This is a retrospective review study to investigate changes in carbapenem consumption and to evaluate the proportion of inappropriate empirical use of carbapenem in the months of September and October of 2009, 2011, and 2013 in a single university-affiliated hospital. Total carbapenem use was classified into 3 categories: prophylactic, directed, and empirical. If an empirical prescription was continued without documentation of any eligible etiologic microorganism, we defined this as 'inappropriate' use. We also considered it 'inappropriate' when a patient's culture revealed no pathogen and the patient was initially not in severe sepsis or septic shock and did not have a history of admission to a health-care facility or of colonization with a pathogen eligible for carbapenem within 3 months. The total amount was 48.1, 51.1, and 91.0 defined daily doses/1000 patient-days in 2009, 2011, and 2013, respectively. Empirical use accounted for 78.4% of all prescriptions. The proportion of inappropriate empirical use ranged from 15.0 to 38.9% of the empirical carbapenem prescriptions.