A few weeks ago in LID this marvellous paper, clearly demonstrated the reduction of fluoroquinolone-resistant but not fluoroquinolone-susceptible C. diff infections (CDI) in English hospitals, coined as “the English C. diff miracle”. The CDI decline coincided with the reduction of fluoroquinolone use, but also with a period in which “horizontal” infection control measures, such as hand hygiene, were improved. As the latter would be equally effective in preventing transmission of resistant and susceptible strains, the fluoroquinolone reduction was considered causative for the observed reduction. A very simple model tells us that that is not necessarily the case.
“All models are wrong”, you frequently hear at meetings, but here is one that is useful (see). If for anything, it shows the beauty of infectious disease epidemiology. Modellers, like Ben Cooper, are like contemporary artists, as they reduce the complex reality to the bare minimum of essential elements. Building on earlier work, Esther van Kleef constructed a simple model (11 differential equations) describing the dynamics of a susceptible and a resistant bug being transmitted in the hospital and the community and with patients being discharged and admitted. As in real life, the resistant strain had a competitive advantage (is better transmitted) in hospitals and the susceptible one did better in the community. At baseline, the relative fraction of new cases acquired in hospital was 25% and 2.5% for the resistant and sensitive strains respectively. Improving hand hygiene adherence (from 40% at baseline up to 50% after implementation of the intervention) resulted in a sharp decline of hospital-acquired resistant cases, but not of incidences with hospital-acquired sensitive bacteria. These trends exactly mimic the real life observations reported in LID.
That’s called counter-intuitive, and fortunately the model helps to explain it: If two strains compete unequally in two environments, a transmission-reducing intervention that preferentially targets one environment will have a disproportionate effect on the strain better adapted to that environment. This follows from the fact that the influx of patients into the hospital carrying sensitive bacteria is the dominant factor in maintaining its hospital prevalence, while patient-to-patient spread is largely responsible for maintaining the hospital prevalence of resistant bacteria.
This does NOT prove that reductions in fluoroquinolone use did not cause the observed decline in (fluoroquinolone-resistant) CDI. It simply demonstrates that the alternative fact (horizontal infection control interventions did not contribute) CANNOT be ruled out. A lovely example of the beautiful complexity of infectious disease epidemiology. Miracles do happen and sometimes we just don’t know why.
PS: The paper is now open for peer review, and the authors may have overlooked a fatal flaw.