Introduction

Despite the availability of effective interventions to prevent mother-to-child transmission (PMTCT) of HIV, each year more than 200,000 infants are newly infected with HIV around the world [1], with more than 20,000 in South Africa alone [2]. Antiretroviral therapy (ART) is a critical intervention for both PMTCT and promoting maternal health. By reducing maternal viraemia, ART initiation in pregnancy leads to significant reductions in the risk of vertical transmission during pregnancy, intrapartum and postpartum [3]. The successful use of ART in pregnancy in Europe and North America has led to discussion of the global elimination of new paediatric HIV infections [4]. Based in part on these successes, there have been calls to consider universal ART initiation in all HIV-infected pregnant women [5]. In April 2013 South Africa shifted policy to use triple-drug ART regimens during pregnancy and breastfeeding for PMTCT. The impact of ART on PMTCT is most profound in women who have higher HIV viral load, lower CD4 counts and are eligible for lifelong ART because these women are at the highest risk of vertical transmission [6].

Initiation of ART early during pregnancy is necessary to achieve maximal benefits [7] and each additional week of ART during gestation is associated with a significant decrease in the risk of MTCT [8, 9]. Despite the clear need to maximize the duration of ART during pregnancy, many women in sub-Saharan Africa start ART too late in pregnancy to achieve maximal benefit [9, 10]. In most settings, pregnant women face numerous delays prior to initiating therapy, including time for laboratory-based CD4 cell count enumeration to determine ART eligibility, transfer of eligible women from antenatal clinics (ANC) to dedicated ART services, and clinical assessment and patient preparation before starting treatment [10, 11].

Despite the importance of rapid initiation of ART in pregnancy to maximize the duration of ART received before delivery, there are few interventions for this. Previously, we reported on the design and preliminary outcomes of the rapid ART in pregnancy (RAP) pilot program in Cape Town, South Africa [12]. The programme’s goal was to reduce MTCT through expedited initiation onto lifelong ART in pregnant women who met South African ART eligibility criteria (CD4 count less than or equal to 350 cells/μL or WHO stage III or IV illness) [13].

Preliminary findings on the RAP programmatic outcomes and acceptability [12] have generated interest in the program, yet there are concerns regarding the costs of the programme, especially in light of reductions in donor and national budgets for HIV/AIDS prevention and treatment [14]. Combined with the growing prevalence of HIV positive and ART-eligible patients, this reduction in funding has forced national and provincial health services to provide more services with fewer resources. It is, therefore, critical that programs have demonstrated cost-effectiveness in order to garner support for expansion.

Methods

Overview

We conducted a cost-effectiveness analysis of the RAP program employing the standard methods of the US panel on cost-effectiveness in health and medicine [15, 16]. The analysis compared detailed costs and outcomes from the RAP program with costs, PMTCT and quality-adjusted life years (QALY) saved outcomes from provincial data and academic literature. Both the RAP model and the standard service model are summarized in Table 1. The primary outcomes for this analysis were the programmatic costs, cost-effectiveness ratio and cost-effectiveness thresholds. Sensitivity analyses determined the stability of the results and threshold analyses established performance standards for RAP and other rapid initiation programs.

Table 1 Rapid initiation of ART in pregnancy (RAP) model and base-case time to ART initiation in pregnant women living with HIV
Full size table

Standard Service Model

Health services for HIV-positive, pregnant women in public health services in South Africa were generally provided in separate antenatal and ART services at the time of RAP implementation. Pregnant women were screened for HIV at their first visit to an ANC and individuals testing positive underwent CD4 enumeration and HIV staging to determine eligibility for lifelong ART. CD4 results were returned to the patient 2–4 weeks after the first ANC visit, and pregnant women with a CD4 count at or below 350 cells/μL or WHO stage III or IV illness were referred for ART [13]. Upon arrival at the ART clinic, women underwent clinical work-up and patient preparation which generally led to an additional 2–4 week delay prior to ART initiation. Overall, women were initiated approximately 4–6 weeks after initial diagnosis in standard services [12, 17].

RAP Service Model

The RAP model has been described in detail previously [12]. Briefly, the RAP program was a package of services designed to reduce the delays from HIV diagnosis to ART initiation in pregnancy through integrated services and expedited initiation. In this model, pregnant women were screened for HIV at their first ANC visit, per the standard of care. All women who tested HIV-positive had blood drawn for CD4 count enumeration and were referred to an on-site, mobile, once-weekly RAP unit on Friday of the same week. CD4 results were confirmed or determined, depending on whether National Health Laboratory Service results had been returned by the Friday visit-using a rapid, point-of-care, Alere Pima™ CD4 test. Women who were ART-eligible were counseled and initiated onto ART the same day with safety blood results and adherence reviewed the following week. Women attended RAP services through 6–8 weeks post-partum before transfer to standard ART services. Thus, ART-eligible women in RAP services commenced lifelong ART approximately 4–5 weeks earlier than women in standard services through the package of immediate referral to on-site ART services, active checking of CD4 results, and counseling and safety blood work concurrent with initiation [12]. Within RAP, 97 % of ART-eligible participants started ART, 91 % of whom initiated on the same day that their treatment eligibility was determined. Participants had no adverse events that required medical intervention [18] and the model was acceptable to both participants and clinic staff [12].

Cost Parameters

Personnel, facility, treatment, laboratory and equipment costs were obtained from programmatic documents using a micro-costs approach. RAP and standard service costs were calculated using local data from RAP programmatic records and the Provincial Government of the Western Cape. These costs are shown in Table 2. RAP costs were then converted into 2011 US dollars using the purchasing power parity of R5.31: US$1 [19]. The South African rand- US dollar exchange rate has changed considerably since the time of program implementation, but all costs are provided in 2011 dollars as this was when the program was implemented. Costs are based on provision of services to 190 women because RAP initiated this number of women onto ART during the analysis time frame. RAP program costs included one day a week employment of 13 medical, counseling, administrative and management personnel, along with RAP programmatic evaluation expenses. Additional costs included CD4 test cartridges, vehicle petrol, and medical costs such as ART medication and laboratory tests. Programmatic costs were actively monitored during program implementation and maintained in programmatic records.

Table 2 Annual costs associated with RAP services and standard services (190 women served from February 2011–February 2012)
Full size table

Standard services included the same laboratory tests and treatment costs, but did not include RAP staff and programmatic evaluation costs. They also included one less month of ART and one fewer viral load test. This assumes that all women initiated by RAP would have eventually been initiated onto ART at an ART clinic under the standard service model, but this would have occurred 1 month later, as shown by Table 1. The costs for ART services and laboratory testing were obtained from the South African national ARV drug price tender from September 2011. Costs of standard ANC and PMTCT services were not included because it is anticipated that these costs would be the same for both RAP and standard services and, therefore, would be removed in incremental analysis. Additionally, costs for existing resources such as the RAP vehicle and the point-of-care CD4 Pima machine were not included in the base case calculation of RAP costs because they were pre-existing resources and this analysis was of the additional costs associated with implementation of the RAP package of interventions for the RAP cohort. We assumed that RAP services were provided on the margin of existing infrastructure. The costs of existing resources were, however, included in a sensitivity analysis to determine the influence of these costs on programmatic cost-effectiveness.

Perinatal Infections Averted

The number of perinatal infections averted was calculated through comparison of a conservative estimate of the perinatal infection rate in standard services with a conservative estimate of the infection rate documented in RAP services. The perinatal transmission rate of ART-eligible women in a similar Cape Town setting was 5.1 % [9] and in a dual ART and antenatal care clinic in Johannesburg the transmission rate was 5 % [20]. A very conservative estimate of the perinatal transmission rate in ART-eligible women in Cape Town is 4 %. Thus, of 190 pregnancies, we expected 7.6 perinatal infections (4.0 %) with standard services.

This transmission rate was compared to the transmission rate documented by RAP services of 0.9 % (1 positive of 107 infants with polymerase chain reaction HIV test results at time of analysis [18]). A more conservative estimate of the perinatal infection rate with RAP services (2.0 %) was used instead because transmission may have been higher in infants without HIV results. Thus, of 190 pregnancies, we expected 3.8 perinatal infections (2.0 %) with RAP services. Therefore, RAP services were associated with 3.8 (7.6–3.8) averted perinatal HIV infections. A range of values were explored in sensitivity analysis to determine the influence of different perinatal transmission rates on programmatic cost-effectiveness.

QALY Saved

For the base case analysis, QALY saved were determined based on the calculation of 16.88 discounted QALY saved per perinatal infection prevented with an infant on ART in South Africa calculated by Soorapanth et al. [21]. QALY saved were discounted at a 3 % rate which is a frequently used rate. While it is anticipated that both standard PMTCT services and RAP had benefits for maternal outcomes and QALYs, these benefits were not included in this analysis for either service. The QALY saved were varied widely in sensitivity analyses.

Treatment Costs for Perinatal Infection

Treatment costs for each perinatal infection were adapted from the discounted lifetime cost for an HIV-infected child with ART in South Africa, determined by Soorapanth et al. [21]. Costs were discounted at a 3 % rate. Soorapanth et al. [21] presented the $10,969 cost in 2003 US dollars. These dollars were adjusted to 2011 dollars using the whole consumer price index for All item Urban Consumers (CPI-U) 1982-84 = 100 (Unadjusted) [22]. The 2011 discounted lifetime cost for an HIV-infected child in South Africa was $13,410 ($10,969*$224.939/$184.0) and comprised the base-case analysis. This cost is very similar to the annual cost per pediatric patient in the first 2 years of care in South Africa, as described by Meyer-Rath et al. [23]. Soorapanth’s et al. [21] low and high estimates of treatment costs ($2,000 and $20,000, respectively), were also adjusted to 2011 dollars and included in sensitivity analyses.

Cost-Effectiveness Analysis

A retrospective, cost-effectiveness analysis was conducted on the RAP program from the payer perspective. This perspective was selected to evaluate the feasibility of future expansion of RAP services for the South African Department of Health. The time frame for analysis was the year that RAP was implemented from February 2011–February 2012, during which time 190 women were initiated by RAP onto ART.

The cost-effectiveness of RAP, compared to standard services, was determined using the formula, incremental cost-effectiveness ratio (ICER) = \(\frac{{{\text{C}} - {\text{AT}}}}{\text{AQ}}\). C was the additional cost of RAP services, calculated by subtracting the standard service cost from the total cost of RAP services. A was the number of additional perinatal infections averted in RAP services, calculated by subtracting the number of perinatal infection in RAP services from the number in standard services. T was the treatment cost for a perinatal infection and Q was the QALY saved by averting one perinatal HIV transmission.

The World Health Organization (WHO) has developed recommendations on the use of per capita gross domestic product (GDP) to determine locally appropriate, maximum willingness to pay per QALY saved. Based on this definition, an intervention is cost-effective if its incremental cost-effectiveness ratio, compared to an alternative, is less than three times the per capita GDP per QALY, and it is highly cost-effective if it is less than the annual per capita GDP per QALY [24]. The estimated 2011 per capita GDP for South Africa was $11,100 [25]. Thus, interventions that cost less than $11,100 per QALY saved in South Africa were considered highly cost-effective and those less than $33,300 per QALY saved were considered cost-effective. It is, however, also important to consider the cost per QALY saved, relative to other similar interventions.

Results

One year of RAP services cost $880 per woman ($166,560 for 190 women) and 1 year of standard services cost $220 per woman ($41,430 for 190 women), as shown in Table 2. In the base-case, the RAP model cost $659 more per woman served ($125,130 for 190 women) than standard services and averted a total of 3.8 perinatal HIV infections (Table 3). The overall base case, cost-effectiveness ratio was $1,160 per QALY saved, indicating that the RAP program was highly cost-effective by the WHO GDP-based standards for South Africa [22].

Table 3 Base case parameters and cost-effectiveness ratio
Full size table

Sensitivity Analyses

A range of sensitivity analyses were conducted in which all parameters were varied through plausible ranges to determine the stability of the cost-effectiveness results. These analyses are summarized in Table 4. The effect of each of the parameters in the model was evaluated. This included decreased infections averted, increased RAP costs, and reduced QALY saved. The effect of Soorapanth et al. [21] high and low estimates of perinatal infection treatment costs were also evaluated. The cost per a QALY saved remained below the three times the GDP threshold ($33,300) in all of the analyses.

Table 4 Sensitivity analyses for parameter assumptions
Full size table

Threshold Analysis

Thresholds for each of the parameters included in the model were then determined for the RAP model. This was done to set performance standards for this and future programs to expedite ART in pregnancy. The cost-effectiveness threshold was determined by calculating the value for each parameter at which $11,100 = \(\frac{{{\text{C}} - {\text{AT}}}}{\text{AQ}}\) since this was the definition of highly cost-effective used in this analysis for South Africa. Cost-saving threshold analyses were also done for the parameters in the numerator of the cost-effectiveness ratio (program costs, infections averted, and treatment costs per perinatal infection). The cost-saving threshold is the value at which the program savings outweigh the program costs ($0 = C−AT).

The results of these calculations are shown in Table 5. The RAP program is highly cost-effective if it reduces MTCT by at least 0.33 % more than standard services; if at least 1.76 QALY are saved per perinatal infection; if RAP costs do not exceed standard services by more than $4,020 per woman; and if perinatal infection treatment costs are much less expensive- actually cost the health system more than not treating infants. The RAP program is cost-saving if it reduces MTCT by at least 4.91 %; if RAP costs do not exceed standard services by more than $270 per woman; and if each perinatal infection treatment costs at least $32,930.

Table 5 Cost-effectiveness and cost-saving threshold analysis for parameters
Full size table

Discussion

This cost-effectiveness analysis found that the package of RAP services for expedited ART initiation during pregnancy was highly cost-effective, compared to standard PMTCT services in Cape Town. At $1,160 per QALY saved, the cost-effectiveness of RAP services is comparable to other biomedical HIV prevention interventions in South Africa such as achieving 75 % ART utilization in treatment eligible individuals ($1,149 per QALY saved), provision of oral pre-exposure prophylaxis to half of all infected persons ($9,009), annual HIV screening and ART utilization by 75 % of infected persons ($1,033/QALY), and medical male circumcision of 75 % of men within 5 years (cost-saving) [26]. Threshold analyses demonstrated that the intervention required minimal reductions in perinatal infections averted beyond standard care (a reduction in perinatal infection of 0.3 %) in order to be highly cost-effective, but a relatively high number of perinatal infections averted beyond standard care (a reduction in perinatal infection of 4.9 %) in order to be cost-saving. Given the relatively low levels of vertical HIV transmission in Cape Town, it is unlikely that a 4.9 % reduction is achievable. A reduction in perinatal transmission by 0.3 % is, however, a highly achievable service standard for reaching cost-effectiveness and likely underestimates the impact of the RAP program.

In mid-2013, South Africa shifted PMTCT policy to provide triple-drug antiretroviral regimens to all HIV-infected pregnant women during pregnancy and breastfeeding, continuing as lifelong ART for women with advanced HIV disease. Under this new approach, some aspects of the RAP intervention, in particular rapid CD4 cell counts, become less important components of effective PMTCT services. However this study is critical in demonstrating that, given the high costs associated with perinatal HIV infection and subsequent pediatric HIV care and treatment, interventions that can expedite ART initiation are likely to be highly cost-effective. Here, RAP service costs were low as compared to the QALY saved, despite the program’s fairly high staff and evaluation costs. Many of these costs were fixed which means that the cost per woman would decrease as service usage increases. Additionally, such costs would be reduced by integration of ART into routine ANC services since the RAP service and formal evaluation costs would be removed. ANC staff members would require training on provision of ART and the assistance of additional peer counselors, but it is anticipated that this could be done for less than $270 per women served, suggesting that integration may be cost-saving. Furthermore, the RAP model remains highly cost-effective at a cost of $4,020 per woman served above standard services, which is likely to be a very achievable cost threshold. These findings, along with the recent changes in South African PMTCT policy, underscore the importance of evaluating the costs and staff capacity for ART initiation in antenatal care in order to determine whether this threshold has been met.

This analysis has three primary limitations. First, data for several key parameters were not available, and in these cases historical estimates were used instead. These include perinatal transmission rates among ART eligible women and estimates of life expectancy and lifetime costs among perinatally infected children. While national perinatal transmission rates are available, these rates underestimate transmission in ART-eligible women, who are likely to have higher HIV viral loads. There have also been substantial advancements in ART medication and access in the past 9 years since the publication of the cost and QALY parameters used in this analysis [21]. Additionally, documentation of RAP’s effectiveness was limited by missing PCR results and we were unable to determine the effect of each additional week on treatment. Although our study is necessarily limited by the quality and appropriateness of the parameters used, we conducted extensive sensitivity analyses and found the policy conclusions to be unchanged at all plausible values for these parameters. An important area for future research is rigorous comparison of the effectiveness of expedited ART during pregnancy to standard services and of the contribution of each additional week on ART towards the cost-effectiveness of rapid initiation. Second, interpretation of the cost-effectiveness ratio is complicated because there is little consensus as to society’s willingness to pay in the developing world. Nonetheless, we used the generally accepted standard based on per-capita GDP [24]. Third, this analysis considers only averted perinatal infections and infant QALY saved. Earlier initiation onto ART will, however, have additional benefits, such as reduced heterosexual transmission to partners [27] and improved maternal health outcomes, resulting in additional QALY saved. Inclusion of maternal benefits is beyond the scope of this analysis, but it is expected that these benefits would increase the demonstrated cost-effectiveness of RAP services.

There is a clear need for interventions that are both effective and cost-effective in order to reduce vertical HIV transmission. Our analysis builds on a growing literature on the cost-effectiveness of the provision of PMTCT services in the African context [28–34]. We also extend this previous work by evaluating a package of interventions, addressing timely questions about the integration of multiple services (ART and PMTCT) within antenatal care. The findings from this analysis highlight the considerable potential of South Africa’s new program to rapidly initiate pregnant women onto ART in ANC. Rapid initiation onto ART services in pregnancy has the potential to help eliminate MTCT through increased duration on ART during pregnancy and reduced maternal HIV viraemia in women at the highest risk of transmission [12]. This study demonstrates that such rapid initiation programs are also highly cost-effective and have the potential to be cost-saving when ART services are integrated into antenatal care.