Technical efficiency of health production in Africa: A stochastic frontier analysis

Abstract

Background: Inefficiency is widespread in health systems all over the world. The World Health Organization (WHO) estimates that 20%-40% of the global health spending is wasted. In African countries, inefficiency of this magnitude will seriously hamper progress towards achieving universal health coverage and other health system goals. It is thus, significant to assess the efficiency of health systems over time in order to set the ground for identifying the contextual factors leading to inefficiency and design appropriate efficiency-enhancing measures.

Methods: Using panel data for the years 2000, 2005, 2010, and 2015, the study employs a time-variant stochastic frontier production function to assess efficiency. The input measure used is current expenditure per capita in purchasing power parity (Int$) terms and the measure of output is health-adjusted life expectancy (HALE). Moreover, mean years of schooling, GDP per capita in Int$, and out-of-pocket payment as a share of current expenditure on health were used as technical inefficiency effect variables. Data were analyzed using Frontier Version 4.1.

Results: The mean technical efficiency scores were 79.3% in 2000, 81% in 2005, 85.6% in 2010 and 88.3% in 2015. Over the four periods of time, Cabo Verde registered the highest technical efficiency scores, while Eswatini and Sierra Leone had the lowest. The minimum technical efficiency scores were 58.7% (in 2000), 59.1% (2005), 67.4% (2010) and 71.8% (2015). These indicate that despite improvements, there is a significant degree of technical inefficiency. Most of the countries among those in the bottom 10% efficiency scores are countries in Southern Africa, which in 2015 had a very high prevalence of HIV among adults, compared to the top 10%, which had prevalence rates of less than 0.1%.

The mean efficiency score increased progressively over time – a nine percentage point increase between 2000 and 2015. The elasticity of current health expenditure was positive (0.06) and statistically significant. All the technical inefficiency variables had no statistically significant effect.

Conclusions: Over the period of time covered in this study, there was some improvement in the average technical efficiency scores. However, there was also marked inefficiency in many countries, which is likely to hamper their progress towards universal health coverage and other health system goals. In a context where health spending is too low to provide needed services, it is imperative to address the causes of technical inefficiency and produce more health for the money. Furthermore, low-performing health systems should learn from their relatively high-performing peers.