DISCUSSION
In this retrospective cohort study of all respiratory cultures from
children with tracheostomies, cultures obtained in children with a
diagnosis of ARI had a 29% higher odds of overall organism isolation
compared to children without ARI. Although many organism species were
isolated in both groups, ARI diagnosis was only associated with
isolation of 4 specific organisms (H. influenzae , M.
catarrhalis , S. pyogenes , and S. pneumoniae) , all of
which are expected in usual bacterial respiratory infections. After
controlling for patient-level factors and serial culture sampling,
isolation of P. aeruginosa was not associated with ARI diagnosis,
suggesting this organism may not be considered causative of ARI and may
not require acute treatment. Furthermore, respiratory cultures performed
poorly as a screening test for bARI, with low sensitivity and moderate
specificity. Taken together, our data suggest that the utility of
respiratory culture testing to diagnose, direct treatment for, or even
screen for ARI and bARI among children with tracheostomies is limited.
The clinical interpretation of respiratory cultures in children with
tracheostomies is highly complex, both during states with ARI and
without ARI, and is becoming more nuanced as epidemiology
evolves.7 The respiratory tract is not a sterile site
and many factors influence the dynamic composition of oropharyngeal
flora in children with and without artificial
airways.8,17,18 Chronic bacterial colonization of the
respiratory tract is a presumed entity, but is difficult to define and
identify in practice.19-21 Emerging studies are
illuminating the complex interaction of the respiratory tract flora with
viruses (i.e., Haemophilus and Moraxella “blooms”),
colonizing bacteria, inflammatory biomarkers, and the microbiome-host
interplay.10,22-26 Thus the traditional, reductionist
notion of attributing an ARI to a single bacteria, virus, or even single
group of pathogens is increasingly recognized as
inadequate.9,27
In our study, respiratory cultures obtained during ARI had higher
adjusted odds of identifying a bacterial organism, a finding which seems
to support some role of respiratory culture testing in diagnosing ARI.
Presumably, the identified organism is a pathogenic cause of ARI in many
of these circumstances, though the nuanced reasons stated above and
retrospective studies in related populations would argue the isolation
of some organisms may still be circumstantial.11,19,28Despite this seemingly straightforward association, this study
highlights other perplexing and contradictory epidemiologic culture
findings. One the one hand, the majority of respiratory cultures in
children with ARI (76.9%) in our study were negative; such a finding
could be the result of multiple factors including an underlying viral
etiology of ARI in many of these children (who would not be expected to
have positive bacterial cultures), symptomatic microbiome
shifts27,29 or bacterial
“blooms”,10 inadequate culture sampling, or a
generally low yield of the respiratory culture test. In contrast, a
number of positive respiratory cultures were identified among children
without ARI; in the absence of a clinician diagnosis of ARI, positive
cultures may indicate clinician interpretation of chronic bacterial
colonization of the respiratory tract.19,20,30,31These contradictory culture findings suggest that respiratory culture
positivity alone has unclear diagnostic yield in differentiating acute
bacterial infection, viral infection, and chronic colonization.
During ARI in children with tracheostomies, clinicians may use
respiratory culture results to inform antibiotic prescribing decisions
including continuation, modification, or discontinuation of therapy.
Increased respiratory culture acquisition has been associated with
increased antibiotic use among children’s
hospitals.32,33 In our study only 4 species remained
associated with ARI on adjusted, multinomial analysis. These species,H. influenzae , M. catarrhalis , S. pyogenes , andS. pneumoniae , are all expected organisms implicated in usual
bacterial respiratory infections among other
populations.34 Evidence-based treatment guidelines
typically recommend empiric antibiotic therapy directed at these
pathogens in uncomplicated cases, without need for
organism-identification testing. The association of ARI with this group
of organisms challenges the utility of respiratory cultures in directing
antibiotic treatment among children with tracheostomies.
Adjusted analysis demonstrated that ARI diagnosis was not associated
with isolation of P. aeruginosa and S. aureus . Isolation
of these two species is thus more likely related to patient factors
(e.g., ventilator dependence, time with tracheostomy tube), and
recurrent sampling, as opposed to true ARI. With an equal likelihood of
identification both with and without ARI, our findings imply that these
organisms may not be causative of ARI but instead be consistent with
chronic colonization. Given the weak association of P. aeruginosaidentification with bARI diagnosis, there may be some clinician
interpretation of this organism as additionally disease-causing in
certain circumstances. Both P. aeruginosa and S. aureusare known to cause biofilm formation of airway devices, a fact which may
or may not be the same as airway
colonization.30,31,35,36 This study questions the
necessity for reflexive anti-pseudomonal and anti-staphylococcal
antibiotic treatment of positive cultures in some clinical situations,
such as in the absence of infectious symptoms or when viral
mono-infection is suspected. How these potentially colonizing airway
bacteria may contribute to ARI susceptibility and frequency is not
known.27
As a screening test for bARI, the respiratory culture overall
demonstrated unfavorable test characteristics.37Respiratory cultures correctly identified only 27.2% of children
diagnosed by clinicians with bARI (sensitivity), and had only a 32.9%
probability of identifying children with bARI (PPV) with a high false
positive rate. The culture had a somewhat better probability of
identifying children without bARI (85.4% specificity, 81.7%
NPV).37 In this case in which both false positive and
false negative test results are common, and both are associated with
measurable harm including antibiotic overuse, these test statistics are
unsatisfactory. For the clinician at bedside, the positive and negative
likelihood ratios for respiratory cultures here are modest at best, and
suggest respiratory cultures are not particularly clinically useful as
bARI screening tests. This is corroborated by recent evidence of poor
concordance between tracheal aspirate and BAL
cultures.38
Further evaluation of treatment approaches and clinical outcomes for
both positive and negative cultures in relation to ARI diagnosis are
needed. In the absence of clinical guidelines to direct clinician
ordering or interpretation of respiratory cultures, clinicians order and
interpret respiratory cultures in different ways.5,11Clinicians may interpret respiratory culture results in the context of
the individual patient, the clinician’s experience, institutional norms,
and knowledge of local microbiology lab practices. However, studies have
demonstrated wide inter-institutional variability in culture ordering
practices, microbiology lab processing, culture results (both positivity
rate and organism type), and clinician response to
results.5,7,39 Variability in sample quality and lab
processing may also contribute, as has been documented in endotracheal
tube cultures.39 The consequence of this variability
is clinical inconsistency in diagnosing ARI,5 use of
and types of antibiotic treatment,32 and probably also
patient outcomes including severity, length of illness, and respiratory
support needs, all of which limit research and improvement efforts.
This study has several limitations. The use of diagnostic codes to
define ARI and bARI predictor status could lead to misclassification of
predictor for children with and without these conditions. Although we
restricted codes to those most consistent with true infection, it is
possible that cultures were incorrectly classified. Furthermore, the
association of positive bacterial cultures with ARI cannot imply
bacterial causation of these infections; our retrospective study was not
designed to explore causation of respiratory bacteria in ARI. The
retrospective design of this study also creates the potential for
residual confounding; there may be other clinical or demographic factors
influencing respiratory culture acquisition and interpretation that we
are unable to capture with our discrete dataset. Furthermore, our
center’s results may not be generalizable to children with
tracheostomies at other institutions. Our institution’s positive culture
prevalence is lower than that observed at other
institutions;15,40 this is consistent with prior
internal studies, and is hypothesized to be related to differences in
our population and/or local factors (e.g. infection control policies,
lab reporting procedures).