Study objective: To determine the sensitivity, specificity, and positive and negative predictive values of blood cultures obtained through a central vein catheter compared with peripheral venipuncture.
Design: Prospective cohort study.
Setting: A medical ICU (19 beds) from a university-affiliated urban teaching hospital.
Patients: Between February 2001 and October 2001, 300 paired blood culture specimens were obtained from 119 patients (2.52 paired cultures per patient).
Intervention: Prospective patient surveillance and data collection.
Measurements and main results: Thirty-four paired culture results (11.3%; 95% confidence interval, 7.8 to 14.8%) were accepted as true-positives representing a true bacteremia. The sensitivity of catheter-drawn and peripheral venipuncture samples was 82.4% and 64.7%, respectively, and specificity was 92.5% and 95.9%. The positive predictive value was 58.3% for catheter-drawn samples and 66.7% for peripheral venipuncture samples, and the respective negative predictive values were 97.6% and 95.5%.
Conclusions: In critically ill medical patients, the negative predictive value of blood samples obtained by catheter draw or peripheral venipuncture for suspected bloodstream infection is good. However, the sensitivity of blood samples obtained by either catheter draw or peripheral venipuncture alone is not adequate to recommend the elimination of blood samples obtained from the other site. Clinicians should also be aware that additional blood samples may be necessary when interpreting positive blood culture results for common skin or central vein catheter contaminants.