Background Antimicrobial stewardship guidelines emphasise the importance of benchmarking hospital antimicrobial drug use in order to improve patient outcomes. However, benchmarking strategies are still in their infancy with several methodological limitations.

Purpose Benchmarking antibiotic use, cost and prevalence of nosocomial infections (NI) in 7 surgical and neurosurgical wards of 3 hospitals in 2014.

Material and methods Consumption and cost of antibiotics and NI prevalence were measured in the different wards. For risk adjustment, the supposed correlation from the literature between antibiotic consumption and casemix index (CMI) was tested with regression analysis.

Results A wide heterogeneity was found in antibiotic consumption (20–64 DDD/100 patient days; 120–730 DDD/100 admissions) and costs between the different wards. Wards using the most and least antibiotics differed when measured in the 2 metrics. In 1 ward, 19 NI/100 admissions were revealed, which was remarkably higher compared with the others (0.91–6.89 NI/100 admissions). Significant interhospital differences were detected in CMI, patient days, number of admissions and average length of stay. We found no correlation between antibiotic consumption and CMI (correlation coefficients, CMI and DDD/100 patient days -0.02; CMI and DDD/100 admissions -0.17).

Conclusion The heterogeneity in antibiotic consumption and costs might be caused by several factors: the measured interhospital differences may be influenced by variations in average length of stay, number of occupied beds and patient casemix. The ideal metric of antibiotic use is still under investigation. We suggest using both DDD/100 patient days and DDD/100 admissions. In the ward with the remarkably higher prevalence of NI, the critical appraisal of the effectiveness of local infection control practice seems to be essential. Recent risk adjustment methods, such as regression analysis with CMI, cannot be validated because these oversimplify the complex risk adjustment process. Other methods need electronic patient records, which are still rare in hospitals. Thus we suggest a novel method for adjusting risks in benchmarking. In all wards the risk factors for NI (eg, days of central venous catheters, days of mechanical ventilation) and comorbidities which influence antibiotic consumption (eg, patients with renal impairment, immunosuppressed patients) should be determined and summed, and then quantified (‘scored’) with the results of relevant good quality published studies.

No conflict of interest.