Comparison of methods for measuring antibiotic consumption in an intensive care unit

Hospitals worldwide are working on minimizing unnecessary use of antimicrobials. To assess actual changes of antimicrobial usage, correct and precise measurements are necessary. This study aimed to compare data on the purchase of antibiotics from the pharmacy and the administration of antibiotics to patients, respectively, in an intensive care unit (ICU). Data were obtained from the Neurointensive Care Unit (NICU) at Rigshospitalet, Denmark. During a 23-month period, comprising 10 770 bed-days (BD), the ward purchased 16 908 defined daily doses (DDD) of antibiotics from the pharmacy, and 15 130 DDD and 41 304 individual doses were administered. Intraclass correlation coefficients (ICCs) were calculated; control and runcharts and a Bland-Altman plot were constructed. Pharmacy sales and drug administration data showed no systematic variation over time with a monthly overestimation of pharmacy sales data of 10% (95% confidence interval (CI), 6.20-14.3%) for all antibiotics, and 7% (95% CI: 1.81-11.1%) for broad-spectrum antibiotics. The antibiotic consumption, without bed-days, has a clinically acceptable ICC of >0.70 and no systematic difference is suggested by the Bland-Altman plot. In this study of a large NICU, whose antibiotic consumption varied at random, pharmacy sales data were an acceptable approximation of the actual summarized drug consumption.