To achieve national and international targets for treatment of paediatric HIV/AIDS, provision of paediatric antiretroviral therapy (ART) must continue to expand in low-income and middle-income countries. In South Africa, as in many countries, access to paediatric treatment services has lagged behind that of adult services, with reported coverage of just 36% of the estimated 300 000 treatment-eligible South African children in 2010, compared with adult coverage of 55% of the approximately 2.5 million treatment-eligible adults . In response, South Africa's National Strategic Plan for HIV/AIDS sets a target of initiating 90% of eligible children on ART and of ensuring that 85% of them remain on treatment by 2016 .
According to a recent meta-analysis and several cohort studies, children on ART in Africa experience high survival and large improvements in immunological function and growth over the first 1–3 years on ART [3–9]. In a pooled analysis of data from multiple African countries, 82.3% of children remained alive and in care 24 months after treatment initiation . In the Western Cape Province of South Africa, 82% of children remained in care after 3 years, with the proportion of patients with viral suppression estimated at 73–75% and the proportion with CD4 percentages more than 20% increasing from 58% after 6 months of treatment to 83% after 36 months . In South Africa's Gauteng Province, 83% remained in care after 36 months, and 96.2% of those remaining in care at 24 months were virally suppressed .
Encouraging as these findings are, securing the additional funding needed to increase coverage among children, at a time of flat or declining donor budgets, will continue to be a challenge. Essential to meeting that challenge will be accurate information on the actual costs of delivering paediatric ART. There is only one published estimate of the costs of providing paediatric ART in resource-constrained countries available to policy makers; however, South Africa is not included in its analysis . Decisions about the scale-up of paediatric ART in South Africa and other countries are, thus, being made with incomplete information about their consequences. The objective of this study was to estimate and analyze the average cost of providing paediatric ART, stratified by patient outcome. This information will assist policy makers, treatment providers, and funding agencies to understand the factors that influence paediatric ART costs and outcomes, estimate resource needs, and improve the efficiency of the national treatment programme.
Settings and cohorts
We collected data on outpatient resource use during the first 24 months after ART initiation at two paediatric ART clinics in Gauteng Province: Empilweni Services and Research Unit (ESRU) at Rahima Moosa Mother and Child Hospital, with data collected for 2005–2009; and Harriet Shezi Children's Clinic (HSCC) at Chris Hani Baragwanath Hospital , with data for 2007–2009. The sites were both located in public sector academic hospitals in urban areas, but they differed sharply in patient volume. At ESRU, the number of children on ART increased from 518 to 1253 over the study years from 2005 to 2009. At HSCC, it grew from 1617 to 2434 over the study years from 2007 to 2009. Both sites followed South African guidelines for treatment initiation and antiretroviral regimens . During most of the study period, ART initiation criteria included more than two hospitalizations per year or prolonged hospitalization (>4 weeks) for HIV-related disease; modified WHO Stage 3 or 4 disease; or a CD4 percentage less than 20% in a child under 18 months old or less than 15% in a child over 18 months old, both irrespective of disease stage. The recommended paediatric regimens consisted of stavudine, lamivudine, and lopinavir/ritonavir for children initiated at less than 3 years of age and of stavudine, lamivudine, and efavirenz for children initiated at 3 years or older .
We enrolled a consecutive sample of 150 children in each clinic who had initiated ART below the age of 13 years, based on a register of patients in care by March 2008 (ESRU) and January 2009 (HSCC). Patients who had been enrolled in a clinical trial, initiated while admitted in a hospital ward, or transferred out of the clinic during the first 2 years after treatment initiation were excluded, as the resource use noted in their outpatient files was likely incomplete or, in the case of participation in a clinical trial, not representative of routine care.
Approval for this study was granted by the Human Research Ethics Committee of the University of the Witwatersrand and the Institutional Review Board of Boston University Medical Campus.
Resource use and unit cost
The methods used to analyse the costs and outcomes have been described previously . Patient-level data were collected from patient files. This included the number of patient visits and consultations by type of health worker (doctor, nurse, pharmacist, social worker, counselor, dietician), amount and type of antiretroviral drugs and nonantiretroviral drugs dispensed, and laboratory and other investigations, including radiograms, electrocardiograms, and ultrasounds. We reviewed clinic and hospital accounts for the number and salaries of staff, the quantities and prices of clinic equipment and supplies, and infrastructure, utilities, clinic administration and management, maintenance, and security costs. Drug unit cost data were collected from the nearest government drug depot and laboratory costs from the National Health Laboratory Service. All other unit costs came from financial records or interviews with administrative staff.
Fixed costs, including the cost of staff who do not provide direct patient care, building, equipment and supplies, were summed per year and divided by the total number of patient-months of care provided in that year to generate an average fixed cost per patient-month of care. This amount was then allocated to each study subject per month the subject remained in care. Buildings were valued based on market rental rates for similar structures in the neighbourhoods surrounding the sites; equipment was depreciated according to standard South African accounting practices . Costs for resources not reported in site medical records, such as inpatient care, costs above the level of the treatment facility (e.g. government costs of oversight and training), and costs to the patients themselves were excluded from the analysis.
Resource usage was analysed from the provider perspective. Cost data were from 2009 and were converted to US dollars at a rate of USD1 = ZAR8.28, the average exchange rate for 2009. Data were collected using CSPro version 3.3 and analysed in SAS version 9.1.
Using a paediatric adaptation of a previously published methodology , each individual was assigned to a single outcome on the basis of vital status, attendance status, laboratory results, or the presence of a new or recurring WHO stage 3 or 4 event 12–24 months after initiating ART. The criteria for defining the outcomes, which were mutually exclusive, were assigned using a hierarchical decision process that takes into account the variability and timing of available information (Table 1). To cope with inconsistent timing of visits and laboratory tests, information reported in the patient's medical record within 3 months on either side of the 12–24-month endpoint was used to assign an outcome.
Patients who died or stopped attending the study clinic during the 12–24-month period were classified as ‘no longer in care’. ‘Stopped attending’ was defined as not having returned for a scheduled visit during the 3 months before the end of the 12–24-month period. Among those still in care, any patient having a new or recurrent WHO stage 3 or 4 event at the most recent visit (excluding irreversible conditions) or whose most recent weight was the same as or had dropped below the weight at initiation was considered ‘in care but not responding’. For children who remained in care and did not meet these event or weight criteria, viral load and CD4 cell counts or percentages (depending on age) were considered for those whom these test results were available. Patients whose medical record reported a detectable viral load, defined as more than 400 copies/ml, at 9–15 or 21–27 months after ART initiation were classified as ‘in care but not responding’; those whose viral load was undetectable were classified as ‘in care and responding’. For patients for whom no viral load test was reported, a CD4 percentage or cell count in month 9–15 or 21–27 was used. Patients whose CD4 percentage or cell count showed an increase from baseline were defined as in care and responding. If neither viral load nor CD4 cell count results were reported, but the child remained in care and did not have a current WHO Stage 3 or 4 event and the most recent weight, where available, was above the weight at initiation, a default outcome of in care and responding was assigned.
Cost per patient
For each type of patient outcome and site, the mean and median cost per patient for the first and second years after treatment initiation were calculated. For purposes of the cost analysis, 24-month outcomes were used for patients whose outcomes changed between 12 and 24 months. We evaluated the breakdown of average cost per patient among the main resources used (drugs, lab tests, clinic visits, fixed costs) and considered changes between the first and second years after initiation. Finally, the mean cost to produce a patient in care and responding at 12 months after ART initiation was calculated by dividing all costs for all the patients in the sample by the number of patients in care and responding at 12 months, and a similar calculation was performed for 24 months using 24-month outcomes and total costs.
Characteristics of study samples
Characteristics of study participants at treatment initiation are described in Table 2. After enrolment, 12 children at HSCC were found to have been initiated while admitted for inpatient care and were subsequently excluded from the study, producing a final sample size of 138 at HSCC and 150 at ESRU. Patients at HSCC were on average older at initiation than those at ESRU and had slightly (although not significantly) higher starting CD4 percentages. In part as a consequence of the older cohort, HSCC was more likely to have used efavirenz, rather than lopinavir/ritonavir in its starting antiretroviral regimen. No patient at either site was prescribed second-line regimens containing didanosine during the study period.
The percentage of children with each outcome and the indicators used to assign the outcomes are shown in Table 3. As noted above, outcomes were assigned solely on the basis of information available in existing medical records. For Year 1 at ESRU, for example, 30% of patients were categorized as in care and responding on the basis of an undetectable viral load, but only 9% were categorized as in care but not responding due to a detectable viral load. This simply indicates that only 39% of patients in the sample remained in care, did not have an unacceptable clinical condition, and did have viral load results reported in the relevant time period, following the definitions in Table 1. Patients who were in fact virally suppressed at 12 months but did not have a viral load result reported in their records could not be assigned an outcome on this basis. Except for any who had an insufficient CD4 cell response, these patients would still have been categorized as in care and responding, but their outcome would have been assigned by default (not meeting the criteria for any other outcome), rather than as a result of a viral load result.
Patients in the cohort at HSCC did better throughout the study period, with 91% of patients in year 1 and 80% in year 2 in care and responding to treatment. At ESRU, only 62% of patients were classified as in care and responding at the end of year 1 and 68% at the end of year 2, with a much higher percentage no longer in care than in HSCC (24 and 27% versus 5 and 6% at 12 and 24 months, respectively). Importantly, at ESRU two-third of the children with a 12-month in care but not responding outcome recovered during the second year on treatment. The 12-month in care but not responding outcomes were mostly due to intermittent increases in viral loads and resulting unacceptably low CD4 cell counts/percentages which later subsided without necessitating a switch of regimens, resulting in an in care and responding result at 24 months. This did not occur at HSCC, where several children developed newly detectable viral loads between 12 and 24 months on treatment.
Resource utilization and costs
The top panel of Table 4 describes average resource utilization per year in care for each study site. Across both study sites and years after ART initiation we estimated an average cost per patient-year in care of US$ 693, regardless of outcome. As reported in the lower panel of Table 4, the average cost per patient remaining in care and responding at 24 months was US$ 826 in year 1 and US$ 717 in year 2 at ESRU, and US$ 678 and US$ 782 in years 1 and 2, respectively, at HSCC. At ESRU the second year on treatment was slightly less expensive for all outcome categories, due mainly to fewer visits to all types of staff and a halving of fixed costs as a result of increasing facility scale coinciding with the second year on treatment most of the cohort. In contrast, at HSCC the second year on treatment was more expensive across all outcome categories, largely as a result of a doubling of the staff contingent in 2008, which coincided with the second year on treatment for most of the cohort, to compensate for past understaffing and prepare for an anticipated future increase in patient numbers. The cost of the antiretroviral medications comprised a third to half of the total cost, with higher antiretroviral costs at ESRU where patients were younger at initiation, and thus more likely to be treated with lopinavir/ritonavir than at HSCC. At both sites, patients made substantially more clinic visits in the first year after treatment initiation than in the second, reflecting the need for more frequent monitoring in the period immediately after initiation. By the second year, most patients who are still in care have stabilized and require less frequent clinical monitoring.
The average cost per patient initiated on treatment, as shown in Table 4, was substantially lower than the cost per patient with an in care and responding outcome, due to the large number of patients who did not remain in care for the full 12–24 months. If outcomes improve, the total cost of treatment each hospital's population of paediatric patients will, thus, increase from current expenditure levels. The total cost of producing a patient in care and responding to treatment was US$ 1117 and US$ 829 in years 1 and 2 in ESRU, respectively, and US$ 819 and US$ 937 in HSCC. This measure, which allows us to summarize cost and outcomes of each clinic in a single metric, divides the total cost accrued by each clinic cohort by the number of in care and responding patients produced by each clinic. Because the patients in our sample in ESRU had better outcomes in year 2 than in year 1, the production cost decreased by 26% during the second year, whereas in HSCC it increased by 14%, mostly due to the increase in staff cost that was not offset by changes in patient outcomes.
Comparison with cost of adult antiretroviral treatment
At a public sector, hospital-based adult ART clinic in the same province of South Africa, located just 2 km from one of the paediatric sites, average costs for adult patients remaining in care (responding or not) in 2009 USD were US$ 802 for year 1 and US$ 795 for year 2, or a total of US$ 1597 for the 2-year period (based on , updated to 2009 unit costs) (see Table 4). Paediatric treatment at our study sites was, thus, less expensive than adult treatment at a comparable site.
South Africa's public sector began to offer ART to HIV-infected children in South Africa nearly a decade ago and had some 152 000 children on treatment by mid-2011 , representing fully a third of all paediatric ART patients in low-income and middle-income countries . Despite this, there are still no empirical estimates available of the actual cost of providing paediatric ART at typical treatment delivery sites in the country. In this article, we report the average cost per child initiated on ART and per outcome achieved during the first 2 years after treatment initiation for two large paediatric ART clinics in Gauteng Province, where around 22% of the paediatric ART population is cared for. Our results will assist programme managers, policy makers, and funding agencies to improve the accuracy of their planning and budgeting, as well as helping paediatric clinics understand and improve the quality of care they provide.
We found that 2 years after treatment initiation, 83% of children in the overall study sample remained in care and 76% were classified as in care and responding to treatment. These outcomes are consistent with previous studies’ findings for paediatric cohorts in South Africa. At comparable adult clinics in South Africa, studies using identical methodology have estimated rates of retention in care at 12 months of 71% . The aggregate 12-month retention rate for children in this study, 85%, suggests that paediatric clinics are achieving outcomes comparable to or better than adult clinics.
Across both of our study sites and years after ART initiation, without regard to outcome, we estimated an average cost per patient-year in care of US$ 693. Although costs did vary by site and year, the average over the 24-month period for patients remaining in care, and thus having complete follow-up differed by less than 6%, at US$ 1547 for the 2-year period in ESRU and US$ 1460 in HSCC. The costs we estimated for paediatric treatment were slightly less than the cost of adult treatment at a nearby facility. The only published study of paediatric and adult ART costs, which considered three African countries but not South Africa, reported that paediatric treatment was considerably more expensive than adult treatment in Ethiopia and Nigeria, but cost much less than adult treatment in Uganda . In that study, cost differences between adult and paediatric treatment nearly vanish when antiretroviral drug procurement is excluded, suggesting that antiretroviral drug costs are the most important driver of the total cost of treatment. Our findings are also explained in part by differences in antiretroviral drug costs. In our study, one of the reasons for the modest cost of paediatric ART may be the lower cost of paediatric antiretroviral regimens (except for those containing lopinavir/ritonavir) compared with adult regimens.
Our study had a number of limitations. First, the study included only two sites, both of which were paediatric ART clinics based at large, urban, academic hospitals in Gauteng Province. They may, thus, not be representative of the costs or outcomes to be expected at smaller sites, primary healthcare clinics, or other provinces or settings. Second, findings were based on relatively small sample sizes at each site. Third, results represent average, not marginal, costs and may, therefore, not reflect the cost of further programme expansion. Fourth, inpatient costs were not included, because ART clinic records did not report inpatient care. Previous work by the study team at ESRU suggests that inpatient care may comprise an important share of the total cost of providing HIV care to children , although the relative contributions of inpatient and outpatient care to total cost are unknown and may depend as much on the accessibility of paediatric inpatient facilities as on actual need for inpatient care. Children who transferred to other treatment facilities during the first 24 months after treatment initiation or who initiated ART while admitted for inpatient care were also excluded and may have differed in costs or outcomes from our study sample. Last, and importantly, the patients included in this study initiated ART prior to the adoption of early infant diagnosis and treatment strategies in South Africa. The fact that few very young children or infants were enrolled into this study is a result of this and could potentially alter average cost in the future.
Despite these limitations, this study offers policy makers and funding agencies the first empirical estimates of the actual cost of providing HIV/AIDS treatment to children in South Africa. Costs in the two clinics were comparable both to each other and to those of adult ART. This is of importance for future policy development and counters the assumption that paediatric ART provision will be substantially more expensive than adult ART due to the different drug formulations and higher potential staff and time needs. We did not see evidence of either in our study. The advanced age (4–6 years) and low CD4 percentage (13–14%) at initiation, however, suggests that few very young children accessed ART in either clinic between 2005 and 2008 and that those who did were already very ill. Implementing the new WHO guidelines for early initiation of paediatric ART , already part of South Africa's new National Strategic Plan for HIV , is, thus, a high priority.
Author contributions: G.M.R., L.L. and S.B.R. designed the study. G.M.-R. wrote the first draft of the article. A.B., B.N., K.T., H.M., L.F. and A.C. contributed data and analysis. A.B., L.L. and S.B.R. contributed data interpretation. All authors helped draft and revise the article and read and approved the final text.
The authors thank all patients contributing data to this analysis as well as staff at both clinics and hospitals dedicating time and effort to the data collection, in particular Annie Jordaan, Vincent Kgagkadi, Shobna Sawry and Languta Sibiya. We are grateful to Busisiwe Sithole who provided an outstanding effort in collecting patient-level data.
Funding for this study was provided by the United States Agency for International Development (USAID) under Award 674-A-00–09–00018–00 to Boston University. The funder had no role in study design, data collection or analysis, decision to publish, or preparation of the manuscript. All opinions expressed herein are those of the authors and do not necessarily represent the views of the funder or study sites.
Source of support: US Agency for International Development.
Previous presentation of findings: partial results of this study (preliminary data from one study site) were presented in Abstract 685, 18th Conference on Retroviruses and Opportunistic Infections, Boston, February 27–March 2, 2011.
Conflicts of interest
The authors declare that they have no conflicts of interest.
1. World Health Organization. Global HIV/AIDS response: epidemic update and health sector progress towards universal access: Progress report 2011
. Geneva, Switzerland: WHO; 2011.
2. Department of Health, Republic of South Africa
. National strategic plan for HIV/AIDS, tuberculosis and STIs 2011–2016
. Pretoria, South Africa
: Department of Health: 2011.
3. Ciaranello AL, Chang Y, Margulis AV, Bernstein A, Bassett IV, Losina E. Walensky RP. Effectiveness of paediatric antiretroviral therapy in resource-limited settings: a systematic review and meta-analysis
. Clin Infect Dis
4. Bock P, Boulle A, White C, Osler M, Eley B. Provision of antiretroviral therapy to children within the public sector of South Africa
. Trans Roy Soc Trop Med Hyg
5. Reddi A, Leeper SC, Grobler AC, Geddes R, France KH, Dorse GL, et al. Preliminary outcomes of a paediatric highly active antiretroviral therapy cohort from KwaZulu-Natal, South Africa
. BMC Pediatr
6. Eley BS, Nuttall J, Davis MA, Smith L, Cowburn C, Buys H et al.Initial experience of a public sector antiretroviral treatment programme for HIV-infected children in Cape Town, South Africa
[abstract]. Abstract TuPeB4412, XVth International AIDS Conference
; 11–16 July 2004; Bangkok, Thailand; 2004.
7. Meyers T, Yotebing M, Kuhn L, Moultrie H. Antiretroviral therapy responses among children attending a large public clinic in Soweto, South Africa
. Pediatr Infect Dis J
8. KIDS-ART-LINC Collaboration. Low risk of death, but substantial program attrition, in paediatric HIV treatment cohorts in Sub-Saharan Africa.J Acquir Imm Defic Syndr
9. Boulle A, Bock P, Osler M, Cohen K, Channing L, Hilderbrand K, et al. Antiretroviral therapy and early mortality in South Africa
. Bull WHO
10. Menzies NA, Berruti AA, Berzon R, Filler S, Ferris R, Ellerbrock TV, Blandford JM. The cost of providing comprehensive HIV treatment in PEPFAR-supported programs
11. Department of Health South Africa
. National antiretroviral treatment guidelines 2004.
Pretoria, South Africa
: Department of Health; 2004.
12. Rosen S, Long L, Sanne I. The outcomes and outpatient costs of different models of antiretroviral treatment delivery in South Africa
. Trop Med Int Health
13. Accounting Standards Board South Africa
. Standard of generally recognised accounting practice: property plant and equipment (GRAP 176)
. Pretoria, South Africa
: Accounting Standards Board South Africa
14. Johnson LF. Access to antiretroviral treatment in South Africa, 2004-2011.S Afr J HIV Med
15. Meyer-Rath G, Violari A, Cotton M, Ndibongo B, Brennan A, Long L, et al
. The cost of early vs. deferred paediatric antiretroviral treatment in South Africa: a comparative economic analysis of the first year of the CHER trial [abstract].
Abstract THLBB103. XVIII International AIDS Conference
, 18–23 July, Vienna, 2010.
16. World Health Organization. Antiretroviral therapy for HIV infection in infants and children: towards universal access. Recommendations for a public health approach.
2010 revision. Geneva, Switzerland: World Health Organization; 2010.