**Statistical analysis**

The numbers of admissions and patient-days were collected to calculate incidence rates. The incidence rate for each hospital was calculated as the number of candidemias per 1,000 admissions, whereas the overall incidence was determined using summed denominators of patient-days and admissions to calculate pooled mean rates. The data generated during the year of the surveillance on the different risk factors, underlying diseases, morbidity and mortality were recorded in a Microsoft Access 2003 (Microsoft Corporation, Redmond, WA) based case report database. Categorical data were analyzed using Chi-square or Fisher's exact tests as appropriate, and continuous variables were compared using the t-test or Wilcoxon test according to the significance of the normality test. Spearman rank-order correlation was used to measure the relationship between the MICs of fluconazole and voriconazole. We performed univariate and multivariate analysis of factors associated with candidemia caused by isolates with decreased susceptibility to fluconazole. Variables significant at *p-*values of less than 0.05 by univariate analysis were included in a multivariate model using a repeated measures logistic regression model (backward and forward). Data were analyzed using the SPSS 11.0.1 software (SPSS, Inc. Chicago, IL) and Stata 8.0 (Stata Corporation, Lenexa, TX).
