Estimating non-communicable disease treatment costs using probability-based cost estimation

The burden and impact of non-communicable diseases (NCDs) are well documented, accounting for 70% of premature deaths globally. In Sub-Saharan Africa, rising NCDs are estimated to account for 27% of mortality by 2020, a 4% increase from 2005. This increase will inevitably lead to a higher demand for NCD treatment services, exerting pressure on limited public financial resources. To get a sense of the resources required to treat NCDs, it is necessary to estimate the costs associated with the diagnosis, treatment and management thereof. Typically, in estimating costs for health services, countries use historical patient level data combined with demographic trend data and non-patient level data to arrive at estimated future costs. This methodology relies heavily on the availability of data from a wide variety of sources stretching beyond the health sector. Low-and-middle-income countries often lack the requisite data and are compelled to use less efficient ways to determine resource allocation. This study explores the use of probability-based cost estimation to estimate the cost of delivering NCD treatment services in South Africa, one such data-poor environment. Probability-based cost estimation, in combination with deterministic cost estimation, is used in arriving at a cost estimate for NCD treatment services at primary healthcare facility level. On its own, deterministic cost estimation can determine total costs, provided all the input variables are known. This is not always possible because of the lack of one or more input variables. In most instances, the lacking input variable is the quantities at which specific conditions will be treated. This problem is addressed by using probability-based cost estimation through which a mean cost is calculated and applied to the target population as a whole, eliminating the need for quantities per condition. Thus, this model contains both deterministic and probabilistic cost estimation elements.