Global antibiotic consumption: A modelling study

Abstract

Background: Antibiotic resistance is a major threat to human health. Antibiotic use is an important driver of resistance; understanding patterns in global usage is essential for combatting limited access and excessive use of antibiotics, and controlling resistance. Existing studies focus on specific countries or regions, we aim to develop a global understanding of antibiotic consumption. Methods and materials: Contemporary antibiotic consumption data covering 46/69 high income (HICs) and 32/135 low and middle income countries (LMICs) were obtained from various sources. We calculated the daily defined doses (DDD) of antibiotics per 1000 population based on the WHO methodology and gap-filled the data using multiple imputation. For the remaining HICs we imputed the total DDD/1000 population of antibiotics for systemic use (J01) using multiple imputation. Antibiotic consumption data were sparser in LMICs, requiring a more extensive modelling framework. Data on the proportion of children with lower respiratory tract infections (LRI) receiving antibiotics were extracted from household surveys and modelled using model-based-geostatistics, informing spatial pattern of antibiotic access across LMICs. These estimates were used (with additional variables) in a stacked generalisation and spatial-temporal Gaussian process regression model to estimate the DDD/1000 population of J01 antibiotics across LMICs. Models were validating using five-fold cross-validation. Additionally, we calculated the types of antibiotics consumed as a proportion of antibiotic consumption. Results: Results are presented as DDD/1000 population/day. The median was 15, with the highest estimates in Mongolia (43), Greece (41), Tunisia (40) and Turkey (40). The lowest were in Indonesia and Philippines (both 5). High antibiotic use in children with LRI was noted in the Middle East, Southeast Asia; use in West Africa was low. Types of antibiotics used varied by country. Penicillins were the most commonly used, with amphenicols and aminoglycosides the least. Conclusion: By combining data from multiple sources and using spatial-temporal models, we provide an estimation of global antibiotic use. We highlight countries with excessive consumption and those with limited access to antibiotics. Surveillance of antibiotic consumption, as part of national action plans requires improvement, especially across LMICs, to fully understand patterns of usage. Judicious use of antibiotics needs to be promoted and implemented globally to curtail rising resistance.