This post is best read with a glass of wine…

As a profession, I think we are really not very good at measuring antibiotic resistance/antibiotic susceptibility patterns…

We are very happy to proclaim at the start of presentations “Antibiotic resistance is increasing”, or “In the era of increasing antibiotic resistance.” without providing any data to support this claim.

We need to move away from this type of talk. We are after all, scientists, not politicians.

However antimicrobial resistance surveillance is deceptively difficult. Here are a few reasons why high quality surveillance data is hard work, and requires a lot of thought and planning…

• It is actually the trends that are critical:- There is a big difference between providing an annual antibiogram to clinicians, and presenting graphs which show changes in antimicrobial susceptibility over time. For example a GP looks at an annual antibiogram provided by the microbiology laboratory and sees that organism X has a resistance rate to antibiotic Y of 10%. Doesn’t sound too bad and certainly a viable treatment option. But if we knew that the resistance rate was 5% last year and 2% the year before that, then we have a problem. It seems obvious, but antimicrobial resistance surveillance is all about trends, not snapshots.
• Too many permutations:-  There are more than 50 different commercially available antibiotics, and many hundreds of microrganisms identifiable on MALDI-TOF. So the number of antibiotic:microbe combinations is well in to the thousands. So which ones should be measured? The obvious ones are those that are commonly encountered and used, e.g susceptibility of E. coli to trimethoprim, ones that are clinically very important, e.g. susceptibility of Streptoccus pneumoniae to penicillin, or those of great public health importance, e.g. susceptibility of E. coli to meropenem. The ones to be avoided is where the combination falls outside these groups, particularly those where the numbers seen are insufficent to get meaningful data. e.g. Selenomonas spp. susceptibility to ciprofloxacin.  The important thing here is to decide on the microbe:antibiotic combinations to be measured before you start your surveillance program. Otherwise the data is open to exploitation, with people picking antimicrobial resistance surveillance data to suit their particular agenda. e.g. Antibiotic resistance is increasing because Microbe A is becoming more resistant to antibiotic B, and ignore the fact that Microbe C is becoming less resistant to antimicrobial D…
• Different Definitions:- Because there are different antimicrobial testing standards out there, e.g. CLSI, EUCAST, CDS, etc., one person’s definition of resistant may not be the same as anothers… It is very important to ensure everyone has the same “definitions” of what is resistant and what is not resistant, before you even start. Otherwise you are on a hiding to nothing…
• Politics:- Everybody has their own wishes and desires, and it is no different when it comes to measuring antibiotic resistance surveillance. Everybody wants to look at different antibiotic:microbe combinations, use different testing methodologies, present the data differently. This can cause problems with not only the accuracy of the data, but also in getting any surveillance data at all, when multiple laboratories are required to work together. When antibiotic resistance data requires the political co-operation of different countries then the difficulties move onto a whole new level altogether. Despite there being a willingness to work together, getting multi-national agreement on surveillance is a monumental task.
• The goalposts get moved. Every so often the breakpoints get changed, for various reasons, so that an isolate that was once susceptible can become resistant (on paper), and vice versa. This is why using MIC values for surveillance purposes is so important, as it is an objective measurement which has no interpretation applied to it. This facilitates the acquisition of accurate surveillance data over many years.
• Memory is erased. Sometimes when the laboratory information sytem (LIS) in a microbiology laoratory gets changed, a lot of the historical susceptibility data can get lost, either because it is not compatible with the new system, or not thought to be important enough to keep. Although electronic storage of laboratory data has been around for at least 20 years, as far as I am aware many microbiology laboratories do not have 20 years of data, for exactly this reason. It is a very important point to consider when considering a change of LIS.
• Biases:- So many things can lead to bias in the surveillance data… Participation bias, sampling bias, patient cohort bias, testing bias, etc.. The list is virtually endless. All these things need to be considered and corrected for as best as possible when performing antimicrobial resistance surveillance.

So it is not easy, by any stretch of the imagination.

Good antimicrobial stewardship programmes should be based on having sound, standardised and objective baseline antimicrobial resistance data against which any interventions can be audited.

The other big area of surveillance which is essential to antimicrobial stewardship programmes is antimicrobial usage data. This data goes hand in hand with antimicrobial resistance surveillance.

Although it’s easy to talk about increasing antibiotic resistance, it is actually very difficult to measure properly…

Michael