No biological marker is highly specific of infection and currently available, especially for bacterial infection. The ideal marker would be easy to perform, rapidly, inexpensive, and correlated with the severity and prognosis of the infection.

decreased in eosinophil count (EC) is unspecific of a particular clinical picture and may support a systemic inflammation, whereas the deeper the eosinopenia is, the darker is the prognosis in ICU.

The duration of eosinopenia is not clearly documented, but it has been recently shown that EC tends to normalization, rapidly after appropriate and effective antimicrobial therapy in case of bacterial infection among adults patients hospitalized in a medicine ward. In the light of this findings, Terradas et al. described that EC returned back to normal between the day 2 or day 3 in survivors, indicating a potential interest as a predictive marker of the evolution among hospitalized patients.

To the best of our knowledge, no work has studied eosinopenia as a prognostic marker of mortality during bacterial infections in the elderly patients in a hospital setting. Our study aims to evaluate the prognosis value of the EC in a geriatric unit of tertiary care hospital.

We performed an observational, retrospective single-center study in a teaching hospital of Paris area (Ambroise Paré Hospital in Boulogne-Billancourt). The hospital information system that is routinely completed by healthcare staff for the financing of hospital activity (Programme de Médicalisation des Systèmes d'Information - PMSI) was used to identify eligible patients, i.e. those who had been hospitalized in acute geriatrics ward between January 1 and December 31, 2018 with a diagnosis or a suspicion of bacterial infection. Information about bacterial infection was then checked in the medical record of the patient. Infections of interest were pulmonary, urinary, digestive, biliary, cutaneous, cardiac, and central nervous system infections, as well as bacteremia.

In case of multiple stays over the study period, only the last one was included in the analysis.

In total, over this 12-month period, we analyzed the stays of patients affected by 126 father codes (entitled "family" of pathology) which were sometimes broken down into child codes (pathologies corresponding to these groups).

The database had been declared to the French Authority for Data Protection (Commission Nationale de l'Informatique et des Libertés - CNIL) via the Assistance Publique - Hôpitaux de Paris (AP-HP) (whose registration number is the 2216836).

Data were collected retrospectively by a single investigator, from the patient's medical record on Agfa® Orbis software. The same software made it possible to consult the totality of the biology, the images as well as the treatments administered throughout the stay.

In this study, eosinopenia is defined by an eosinophil count under 100 eosinophils/mm3 based on our previous studies.

Considering D0 as the date of the start of diagnosis by a clinician in the hospital, the other four dates were between D1 and D7.

The patients were separated into two groups: a group of patients who died during hospitalization and a group of patients released alive from their stay. If the patients were still hospitalized 30 days after their admission, they were classified in the group of "living" patients since they had not died on D30.

Inclusion Criteria:

• Eligible patients were hospitalized in acute geriatrics ward between January 1 and December 31, 2018 with a diagnosis of bacterial infection coded in the medical chart and completed by healthcare staff for the financing of hospital activity.
• Infections of interest were pulmonary, urinary, digestive, biliary, cutaneous, cardiac, and central nervous system infections, as well as bacteremia.
• A White blood count cell with eosinophil count available at day 0 from admission, day 3 +/-1 day

Exclusion Criteria:

• bone and joint infections because of specificities in the management of these infections (e.g. surgical procedures)

• Disease that could influence the eosinophil count or that could be the cause of diagnostic errors:

  • Acquired immunosuppression: HIV associated with a CD4 count of less than 200/mm^3, immunosuppressive treatments (corticosteroid therapy at a dose ≥10 mg/d prednisone equivalent, anti-cancer chemotherapy, methotrexate etc.)
  • Previous known haematological disorders

• Discrepancies between the hospital coding of International Classification of Diseases (ICD-10) and the diagnosis of bacterial infection into the medical chart • Patients already on antibiotic therapy for more than 48 hours before the inclusion

Elderly patients admitted in acute geriatrics, independently of their age, for a bacterial infection according to the ICD-10 and medical chart