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 or Fisher’s exact test for categorical variables and Wilcoxon rank sum test for continuous variables.
To examine the independent association between diagnosis of ARI or bARI and isolation of any organism on respiratory cultures, as well as specific isolation of P. aeruginosa and S. aureusisolation (the two most frequently isolated organisms), generalized estimating equations were used including potential confounders while accounting for within-patient clustering. Patient demographics included age at culture collection, time with tracheostomy at culture collection, race, ethnicity, and insurance type; measures of clinical complexity included number of complex chronic conditions, BPD or CLDI diagnoses, and baseline ventilator use; clinical variables included respiratory culture source, as well as location and department sending specimen.
To examine the independent association between diagnosis of ARI and all specific organism isolations simultaneously, a multinomial logistic regression equation was calculated including the above potential confounders, accounting for within-patient clustering, and additionally accounting for culture-level clustering among polymicrobial culture results.
To evaluate the test characteristics of the respiratory culture, sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) were calculated using clinician-coded diagnosis of bARI as the reference standard and indicator of true bacterial disease. Test characteristics were calculated for the respiratory culture overall and by specimen type (i.e., TA, BAL).
Secondary analyses were conducted examining results of each respiratory culture source (TA and BAL) individually. As isolation of P. aeruginosa in both predictor groups may signal these children have chronic colonization, subgroup analysis was conducted examining children with chronic P. aeruginosa isolation. Additional subgroup analysis excluding presumed surveillance BAL cultures without diagnosis data was also conducted.
All analyses were performed with R v4.1.1 (Vienna, Austria).16 P -values <0.05 were considered statistically significant. This study was approved by the CCHMC Institutional Review Board.