Discordant Clostridioides difficile diagnostic assay and treatment practice: a cross-sectional study in a tertiary care hospital, Geneva, Switzerland

The decision was made to include all eligible patients and no formal sample size calculation was performed. Instead, we restricted the number of investigated parameters before any confirmatory analysis. Based on the "10 events per variable" rule of thumb, we limited the number of parameters investigated to eight factors selected among known risk factors and clinical characteristics compatible with CDI. Patient characteristics and CDI risk factors were described overall and by treatment for CDI and reported as frequencies and percentages. A multivariate logistic regression model using a backward stepwise method was performed to determine which parameters were independently associated with CDI treatment. At each step, starting from all eight parameters, the variable with the highest p-value on the likelihood ratio test was removed from the model until all remaining factors were statistically significantly associated with CDI treatment at a two-sided level of 5%. Sensitivity analyses were performed to assess the robustness of the results when deceased patients were a) excluded from the analysis and b) considered as not treated. Missing data were systematically removed from analyses. Statistical significance was assessed at a two-sided 0.05 level for all analyses. All statistical analyses were performed using Stata software, version 15 (StataCorp, College Station, TX).