Comparison of semi-quantitative and quantitative methods for diagnosis of catheter-related blood stream infections: a systematic review and meta-analysis of diagnostic accuracy studies

Abstract

Catheter-related blood-stream infections (CRBSIs) are the most common healthcare-associated blood-stream infections. They can be diagnosed by either semi-quantitative or quantitative methods, which may differ in diagnostic accuracy. A meta-analysis was undertaken to compare the diagnostic accuracy of semi-quantitative and quantitative methods for CRBSI. A systematic search of Medline, Scopus, Cochrane and Embase databases up to January 2020 was performed and subjected to a QUADAS (quality assessment of diagnostic accuracy studies 2) tool to evaluate the risk of bias among studies. The pooled sensitivity and specificity of the methods were determined and heterogeneity was evaluated using the χ2 test and I2. Publication bias was assessed using a Funnel plot and the Egger's test. In total, 45 studies were analysed with data from 11 232 patients. The pooled sensitivity and specificity of semi-quantitative methods were 85% (95% CI 79–90%) and 84% (95% CI 79–88%), respectively; and for quantitative methods were 85% (95% CI 79–90%) and 95% (95% CI 91–97%). Considerable heterogeneity was statistically evident (P < 0.001) by both methods with a correspondingly symmetrical Funnel plot that was confirmed by a non-significant Deek's test. We conclude that both semi-quantitative and quantitative methods are highly useful for screening for CRBSI in patients and display high sensitivity and specificity. Quantitative methods, particularly paired quantitative cultures, had the highest sensitivity and specificity and can be used to identify CRBSI cases with a high degree of certainty.

A likelihood‐based sensitivity analysis for publication bias on the summary receiver operating characteristic in meta‐analysis of diagnostic test accuracy

“…In many meta-analyses of diagnostic test accuracy, authors assessed publication bias by presenting a funnel plot for the log-transformed DOR (lnDOR). 17,18 Statistical properties of the funnel-plot-based methods were examined in several articles. By employing the lnDOR, Deeks et al 19 conducted simulation studies to evaluate the usefulness of the Begg, Egger, and Macaskill tests of funnel plot asymmetry.…”

“…An intuitive idea is to utilize the graphical methods of the funnel plot and the trim‐and‐fill method for some univariate diagnostic measures such as the (log‐transformed) DOR. In many meta‐analyses of diagnostic test accuracy, authors assessed publication bias by presenting a funnel plot for the log‐transformed DOR (lnDOR) 17,18 . Statistical properties of the funnel‐plot‐based methods were examined in several articles.…”