Category Archives: Data Interpretation

“Cohort Bias”

At my laboratory we only test nasal Staphylococcus aureus susceptibility to mupirocin in the following two circumstances:

  • Prior to joint replacement surgery as part of a Staphylococcal decolonisation bundle
  • In patients where the clinical details state recurrent skin infections

In the patients about to get their joints replaced, our nasal Staph aureus resistance rate to mupirocin is 3%. Not surprising really. This is a generally older cohort, less likely to suffer from impetigo and skin boils etc., and thus less likely to have been exposed recently to mupirocin.

In the patients who have recurrent skin infections, our nasal Staph aureus resistance rate to mupirocin is 15%. This is not surprising either. This cohort is generally young, and due to their clinical history are much more likely to have been exposed to a lot of mupirocin. As a cohort, they potentially have a lot of physical contact with each other (in kindergartens, in the school playground, on the sports field, in cinemas, backs of cars etc..,) facilitating cross-transmission.

This is a good demonstration of how much antibiotic resistance can vary, depending on what population you are looking at. 3% is markedly different from 15% and management of these different rates might be very different from an antibiotic stewardship point of view.

It also reflects the difficulties in measuring antibiotic resistance and then how to report such results in a meaningful manner.

We like to simplify things, and have just one result regardless of what biases might be at play. Measuring resistance rates is complicated enough due to the sheer number of microbe-antimicrobial combinations that can be permutated. To add another level of complexity by calculating different values for any one microbe-antimicrobial combination is too much for most of us to handle! 

But sometimes the difference in values between different population cohorts (as demonstrated above) is just too much to be ignored…



I get the occasional anxious phone call from clinicians concerned about the “rising rates ” of trimethoprim resistance to E. coli…

Not being entirely convinced, I did a (20 year) search for E. coli resistance to trimethoprim at my lab, analysing over 2 million isolates, and came up with the following graph.



I couldn’t work out how to insert a trendline into the graph (I am so useless…), but I think you will agree that it is going to be fairly flat.

The antibiotic apocalypse is not arriving in New Zealand anytime soon. In fact the whole concept of “antibiotic resistance” as perceived by the public is horribly generic and oversimplified…

This example above of course is just one microbe/antimicrobial combination out of many hundreds that could have been analysed, but the observation did highlight a couple of things to me:

  • If antibiotic usage is relatively constant in a population over a prolonged period of time, then antimicrobial resistance does not necessarily rise inexorably. (q.e.d.)
  • Always back your claims up with objective data wherever possible. It is the trends which are critical in the surveillance of antibiotic resistance. We are lucky that at my lab we can now search back through 20 years of electronic data. Before 1996 the data was paper based (and likely lost in a basement or incinerated by now!)

If you did a similar exercise for all the possible microbe/anti-microbial combinations (I just might if the Christmas holidays are quiet!), you will find some trends that are upwards, some that are static, and some where the resistance rate is trending downwards.

A bit like Twitter really….

So when someone says to you. “Antibiotic resistance is increasing all the time. In 10 years time, all infections will essentially be untreatable” (I really detest this type of generic, off the cuff, unsubstantiated statement…)

…you should respond with something along the lines of “Exactly which microbe and antimicrobial combination are you talking about?” and “Show me your data…”.

Some infections will be, and already are, untreatable (mostly due to extreme and focused selection pressure), but the chances of a whole bacterial species becoming pan-resistant are remote. There are two main reasons for this. i) Bacteria survive in open systems, and ii) Bacteria need to expend energy to become resistant.

But these are other stories altogether…


Back to Basics

Working on the urine bench always brings to the front of my mind the lack of understanding that some clinical staff have of the basics of microbiology and the commonly used abbreviations – Note that I said some and not all.

We frequently see request forms with statements such as “MSU Bag urine” – which one is it? A MSU or a Paediatric Bag collect? or “MSU” but with clinical details that the patient has a “permanent SPC” – again is it an MSU or a Catheter specimen? or most commonly “MSU” written under tests required which in my opinion is being mixed up with the abbreviation MC&S (short for microscopy, culture and sensitivities). MSU is not a test but a specimen type. It is an abbreviation of Midstream Urine as opposed to other urine types such as CSU or Catheter spec urine, EMU or early morning urine, Aspirated urine or just a random collect or clean catch collect.

This may seem to be a bit pedantic but the type of specimen we receive in the laboratory has a bearing on what and how we report out to the clinician and whether or not we deem it relevant to report sensitivities. For example, a true MSU should have less chance of having cellular and/or bacterial contamination as opposed to a clean catch urine which often has larger numbers of epithelial cells present and often vaginal contamination if collected from female patients. Also, growth from a catheter specimen is less likely to get sensitivities reported due to the fact that the presence of bacteria and/or cells is often reflective of colonisation rather than infection and changing the catheter, if in situ for a longer term, will often be more effective than antibiotic treatment. Again this is where it is important for the laboratory to be given relevant clinical details so we know if the patient is showing signs of systemic involvement in which case antibiotics will be reported, or if the catheter is merely an in/out catheter for collection purposes as opposed to a long-term solution for a tetraplegic.

It is also important that the correct urine type be collected for the right purpose. If a clinician is wanting testing for TB then an MSU is not going to be sufficient and will be rejected by the laboratory. They will need to ensure a full early morning collect is sent through to the lab so that it can be further concentrated to optimise the chances of isolating any Mycobacteria present. Likewise, random urine collects, although adequate for screening purposes is not the preferred specimen for accurate biochemical dipstick testing or bacterial isolation due to the fact that the potential exists for dilution if the patient has recently consumed fluids.

All this information assists us as laboratory workers to perform our job to the best of our ability and to put out results that are relevant to patient treatment so it is important for clinical staff to understand the differences and to ask themselves what they are wanting to achieve from their request from the laboratory and then together we can maximise the outcomes for patients.