[Evaluation of eosinopenia as an infection marker in critical care patients]

To evaluate eosinopenia as an early marker of infection.

Retrospective cohort study.

Medical-surgical ICU patients with high severity scores.

Data on days 1-5: Demographic data, diagnosis, clinical repercussion, mechanical ventilation, clinical development, length of stay, APACHE II, leukocytes, SOFA and lactate. Patients divided into two groups: with and without infection. ROCs (receiver operator characteristic) curves were plotted and best point for discriminative values determined.

244 patients were included: 22.5% with infection. 52.9% medical, 22.5% surgical and 24.6% polytrauma patients. APACHE II: 14.9+/-8.9. In a logistic regression model of infection (dependent variable infection), the independent variables were: APACHE II, SOFA, monocytes and eosinophils. The ROC curve for eosinophils on the first day: area of 0.72; the best cut off value is 10 eosinophils/microl, with sensitivity (S): 64.8% and specificity (Sp): 70.9%. In medical patients, the area under curve is 0.80, with ideal cut off value of 9 eosinophils/microl; in surgical patients is 0.53, with a cut off ideal value of 54. We combined eosinophils and monocytes: a cut-off value of 9 eosinophils/microl in medical patients with >400 monocytes/microl, has: S: 86.7%, Sp: 74.7%, a positive predictive value (PPV) of 40.6% and a negative predictive value (NPV) 96.6%; in postsurgical patients with <400 monocytes/microl and a cut-off value of 54 eosinophils: S: 100%, Sp: 20%, PPV: 52.9% and NPV: 100%.

In a medical-surgical ICU, the capacity to discriminate infection through examining eosinopenia is not high. It could be useful to rule out infection if we combined eosinopenia with monocytes count.