Bulletin of the World Health Organization

Scale-up of a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa: does rapid expansion affect patient outcomes?

Portia C Mutevedzi a, Richard J Lessells a, Tom Heller b, Till Bärnighausen a, Graham S Cooke a & Marie-Louise Newell a

a. Africa Centre for Health and Population Studies, PO Box 198, Mtubatuba, 3935, South Africa.
b. Hlabisa Hospital, Hlabisa, KwaZulu-Natal, South Africa.
C. orrespondence to Portia C Mutevedzi (e-mail: pmutevedzi@africacentre.ac.za).

(Submitted: 03 July 2009 – Revised version received: 09 December 2009 – Accepted: 11 December 2009 – Published online: 10 May 2010.)

Bulletin of the World Health Organization 2010;88:593-600. doi: 10.2471/BLT.09.069419


South Africa is home to 5.7 million people living with human immunodeficiency virus (HIV) and acquired immunodeficiency syndrome (AIDS), or approximately 1 in 6 of the world’s HIV+ population.1 It also has the largest public-sector antiretroviral therapy (ART) programme in the world, with an estimated 460 000 individuals established on ART by December 2007. However, as 1.5 million people living with HIV/AIDS in South Africa are estimated to be in need of ART, this figure only equates to 30% coverage.2 If the ambitious target set under the National Strategic Plan of universal access by 2011 is to be achieved, scale-up of ART services must be intensified.3,4

There is good evidence that ART can be delivered effectively in South Africa, with individual virological and immunological responses to ART equivalent to those in high-resource settings.5 Most published evidence relating to ART scale-up in South Africa, however, has come from urban treatment cohorts or from provinces with primarily urban populations.59 Neighbouring countries in southern Africa have also reported programme outcomes, although again based largely on urban cohorts.10,11 Only the Malawi national programme has reported large-scale data on a predominantly rural population.12 Delivery of HIV treatment and care to rural communities presents unique challenges, and current ART delivery models may significantly limit the accessibility of ART.13 To have the greatest impact on public health, HIV treatment programmes will have to be decentralized and integrated into the existing primary health care system.14 Preliminary evidence from such rural programmes has demonstrated that ART provision in rural communities is feasible, given the appropriate resources and infrastructure.15,16

Concern has been raised that the rapid expansion of HIV services will reduce the quality of care for individuals within the programme as capacity and resources are stretched.17 Characteristics of individuals accessing treatment may change over time, and this in itself may affect overall treatment outcomes.18 Monitoring treatment outcomes is essential to identify constraints or deficiencies in programme performance, and viral load monitoring may provide warning of adherence problems and of possible development of acquired antiretroviral resistance.19

To evaluate a decentralized model of ART delivery in a rural community, we analysed the scale-up of our programme and explored trends in the characteristics of individuals accessing treatment and their outcomes within the first year of treatment.



The Hlabisa HIV Treatment and Care Programme is a partnership between the local Department of Health and the Africa Centre for Health and Population Studies to deliver the South African Comprehensive HIV and AIDS Care, Management and Treatment Plan.20 It is a decentralized programme of HIV treatment and care delivered through a network of 16 primary health care clinics in the Hlabisa subdistrict of Umkhanyakude district in northern KwaZulu-Natal. The subdistrict covers an area of 1430 km2 and has a population of approximately 220 000, mostly living in rural areas.21 HIV infection prevalence peaks at 50.9% among females aged 25–29 years and 43.5% among males aged 30–34 years.22 The programme employs a public health approach to ART delivery aimed at facilitating rapid scale-up of HIV treatment services.14,23 Support for the programme is provided by the President’s Emergency Plan for AIDS Relief (PEPFAR) through the United States Agency for International Development (USAID).


The programme adheres to the national antiretroviral treatment guidelines, which recommend initiation of ART for adults with stage IV disease as defined by the World Health Organization (WHO)24 or a CD4+ lymphocyte (CD4 cell) count of < 200 cells/µl. 25 The first-line ART regimen consists of stavudine, lamivudine and either efavirenz or nevirapine. ART is initiated at primary health care clinics (or at Hlabisa hospital) by a physician after standard pre-ART evaluation and three treatment literacy sessions. Monitoring and ART dispensing are performed by nurses and counsellors at 2 weeks and 4 weeks and then at 4-weekly intervals thereafter. CD4 cell count and HIV load are measured every 6 months.

Data acquisition

Longitudinal individual patient records are maintained at primary health care clinics. After initiation of ART, clinical and demographic information is transferred to a database developed and housed at the Africa Centre for Health and Population Studies (ART Evaluation and Monitoring Information System, or ARTemis), which is then updated monthly with follow-up status. Laboratory results (CD4 cell count and HIV load) are regularly imported into ARTemis. Demographic information is collected biannually and put into the Africa Centre Demographic Information System for approximately 90 000 individuals who are members of households within the Africa Centre Demographic Surveillance Area,26 which covers about one-third of the Hlabisa health subdistrict. Information relating to socioeconomic status and household structure for individuals in ARTemis is therefore obtained through linkage of data sets, for all of which the patient’s national identity number is used. For these analyses, data were taken from the demographic surveillance period encompassing ± 9 months from the date of ART initiation.

Data analysis

Our analyses included all adults (≥ 16 years of age) who initiated ART within the programme between 1 October 2004 and 30 September 2008. Patients already on ART (504) who transferred into the programme were excluded. Patients were categorized into 6-monthly groups on the basis of date of ART initiation and were stratified into four clinic groups: hospital-based (1 clinic); urban/periurban (2 clinics); large rural (≥ 200 patients on ART as of 30 September 2008 – 6 clinics); and small rural (< 200 patients on ART as of 30 September 2008 – 7 clinics). We assessed trends and differences in baseline characteristics across the 6-monthly and clinic strata on the basis of median tests and centile regression for continuous variables and proportions test and logistic regression for categorical variables. To assess possible selection bias associated with incomplete linkage of clinical and population data, we compared baseline and outcome variables between linked and unlinked individuals.

Outcomes (death, loss to follow-up and detectable viral load) at 12 months post-ART initiation were analysed for all patients who initiated ART up to 30 September 2007. Follow-up status was analysed at 12 months, whereas HIV load values were taken at 12 months ± 2 months to allow for variance in collection time points. HIV load was measured at a provincial laboratory using the NucliSens EasyQ® HIV-1 assay (bioMérieux sa, Marcy l'Etoile, France), with a lower detection limit of 25 copies/ml. Loss to follow-up was defined as having no clinic visit for 90 days or more. A total of 84 patients (2.8%) had only a baseline visit before being reported as deceased (n = 29), lost to follow-up (n = 38) or transferred out (n = 17). These patients were allocated 15 days of follow-up – the median time between the baseline visit and their next scheduled monthly clinic visit. Kaplan–Meier survival analysis was used to assess trends in time to death and detectable viral load and Cox regression to assess determinants of these two outcomes.

To account for substantial missing viral load data, independent baseline factors associated with missing data at 12 ± 2 months post-ART were determined through logistic regression at a 25% significance level. The resultant deterministic model was used to generate a propensity-to-be-missing score, defined as the probability for each patient of having a missing viral load reading.27,28 The Cox proportional hazard model for detectable viral load was controlled for the propensity score as a re-weighting adjustment factor. Data were analysed using Stata version 10.0 (StataCorp, College Station, United States of America).

Ethical approval

Written informed consent was obtained from all patients in the programme to allow use of anonymous clinical data in research. Ethical approval was obtained from the Biomedical Research Ethics Committee of the University of KwaZulu-Natal and the Research Office of the KwaZulu-Natal Department of Health for linkage and use of these data.



A total of 5719 adult patients were initiated on ART in the programme between 1 October 2004 and 30 September 2008 (Fig. 1); 2509 patients (43.9%) reported a physical address within the demographic surveillance area at the time of ART initiation and hence were eligible for linkage, and 1804 (72%) of them were successfully linked.

Fig. 1. Cumulative number of patients initiating antiretroviral therapy between October 2004 and September 2008 in a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa
Fig. 1. Cumulative number of patients initiating antiretroviral therapy between October 2004 and September 2008 in a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa
HIV, human immunodeficiency virus.

Despite a 10-fold increase in the number of patients initiating ART (Table 1), the health worker-to-patient ratio only declined by 0.5, as staff increased from 6 in the first half-year of analysis to 99 in the last half–year. The most rapid growth occurred in large rural clinics as decentralization progressed (Fig. 2).

Fig. 2. Number of patients initiated on antiretroviral therapy from October 2004 to September 2008 in a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa, by sex and clinic group
Fig. 2. Number of patients initiated on antiretroviral therapy from October 2004 to September 2008 in a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa, by sex and clinic group
HIV, human immunodeficiency virus.

Baseline characteristics

Females (3883) made up 67.9% of all patients initiated on ART, with no significant differences in proportion of females between clinic groups. Males initiated at a higher median age (37 years, interquartile range, IQR: 32–45) than females (34 years, IQR: 28–41) (P < 0.001) and had a lower median CD4 cell count at initiation (91 cells/µl, IQR: 41–156) than females (128 cells/µl, IQR: 66–179) (P < 0.001). A higher proportion of males (29.6%) than females (18.2%) (P < 0.001) initiated with a CD4 cell count < 50 cells/µl (P < 0.001). Prevalent tuberculosis at initiation was more common in males (20.4%) than females (15.6%) (P < 0.001).

There were no significant differences across clinic groups, except for the median CD4 cell count, which was significantly lower at the hospital clinic (99 cells/µl) than at the urban/peri-urban (114 cells/µl; P = 0.032), large rural (128 cells/µl; P < 0.001), and small rural clinics (117.5 cells/µl; P = 0.023).

Table 1 shows overall baseline characteristics and trends over time.


People who initiated treatment between October 2007 and September 2008 (2709) were excluded, leaving 3010 patients for outcomes analysis. Overall retention at 12 months was 84.0% (95% confidence interval, CI: 82.6–85.3) and mortality was 10.9% (95% CI: 9.8–12.0). The proportion with a detectable viral load declined but remained high at around 23%. Table 2 shows overall outcomes, including trends over time.

At 12 months post-initiation, the 3010 patients had accumulated 2617 person-years of follow-up, with 329 deaths: a mortality rate of 12.6 deaths per 100 person-years (95% CI: 11.3–14.0). The mortality rate peaked in the first 3 months following ART initiation at 30.1 deaths per 100 person-years (95% CI: 26.3–34.5) and fell thereafter to 6.2 deaths per 100 person-years (95% CI: 5.2–7.4). Male gender, lower weight, CD4 cell count < 50cells/µl, low haemoglobin, high creatinine and lack of formal education were significantly associated with mortality (Table 3). Univariably, initiation outside the hospital clinic was associated with lower mortality (with reference to the hospital clinic, hazard ratio, HR: 0.49, 0.62 and 0.51 for urban, small rural and large rural clinics, respectively; P < 0.01), but significance was not maintained when we adjusted for time period. There were no statistically significant differences between people with missing values and the reference group for the respective variables (Table 3).

Of 2527 people recorded as active 12 months post-initiation, 758 (30.0%) had a viral load result recorded at 12 ± 2 months following ART initiation, and 174 of those 758 patients (23.0%; 95% CI: 19.5–25.5) had detectable viral loads (≥ 25 copies/ml). The propensity to be missing decreased from a median probability of 0.96 in the first half-year to 0.52 in the final half-year (P < 0.001). The hazard of having a detectable viral load was 2.8 (95% CI: 1.05–7.85) in the last three half-year periods compared to the three preceding periods. This increased risk lost statistical significance after we adjusted for the propensity score. We found no factors that were significantly associated with having a detectable viral load in the multivariable model.

Comparison of data

The proportion of patients who died was 11.3% in the unlinked group versus 10.3% in the linked group (P = 0.347). The proportion retained was 83.5% versus 84.8% (P = 0.423) in the unlinked and linked groups, respectively, and the proportion with a detectable viral load was 22.9% versus 23.0% (P = 0.987). Although the proportion of females in the unlinked group (66.4%) was slightly lower than in the linked group (71.1%) (P < 0.001), there were no significant differences in other baseline characteristics: CD4 cell count and proportion with CD4 cell count < 50cells/µl, blood creatinine, blood haemoglobin, weight and age.


To meet targets of universal access to ART, efforts to scale up HIV treatment programmes in South Africa need to be intensified. The treatment model described here, a decentralized programme with treatment delivered through the primary health-care system, has allowed rapid scale-up, with 5719 adult patients initiated on ART in a 4-year period. The rapid growth in rural clinics is encouraging and suggests that the availability of care at the local level reduces the common barriers (travel time and cost) to ART uptake. Success in scale-up may be partly attributed to growth in the programme’s financial and human resources, which increased by about 20-fold in the first 4 years as a result of PEPFAR support. Notwithstanding all efforts to scale up financial and human resources, most of the programme emphasis has been on initiating patients onto treatment. Other aspects of HIV care, such as pre-ART monitoring and long-term adherence support, have received less emphasis so far, but these will become more important as programmes mature, and additional resources will be required to maintain delivery of high-quality care.

Sustaining such programmes requires significant inputs of human and financial resources, and it may not be feasible to maintain steep increases. However, at some point a steady-state will likely be reached, and resources can then be focused on sustaining the programme. Task-shifting and reducing the number of times patients attend clinics for drug collection may enhance programme sustainability. Further integration of primary health care systems and HIV treatment and care delivery services will also facilitate programme continuity.

The characteristics of individuals accessing ART through the programme have changed little over time. Unlike the situation in other programmes in South Africa that started before the national roll-out of ART57, the median baseline CD4 cell count of patients served by our programme has remained relatively high since inception in 2004, although it is similar to that reported for other programmes in sub-Saharan Africa in 2005–2006.18 A high proportion of patients continue to start treatment with very advanced disease: 1 in 3 male patients have CD4 cell counts of < 50 cells/µl when they initiate ART. The higher proportion and younger age of females in the programme are consistent with local prevalence data and with other cohorts in South Africa.57,22 However, the difference in median baseline CD4 cell count between females and males (37 cells/µl) is even greater than the mean difference reported for other sub-Saharan Africa programmes participating in the Antiretroviral Therapy in Lower Income Countries (ART-LINC) Collaboration (22 cells/µl).18 The finding that male sex was also independently associated with mortality highlights the need for strategies to engage men in HIV care before the onset of advanced disease.

The overall mortality rate in our programme (12.6 deaths per 100 person–years) is high but still comparable to the rates found in other cohorts in southern Africa.8,2932 Most deaths occur in the first 3 months after ART initiation, as seen in other cohorts, largely as a result of people presenting with late-stage disease, although our rate of 30.1 deaths per 100 person–years during the first 3 months of ART is much higher than reported elsewhere. The contribution of immune reconstitution inflammatory syndrome (IRIS) and/or the lack of routine physician follow-up to this early mortality cannot be ascertained from our data. Linkage with our demographic surveillance data has revealed that lack of education is significantly associated with death among patients on ART. This relationship has been described elsewhere and may indicate a group that requires enhanced support both before and after initiating ART.33

We show relatively low rates of loss to follow-up. This is likely explained in part by the active tracking system within the programme, which includes phone contact by clinic staff, followed, if necessary, by a home visit from a tracker nurse. Our figures for mortality and loss to follow-up might thus reflect the true situation better than those from other programmes in which patients are not actively tracked and where up to half of those lost to follow-up may have died.34,35 It is possible that the decentralized model, with treatment provided closer to patients’ homes, also contributes to this low rate of loss to follow-up. It is encouraging that there is no sign of a consistent increase in loss to follow-up rates despite the rapid scale-up of the programme.

One troubling aspect of our data is the relatively high rate of detectable viral load among patients on ART. There were many missing results, which is likely to be related to several factors: samples not taken at appropriate time points, results not returned from the provincial reference laboratory and faults in data capture. However, by using propensity score reweighting, we have shown that almost 1 in 4 people on ART had a detectable viral load at around 12 months. We used a strict cut-off (< 25 copies/ml), as we believe that this will alert us to deficiencies within the programme, although this approach makes comparison with other programmes using different criteria difficult. Primary drug resistance should not be a significant factor in lack of virological suppression, as levels of primary resistance were low in our setting at the time of the analysis.36 There have been no significant stock-outs of ART in our programme, and the high rate of unsuppressed viral load may therefore suggest suboptimal adherence. We are researching this question in more detail and are learning how to translate successful models of adherence support from neighbouring programmes.37

The similarity between the unlinked and linked groups is as expected, given that most of the lack of linkage in those eligible arose as a result of typographical errors in entering identity numbers into patient records, failure to match the identity number to other personal identifiers, or the fact that patients presented at clinics without identity documents. Since these occurrences probably occurred at similar rates in all groups, we do not believe they influenced outcomes or their associated factors.

Our study had certain limitations. Data collected from routine programme evaluation are limited by operational constraints that might not hamper formal research cohorts. Our description of scale-up and outcomes is limited by a lack of information on individuals enrolled in the programme but not initiated on ART. Data from other programmes suggest that mortality is extremely high in the period after enrolment but before initiation of ART.29,30 It would be extremely useful to evaluate whether our decentralized model based on primary health care improves access to treatment and reduces pre-initiation mortality but increases early post-initiation mortality. By exploring outcomes at 12 months only we may have missed trends in longer-term outcomes, particularly retention and viral suppression.

In summary, we have shown that rapid scale-up of ART delivery through a decentralized primary health care programme is feasible and produces outcomes comparable to those reported elsewhere in southern Africa. The relatively high rates of detectable viral load in patients on ART are troubling, and further scale-up poses significant challenges and will require considerable commitment if we are to reach the goal of universal access.


We would like to thank Colin Newell for development of and assistance with the database, Samukelisiwe Dube for assistance with data-cleaning, the monitoring and evaluation team at the Africa Centre for data collection, and the Department of Health. We would particularly like to thank all the clinic staff whose dedicated work in the face of so many challenges is an inspiration to us all.


The HIV treatment and care programme has received generous support from the United States Agency for International Development (USAID) and the President’s Emergency Plan for AIDS Relief (PEPFAR) under the terms of award no. 674-A-00-08-0001-00. The opinions expressed herein are those of the authors and do not necessarily reflect the views of USAID or the United States Government. Tom Heller and Till Bärnighausen are supported by the Centre for International Migration and Development, Gesellschaft für Technische Zusammenarbeit (GTZ), Federal Ministry of Economic Cooperation and Development, Germany. The Africa Centre is supported by a core grant from the Wellcome Trust.

Competing interests:

None declared.