Statistical Analysis
QI macros add-in for Excel 2020.01 (KnowWare International, Denver,
Colorado) was used to generate the statistical process control charts of
the outcome measures. To adjust for the seasonal variation impacting the
number of patients in the PICU with critical asthma, subjects were
divided into groups of 10. The upper control limit and lower control
limit were calculated as 3 standard deviations above and below the
center line. We considered 8 consecutive points above or below the
center line to represent a special cause variation, prompting a change
in the center line. Subject demographics, clinical characteristics, and
balancing measures were compared among five groups of subjects: the
pre-intervention, initial post-implementation (PDSA 1), inclusion of
patients on continuous albuterol (PDSA 2), rate wean increment change
(PDSA 3), and after the HFNC holiday (PDSA 4) using Kruskal-Wallis tests
for continuous variables and chi square tests for categorical variables.
A multivariable linear regression model was constructed attempting to
control for known confounders (race, sex, intermittent magnesium doses,
and PRISM-III score) a priori to determine our interventions impacts on
HFNC duration. Statistical analyses of the subjects’ characteristics
were performed using Stata17. A cutoff P value of <0.05 was
considered statistically significant. The Standards for Quality
Improvement Reporting Excellence (SQUIRE) 2.0 Guidelines (22) were
followed during the preparation of this manuscript (supplement table 1).