This analysis demonstrates the increasing role played by LTFU over time in the programme outcomes of the South African national ART programme. The rapid pace of ART scale-up in South Africa is evident from the 12-fold increase in this analysis in the number of patients starting ART since 2002, with 63% of all patients initiating ART during 2006 and 2007 alone. While recorded mortality has declined during this period, observed LTFU has increased substantially and presents a major threat to evaluating the effectiveness of the national programme.
Patient retention is a vital measure of the effectiveness of ART services [6
]. Retention in long-term care is complex, especially in low- and middle-income countries [8
], but not a new issue: primary health care services have long faced the problem of patient attrition in providing care for chronic diseases [6
]. A systematic review of ART programmes in Sub-Saharan Africa found large variation in patient retention across programmes, ranging from 46–85% after 2 years on ART [12
]. At the start of the South African national programme, based on experience with other chronic diseases, it was suggested that the ART service may retain 60–80% of patients annually [11
]. Retention in the earlier years of the programme exceeded this expectation: at 2 years, 71% of all patients were still known to be in care, but the steady increase in attrition during the first 12 months on ART in successive years of enrolment is cause for concern.
Mortality is one reason for patient attrition: in this cohort, observed mortality at 12 months was 6.6%, which is comparable with results from other developing countries [13
]. With successive years of enrolment, 12-month mortality decreased. This may be a true decline due to improved coverage of services and patients enrolling with less advanced HIV disease [14
]. It is also plausible that as a programme expands, its ability to accurately ascertain patient deaths deteriorates, and high observed LTFU may be associated with poor mortality ascertainment [15
]. It is likely that our study, based on routine surveillance, underestimates true mortality in these cohorts. Recent corrected mortality estimates for single South African ART cohorts (based on linkage to the national death register) found that at 3 years on ART, corrected cumulative mortality was 12–15% [16
] compared with our uncorrected estimate of 10%. There is an urgent need to improve ascertainment of deaths in low- and middle-income countries [18
Yet even with such underestimation, mortality is not the major reason for patient attrition in large ART programmes in developing countries. The greater threat to the success of the South African ART programme may be the observation of high levels of LTFU, insofar as this outcome reflects patients who have truly left care. The size and pace of ART scale-up may have contributed to observed LTFU. The programme has grown in size dramatically, with our combined cohort increasing enrolment 12-fold over 5 years. Such rapid increases have placed considerable strain on health services that were already overburdened [8
] and may have undermined the programme’s ability to monitor and retain patients in care. During 2007 alone, 33% of patients in this study were enrolled onto ART: compared with the 2002/2003 cohort of patients, they had a 12-fold higher risk of appearing LTFU. In addition, with longer duration on ART, observed LTFU accounted for an increasing proportion of overall programme attrition: from 9% at 6 months to 29% at 36 months on ART.
If the rapid expansion of ART services does increase observed LTFU, the situation may worsen as countries continue to expand access to HIV treatment. Based on 2002 WHO treatment guidelines, adult ART coverage in South Africa was an estimated 40% in 2008 [22
]. In addition, the South African government recently revised its treatment guidelines to include all infected infants <1 year of age, pregnant women with CD4 counts ≤350 cells/µL, patients co-infected with TB [23
]. South Africa and many other countries in sub-Saharan Africa will need to continue to expand services while retaining large numbers of patients in care. This will require strengthening systems for chronic disease care in these countries[6
], where most health programmes are oriented towards episodic illnesses and acute care.
Successfully re-orienting health systems towards long-term chronic care will require a better understanding of the phenomenon of LTFU. Often viewed as a single construct, observed LTFU in an ART cohort more likely represents a range of patient outcomes including patients truly LTFU (ie. lost to care) as well as those classified LTFU through administrative error or inadequate patient monitoring systems [15
]. In a situation of rapid scale-up of ART in resource-limited health systems, the ability to capture and report patient data may become increasingly inadequate [24
]. Indeed, our results suggest that larger cohorts may have become more subject to these challenges in recent years. For example, the apparently sharp increase in observed LTFU among patients enrolled in 2007 is likely to reflect the cumulative burden of increasing patient numbers on both ART services and health informatics systems. This phenomenon may be particularly acute at larger and rapidly expanding ART sites, some of which enrolled up to 50% of their cumulative number of patients in 2007 alone.
Despite the scope of the problem of observed LTFU in ART services in southern Africa, relatively little is known about this phenomenon. These cohorts, which are largely funded by the national Department of Health, report active tracing (ie dedicated resources to undertake one or more of the following: telephone call, home follow-up, physician’s report and/or data linkage [3
]. However, largely due to resource constraints, funding for patient follow-up, particularly at this scale, is limited. There is a small literature on factors associated with patient retention highlighting the possible role of patient preparation [25
], treatment supporters [26
], patient costs [10
], improved databases [24
], community support [28
] and simplified services [6
]. However, research is needed to better understand observed LTFU and the relative contributions of true LTFU (patients dropping out of ART services) versus administrative LTFU (patients who are retained in care but appear LTFU due to problems with data capturing and reporting). In contrast to these individuals, patients who are truly LTFU are likely to be non-adherent to treatment and at higher risk of death [16
]. In addition, they face increased risk of drug resistance to ART, undermining the long-term effectiveness of treatment programmes [12
]. Additional research is needed into the programme-level determinants of LTFU, better characterization of patients classified LTFU and insights into patients’ movements in and out of care.
This is the first report on outcomes from multiple cohorts in the world’s largest antiretroviral therapy programme, and to our knowledge, the largest analysis of individuals starting ART in sub-Saharan Africa. It is strengthened by up to 5 years of patient follow-up on more than 40,000 patients. The results are likely generalisable to the patient population accessing public sector ART in most of South Africa [3
] where 80% of the population rely on the public sector for services [30
]. However, this analysis has several important limitations. As is the case with other large-scale ART programmes based on routine monitoring and evaluation, it is constrained by issues of outcome ascertainment and missing data [24
]. Outcome ascertainment should improve as more cohorts in South Africa link to the death register, presumably increasing observed mortality and decreasing observed LTFU. Data completeness may continue to present a challenge, particularly as programmes continue to expand. WHO staging were the least complete data point in this analysis, yet their inclusion in multivariate analysis impacted on the association between baseline CD4 count and outcomes, highlighting the importance of complete baseline data. Finally, this paper reports on averages across cohorts which may differ in data quality, completeness and outcome ascertainment. Despite these constraints, this analysis utilizes routinely collected data to provide valuable insight into the effectiveness of a huge national programme, and has important implications for South Africa and for other programmes in similar contexts.
In summary, this analysis demonstrates that the South African national ART programme has undergone rapid scale-up over the past 5 years. While recorded mortality has decreased, programme retention has deteriorated, as decreasing patient mortality has been greatly offset by high and increasing levels of LTFU. This increased LTFU may represent true loss to care, but also may be due to increasing difficulty in monitoring patients enrolling into care as well as patient movements in and out of care. These possibilities require further investigation. Innovative, effective strategies are needed to follow and retain patients in large HIV treatment programmes while rapidly expanding access to ART services.