Less haste, more speed: Does delayed blood culture transport time lead to adverse incubation times or yield?

Abstract

Background: Blood culture remains a vital diagnostic tool in the acutely unwell patient. The UK Standards for Microbiology Investigations (SMI) stipulate pre-analytical requirements that are aimed at increasing yield and reducing turnaround time. The requirement to load blood cultures on machines within 4 h has been shown to reduce turnaround time but limited evidence exists as to whether it improves yield.

Methods: We extracted blood culture results, including organism growth, time to detection, location and time of sample collection from 4 hospitals in Southwest England. We then used mixed effects, Bayesian linear and logistic regression models to examine the effect of predictor variables like time to laboratory (TTL) and sent time on the response variable of growth or time to detection. We fitted generalised additive models to explore non-linearity.

Results: 449,191 culture sets were analysed, 398,077 of which had enough data to include in the final analysis. 37,255 sets flagged positive (9.36%) of which 21,330 were considered pathogens. Our primary analysis identified a small decrease in yield with each hours delay in loading (0.997 (95%CrI 0.994-1.001)). This effect was largest in Streptococcus pneumoniae, agalactiae and pyogenes. In our analysis on time on the machine, culture sets spend 10.03 (95%CrI 12.66-7.31) minutes less on incubators for each hour delay. Neither anaerobes nor yeasts showed any loss of recovery from increasing TTL. There was no convincing non-linearity in either of these effects.

Conclusion: There is a marginal loss of growth for every hour a blood culture is left unincubated, with the loss of recovery of Streptococcus pyogenes and other streptococci being most common. There was no evidence of a reduction in Gram-negatives, anaerobes, or yeasts. There was a small decrease in time to detection for delayed sets. This analysis suggests there may be marginal benefit in reducing time to load.

Keywords: Blood culture; Microbiology; Streptococcus pyogenes.

Fig. 1

Flow chart of culture set inclusion.

Fig. 2

A) Density plot showing logarithmic distribution of Time to laboratory per site. B) A density plot showing time to detection per site.

Fig. 3

A) Generalized additive model plot of the estimate against time to laboratory for growth with 95% credible intervals. B) Generalized additive model plot of estimate against time to laboratory for growth with 95% credible intervals, adjusted for location, year, and working day.

Fig. 4

A) Generalized additive model of time to laboratory on time to detection with 95% credible intervals, without adjustment. B) Generalized additive model of time to laboratory on time to detection with 95% credible intervals adjusted for location, working day, organism and year.

Fig. 5

Forest plot for the estimate of growth per organism, adjusted for location, working day and year.

Fig. 6

Forest plot of the estimate on isolates' time to detection from TTL.