Despite advances in supportive care and use of antibiotics, sepsis preserves its importance due to its high mortality and morbidity for neonates. Identifying the causative agents and antibiotic resistance yearly in a neonatal intensive care unit (NICU) helps the physician to choose the most appropriate empirical therapy. In this study we aimed to evaluate positive blood cultures and antibiotic susceptibilities of newborns with proven sepsis during the years 2000-2002 in our NICU. The charts of babies with sepsis were evaluated for clinical characteristics, positive cultures and antimicrobial susceptibilities, retrospectively. Although most of the admitted patients were premature (76.5%), the frequency of proven sepsis was quite low, at 9.1% among 909 newborns. Mortality rate in sepsis was 16%. The most commonly isolated micro-organisms were coagulase-negative staphylococci (CoNS) (31.3%), fungi (19.2%), Staphylococcus aureus (13%) and Klebsiella pneumoniae (10.5%). Methicillin resistance for CoNS was 92.3% and for S. aureus was 72.7%. In the last year, a significant increase in the frequency of Klebsiella pneumoniae (8.3 vs 14.2%), CoNS (27.1 vs 37.1%), Pseudomonas aeruginosa (2.1 vs 8.6%) and fungal infections (18.8 vs 20%) was observed compared to the previous years. An initial empirical antibiotic therapy for late-onset sepsis was designed with teicoplanin + piperacillin-tazobactam/meropenem + antifungal (fluconazole or amphotericin B) as the best combination to cover this spectrum until the culture results arrive. However, this combination is only compatible with our results and may not be applied in all units. Every unit must follow the bacterial spectrum and antibacterial resistance patterns to choose their specific empirical treatment strategy for nosocomial infections.