A prospective cohort study with retrospective validation was initiated in order to assess whether a monthly bacteriological report improves the accuracy in detecting hospital-acquired infections (HAI). The setting was a 14-bed medical intensive care unit (ICU) in a 821 bed French university affiliated hospital. One thousand, six hundred and two patients were admitted during the two-year study period, the mean age was 58+/-19 years, the mean Simplified Acute Physiology Score 2 (SAPS 2) was 34+/-21, and ICU mortality was 14%. The microbiology laboratory sent monthly bacteriological reports of urine samples and central venous catheter (CVC) tips back to the intensive-care unit physician in charge of the HAI surveillance programme. This enabled a comparison to be made between prospectively and retrospectively diagnosed hospital-acquired urinary tract infections (HAUTI) and CVC-related infections (HACVCI), HAUTI were prospectively identified in 51 cases (incidence density=10.03/1,000 days) and 23 more cases were found after receiving the monthly bacteriological report (final HAUTI incidence density=14.6/1,000 days, P<0.05). HACVCI were prospectively recognized in 13 cases (incidence density=4/1,000 days) and eight more cases were discovered (final HACVCI incidence density=6.52/1,000 days,P >0.1). All retrospectively diagnosed HAI occurred during the last 48 h of the patients' ICU stay. We conclude that the routine HAI surveillance programme is reliable, except for the last 48 h in the ICU. The monthly bacteriological report improved the accuracy of the HAI reporting rate.