Blood cultures as a surrogate marker of case-mix for adjustment of hospital antibiotic consumption

Abstract number: 902_p1182

Lamoth F.


Surveillance of hospital antibiotic consumption is an important step towards appropriate use. Consumption is generally adjusted to an indicator of hospital occupancy, e.g. patient-days. Such indicators are readily available, but they do not allow adjustment to the case-mix. We investigated whether the number of blood samplings drawn for culture could be used as a surrogate marker of the burden of infection in a hospital ward, and therefore could serve for adjustment of antibiotic consumption to case-mix.


We first determined one set of clinical and laboratory criteria to define opportunities for prescription of antibiotics (OPA), based on retrospective review of adult patients hospitalised in one university hospital (derivation sample: 178 patients; validation sample: 200 patients). The correlation between OPA and actual antibiotic prescription was good (positive predictive value (PPV) 81%, negative predictive value (NPV) 95%). Then correlation between blood cultures and OPA was established for a medical ward. Finally, we compared antibiotic consumption in this ward measuring it either in defined daily doses (DDD) per patients-day, or in DDD per blood culture, for 16 consecutive trimesters.


Blood cultures had a 70% PPV and a 97% NPV of predicting OPA. Of the 16 consecutive trimesters, 2 were found showing up high antibiotic consumption (>95% CI for the mean) as measured in DDD either per patients-day or per blood culture. For three trimesters antibiotic consumption was high according to one measurement method only. Analysis of random samples of 50 patients per trimester showed that this discrepancy was not due to the instability of blood cultures as marker of OPA. It also confirmed that blood cultures were more accurate for adjustment of consumption since (i) for one trimester with high consumption in DDD per patients-day only, this result was actually due to higher incidence of OPA; (ii) two trimesters with high consumption in DDD per blood culture only had indeed low incidences of OPA, and therefore a probable problem with antibiotic use that was not identified when measured in DDD per patient-day.


Blood cultures are a stable marker of OPA in a medical ward. This allows the identification of periods deserving a detailed investigation given the unexplained high antibiotic consumption.


Session Details

Date: 01/08/2007
Time: 00:00-00:00
Session name: XXIst ISTH Congress
Location: Oxford, UK
Presentation type:
Back to top