Meticillin-resistant Staphylococcus aureus (MRSA) antibiogram: How inaccurate have our estimates been?

•Australia estimates antibiotic resistance in meticillin-resistant S. aureus (MRSA) biannually from small samples.•We tested all specimens since 2005 for resistance in one state in Australia.•All outpatient specimens illustrated that MRSA resistance has been underestimated.•Biannual sampling did not capture sufficient endemic phenotypes.•Highly resistant endemic phenotypes can only be expressed in a full annual sample.

The objective of this study was to determine the accuracy of antimicrobial resistance patterns reported by the Australian Group on Antimicrobial Resistance (AGAR) established using surveys of just the first 100 Staphylococcus aureus isolates from each participating hospital. Patterns of resistance of a survey sample of S. aureus isolates collected prospectively from five Queensland hospitals participating in the AGAR biennial national survey, using the first 100 isolates diagnosed from each test year, were tested. Meticillin-resistant S. aureus (MRSA) antibiograms for five antimicrobials commonly used to treat outpatients established from a sample have underestimated the true level of resistance by 13–21 percentage points. Conversely, inpatient antibiograms have significantly overestimated the resistance level. Random selection of 100 isolates from all isolates did not provide valid resistance patterns for outpatients or inpatients. Nearly 50% of resistance demonstrated in all inpatient isolates and about 45% of resistance in all outpatient isolates was due to AUS-2/3-like, EMRSA-15-like and MRSA unclassified. In conclusion, testing S. aureus, and in particular MRSA, for resistance levels to commonly prescribed antimicrobials is under/over-estimated in Australia because of a faulty annual sampling method that failed to consider the effect of endemic phenotypes (AUS-2/3-like and EMRSA-15-like). MRSA represents one-third of all S. aureus AGAR isolates. Endemic phenotypes bias the antibiogram patterns when small consecutive sampling (first 100 samples) is used and this bias remains even when samples are selected at random. A minimum sample of 6 months of isolates must be used to accurately establish a national antibiogram.