Identifying hospital-level predictors for antibiotic use: a Global Point Prevalence Survey study among Belgian, Philippine and South African hospitals

Abstract

Introduction Understanding determinants that influence antibiotic use is crucial for developing effective stewardship interventions that optimize the use of antibiotics and ultimately mitigate the burden of antibiotic resistance. This study aims to find hospital-level predictors of antibiotic use among hospitals worldwide.Methods Antibiotic use data were collected from hospitals that participated in 2022-23 in the Global Point Prevalence Survey (Global-PPS) basic inpatient and Healthcare-Associated Infection (HAI) modules from three countries with high participation degrees: Belgium, the Philippines and South Africa. A linear mixed model was applied to predict hospital-level prevalence, considering all variables from the Global-PPS as potential predictors. Stepwise forward and backward selection modelling were conducted to identify the best-fitting models per country separately.Results In total, data from 138 hospitals were retrieved, including 19 Belgian, 55 Philippine and 64 South African hospitals. No predictors were shared across the three countries, such as hospital type, proportion of occupied medical beds and proportion of available guidelines. However, the proportions of admitted patients with certain invasive devices were significant predictors for increased odds of hospital-wide antibiotic prescribing, despite the type of invasive device varying between countries.Conclusions Factors related to the patient mix, including the proportion of patients with an invasive device, are associated with increased odds of antibiotic prescribing at hospital level. Considering substantial differences in predictors associated with country-specific patient populations and hospital characteristics, more research is needed to explore additional determinants, such as healthcare system and broader contextual factors that may influence hospital-level antibiotic prevalence.