Introduction: Treatment of cancer has contributed to a growing number of immunocompromised patients with life-threatening nosocomial infections (NI). High mortality with considerable cost is observed when they are admitted to the intensive care unit (ICU). Few studies on infection control and surveillance have been undertaken in this population group., Methods: All patients treated at a six-bed medical-surgical oncology ICU for > 48 hours were prospectively observed for the development of an NI and the influence of device utilization on infection rates. The analysis used the standard definitions of the National Nosocomial Infection Surveillance System Intensive Care Unit surveillance component., Results: From September 1993 through November 1995, 370 infections occurred in 623 patients during 4034 patient-days, for an overall rate of 50.0 per 100 patients or 91.7 per 1000 patient-days. Pneumonia (28.9%), urinary tract infections (25.6%), and bloodstream infections (24.1%) were the main types of infection. The most common microorganisms isolated were Enterobacteriaceae (29.7%), fungi (22.2%), and Pseudomonas aeruginosa (13.2%). The median device utilization ratios were 0.63, 0.83, and 0.86 for ventilator, indwelling urinary catheter, and central venous catheter, respectively. The highest median device-specific associated infection rate was 41.7 for ventilator. The median for the average length of stay was 8.8 days, and the average severity of illness score was 4.0. There was a strong positive correlation between the overall NI patient rate and device utilization (r = 0.56, p < 0.01), average severity of illness score (r = 0.54, p < 0.01), and average length of stay (r = 0.67, p < 0.01). No correlations were statistically significant when patient-days were used in the denominator. Among the devices only the number of central venous catheter days was significantly correlated with infections (r = 0.51, p = 0.01). The NI patient-day rates were progressively higher the longer the patients stayed in the ICU., Conclusions: The high rates reported in this study may reflect a combination of several factors related to the underlying illness, neutrophil count, and exposure to invasive procedures. The adjusted infection rates described here provide specific surveillance data for further interhospital comparisons and also to assess the influence of invasive medical interventions, allowing the implementation of preventable measures to control infections.