Denominators for national estimates of antimicrobial use may not work for herds

Plans for reducing usage of antimicrobial agents at the national level are typically based on target setting: that is, identifying current levels of usage, and specifiying a target for reduction. Target setting requires appropriate measures of usage which are perceived as being representative of amounts of these agents that are actually used, and perceived to be fair in the sense that they do not unduly bias the overall summary measure in favour of some types of usage or some types of herd demographics. In this study, we identified the antimicrobial agents used in some UK cattle herds, compared different measures to determine how they relate to one another, and evaluated the application of the Population Corrected Unit as a summary of herd-size. We extracted data on pharmaceutical sales for 75 cattle herds over the period 2011-2015 in a UK veterinary practice and linked these to farm demographic data pertaining to the herd-sizes from the Government’s agricultural census. The data were screened for products that contained antimicrobials. The results were aggregated at the herd-year level. We then compared three measures of usage: total mg active ingredient, DDDVet & DCDVet. Results demonstrate that there was a very strong linear relationship between these three measures of usage. The mean weight of active ingredient sold per kg of all bovine livestock was 12.8 mg/kg per herd p.a. for beef herds and 16.3 mg/kg per herd p.a. for dairy herds. We estimate that, by weight, only 1.88% of the antimicrobials sold was classified as “Highest Priority Critically Important”(World Health Organization 2017). The use of the Population Correction Unit (PCU) is a method of arriving at a single aggregate measure of herd size that is considered relevant to antimicrobial use. The calculation of PCU is based on the number of animals considered to be relevant, multiplied by a standardised weight for those types of animals. In this study, estimates of usage divided by PCU were highly sensitive to unusually low values for PCU for specific herd-years as a result of the demographic composition in those herd-years. Therefore we propose that estimating usage per PCU is not very appropriate when comparing groups of cattle with very different demographic composition. Conclusion We describe our concerns regarding any comparison of antimicrobial usage between groups of cattle where the demographic composition of the groups are very different because of the sensitivity of such measures to the differences in demographics.