Statistical analysis

Software STATA IC/12.1 was used for statistical analysis. Significance of association between categorical variables was tested by means of Fisher exact test. To demonstrate HCWs' attitudinal differences toward vaccination, non parametric tests were used to compare quantitative variables between categories (Wilcoxon's and Kruskall-Wallis' test) and to evaluate the significance of their trend (Cuzick's test), in particular considering the immunization against influenza.

To test the mediation effect in the relationship between one psychometric variable and the binary variable measuring active immunization against influenza, the STATA function ‘medeff’ was used.⁴⁹ An OLS regression model estimated the coefficients in the paths between continuous predictor and continuous mediators and a probit regression estimated the coefficients between continuous predictor or mediators and the binary outcome. The pathways are described as Average Causal Mediation Effect (ACME, or ‘indirect effect’), as Average Direct Effect (ADE, or ‘direct effect’) both expressed also as percentage of ‘Total effect’.