates of pre-ART attrition from care and early loss from care on ART among ART-eligible PLWH have consistently been estimated at 20%–30% of patients in South Africa [18–21]. What has remained less clear is the impact of guideline changes and shifting delivery systems on attrition and early loss over time. In 2013, WHO published consolidated guidelines on the use of antiretroviral drugs for HIV treatment and prevention across all age groups and populations, that were then revised based on new scientific evidence in 2015 [22]. WHO set forth these guidelines to increase treatment CD4⁺ thresholds, with the explicit goal of reducing transmission between sexual partners [23] and providing health benefits for patients taking treatment [24–26].

As the South African government has moved to incorporate these recommendations, expanding treatment to all PLWH, more people have become eligible to start ART. However, challenges remain in both ART initiation and retention in care, resulting in the failure of these strategies to achieve the expected benefits for the individual, let alone at a population level [27]. Recent research by Haber et al. showed that in a large population of PLWH in South Africa, the early stages of the cascade were the most vulnerable to losses [28]. South Africa has faced the added challenge of transitioning from a well-resourced HIV care program supported by donors to a decentralized system of care though a vast network of public health clinics, while expanding enrollment into ART programs throughout the country.

The impact of policy-level changes on treatment outcomes remains uncertain. In this study, the rate of early loss continued to rise throughout the study period. While the later cohort was significantly healthier than the earlier cohort, the increase in loss from care persisted in adjusted analyses and thus did not appear to reflect differences in CD4 count and WHO stage. We believe this trend may reflect large programmatic shifts in care and clinic crowding, as efforts have continued to expand to accommodate millions more PLWH into treatment programs despite constrained resources and a limited number of healthcare providers [3,29]. In addition, our prior qualitative research [30], along with prior research conducted by Fox et al. [31] and Duff et al. [32], suggests that perceptions of health and illness may be a strong driver of ART decision-making, and that these perceptions may not always be correlated with actual CD4⁺ counts [33]. As such, the goal of getting patients onto treatment earlier may remain a challenge if PLWH perceive themselves to be healthy, especially in settings where clinic crowding remains significant.

Our data have several limitations and a number of strengths. First, we are limited by the fact that these data were accumulated at a single high-volume site. Therefore, it is unclear if these findings are generalizable. Despite this, our findings are consistent with data from other multi-site, large cohorts [27,28,34]. Second, despite our access to staffing numbers, we are unable to formally investigate whether a crowding effect was the true cause of higher rates of loss from care in the later cohort. Third, our sample cannot account for patients who have potentially left the area and accessed care in other provinces. While this is a challenge consistently noted in previous research [35,36], this clinic utilized a robust tracking system when patients did not return for treatment, including active tracing of patients who missed clinic visits through home visits by community care workers. Finally, these data were collected from a site where treatment and testing were available. As such, it remains unclear if the linkage rates observed would be as high if the site was a stand-alone testing site where patients then had to link to a new site for care.

In conclusion, over one-quarter of this well-established ART-eligible cohort did not achieve the long-term benefits of treatment due to early mortality, failure to start ART, or ART discontinuation <16 weeks in care. Early ART discontinuation, which appeared to be independent of CD4⁺ count or WHO stage, likely reflected larger programmatic trends towards higher volume treatment centers that result in clinic crowding. Future interventions should focus on those most at risk for pre-ART attrition and early loss from care as programs continue to expand in an era of treatment for all in South Africa.

Supporting information

Acknowledgments

The authors thank the study participants and the study teams at the Desmond Tutu HIV Centre, University of Cape Town Medical School, Cape Town, South Africa, and Harvard Medical School, Boston, Massachusetts, US.

The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the US National Institutes of Health.