Clinical impact of implementing humidified high-flow nasal cannula on interhospital transport among children admitted to a PICU with respiratory distress: a cohort study

Table 3 Sensitivity analysis for the primary test

	Analysis method	n	Estimated effect		p
	Analysis method	n	Ratio	(95% CI)	p
	Final model	3022	0.64	(0.49–0.84)	0.001
a	Bandwidth 80% (including 2011–2018)	2554	0.66	(0.49–0.88)	0.01
b	Bandwidth 60% (including 2012–2017)	2046	0.74	(0.52–1.04)	0.08
c	Washout-period (excluding 2014–2015)	2253	0.53	(0.34–0.82)	0.004
d	120 monthly time periods	3022	0.67	(0.52–0.86)	0.002
e	Consistent severity (excluding low PIM-2)	2274	0.73	(0.53–1.00)	0.049
f	Additional adjustment for severity score (PIM-2)	3022	0.66	(0.51–0.86)	0.002
g	Model (using matched cohort)	1978	0.78	(0.66–0.92)	0.004
h	Model (using discontinuity regression for transport)	1006	0.70	(0.51–0.95)	0.02
i	Uncensored outcome	3022	0.65	(0.50–0.84)	0.001
j	Excluding extremely influential observations	2787	0.66	(0.53–0.83)	< 0.001

CI confidence interval, PIM paediatric index of mortality
The sensitivity analysis was performed to assess the robustness and resistance of the final model which was designed to obtained the estimated effect by the implementation of humified high-flow nasal cannula on interhospital transport on the primary outcome (length of stay in the intensive care unit)
In the final model, a multivariable linear regression with the log-transformed outcome was used. The study period was Jan2010-Dec2019
^abIn these analyses, children were included for the shorter study duration than the original 10-year study period
^cIn this analysis, admissions from Jan2014-Dec2015 were excluded
^dIn this analysis, instead of the yearly time period, the monthly time period was used with indicator variables for each month
^eIn this analysis, children with 25 percentile or lower severity score were excluded
^fIn this analysis, the model was expanded with an additional severity score
^gIn this analysis, matched cohorts between transported and non-transport children were generated and then the difference in differences between pre- and post-intervention era in two cohorts were calculated
^hIn this analysis, only transported children were included while the model was same as the final model
ⁱIn this analysis, the uncensored outcome was used
^jIn this analysis, we excluded 235 extremely influential observations based on the Difference in Fits (DIFFITS)

ISSN: 1364-8535