Statistical analysis
Continuous variables were described using medians and interquartile ranges (IQR). Categorical variables were described using counts and percentages. Patient characteristics and outcomes were stratified by primary exposure and compared using Chi-square test or Fisher’s exact test for categorical variables as appropriate and Wilcoxon rank sum test for continuous variables. Generalized linear models were used for comparisons in child-level count data while generalized linear mixed models were used for comparisons in culture-level data to account for the correlation within child clusters.
Poisson regression models with robust standard errors were used to examine the association between each predictor group (of dichotomous culture status) and children’s 3-year outcomes, adjusted for covariates including sex, race, insurance type, and number of complex chronic conditions and accounting for correlation within child clusters. Kaplan-Meier survival analyses with log-rank test were used to assess death-free survival. Race was included in analysis due to known racial differences in outcomes among infants with BPD,23 and as a proxy for structural racism and bias.
Analyses were performed with SAS V9.4 (SAS Institute Inc., Cary, NC, USA) and with R v4.1.1 (Vienna, Austria).24P -values <0.05 were considered statistically significant.