Diagnosis of vascular catheter-related bloodstream infection: a meta-analysis

Abstract

Catheter-related bloodstream infections increased in incidence during the past decade, causing significant morbidity, mortality, and excess hospital costs. Absence of inflammation at the catheter site in most cases makes clinical diagnosis uncertain. The relative accuracy and cost-effectiveness of different microbiologic tests for confirming that bloodstream infection is catheter related have remained unclear. A meta-analysis of published studies was conducted regarding the accuracy of diagnostic test methods using pooled sensitivity and specificity and summary receiver operating characteristic (ROC) curve analysis. The cost for each test was estimated by methods published by the College of American Pathologists. Costs of catheter replacement and antibiotic therapy for false positive results were included in the cost per accurate test result. Twenty-two studies evaluating six test methods met inclusion criteria for the meta-analysis. Accuracy increased in ROC analysis for catheter segment cultures with increasing quantitation (P = 0.03) (i.e., quantitative > semiquantitative > qualitative) largely due to an increase in specificity. The highest Youden index (mean = 0.85) was observed with quantitative catheter segment culture, the only method with pooled sensitivity and specificity above 90%. For blood culture methods, there was no statistically significant trend toward increased accuracy. The unpaired quantitative catheter blood culture offered the lowest cost per accurate test result but was only 78% sensitive. In conclusion, quantitative culture was the most accurate method for catheter segment culture, and unpaired quantitative catheter blood culture was the single most cost-effective test, especially for long-term catheters.