Eosinopenia as a diagnostic marker of bloodstream infection in hospitalised paediatric and adult patients: a case-control study.

The objective of this study was to assess whether eosinopenia was a reliable diagnostic marker of bloodstream infection in hospitalised adult and paediatric patients. The design was a case-control study, set in a tertiary adult and paediatric hospital. A total of 157 adult and 85 paediatric patients with bloodstream infection ('cases') were compared to 195 and 94 randomly selected adult and paediatric patients who had clinical suspicion of bloodstream infection but with a negative blood culture ('controls') respectively. Patients with haematological or immunosuppressive disease and control patients who were treated with antibiotics within one week prior to the blood culture were excluded. Eosinopenia, or undetectable eosinophil count (<0.01 x 10(9) or <10/mm3), was more common among the cases than the controls (46.5% vs 21.5%, respectively). The specificity of eosinopenia to predict bloodstream infection in adult patients was reasonable (79%, 95% confidence interval [CI] 74 to 82), but its sensitivity was low (47%, 95% CI 41 to 52). The absolute eosinophil count only had a modest ability to discriminate bloodstream infections from controls in adult patients (area under receiver operating characteristic curve 0.349, 95% CI 0.288 to 0.411). Eosinophil counts had very little overall predictive ability (area under receiver operating characteristic curve 0.448, 95% CI 0.363 to 0.533, P=0.237), and the sensitivity (54%, 95% CI 47 to 61) and specificity (56%, 95% CI 49 to 63) of eosinopenia to predict bloodstream infection in paediatric patients were both low. In the multivariate analyses, only C-reactive protein concentrations and neutrophil counts, but not eosinopenia, were significantly associated with the presence of bloodstream infection in both adult and paediatric patients. The presence of eosinopenia can be considered as an inexpensive warning test for bloodstream infection in hospitalised adult patients so that further investigations can be initiated. An absence of eosinopenia is, however not sensitive enough to exclude bloodstream infection. C-reactive protein concentrations and neutrophil counts were both better markers of bloodstream infection than eosinopenia in hospitalised paediatric and adult patients.