Summary Increased severity of illness among hospitalised patients and an ageing population have led to an increased incidence of hospital acquired infections and represent a significant challenge to the clinician in terms of managing infections. The collateral damage which can occur with antibiotic therapy is also an important consideration when initiating empirical antibiotic therapy, particularly in patients who are seriously ill or immunocom-promised. Collateral damage is the term used to describe the adverse ecological effects of antibiotic therapy, such as the selection of drug-resistant organisms, and the adverse events associated with antibiotic therapy such as Clostridium difficile disease. Antibiotic use and ineffective infection control have been implicated in the development and spread of resistant Gram-positive and Gram-negative bacterial pathogens which are associated with increased mortality and morbidity, prolonged hospitalisation and increased costs. Carbapenem consumption and mechanical ventilation have been linked to colonisation or infection with problematic organisms including methicillin-resistant Staphylococcus oureus, Pseudomonas aeruginosa and Stenotrophomonas maltophilia , while cephalosporin use has been associated with evolution of infections due to vancomycin-resistant enterococci (VRE) and Gram-negative bacilli producing extended-spectrum β-lactamases (ESBL), and to colonisation or superinfection with Clostridium difficile. The safety profile of antibiotics must also be taken into consideration when selecting therapy, and single broad-spectrum agents may provide excellent coverage with a low risk of adverse events. The use of single agents may be associated with lower costs, improved ease of administration and fewer drug-drug interactions. However, in an environment of increasing resistance, initial aggressive therapy may be required to avoid excessive mortality and morbidity. Ideally antibiotic therapy should be directed by culture and knowledge of local susceptibility patterns. Before culture results are available therapy may need to be initiated empirically to cover the likely pathogens. In neutropenic patients with fever the current guidelines recommend the use of empirical therapy at the onset of fever for all patients. Where no aetiology is identified, antibiotic therapy should continue for at least 2 weeks while aggressive attempts are made to define the source of fever. When the aetiology of infection has been identified, therapy should be adjusted to provide optimal treatment with the best safety profile and lowest cost. The principal of avoiding collateral damage provides a useful framework for selecting antibiotics for empirical therapy in today's changing environment.