Background: Antimicrobial resistance is a major threat. The prevalence of multidrug resistant organisms (MDROs) in hospital wards depends both on the local antibiotic selection pressure and MDRO introduction from other wards or the community. Understanding the relative contribution of these factors is essential for designing effective antimicrobial stewardship and infection control strategies. Here we adopt a metapopulation approach to compare the impacts of local selection and inter-ward introduction on the prevalence of various MDROs and their non-MDR counterparts. Materials/methods: Data were collected from October 2016 to October 2017 from 523 wards in 4 hospital groups in Lyon, France. We considered six bacterial groups: cephalosporin-resistant Escherichia coli and Enterobacter cloacae and carbapenem-resistant Pseudomonas aeruginosa, and their non-MDR variants. The main model outcome was the number of patients infected by each bacterial group in each ward. Local selection pressure was estimated from the number of delivered defined daily doses of antibiotics. As a proxy for inter-ward introduction, we counted patients transferred from wards harbouring the same bacterial group. We used a modified Hanski metapopulation model with Poisson regression controlling for sampling bias, to examine associations of the prevalence of each bacterial group with antibiotic consumption and patient transfers. Results: Antibiotic consumption consistently predicted prevalence (P<10-16), but the proportion of model deviance explained varied between groups, from 5.6% in non-MDR E. coli to 22.3% in MDR E. cloacae. For E. coli and P. aeruginosa, inter-ward introduction was as important as antibiotic consumption for predicting non-MDR prevalence (P<10-16); for MDR prevalence its contribution was ~5 times less than that of antibiotic consumption (P<0.05). For both MDR and non-MDR E. cloacae, inter-ward introduction did not predict prevalence (P>0.05). Conclusions: The impact of antibiotic consumption on MDRO prevalence strongly outweighs that of inter-ward introduction. The lack of impact of introduction on E.cloacae prevalence suggests that the selection of both MDR and non-MDR variants occurs locally in individual wards and is strongly dependent on antibiotic consumption. Hence, patient isolation might be less effective for controlling E. cloacae spread compared to P. aeruginosa or E. coli. Controlling resistance requires adapting strategies to each bacterial taxon.