Results
There were 318 patients who met inclusion. The mean time for follow-up
was 2.3 years (range = 0.1 to 7.8 years). The final status at the end of
the follow-up period was N=140 (44%) alive with a tracheostomy in
place, N=96 (30%) decannulated, N=61 (19%) died with a tracheostomy in
place, and N=21 (6.6%) lost to follow up.
Tracheostomies were placed at a median age of 6.9 months (IQR: 4.1 –
49.2 months). The population was N=170 (53%) male and N=148 (47%)
female. There were N=175 (55%) White, N=105 (33%) Black, N=15 (4.7%)
Asian, and N=23 (2.7%) other. There were N=96 (30%) who identified as
Hispanic. The preferred language was predominately English (N= 270,
86%) followed by Spanish (N=39, 12%). For further details, please
refer to Table 1 .
Socioeconomic findings for the population were based on zip-code and
county level measurements obtained from the Opportunity Insights’
databases on Social Capital and Opportunity Indices. The median (IQR)
range for the economic connectiveness ratio was 0.73 (0.58-0.92); the
support ratio was 0.85 (0.81-0.94); the fraction of single parents was
0.37 (0.28 – 0.38), and median household income was $53,442 ($49,529
– $62,463).
Table 2 shows associated conditions for the patient population.
Notable findings include 41% (N=121) short gestation; N=72 (23%)
having a history of respiratory distress syndrome; and N=133 (42%)
being diagnosed with pulmonary hypertension. Other associated conditions
included cardiac conditions (42%), birth hypoxia (11%), and
tracheobronchomalacia (7.5%).
The median length of stay for the index stay when the tracheostomy was
placed was 101 days (IQR: 52-166 days). The total number of patients
discharged on a ventilator was N=272 (86%). The median time to
ventilator liberation and eventual decannulation was 2.3 and 1.9 years,
respectively. The median time to death for patient who expired with a
tracheostomy in place was 0.70 years. The level of neurocognitive
disability was considered severe in N=170 (59%) of children.
The number of children with BPD was N=136 (43%). Almost all were
discharged on a ventilator. There were several statistically significant
differences between children with tracheostomies with and without BPD.
Children with BPD tended to be younger at the time of tracheostomy
placement (5.2 months vs. 24.5 months, P <.001), have
shorter gestational ages (28 weeks vs. 38 weeks,P <.001), and weighed less at tracheostomy placement
(1.1 vs. 2.9 kg, P <.001). A higher proportion were
Black (44% vs. 25%, P =.004). The level of social support by
neighborhood level was slightly decreased among children with BPD (0.83
vs. 0.86, P =.021). Severe disabilities such as non-ambulant
cerebral palsy or profound sensorineural hearing loss were also more
common among those with BPD (63% vs. 57%, P =.003). See Table 3
for additional details.
Unadjusted estimates of time to mechanical ventilator liberation for
children with BPD, compared to non-BPD children, showed a hazard ratio
of 0.81 (95%, CI 0.63 – 1.04, P =.10). A confounder-adjusted
estimation also did not find any variable associated with time to
ventilator liberation. Additional analysis showed children with BPD
spent a median of 2.92 years (IQR 1.60 – 4.03, P =.003) on
mechanical ventilation compared to 1.84 years (IQR 0.84 – 3.60,P =.003) for children without BPD. See Table 4 for
further details.
Unadjusted estimation of time to decannulation for children with BPD
yielded HR=0.92 (95% CI, 0.62 – 1.38). The adjusted model found the
time on a ventilator interacted with BPD (see Table 5 ). The
model suggests that children with BPD on a ventilator are more likely to
take longer to decannulate than a child on a ventilator who does not
have BPD. Using the 25th, 50th, and
75th percentiles for time to decannulation, the median
times to decannulation would be 2.6, 3.2, and 4.5 years for BPD and 1.8,
3.1, and 5.8 years for non-BPD children. The model’s goodness-of-fit is
presented in Figure 1 . In addition, statistical significance
held when using jackknife regression variance components estimation for
model validation suggesting that outliers did not have a substantial
impact on regression coefficients.
The unadjusted survival analysis for time to mortality found that the HR
for BPD was 0.58 (95% CI, 0.34 – 0.98). The adjusted model includes
pulmonary hypertension as a variable of significance (aHR = 2.5, 95%
CI: 1.43 – 4.40) while reducing the confidence interval of the effects
of BPD on mortality (aHR = 0.38, 95% CI: 0.22 – 0.69). The model did
not reveal a statistical interaction between the two variables
suggestive of an additive effect (Table 6) . Goodness-of-fit
plots (Figure 2 ) and jackknife regression of the model
suggested a good fit. Additional subgroup or sensitivity analyses were
not performed.