Two months ago I blogged on the reduction of fluoroquinolone-resistant but not fluoroquinolone-susceptible C. diff infections (CDI) in English hospitals, coined as “the English C. diff miracle”. A very simple model challenged the conclusion that this was caused by a reduction in fluoroquinolone use in hospitals. A debate on the assumptions of the model now provides the perfect outline for a PhD thesis.

The model and results were published on a preprint server and as letter to the editor in LID, allowing the authors to respond appropriately. Such exchanges of ideas sometimes read as a pro-con debate, or in boxing terms “float like a butterfly, sting like a bee”. In this case the floating is emphasizing the real-life complexity of C. diff epidemiology and the stings are directed towards some of the model assumptions, leading to the – well-known – conclusion that “all models are wrong, but some are useful” (We need a new quote). Now, let this debate be a starting point, rather than a final round.

Indeed, model assumptions are – by definition – things that may be wrong, but – (un)fortunately – nobody knows for sure. Some assumptions are equally wrong as irrelevant for the model results (indicating that even a seemingly simple model can be unnecessarily complex), but some address fundamental epidemiological characteristics. Elucidating these would enormously enhance our understanding of infectious diseases epidemiology and infection control.

Three of these are:

1.    Do fluoroquinolone-susceptible strains have a transmission advantage in the community (compared to fluoroquinolone-resistant strains)? This relates to the beauty of co-existence (see here). In fact, the biological observation is that resistant strains fail to completely replace susceptible ones, but how they interact/compete remains to be determined.
2.    Are infections caused by fluoroquinolone-resistant and fluoroquinolone-susceptible strains mutually exclusive? The answer is probably NO, but any protective effect of carriage with susceptible bacteria against infections with resistant strains could majorly change epidemiology. Imagine, that we could protect humans against infections with resistant bacteria through intentional colonization with susceptible strains.
3.    Do fluoroquinolone-resistant and fluoroquinolone-susceptible strains differ in their duration of asymptomatic carriage in hospital and community settings? Needed to quantify the relevance of the feedback loop: resistant bacteria acquired in the hospital, persisting after discharge and reintroduced in the hospital with readmission.

Each question seems very easy to answer. So, go ahead.

If you’re not interested in CDI (understandable), just replace it by MRSA, ESBL, CRE, CPE, E. coli, K. pneumoniae, mcr-1, etc, etc.

And for those trying to criticize a mathematical model with a catchy quote, think of this one: “Describing complex, poorly understood reality with a complex, poorly-understood model is not progress” (J. Maynard Smith, 1920-2004).