Estimating Cost Functions for Resource Allocation Using Transmission Models: A Case Study of Tuberculosis Case Finding in South Africa

Cost functions linked to transmission dynamic models are commonly used to estimate the resources required for infectious disease policies. We present a conceptual and empirical approach for estimating these functions, allowing for nonconstant marginal costs. We aim to expand on the current approach which commonly assumes linearity of cost over scale. We propose a theoretical framework adapted from the field of transport economics. We specify joint functions of production of services within a disease-specific program. We expand these functions to include qualitative insights of program expansion patterns. We present the difference in incremental total costs between an approach assuming constant unit costs and alternative approaches that assume economies of scale, scope and homogeneous or heterogeneous facility recruitment into the programme during scale-up. We illustrate the framework’s application in tuberculosis, using secondary data from the literature and routine reporting systems in South Africa. Economies of capacity and scope substantially change cost estimates over time. Cost data requirements for the proposed approach included standardized and disaggregated unit costs (for a limited number of outputs) and information on the facilities network available to the program. The defined functional form will determine the magnitude and shape of costs when outputs and coverage are increasing. This in turn will impact resource allocation decisions. Infectious diseases modelers and economists should use transparent and empirically based cost models for analyses that inform resource allocation decisions. This framework describes a general approach for developing these models. •Mathematical and economic modeling is being used to inform how disease control resources are allocated and what policies are adopted. These models specify complex nonlinear relationships between service coverage and impact. Nevertheless, functions describing the relationship between costs and service volume have typically been less sophisticated, assuming constant marginal costs of expanding service coverage. This approach runs counter to the theoretical understanding of the way costs behave when scaling up.•We propose an alternative approach. We developed a mechanistic framework to estimate total costs for inclusion in model-based economic evaluations using a combination of secondary data from small-scale costing studies and routine reporting systems. We provide a pragmatic, mechanistic