Objective: To investigate how many nosocomial infections would be missed if surveillance activities were restricted to patients having either microbiology reports or antibiotic administration.
Design: Analysis of data from a large prevalence study on nosocomial infections (Nosocomial Infections in Germany-Surveillance and Prevention).
Setting: A total of 14,966 patients were investigated in medical, surgical, obstetric-gynecologic, and intensive-care units of 72 German hospitals representatively selected according to size. Five hundred eighteen patients (3.5%) had at least one nosocomial infection. Microbiology reports were available for 56.6% of these patients on the prevalence day, and 86.3% received antibiotics.
Results: Only 31 nosocomially infected patients (6%) would have been missed by using either microbiology reports or antibiotic treatment as an indicator. These indicators of nosocomial infections had a high diagnostic sensitivity for nosocomial pneumonia (98.8%), urinary tract infections (96.3%), and primary bloodstream infections (95.3%), but a lower sensitivity for wound infections (85.4%). Thus, 97.4% of all nosocomial infections were found with this method in intensive-care units and 96.1% in medicine units, but only 89.7% in surgical departments. In 9 (12.5%) of 72 hospitals, the overall sensitivity would have been <80% using a combination of the two indicators. For this reason, the situation in one's own hospital should be checked before using this method.
Conclusions: After checking the situation in one's own hospital, the "either-or" approach using the two indicators "microbiology report" and "antibiotic administration" can be recommended as a time-saving measure to diagnose pneumonia, urinary tract, and primary bloodstream infections. For wound infections, additional information obtained by changing dressings or participating in ward rounds is necessary to achieve satisfactory sensitivity in the surveillance of nosocomial infections. Of course, it is necessary that the surveillance staff discard all false positives to ensure a satisfactory specificity.