Bacteremia was defined as growth of any pathogen Vorinostat clinical in the blood culture. The isolation of coagulase-negative staphylococci from the blood culture was considered to indicate contamination and thus absence of bacteremia.Statistical analysisDescriptive analysis included means or percentages with 95% confidence intervals (CIs) or medians and ranges, as appropriate. Missing values of categorical variables were considered to indicate the absence of that characteristic. This was applied for shaking chills (n = 66) and costovertebral tenderness (n = 18). Univariate analysis was performed using the Student’s t-test or Mann-Whitney U test for continuous variables and Chi-square tests for categorical variables. Covariates found to be associated with bacteremia on univariate analysis at a level of significance P < 0.

2 were eligible for inclusion in a multivariate logistic regression model using a backward selection procedure [21]. Measures for association were expressed as odds ratios (ORs) for disease with their 95% CIs for categorical variables. We tested the following three models: 1). A clinical model including clinical variables only; 2). A clinical model added with the PCT value; 3). A model based on PCT only. The predicted probabilities of bacteremia (Pbac) in any patient for the different models were calculated by using the following regression equation: ln (Pbac /(1- Pbac)) = intercept + ��-coefficient * variable, where the intercept and ��-coefficient are obtained from logistic regression analysis.

We constructed receiver operating characteristic (ROC)-curves for the different models using Pbac as the test variable and bacteremia (yes/no) as state variable. The discriminative power and the diagnostic performance of the prediction models were compared by calculating the area under the curve (AUC) of the ROC-curve and by Nagelkerke’s R2. In addition, for the clinical models, based on the ��-coefficient, points were assigned for each predictor and different cutoff values were used to calculate corresponding sensitivity, specificity, positive and negative predictive values (PPV, NPV) and likelihood ratios for predicting bacteremia were calculated. For PCT, different cutoff values were tested, according to the instructions by the manufacturer for diagnosis of bacterial sepsis or lower respiratory tract infection; the cutoff value corresponding with a sensitivity of 95% and highest specificity was chosen for further analysis.

A P-value < 0.05 was considered indicative for statistical significance. SPSS software (SPSS Inc., Chicago, Ill, USA; version 17.0) was used for statistical analysis.ResultsPatient characteristics and microbiological resultsOf 728 patients screened for eligibility, 642 met the inclusion criteria and were included in the study of which 581 were evaluable with concurrent blood cultures Dacomitinib and PCT measurements at baseline.