Background: Excessive antibiotic use increases the risk of development and dissemination of bacterial resistance. A comparative analysis of the correlation between hospital antibiotic consumption and rates of bacterial resistance is needed for a better understanding of the complex relationship between antibiotic use and resistance. Apart from economic and market research studies, estimates of antibiotic consumption in German hospitals, however, are not available. Methods: In a pilot project (INTERUNI-II), retrospective data from eight university hospital pharmacies covering the period 1998, 1999, and 2000 were collected to obtain estimates for the antibiotic use densities in the medical services of teaching hospitals. Antibiotic use densities were expressed as prescribed daily doses per 100 occupied bed days (PDD/100). The definition of prescribed daily doses was according to guidelines for antimicrobial therapy in adults with normal renal and hepatic function used in the participant hospitals. Means and ranges of antibiotic use densities were separately assessed for medical intensive care units (MICU), hematology-oncology services (HEMONC), and other medical services (OTHER MED). Results: Mean antibiotic use density in internal medicine was 55.2 PDD/100 overall, ranging between 39.4 and 75.8 PDD/100 in the eight participant hospitals. In seven hospitals antibiotic use density increased during the years of observation. Antibiotic use was higher in MICU areas (3-year average, 122.3; range, 98-167 PDD/100) than in HEMONC (3-year average, 86.9; range, 67.8-129.4 PDD/100) and OTHER MED areas (3-year average, 42.8; range, 31.7-50.6 PDD/100). There was an increasing use of oral antibiotics resulting in a substantial proportion of oral agents among all antibacterial drugs outside MICU areas (year 2000, HEMONC, range, 36-74% of all PDD; OTHER MED, range, 43-59% of all PDD). β-lactam antibiotics were the most frequently prescribed drugs (3-year average, 22.6 PDD/100). 56% of β-lactam PDD belonged to the class of broad-spectrum β-lactams (ranges, MICU, 49-82%; HEMONC, 61-89%; OTHER MED, 24-58%). Fluoroquinolones were the second most prescribed drug class (3-year average, 13 PDD/100). They were most frequently used in HEMONC (3-year average, MICU, 14.5; HEMONC 26.5; and OTHER MED 8.6 PDD/100, respectively). There was considerable variation between participant hospitals in the use of specific drug classes in given patient care areas. Conclusion: This retrospective study showed significant variation in overall and specific antibacterial drug class use between German teaching hospital medical services and defined patient care areas. Given the variation in the obtained estimates, targeted prospective hospital antibiotic use surveillance with fast data acquisition and analysis might offer an excellent opportunity to evaluate the impact of differences in antibiotic use and of revised therapy guidelines on the evolution of nosocomial bacterial resistance.