Statistical Analysis
We summarized the demographic data among CF hospitalizations by age, sex, race, geographic region, payer, hospital location and bed size and compared those with and without a co-occurring diagnosis of CDI. Continuous variables were summarized using means and standard deviations while categorical variables were expressed in proportions. We then compared demographic and clinical characteristics between patient admissions for CF with and without CDI using non-parametric bivariate tests as appropriate.
Next, we summarized outcomes of in-hospital mortality, LOS, and healthcare expenditures overall and by presence or absence of CDI using bivariate statistics. We then fit multivariate models to determine the independent associations between CDI and study outcomes, adjusting for variables associated with CDI at p<0.1 on bivariate analyses (calendar year, sex, payer and hospital location/teaching status). For mortality, we fit logistic regression models to estimate odds ratios and 95% confidence intervals. To accommodate the skewed distributions for outcomes of LOS and hospital charges, we used general linear models with gamma distribution and logarithmic transformation. We reported on the percentage change from the referent along with 95% confidence interval.
We next evaluated trends in the proportion of CF hospitalizations with co-existing C. difficile over time between the years 1997 to 2016 using the chi-square test of trend.
All analyses were conducted using SAS 9.4 (Cary, North Carolina).