Pediatric Delirium and Associated Risk Factors
Normality tests were first performed to assess whether continuous covariates were normally distributed. If covariates were normally distributed, t tests were used. For covariates not normally distributed, the nonparametric Wilcoxon rank-sum test was used to compare the median differences of covariates by delirium status (yes or no). For discrete covariates, the chi-square test and Fisher exact test were used to compare the frequencies/proportion of covariates by delirium status. Multivariable logistic regression was performed to evaluate the independent associations between potential confounding factors and risk factors with delirium status. Any bivariate association that achieved a p value of less than 0.2 was entered into the multivariate model. The odds ratios (OR), 95% CIs, and p values of the covariates were reported. In order to correct for more than one delirium diagnosis within some individuals, generalized estimating equation (GEE) analysis was performed to determine if results obtained using the standard logistic regression analysis materially changed. All statistical tests were two-sided, and p value of less than 0.05 was considered statistically significant. All analyses were performed in SAS version 9.3 (SAS Institute, Cary, NC) and figures generated in STATA 13 (StataCorp LP, College Station, TX).
Statistical Methods
Normality tests were first performed to assess whether continuous covariates were normally distributed. If covariates were normally distributed, t tests were used. For covariates not normally distributed, the nonparametric Wilcoxon rank-sum test was used to compare the median differences of covariates by delirium status (yes or no). For discrete covariates, the chi-square test and Fisher exact test were used to compare the frequencies/proportion of covariates by delirium status. Multivariable logistic regression was performed to evaluate the independent associations between potential confounding factors and risk factors with delirium status. Any bivariate association that achieved a p value of less than 0.2 was entered into the multivariate model. The odds ratios (OR), 95% CIs, and p values of the covariates were reported. In order to correct for more than one delirium diagnosis within some individuals, generalized estimating equation (GEE) analysis was performed to determine if results obtained using the standard logistic regression analysis materially changed. All statistical tests were two-sided, and p value of less than 0.05 was considered statistically significant. All analyses were performed in SAS version 9.3 (SAS Institute, Cary, NC) and figures generated in STATA 13 (StataCorp LP, College Station, TX).