Discussion
In this single-center retrospective cohort study of infants and young
children with BPD and tracheostomies, identification of pathogenic
respiratory bacteria including P. aeruginosa on respiratory
culture was not associated with continued need for any respiratory
support (e.g., oxygen, positive pressure) at 3 years post-tracheostomy
placement. However, P. aeruginosa identification was associated
with decreased probability of decannulation by 3 years. Furthermore, the
small subpopulation of children with chronic P. aeruginosa had a
distinctly lower survival probability as compared with children with no
bacterial isolation or only sporadic P. aeruginosa isolation, an
association that was not explained by differences in medical complexity.
Our findings suggest that respiratory tract P. aeruginosa , when
identified sporadically, may not influence long-term respiratory
dysfunction, but does delay decannulation. Findings also raise the
question of whether chronic P. aeruginosa is inherently harmful
or if it is representative of other important sequelae driving
mortality.
The lack of association between presence of respiratory bacteria,
including P. aeruginosa , and continued use of respiratory support
at 36 months post-tracheostomy placement suggests that identification of
pathogenic bacteria, whether during ARI or during surveillance testing,
may not directly cause or propagate respiratory dysfunction. This is in
contrast to what is known among other populations (i.e., cystic
fibrosis, COPD), in which such pathogenic respiratory bacteria has been
demonstrated to decrease lung function over time, as directly measured
by pulmonary function testing and indirectly measured by disease
exacerbation risk.14-18,25-30 This lack of association
with ongoing respiratory support among children with tracheostomies is
particularly important because it was identified in spite of the fact
that children with pathogen identification had far more cultures
obtained than their peers without pathogen identification. In additional
analysis of what factors do drive untimely respiratory support weaning,
not surprisingly, a child’s number of complex chronic conditions was
associated with continued need for respiratory support in this cohort.
While this finding has been described elsewhere,31 the
interaction between complexity and pathogenic organism identification
warrants further study, as does the role of bacterial quantity and
timing in relation to outcomes.
In contrast, children with any identification of P. aeruginosahad only half the probability of achieving decannulation by 3 years
post-tracheostomy, suggesting delayed time to decannulation. This effect
was additive, with each instance of P. aeruginosa identification
decreasing the associated decannulation probability by 35%. This
important association was not observed for the collective predictor of
any respiratory pathogen identification, but instead was specific toP. aeruginosa .
In the setting of P. aeruginosa isolation specifically, we
hypothesize that children with P. aeruginosa identification may
be more prone to chronic respiratory symptoms, and therefore be poor
candidates for decannulation. This group may also have more frequent
culture testing obtained, which, if cultures are positive and are being
treated with antimicrobials, could lead to delays. It is also possible
that children with other respiratory causes for delayed time to
decannulation, such as comorbid airway malacia, obstructive sleep apnea,
subglottic stenosis, or other anatomical considerations are more likely
to have P. aeruginosa identified on respiratory cultures. There
is some limited evidence of an association between malacia and stenosis
with non-pseudomonal respiratory bacteria in other pediatric
populations, but it is not clear if P. aeruginosa exerts a direct
or indirect influence on airway anatomy, or if this effect would be
enough to delay decannulation or lead to a requirement for surgical
airway reconstruction.32-34
Furthermore, clinicians may perceive an elevated risk of future
respiratory illness due to P. aeruginosa and thus hesitate to
swiftly decannulate following liberation from respiratory support. Our
prior analysis from this cohort identified that P. aeruginosaisolation is not associated with episodes of clinician-diagnosed ARI
when controlling for repeated testing and
colonization;21 thus the decreased probability of
achieving decannulation with P. aeruginosa is not likely
explained by direct delays from culture-positive pseudomonal ARI
diagnoses or treatments. However, the risk of future ARIs after priorP. aeruginosa is unknown. Similarly, clinicians might perceive a
risk of respiratory dysfunction related to this organism (as is
well-documented in cystic fibrosis, but not supported by our data here)
and therefore proceed with caution towards decannulation.
In addition to potential respiratory-related reasons for delayed
decannulation, active non-respiratory comorbidities could increase a
child’s risk for P. aeruginosa , as has been demonstrated in
adults with chronic disease,35-37 and may
simultaneously de-prioritize or delay decannulation proceedings.
Although we controlled for comorbid conditions broadly in the form of
number of CCC categories in analysis, specific relevant comorbidities
may be unaccounted for. There may also be important social, family, or
healthcare factors influencing time to decannulation that do not affect
time to wean respiratory support. At our institution, steps to wean
respiratory support can be advanced in the outpatient or telemedicine
settings. However, some steps for decannulation readiness evaluations
necessitate hospital-based care (e.g., sleep studies, airway
evaluations), which is more difficult and time-intensive to coordinate
and requires children be illness-free for longer periods.
Although the prevalence of chronic P. aeruginosa identification
in our cohort was very low – lower than reported
elsewhere19,20 – the 50% mortality rate identified
in this group over the 3 years post-tracheostomy placement warrants
attention. Whether this high mortality is a direct effect of P.
aeruginosa itself or an indirect effect is unclear from our data, which
lacks causes of death. Respiratory infections and antibiotic use are
both associated with respiratory and gastrointestinal dysbiosis, which
in turn each cause diminished immune system functioning and increased
risk for further acute infection of various types, including ARI, among
both murine models and humans.38-41 It is furthermore
possible that this group of children died of pseudomonal ARIs occurring
after pseudomonal colonization; increased frequency of purulent
tracheobronchitis and pneumonia have been documented among adults with
tracheostomies who have tracheobronchial Gram-negative organism
colonization.42 On the other hand, children who are at
risk for chronic P. aeruginosa due to immunocompromised status,
comorbidities, or high hospital utilization are also likely at higher
risk for mortality related to these non-pseudomonal factors, as has been
reported in other populations.37,43-47
This study demonstrated a similar overall mortality rate to that
reported in other studies of infants and young children with severe BPD
and tracheostomies.9 In survival analysis, we
identified that most deaths occurred in the first 1 year following
tracheostomy placement. In our adjusted analyses, children of non-White,
non-Black race had a nearly threefold higher risk of 3-year mortality
across predictor groups as compared with white peers; notably, the other
covariates of sex, insurance type, and CCCs included in this model did
not reach statistical significance. This racial disparity in such an
important outcome raises concern for inequitable access to care,
barriers to care utilization, inadequate cultural sensitivity or
accommodations, or other socioracial factors that could underly this
difference. This disparity follows some of the known racial differences
in premature birth, a key risk factor for development of
BPD.48,49 More distally, Lewis et al. recently
identified a 2-fold increased risk of the shorter-term outcome of NICU
mortality among non-White infants with severe BPD in a multicenter
study;23 there was no racial difference in probability
of tracheostomy placements in that study. Our findings provide some
evidence that these racial mortality differences, likely the result of
racist or biased institutional practices, persist beyond the NICU time
course, but further study into causes and implications of longer-term
outcome differences along racial lines is warranted.
This study has several limitations. Our study did not differentiate
between bacteria identified during acute respiratory illness from
bacteria identified in surveillance culture testing (i.e., during
wellness). Although at our institution, children with tracheostomies are
unlikely to seek medical care outside of CCHMC, it is possible that our
study excluded respiratory cultures obtained from external sites,
leading to potential rare misclassification of predictor status. The
mortality rate of our population was higher than expected; our primary
outcome was defined only for alive children, leading to possible
selection bias for this analysis. The retrospective design of this study
also creates potential for residual confounding, in which other clinical
or demographic factors influencing respiratory pathogen detection and
outcomes are not captured by our dataset. Furthermore, our center’s
results may not be generalizable to other institutions; our
institution’s positive culture prevalence is notably lower than that
observed at other institutions,20,50 but consistent
with our prior internal studies.21,44 This lower
prevalence is hypothesized to be related to differences in our
population and/or local factors (e.g., infection control policies, lab
reporting procedures). Similarly, as noted above, the number of children
meeting chronic P. aeruginosa identification criteria was very
small which limits the conclusions from that subgroup’s analyses.