Methodological validation of monitoring indicators of antibiotics use in hospitals

Background: For several years now, the French national recommendations have been trying to set up a surveillance system in hospitals to link data on antibiotic resistance and data on the use of antibiotics, particularly for certain 'micro-organism/antibiotic' pairs. The indicators recommended in the lastest newsletter of the Direction Generale de la Sante (French Public Health Department) for monitoring the consumption of antibiotics were the number of days of treatment or the number of defined daily doses (DDD), both (in)directly related to the number of days of hospitalisation and/or the number of patients hospitalised. Objective: The aim of this study was to compare the actual number of days of treatment, which is an observed indicator, with two indicators calculated on the basis of the ODD and the DPD (daily prescribed dose), both in terms of feasibility of collection and the relevance of the information generated. Materials and methods: For several hospital care units, the 'length of exposure' to a given antibiotic was determined by four different indicators: two actual observed indicators [the patient's medical file (reference) and the named-patient based, computerised dispensing system from the central pharmacy] and two derived calculated indicators [obtained by dividing the number of grams prescribed by the ODD or by the DPD]. Results: The average incidence density of antibiotic treatment (length of exposure per 1000 days of hospitalisation) obtained by the calculated indicators was higher than that obtained with the observed reference (+ 52% for the ODD and + 33% for the DPD) but lower than that obtained with the second observed indicator (computerised system) (-10%). The differences were large and random (high variability depending on the hospital department, the antibiotic and the administration route; variations in both directions: actual length of treatment longer or shorter than the calculated length of treatment). Conclusion: The question which indicator should be chosen is inconclusive for the evaluation of the selection pressure exerted by an antibiotic. The two indicators proposed in the newsletter (observed indicator and calculated indicator) seem to be complementary for use in a regional or national network to monitor resistance and consumption of antibiotics. Each hospital should validate the indicators and define for itself which indicator is most appropriate for estimating the actual length of antibiotic exposure. This may imply different indicators for different units, antibiotics or even administration routes within one particular hospital setting. Once validated the hospital has a powerful tool generating data that can be linked to resistance data.