The determinants associated with severe outcome and death from Escherichia coli bloodstream infections (BSI) remain poorly understood. The epidemiology of E. coli BSI has recently changed dramatically with the global emergence of multiresistant strains producing extended-spectrum ß-lactamases (ESBL). Outcome is worse in case of ESBL-E. coli, which may be due to the intrinsic virulence of ESBL-E. coli or to a delayed adequate empirical antibiotic therapy because of multiresistance. Predicting the severity of an infection as soon as the initial clinical assessment is of major importance to provide the best care, while limiting unnecessary hospitalizations and costs. Yet, no simple clinical score exists to predict the severity of E. coli infections.

In a translational approach, the investigators will include during a maximum of one year 500 adults with E. coli BSI hospitalized in 7 hospitals in the Paris area, France. Precise clinical data will be collected at inclusion and 28 days after inclusion or upon patient's discharge (if before day 28). The primary endpoint of the study is death from E. coli BSI at day 28.

The first aim is to determine risk-factors for death at day 28, including clinical and bacteriological factors (determined by WGS) in the era of the global emergence of ESBL producing E. coli. The second aim is to determine virulence characteristics of ESBL strains both at the genome and phenotypic level thanks to a mouse model of septicaemia, and compare them to the clinical data. The third aim, will establish and evaluate a simple clinical severity score (named COLISCORE), in order to help clinician evaluate patients' severity upon initial clinical evaluation and particularly to detect patients at risk of severe outcome. The ultimate goal of this work is to have a clinical impact on patients' management, by understanding the determinants of patient severity due to E. coli BSI in the context of current major epidemiological changes.

This project will be conducted as a collaboration between the ASSISTANCE PUBLIQUE HOPITAUX DE PARIS (or Paris teaching hospitals) and the Infection, Antimicrobials, Modelling and Evolution (IAME) research group, a joint appointment from Paris DIDEROT University, Paris Nord University and the INSTITUT NATIONAL DE LA SANTE ET DE LA RECHERCHE MEDICALE (INSERM).

This study is a prospective multicentric non interventional observational, study, which will include up to 500 patients from 7 hospitals.

Study Centres:

Seven hospitals affiliated with the IAME research group and included in the ASSISTANCE PUBLIQUE HOPITAUX DE PARIS network will participate. The seven hospitals are all tertiary care teaching hospitals from Paris or the close suburbs, part of the ASSISTANCE PUBLIQUE HOPITAUX DE PARIS network. They account for a total of 4230 acute-care adult beds. A total of 780 E. coli BSI occurred in these centres in 1 year with 15% of ESBL producing strains.

This study will include adult patients hospitalized in one of the seven participating centres based on the inclusion and exclusion criteria listed below. A total of 14 clinical investigators (CI) and microbiology investigators (MI) will be responsible for patients' inclusion and data collection in each of the 7 study centres.

Visits.

Two visits will be organised:

Visit 1: day of the patient's inclusion in the study (=day of the results of the first positive blood culture)

Visit 2 : day the patient leaves the hospital or day 28 after Visit 1 if they are still hospitalized.

At visit 1, the E. coli strain responsible for the BSI as well as any other positive sample will be conserved and sent to a central laboratory for full genome sequencing of the E. coli strains (IAME Laboratory, University Paris DIDEROT). The investigators will analyze the molecular characteristics of strains using high-throughput sequencing techniques to try to highlight virulence factors and will compare the results with clinical data.

Endpoints

The primary endpoint of this work is death at day 28 (or visit 2).

Multiple secondary endpoints will be studied at day 28 including length of hospital stay and the need for intensive care.

Statistics

For the primary endpoint the investigators will analyse the risk factors associated to death at day 28 (yes/no) using logistic regression. The clinical risk factors achieving a P value < 0.10 will then be entered into the multivariate logistic regression model. A backward selection method will be used to obtain a model in which all clinical risk factors have a P value < 0.05. Second, the link between death and various bacteriological factors will be studied using a similar approach. Third, a final multivariate logistic regression will be performed adding bacterial factors to the final model with clinical risk factors and performing a backward analysis. The investigators will also perform an exploratory analysis to study the association between all the E. coli genes and death, in a model adjusted with the important clinical factors as defined above.

The investigators will use the results from the multivariate analysis to elaborate a clinical severity score for E. coli BSI, called the COLISCORE. The objective of the COLISCORE is to predict the patients' severity and outcome upon admission and help the clinician's initial management decisions. This score should be simple and easy to use at the patients' bedside.

Inclusion Criteria:

• Isolation of E. coli from 1 or more set of aseptically inoculated blood culture bottle collected in one of the study centres
• Age ≥ 18 year-old

Exclusion Criteria:

• Receiving vasopressors before the onset of the bacteraemia
• Patient previously included in the study
• Patient's opposition

Patients with Escherichia coli bloostream infection