In recent years, pricing of medicines and health commodities has become more transparent and readily available through resources including the International Drug Price Indicator Guide and the Global Fund Price and Quality Reporting mechanism [1,2]. Although this transparency has allowed for more accurate cost estimates, supply chain costs for delivering specific products—are still poorly understood and undefined [3,4]. Previous studies demonstrated that supply chain costs account for a significant proportion of total programme costs [5-7]. Surveys have found that when measured as a percentage of the total acquisition cost, supply chain costs ranged from 1% for contraceptives in Bangladesh to 13% for essential health commodities in Ghana [8], 18% for ACT in Malawi and up to 44% for large-volume bed nets in Liberia [9]. Such “ballpark” estimates frequently determine the budget presented to donors [10] and are often insufficient for delivering the products to health facilities. Author analysis of Global Fund proposals during Rounds 8–10 found that between 3 and 14% of total procurement costs are earmarked for procurement and supply chain management (PSM) functions in Africa [11]. Given that donors are placing a greater emphasis on value for money and that commodities consist about a quarter of all health care costs [12], using accurate supply chain cost estimates is key to improving access to medicines and health commodities [13].

Estimating the supply chain costs of a public health system is complex for several reasons. Most public health supply chains, particularly those in low- and middle-income countries, do not systematically track supply chain costs and rarely keep records on expenditure by each supply chain function. Additionally, multiple stakeholders including donors, government agencies, and international organizations are responsible for different aspects of the supply chain. Finally, the supply chain costs are often integrated within other costs and capturing the costs from each stakeholder, particularly indirect costs, is an onerous task. Anti-malarial medicines are often distributed using a combination of malaria-specific supply chains and integrated systems, which further complicates estimates. Supply chain costs in each country also depend on the typology of the system, number of levels in the distribution system, and the types of fixed and variable costs inherent in the structure [14].

In most malaria-endemic countries, access to ACT and RDTs for malaria diagnosis and treatment remains well below targets despite substantial increases in international disbursements to malaria-endemic countries between 2002 and 2012. These increases have slowed in recent years [15,16] and the Global Fund’s new funding model is placing a greater emphasis on value for money [17], making better cost estimates even more important.

The objective of this study was to estimate the supply chain costs for ACT and RDTs from the central to the peripheral levels of the public sector health systems in Benin and Kenya. The two countries were selected to allow for east–west geographical diversity in Africa and to allow for the anomalies in supply chain design in Anglophone and Francophone countries. The countries are also ranked differently on the World Bank logistics performance index (LPI), a composite estimate of road network maturity. Benin has an LPI of 2.85, places 83 out of 155 countries surveyed and is the highest ranked low-income country. Kenya has an LPI of 2.43 and ranks 122 out of the 155 countries. In addition, the SIAPS project presence in both countries facilitated the logistics associated with data collection and follow-up.

Benin

Warehousing and distribution is carried out by La Centrale d’Achat des Medicaments Essentiels et des Consommables Médicaux (CAME), an independent, not-for-profit organization (Figure 1). Procurement of ACT and RDTs is done through international procurement agents and coordinated by the Programme National de Lutte Contre le Paludisme (PNLP). ACT and RDTs are stored in a dedicated malaria warehouse and are collected by zonal depots or sent to one of two regional depots in Parakou and Natitingou using CAME trucks. Regional stores pick up their stock from the CAME depots and health centres and hospitals collect stock from the regional stores. The national hospital, the Centre National Hospitalier Universitaire (CNHU), and departmental hospitals supply themselves directly from CAME Cotonou. Product quality control is carried out by the Laboratoire National de Contrôle de Qualité (LNCQ) upon arrival in-country and the costs for this are borne by the public sector.

Kenya

The Kenya Medical Supplies Agency (KEMSA), a parastatal organization, manages procurement, warehousing, and distribution of medicines and health products in Kenya. It also stores and distributes products procured by other donors directly to health facilities. KEMSA has two warehouses in Nairobi and since 2013, the county governments procure commodities directly either from KEMSA or other suppliers (Figure 2). At the time of this study, however KEMSA was the central procurement agency distributing ACT and RDTs to hospitals, health centres, and dispensaries using a privately contracted transport provider. The costs thus presented in this paper represent this structure. The only quality control costs incurred by the government is post-market surveillance for ACT carried out by the national Pharmacy and Poisons Board (PPB) in collaboration with the National Quality Control Laboratory (NQCL) and the Department of Malaria Control (DoMC).

This study identified the structural and operational supply chain costs associated with distribution of ACT and RDTs in Benin and Kenya. The cost of procurement, distribution, and product quality control was estimated using an output metric of cost per USD value of an ACT and RDT unit distributed. Other studies have used a variety of cost metrics: costs per USD value [20-22]; costs per kilogram [22], and costs per kilometre [20-22]. These studies estimated costs to be USD 0.63–1.15 for vaccines in Mozambique [21], USD 0.24–2.00 for general products in Nigeria [20]; and 7.6–24% for antiretrovirals, essential medicines and vaccines in Zambia, Uganda and Niger, respectively [23-25].

The findings indicate that supply chain costs play a significant role in the total cost of delivering ACT and RDTs. A few cost drivers stand out in both Benin and Kenya - labour, transport, and utilities. At the level of storage, labour and utilities stand-out accounting for up to 71% of the total distribution costs of ACT in Kenya.

It should be noted that labour cost estimates for ACT and RDT stock-management were based on self-reported hours during interviews with heath workers, a process likely to be associated with reporting bias. However, despite this, these findings are consistent with those from other studies [22,23,25], although in the case of vaccines, high cold chain and equipment costs make total supply chain costs higher and, therefore, labour becomes a smaller percentage of total supply chain costs (Lydon et al. 2014 [3]). In addition, labour costs may be co-funded by the countries themselves and often not included in proposals to donors. The costs in this paper, however, represent the total resources that countries are dedicating to distribution.

The analysis also established that supply chain costs per ACT unit varied by the level of the system with costs increasing as the product reaches the last mile. In both countries, despite having very different health system structures, the costs tended to be the highest at the level of the health centre or dispensary in comparison to the central or intermediate storage depots. A further breakdown of the health facilities by hospitals and primary health clinics illustrated that health clinics were much more expensive than hospitals for supply chain functions per unit of product. Although the overall operating costs of a hospital were higher, clinic storage costs were approximately six times the cost of storage at hospitals for ACT and three times for RDTs mostly due to the higher volume of ACT and RDTs flowing through hospitals. These findings are consistent with those from other studies [7,21,25,26]. Central storage costs for anti-malarials were much higher in Benin possibly due to the use of dedicated warehouses and labour for their management. Similarly in Kenya, operating costs were found to be about three times higher for ACT and RDTs in rural facilities than urban ones. The total volume procured by urban facilities was five times greater than that procured by rural facilities and higher transport costs to rural facilities may play a part in this observation.

The findings also illustrated the importance of product volumes as measured by product dimensions in estimating costs of the supply chain. A greater portion of the storage costs are fixed and reliant on volume; therefore, when considering the costs attributed to a unit of product, as volume increases, costs proportionately decrease. In the case of quality control, costs are often a rate attached to the batch size tested. Cost per unit of product will therefore stay relatively stable despite fluctuation in total volume. The study provided evidence that using a percentage of the acquisition costs to estimate supply chain costs can be inaccurate. This has major implications for the way costs for procurement and SCM are currently estimated. These findings indicate that using “ballpark” percentages to estimate supply chain costs, may cause shortfalls in resource allocations for supply chain functions. Countries should be supported to generate more accurate estimates in the future.

The authors note that the costs described in this study are for the existing public sector supply system, including current inefficiencies [27-30]. This work does not separate inefficiencies but rather builds them into the cost analysis. The study does not identify such inefficiencies in the system, create an optimization model or suggest what an ideally operating system should cost. Doing so would require additional data and analysis on service-level indicators including product availability, quality, and price. While this work moved costing closer toward transparency and accuracy, it is only a glimpse of supply chain operations Kenya and Benin. A more complete picture needs routine data collection with sufficient time-lapsed data points to allow statistical conclusions to be made. In addition to greater cost transparency, this could lead to a richer understanding of operational needs. This cost data however provides valuable information for these additional analyses including information for economic evaluations.

This study has a few limitations. Although the proportional cost of the supply chain was obtained in this research, it is specific to the ACT and RDTs in distributed in Benin and Kenya. These percentages are not to be applied generally by a CMS across all the products of varying value and volume. Two additional considerations that should be made are time allocation, which is the driving force behind labour cost, and specific distribution requirements, such as those for a cold chain. Secondly, the sampling methodology was not intended to be statistically representative of the entire country but was rather directed at identifying the key variables that may influence supply chain cost which may have introduced a level of selection bias.

While an accurate costing of the supply chain is important for budgeting and planning, conducting a micro-costing exercise is a highly resource- and time-intensive effort. Given that the supply chain costs vary according to the acquisition cost, a better methodology is needed for estimating costs rather than using a fixed percentage. Three options should be considered in place; one option is to use different weights for the costs of the supply chain for high and low-value products. A second option would be to use algorithms of different handling fees based on budget line items [10], which would entail separating products and allocating different proportions to each group. Lastly, the authors recommend using a simple extrapolative model to estimate distribution costs at country level. This remains a priority for researchers and implementers working in this arena.

Unpacking the supply chain costs has several advantages. First, countries and donors can identify the major cost drivers where a more in-depth analysis may be required to identify opportunities for improved efficiency particularly if major discrepancies between similar countries are found. Second, it allows donors and countries to separate the costs for the supply chain tiers and functions and allows identifying areas of support based on an empirical evaluation of costs. Third, it allows countries to demonstrate how government funds are being leveraged for procurement and SCM functions in proposals. The study provided information that can be used by policymakers to understand the costs of the supply chain for ACT and RDTs and to advocate for the efficient allocation of resources for future programme needs.

In conclusion, product volumes should be considered when costing distribution rather than value. Countries should avoid using percentages of total acquisition costs to estimate the cost of product distribution, and planners and policy makers can use the information on cost drivers to conduct a more detailed analysis to determine how to increase efficiency. Given that countries should use accurate distribution cost data when budgeting for malaria interventions and that costing studies are expensive and time-intensive, a model to generate more accurate estimates in the future, by estimating costs according to volume rather than value would be embraced by stakeholders. Furthermore, additional country-level studies are needed to capture country-specific costs and to draw inferences across different supply chain structures.