Costing RTS,S introduction in Burkina Faso, Ghana, Kenya, Senegal, Tanzania, and Uganda: A generalizable approach drawing on publicly available data

Recent results from the phase 3 trial of RTS,S/AS01 malaria vaccine show that the vaccine induced partial protection against clinical malaria in infants and children; given the high burden of the disease it is currently considered for use in malaria endemic countries. To inform adoption decisions the paper proposes a generalizable methodology to estimate the cost of vaccine introduction using routinely collected and publicly available data from the cMYP, UNICEF, and WHO-CHOICE. Costing is carried out around a set of generic activities, assumptions, and inputs for delivery of immunization services adapted to a given country and deployment modality to capture among other factors the structure of the EPI program, distribution model, geography, and demographics particular to the setting. The methodology is applied to estimate the cost of RTS,S introduction in Burkina Faso, Ghana, Kenya, Senegal, Tanzania, and Uganda. At an assumed vaccine price of $5 per dose and given our assumptions on coverage and deployment strategy, we estimate total economic program costs for a 6–9 months cohort within $23.11–$28.28 per fully vaccinated child across the 6 countries. Net of procurement, costs at country level are substantial; for instance in Tanzania these could add as much as $4.2 million per year or an additional $2.4 per infant depending on the level of spare capacity in the system. Differences in cost of vaccine introduction across countries are primarily driven by differences in cost of labour. Overall estimates generated with the methodology result in costs within the ranges reported for other new vaccines introduced in SSA and capture multiple sources of heterogeneity in costs across countries. Further validation with data from field trials will support use of the methodology while also serving as a validation for cMYP and WHO-CHOICE as resources for costing health interventions in the region.


RTS,S vaccine against
Plasmodium falciparum malaria has demonstrated moderate levels of efficacy in phase 3 trials in Africa and is currently considered for use within the Expanded Programme on Immunization (EPI) in endemic countries [1]. Tested in children (5-17 months) and young infants (6-12 weeks) the vaccine has shown high initial efficacy, but its protection waned quickly with efficacy against clinical disease at 36.3% and 25.9% depending on age at immunization [2,3]. The vaccine is thus evaluated as an additional tool for preventing clinical disease in children, not a replacement for existing malaria preventive, diagnostic, and treatment measures [4]. Despite being partially effective, modelling studies predict RTS,S to have a substantial public health impact on disease burden [5].
With a positive scientific opinion on vaccine efficacy and safety issued by the European Medicines Agency earlier this year, the WHO is expected to follow-up with a policy recommendation on the use of RTS,S [4]. As countries, donors, and international organizations consider RTS,S introduction, data on program costs are needed. Combined with effectiveness, cost data allow policy makers to assess the value of this new intervention in the context of a malaria control strategy [6,7]. Decision-making at the country level is further concerned with financing and feasibility of mobilizing and maintaining the level of resources to support the new vaccine [8,9].
To inform these decisions the study proposes a generalized methodology to estimate costs of RTS,S introduction in the EPI program. We apply it to assess these costs in Burkina Faso, Ghana, Kenya, Senegal, Tanzania, and Uganda. Costing is implemented around a set of generic and easily modifiable assumptions describing vaccine introduction; these capture among other factors the structure of the EPI program, distribution model, geography, and demographics particular to the setting. Our findings illustrate the broad ranges for costs of introduction dependent on the level of spare capacity to accommodate the vaccine. We show that differences in EPI structure, cost of labour, and intensity of use of resources within the system translate into significant differentials in cost of vaccine delivery between settings.

Vaccine presentation and deployment strategy
We base our assumptions about vaccine presentation on [10,11]. RTS,S is a monovalent lyophilized vaccine reconstituted with an adjuvant; both require cold chain storage at 2-8 • C. The vaccine and the diluent are clipped together with a packed volume of 9.68 cm 3 per two-vial package. Two doses per vial are yielded after reconstitution.
Clinical trial data suggest that at least three doses are required for protection against malaria. Following the trial design [11], we evaluate four immunization schedules: a program targeting infants at 6, 10 and 12 weeks of age (6-12 weeks); children at 6, 7.5 and 9 months (6-9 months); and a four dose schedule in the two age groups including vaccination at 18 months after the 3rd dose. We assume national roll-out to scale through routine outlets with immunization schedule tied to DTP. For 6-9 months implementation we assume first and third doses to be administered along with vitamin A and measles vaccine and treat these as routine immunization visits for purposes of costing. New, out-of-routine schedule, visits are assumed for the second dose in 6-9 months and fourth doses.

Perspective
A broad provider perspective is adopted in this evaluation; all resources required to introduce the RTS,S into the national program are included in the analysis.

Scope
We estimate both the economic and the financial costs of introducing the malaria vaccine into the EPI. Financial costs represent actual expenditures on goods and services. Economic costs "define costs in terms of the alternative uses that have been forgone by using a resource in a particular way"; these include, in addition to the financial costs, a valuation of resources that do not have financial transactions (i.e. donated goods and services or capital goods, health care resources diverted from other uses or shared with other health programs, and inputs whose prices are distorted [12]). Given paucity of data on the level of existing capacity in the system, the financial costs are estimated under an assumption of 100% spare capacity to accommodate the vaccine. In reality, however, some countries might need to invest in scaling-up across a range of service inputs be it cold chain, vehicle fleet, or labour to deliver the new intervention. Assumptions for capacity scale-up, while not used to produce estimates here, are shared in Supplementary File S1. The economic costs are implicitly evaluated under the assumption of no spare capacity. Taken together, the two sets of estimates give a wide but informative range for potential vaccine introduction costs.

Assumptions for delivery of immunization services
We defined a set of essential activities for RTS,S introduction based on WHO [7,13,14], USAID [15,16], and other guidelines on immunization [17,18]. Assumptions about operational aspects of the program were further informed by published micro-costing studies that evaluated the introduction of new vaccines in lowincome countries [19][20][21][22][23]. No campaigns or additional outreach activities outside of the routine EPI delivery were considered. For deployment modalities requiring out-of-routine schedule visits we assumed lower coverage rates and adjusted service delivery assumptions to reflect the longer time needed to administer the vaccine, increased IEC, and supervision to maintain coverage. Assumptions on activities costed and key inputs are presented in detail in Table 1. Scenarios by schedule are highlighted in Supplementary Table S1.

Resource lists
Resource lists were populated following the activities defined for vaccine introduction. These were adapted to countries using comprehensive Multi-Year Plans (cMYP) for immunization [24][25][26][27][28][29]. The latter were particularly useful in identifying staff categories, equipment, and quantities of resources used by the EPI program at each level. Resource lists and prices for each country are documented in Supplementary File S2.

Input prices and unit costs
Data on input prices and unit costs came from several sources. Information on wages by level of EPI staff, per-diems, and some other line items were taken from cMYP [24][25][26][27][28][29]. We also used data from the UNICEF [30] for prices of immunization supplies, and related equipment. Additionally data from the WHO-CHOICE databases were used to cost facility rental, hotel rates, fuel, and other commodities [31]. Prices of commodities obtained from the international price lists were adjusted for freight, insurance and wastage [13].
Shared inputs were attributed to RTS,S based on the direct allocation [12]; except for vaccinators whose contribution to the new intervention was costed based on time required to administer the vaccine. Similarly, cost of cold chain and vehicles were allocated to RTS,S based on use; for these inputs use refers to the volume of the vaccine and immunization supplies and time for storage or distance over which these were transported or stored.
Cost of capital items including vehicles, facilities, and equipment was annualized over the respective estimated useful life. Expenditures associated with activities held in the introductory stage were considered capital goods and were annualized and discounted over 5 years at 3%.

Algorithm for calculating cost of immunization
For each activity outlined in Table 1 we defined formulas to combine price and unit cost data with assumptions on resource use. These start with the general representation of an activity in terms of cost components, break it down to micro inputs and detail how unit costs are combined with quantities to obtain total cost per activity. Formulas are presented in Supplementary File S3; these are generalized and could be adapted to alternate assumptions on service delivery. Costs are estimated for a single cohort of surviving infants and are reported in terms average annual, per FVC, and per dose administered metrics.

Sensitivity analysis
The most critical assumptions made when estimating cost of vaccine delivery relate to coverage, wastage rates, and use of labour at service point. These parameters were varied over an inclusive range while keeping all other inputs at base values. Assumptions on wastage, discount rate, and time to administer the vaccine were generic; country data were used in the baseline for all other inputs varied. Resulting estimates of cost of vaccine delivery are summarized by country in tornado plots.

Overview of key demographic, coverage, and EPI inputs by country
Differences in country cohorts and the EPI system detailed in Table 2 help interpret the level and variation in costs across countries. The cohort size of Tanzania of 1.7 million infants is about 4 times as large as that of Senegal. There is variation not only in the level of coverage achieved but also in the level of coverage sustained between the doses. In Uganda coverage at third dose is 8 percentage points lower than the first dose; the drop-off is about 3 percentage points in Ghana and Senegal. There is variation in wages at facility level: nearly $800 per month are reported for Kenya, $421 and $115 per month for Tanzania and Burkina Faso, respectively. Interestingly the ordering of countries changes when we look at wages at higher levels of EPI; both at central and district levels highest wages are reported for officers in Ghana. Finally, there are differences in the number of days and number of EPI staff conducting supervision visits across countries; these vary from 0.5 to about 10 days per month with as few as 4 to as many as 9 officers per district involved in supervisory capacity.

Cost of RTS,S immunization
Unless noted, costs are presented for the 6-9 months schedule assuming a vaccine price per dose of $5; estimates for other implementation strategies are reported in Supplementary Tables S2-9.

Differences in cost of RTS,S immunization by country
Country differences in cost of vaccine delivery are a function of program design, coverage, EPI structure, number of antigens in the EPI schedule, and prices, the most critical of which is cost of labour. The lowest economic cost of delivery is estimated for Burkina Faso at $0.72 and highest for Kenya -at about $2.34 per dose administered ( Fig. 1 A). Over a three-fold difference between the countries is largely driven by wage differentials: compared to Burkina Faso, wages of EPI officers in Kenya are nearly five times higher; the differentials persist across all distribution levels although at lower levels wage differences are smaller ( Table 2). In addition, compared to other countries, Kenya has one of the lowest projected coverage rates resulting in a lower denominator and, consequently, a lower base over which the fixed costs, including introduction investment, are allocated. When summarized in terms of economic cost per FVC, variation between countries is much smaller: the lowest estimate is for Burkina Faso at $23.11 and the highest -for Kenya at $28.28 (Fig. 1 B). Convergence in costs at this level is due to differences in coverage between countries; cost per FVC increases steeply with drop-off between doses.

Differences in cost of RTS,S immunization by schedule
Differences in cost by schedule are similar across the 6 countries; these are illustrated in Fig. 2 for Tanzania. For 6-12 weeks and 6-12 weeks 4 dose implementations, the cost of delivery is lowest; higher costs are estimated for 6-9 months modalities, although at this level differences between strategies are modest. When summarized as cost per FVC, program costs for modalities including doses outside of the routine schedule are significantly higher. The much higher cost per FVC for boosting schedules is again due to coverage assumptions, namely we assumed 80% of 3rd dose coverage for 4th dose, resulting in a lower denominator for these strategies.

Cost drivers of RTS,S immunization
Resource requirements for each program component as a proportion of average annual delivery costs are illustrated in Fig. 3. Costs at the facility level associated with the immunization visit account for the largest proportion of total delivery costs. The relative weight of other inputs varies across settings with differences across input categories mainly driven by differences in the structure of EPI program (levels of cold storage, number of staff at each unit, etc.), resource use, and wages. For instance, in Kenya labour heavy vaccination activities account for almost 60% of total delivery costs compared to only about 35% in Burkina Faso where, as discussed, reported wages are significantly lower. Activities such as supervision, monitoring and introduction incorporate heterogeneities in wage structure across countries as well as levels of resource use. The latter is illustrated with supervision activities; in Tanzania supervision accounts for nearly 16% of total program costs based on reported 4-14 days per month devoted to the activity across EPI levels; in contrast, in Kenya and Burkina Faso an average of only 3 days per months are allocated to supervision.

Sensitivity analysis
Results of the sensitivity analysis show that cost of service delivery changes almost proportionally with immunization coverage (Fig. 4). These are sensitive to assumptions on time to administer the vaccine both within and outside of the EPI schedule; if instead of 5 min to administer an additional vaccine dose [19,32] 15 min are required, to allow, for instance, for information and incentivisation, service costs would increase by as much as 25%. Use of EPI shared resources including cold chain, vehicles, and program management are among other influential parameters.

Discussion
Costing an intervention that has not been implemented is a difficult undertaking further complicated by the absence of country   policy with respect to its use. This study presented an approach to estimate the prospective costs of RTS,S introduction that can be easily adapted to the specifics of the vaccine presentation, country EPI program, and implementation strategy. One of the main advantages of the methodology is that it can be implemented using routinely collected and publicly available. Estimates presented should be interpreted in the context of assumptions made on vaccine price, coverage, delivery modality, and data constraints. While we made every attempt to adapt these to country settings, lack of detail on operational aspects of the program might have resulted in a more normative distribution model. Aside from concerns about relevance of activities costed for a given setting, more general concerns about the quality of cMYP reports, borrowing of data and extrapolation across countries to fill in the data gaps introduced additional uncertainty.
One previous study assessed prospective costs of RTS,S introduction via routine EPI in Tanzania [19]; these were estimated using a similar methodology and relying on local data including system capacity use. The incremental cost of vaccine introduction was estimated at $0.66 per dose, which is lower than $0.93 per dose estimated in our study for Tanzania, but comparable given differences in scope. The contribution of the main cost drivers to the vaccine delivery costs were of the same order of magnitude in the two studies. Consistent with these earlier analyses, delivery costs account only for a small proportion of total program costs. At a vaccine price of $5 per dose about 95% of financial and 84% of economic costs are accounted for by vaccines and immunization supplies.
Although comparability with other antigens is limited given differences in vaccine properties, deployment strategies, immunization rate, etc., costs estimated here for RTS,S delivery are similar to other vaccines recently introduced in the region. De la Hoz-Restrepo [33] Fig. 4. One-way sensitivity analysis on average annual economic cost of RTS,S immunization deployed via 6-9 months schedule in Burkina Faso, Ghana, Kenya, Senegal, Tanzania, and Uganda: Service delivery (USD, 2013). "Vaccine wastage" low value of 5% and a high of 25% are tested. "Coverage" under low scenario is assumed to be 75% of baseline value, at high scenario it is taken to be 125% of baseline. "EPI allocation" and "RTS,S allocation" of shared inputs are scaled down to 75% of baseline value and up to 125% under high scenario. "Time per dose EPI" is varied from 5 min at baseline to 3 and 15 min under low and high scenarios respectively. "Time per dose outside EPI" is varied from 7 min at baseline to 5 and 15 min under low and high scenarios respectively. "Days outreach" under low scenario are set to 2 and to 15 under high scenario. "Discount rate" values of 0% and 10% are evaluated. the operational strategy for RTS,S introduction estimates could be revised accordingly.
Costs estimated by the study are of interest to a diverse set of stakeholders. At country level these provide baseline values of resources required to introduce and maintain the intervention. Information particularly vital for Gavi graduating countries that will assess not only the additional resource needs for the new antigen but also the increased cost-sharing for provision of immunization for the current schedule and sustainability of the new schedule in the longer term. We show that delivery costs at country level are substantial; for instance, in Tanzania these could range between $0.8 and $4.1 million per year depending on scope of costing and level of spare capacity in the system.
For donors and global institutions supporting immunization programs the analysis presents some initial estimates of resource needs for vaccine introduction; moreover it highlights the extent of resource use beyond procurement supplied by countries when introducing the intervention. Donors, in particular Gavi, might be interested in the cost of particular program components like vaccine introduction investment to gauge the size of introduction grants to support RTS,S in endemic countries. We show that the costs are higher in particular if vaccine is delivered outside of routine schedule and if new visits would be required. For the 6 countries, we estimate introduction costs between $0.76 and $1.72 per infant in the surviving cohort; these are comparable to levels of support currently awarded by Gavi for new vaccine introduction: $0.80 per child for vaccines delivered to infants and $2.40 per girl for HPV vaccine [35].

Conflict of interest
This work was supported with a research grant from PATH -Malaria Vaccine Initiative -a partner contributing funding, through the Bill and Melinda Gates Foundation, to RTS,S development; KG was supported with this grant. No funding bodies had any role in the study design, data analysis, decision to publish, or preparation of the manuscript. Other contributors have no conflicts of interest to report. at WHO. Farzana Muhib, Deborah Atherly, Clint Pecenka, and Carla Botting at PATH. John Koku Awoonor with the Ghana Health Services for helping validate assumptions on vaccine service delivery.

Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.vaccine.2015.10. 079.