Antibiotic resistance in hospitals: a ward-specific random effect model in a low antibiotic consumption environment

• Published in: Statistics in medicine Vol. 32; no. 8; pp. 1407 - 1418
• Main Authors: Aldrin, Magne, Raastad, Ragnhild, Tvete, Ingunn Fride, Berild, Dag, Frigessi, Arnoldo, Leegaard, Truls, Monnet, Dominique L., Walberg, Mette, Müller, Fredrik
• Format: Journal Article
• Language: English
• Published: England Blackwell Publishing Ltd 15.04.2013; Wiley Subscription Services, Inc
• Subjects: Anti-Bacterial Agents - administration & dosage; Antibiotics; binomial regression; Computer Simulation; Drug resistance; Drug Resistance, Bacterial; Hospitals; Humans; Medical statistics; Models, Statistical; Norway - epidemiology; Poisson regression; Pseudomonas aeruginosa - drug effects; Pseudomonas aeruginosa - isolation & purification; Pseudomonas Infections - drug therapy; Pseudomonas Infections - epidemiology; Pseudomonas Infections - microbiology; Regression analysis; resistance; Retrospective Studies; scenario analysis; Simulation; Norway
• ISSN: 0277-6715; 1097-0258
• DOI: 10.1002/sim.5636

Association between previous antibiotic use and emergence of antibiotic resistance has been reported for several microorganisms. The relationship has been extensively studied, and although the causes of antibiotic resistance are multi‐factorial, clear evidence of antibiotic use as a major risk factor exists. Most studies are carried out in countries with high consumption of antibiotics and corresponding high levels of antibiotic resistance, and currently, little is known whether and at what level the associations are detectable in a low antibiotic consumption environment. We conduct an ecological, retrospective study aimed at determining the impact of antibiotic consumption on antibiotic‐resistant Pseudomonas aeruginosa in three hospitals in Norway, a country with low levels of antibiotic use. We construct a sophisticated statistical model to capture such low signals. To reduce noise, we conduct our study at hospital ward level. We propose a random effect Poisson or binomial regression model, with a reparametrisation that allows us to reduce the number of parameters. Inference is likelihood based. Through scenario simulation, we study the potential effects of reduced or increased antibiotic use. Results clearly indicate that the effects of consumption on resistance are present under conditions with relatively low use of antibiotic agents. This strengthens the recommendation on prudent use of antibiotics, even when consumption is relatively low. Copyright © 2012 John Wiley & Sons, Ltd.