Limitations:
We acknowledge several limitations in our study. Administrative data
sources may contain miscoded or inaccurate information. For example, the
use of ICD-9 code to study CDAD may potentially misclassify CDAD if
physicians or hospital coders did not code for this diagnosis or if test
results were not back at the time of hospital discharge. However, the
ICD-9 code (ICD-9-CM 008.45) for C. difficile has been described
and validated previously with 78% sensitivity and 99.7% specificity
for correctly identifying and classifying admissions.
An administrative definition of C. difficile may also include
some patients who had colonization rather than true disease; however,
patients in our study were ill enough to require hospitalization, and so
this proportion may likely be smaller. In addition, it is a
well-established fact that CF patients have high rates of colonization,
and therefore testing may be limited to cases where there is only a high
index of suspicion. On the other hand, given CDI’s atypical presentation
in CF with lack of diarrhea, there are chances that this diagnosis may
either have been overlooked, or a formed stool sample may have been
rejected by the lab, causing an underestimate of the disease burden.
Administrative databases also offer limited ability to adjust for the
severity of underlying CF disease or use of concomitant medications
(immunosuppressive drugs, antibiotics, CFTR modulators, mucolytics). We
recognize the possibility of detection bias, with patients with more
severe disease being more likely to be tested for C. difficile .
The possibility of this bias (i.e., more frequent testing for C.
difficile among those hospitalized patients with more severe disease)
playing a role in our results cannot be completely ruled out as an
explanation for the increasing disparity in outcomes of hospitalized CF
patients with and without C. difficile infection. We are also not
able to determine time of onset of C. difficile infection
relative to the hospitalization.