In clinical practice, early recognition of bacteremia leads to prognostic improvement. Recently, cell population data (CPD) from the Sysmex XN-series hematology analyzer has attracted attention as a new method for the early diagnosis of bacteremia, but its usefulness in clinical practice remains unclear. We focused on the fluorescent light distribution of the neutrophil area (NE-WY) and used machine learning to determine whether predictions could be improved. Among adult patients with clinically suspected bacteremia at our hospital, 533 who did not meet the exclusion criteria were enrolled. Propensity score matching was performed for bacteremia (n = 106) and non-bacteremia (n = 427) cases. NE-WY showed the largest difference between the two groups (AUC, 0.768; sensitivity, 73.6%; specificity, 67.9%; cutoff value, 686.5(no unit)). In machine learning, the first branch of the decision tree was NE-WY, and the cutoff value was set at 689.5(no unit). Support vector machines were used to examine multiple variables, but there were no significant differences relative to NE-WY alone. This is the first report to demonstrate that NE-WY is useful for detecting bacteremia by analyzing CPD using machine learning. The combination of CPD and machine learning is expected to produce new results.
Horie et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: