OBJECTIVES: To consider reliable sources of data to assess epidemiology of hospital-acquired infections (HAI) in intensive care units (ICU). METHODS: In order to get epidemiological data, a literature review (2002-2009) focused on bloodstream infection (BSI), catheter related BSI (CR-BSI) and ventilator-associated pneumonia (VAP) in hospital ICU in the US was performed using EMBASE database, completed with data from Centers for Disease Control & Prevention (CDC). US hospital databases (2007) were also considered to put findings in perspective: Premier PerspectiveTM Hospital Database (PPHD) and State Inpatient Database (SID: part of healthcare cost and utilization project (HCUP)) providing ICU data for this study. RESULTS: The rate of ICU hospitalized patients in the US databases (9.4%) was within the results in literature (8%-15%). The proportion of ICU patients with devices was higher in literature than in SID and PPHD databases with for central line catheters: 48%, 19%, 26.2% and for mechanical ventilation: 33%, 21.9% and 15% respectively. CR-BSI were reported in 4.8% of ICU patients with central venous catheter, in literature (1.4 - 5.5 /1,000 central line days), 9.4% in SID (5.7/1,000 catheter/central line days) and 33.8% in PPDH*. Rate of VAP in ICU was estimated at 2.1-10.7/1,000 ventilator days in literature and 12.2/1,000 ventilator days in SID. Proportion of VAP amongst ICU mechanically ventilated patients was similar within the databases (12%). BSI attributable mortality was comparable in the literature (12-25%) and hospital database (24.7% in PPHD). Attributable mortality for VAP was not found in the literature. This latter was estimated from the hospital database (19.6% in PPHD). CONCLUSIONS: Different sources are available to estimate the prevalence of HAI in ICU. Results should be interpreted with caution due to methodology limitations (e.g. HAI case definitions). Nevertheless, real-life databases appear appropriate to estimate attributable mortality of HAI in ICU. * central venous/line catheter