Antibiotic prescription is frequently inappropriate in hospitals. Our objective was to evaluate whether the quality of antibiotic prescription could be measured using case vignettes to assess physicians' knowledge.

The study was conducted in 2 public teaching hospitals, where 33/41 units and 206/412 physicians regularly prescribing antibiotics to inpatients agreed to participate. A cross-sectional survey of knowledge was performed using 4 randomly assigned case vignette sets. Curative antibiotic prescriptions were then evaluated using standard criteria for appropriateness at initiation (day 0), after 2–3 days of treatment (days 2–3), and at treatment completion. We compared knowledge of the physicians with their observed prescriptions in the subset of 106 physicians who completed the case vignettes and prescribed antibiotics at least once.

The median global case vignette score was 11.4/20 (interquartile range, 8.92-314.3). Of the 483 antibiotic prescriptions, 314 (65%) were deemed appropriate at day 0, 324 (72%) on days 2–3, and 227 (69%) at treatment completion. Prescriptions were appropriate at all 3 time points in only 43% of patients. Appropriate prescription was positively and independently associated with emergency prescription on day 0, documented infection on days 2–3, and 1 of the 2 hospitals at treatment completion. In addition, appropriateness at the 3 evaluation times was positively associated with prescription in anesthesiology or the intensive care unit. Case vignette scores above the median were significantly and independently associated with appropriate antibiotic prescription on days 2–3 and at treatment completion.

Case vignettes are effective for identifying physicians or hospitals whose knowledge and practice regarding antibiotic prescription require improvement.

To evaluate whether the standardized incidence ratio (SIR) is a more reliable tool for comparing rates and temporal trends of surgical site infection (SSI) in surgery wards than the incidence rate among patients with an National Nosocomial Infections Surveillance system (NNIS) risk index category of 0.

Observational, prospective cohort study in a sequential SSI surveillance system.

Volunteer surgery wards in a surveillance network in northern France that annually conducted SSI surveillance for 3 months from 1998 to 2000.

The incidence rate was the number of SSIs divided by the number of patients included, stratified by the NNIS risk index category. SIR was the observed number of SSIs divided by the expected number computed using a multiple regression model.

Overall, 26,904 patients in 67 surgery wards were enrolled. Between 1998 and 2000, the SSI incidence rate among patients with NNIS risk index category 0 decreased from 2.1% to 1.4%, which was a 33% reduction (P = .002). The SIR decreased from 1.2 (95% confidence interval [CI], 1.1-1.3) to 0.8 (95% CI, 0.7-0.9), which was a 20% decrease per year and an overall 33% reduction. The number of SSIs was significantly higher than expected in 17 of 201 surveillance periods over the 3 years. The classification of the wards according to the 2 indicators over the 3 years showed that wards with a high SIR did not consistently have the highest SSI incidence rate among patients with NNIS risk index category 0, partly because the type of surgical procedure and the duration of follow-up are not taken into account in the NNIS risk index.

SIR should be considered a reliable indicator to estimate the reduction in SSI incidence that results from implementation of infection control policies and for comparison of SSI rates between wards.