This prospective, observational multicentre (n=24) study investigated relationships between antimicrobial choices and rates of empiric appropriate or adequate therapy, and subsequent adaptation of therapy in 171 ICU patients with severe nosocomial infections. Appropriate antibiotic therapy was defined as in vitro susceptibility of the causative pathogen and clinical response to the agent administered. In non-microbiologically documented infections, therapy was considered adequate in the case of favourable clinical response <5 days. Patients had pneumonia (n=127; 66 ventilator-associated), intra-abdominal infection (n=23), and bloodstream infection (n=21). Predominant pathogens were Pseudomonas aeruginosa (n=29) Escherichia coli (n=26), Staphylococcus aureus (n=22), and Enterobacter aerogenes (n=21). In 49.6% of infections multidrug-resistant (MDR) bacteria were involved, mostly extended-spectrum beta-lactamase (EBSL)-producing Enterobacteriaceae and MDR non-fermenting Gram-negative bacteria. Prior antibiotic exposure and hospitalisation in a general ward prior to ICU admission were risk factors for MDR. Empiric therapy was appropriate/adequate in 63.7% of cases. Empiric schemes were classified according to coverage of (i) ESBL-producing Enterobacteriaceae and non-fermenting Gram-negative bacteria ("meropenem-based"), (ii) non-fermenting Gram-negative bacteria (schemes with an antipseudomonal agent), and (iii) first-line agents not covering ESBL-Enterobacteriaceae nor non-fermenting Gram-negative bacteria. Meropenem-based schemes allowed for significantly higher rates of appropriate/adequate therapy (p<0.001). This benefit remained when only patients without risk factors for MDR were considered (p=0.021). In 106 patients (61%) empiric therapy was modified: in 60 cases following initial inappropriate/inadequate therapy, in 46 patients in order to refine empiric therapy. In this study reflecting real-life practice, first-line use of meropenem provided significantly higher rates of the appropriate/adequate therapy, irrespective of presence of risk factors for MDR.