Abstract
Efficient logistics systems move essential medicines down the supply chain to the service delivery point, and then to the end user. Experts on logistics systems tend to see the supply chain as requiring centralized control to be most effective. However, many health reforms have involved decentralization, which experts fear has disrupted the supply chain and made systems less effective. There is no consensus on an appropriate methodology for assessing the effectiveness of decentralization in general, and only a few studies have attempted to address decentralization of logistics systems. This paper sets out a framework and methodology of a pioneering exploratory study that examines the experiences of decentralization in two countries, Guatemala and Ghana, and presents suggestive results of how decentralization affected the performance of their logistics systems. The analytical approach assessed decentralization using the principal author's 'decision space' approach, which defines decentralization as the degree of choice that local officials have over different health system functions. In this case the approach focused on 15 different logistics functions and measured the relationship between the degree of choice and indicators of performance for each of the functions. The results of both studies indicate that less choice (i.e. more centralized) was associated with better performance for two key functions (inventory control and information systems), while more choice (i.e. more decentralized) over planning and budgeting was associated with better performance. With different systems of procurement in Ghana and Guatemala, we found that a system with some elements of procurement that are centralized (selection of firms and prices fixed by national tender) was positively related in Guatemala but negatively related in Ghana, where a system of 'cash and carry' cost recovery allowed more local choice. The authors conclude that logistics systems can be effectively decentralized for some functions while others should remain centralized. These preliminary findings, however, should be subject to alternative methodologies to confirm the findings.