Effect of antimicrobial consumption on Escherichia coli resistance: assessment and forecasting using Dynamic Regression models in a French university hospital (2014-2019)

Abstract

IntroductionThe aim of this study was to determine the correlation between antimicrobial consumption (AMC) and antimicrobial resistance (AMR) in Escherichia coli at a hospital level, and assess the capacity of dynamic regression (DR) models to predict AMR for their use in deployment of antimicrobial stewardship programs (ASPs). MethodsA retrospective epidemiological study was conducted in a French tertiary hospital between 2014 and 2019. DR models were used to assess the correlation between AMC and AMR from 2014 to 2018. The predictive abilities of the models were estimated by comparing the predicted data with those observed in 2019. ResultsRates of fluoroquinolone and cephalosporin resistance decreased. AMC increased overall but decreased for fluoroquinolone. DR models highlighted that the decrease in use of fluoroquinolone and the increase in use of anti-pseudomonal activity penicillin with beta-lactamase inhibitor (AAPBI) explained 54% of the decrease in fluoroquinolone resistance and 15% of the decrease in cephalosporin resistance. In addition, penicillin/beta-lactamase inhibitor (PBI) consumption explained 53% of PBI resistance, and beta-lactam use explained 36% of penicillin resistance, with both remaining stable over time. DR models had predictive capabilities with margins of error from 8% to 34%. ConclusionOver a six-year period in a French tertiary hospital, decreasing rates of resistance to fluoroquinolones and cephalosporins were correlated with decreasing use of fluoroquinolone and increasing use of AAPBI, whereas rates of resistance to penicillin remained high and stable. The results indicate that DR models should be used with caution for AMR forecasting and ASP implementation.