More than 80% of patients with upper gastrointestinal bleeding (UGIB) are hospitalized after their visit to the emergency department (ED). However, some of these hospitalizations do not seem justified. Several clinical scores have been developed to classify patients according to their risk of death or need for therapeutic intervention.

The aims of this study are:

1. to describe the characteristics of patients hospitalized for UGIB after their visit to the ED
2. to assess the predictive factors of hospital intervention or death
3. to assess the accuracy of existing prognostic scores to classify patients according their risk of death or need for therapeutic intervention (external validation) and to identify low-risk patients not requiring intervention.
4. Depending on the results, a new score could be derived to identify patients at low risk for intervention or death.

Upper gastrointestinal bleeding (UGIB) is caused by several conditions, with an overall mortality of approximately 10%. More than 80% of patients are hospitalized after their visit to the emergency department (ED) [1]. The hospitalization usually allows an etiological diagnosis to be made by performing an upper gastrointestinal endoscopy (UGE) and a hemostatic treatment if required. However, some of these hospitalizations do not seem justified. Indeed, some patients do not require any intervention (i.e., blood transfusion, endoscopic, radiological or surgical hemostasis) and have no complications (i.e., rebleeding, death). Some of them do not even have a diagnostic UGE during their hospital stay.

Several clinical scores have been developed to classify patients according their risk of death or need for therapeutic intervention, some of them to identify patients with low-risk of complications (rebleeding, death) and/or need for intervention. The most commonly known are the pre-endoscopic Rockall score, the Glasgow-Blatchford Score (GBS) and the AIMS65. Among them, the GBS seems to be the most efficient [2,3]. Its use is encouraged in the latest French, European and international recommendations [4].

Other prognostic scores have more recently been proposed with this purpose (e.g. modified GBS, CANUKA, H3B2, C-Watch, Harbinger), with interesting results but scarce or no external validation.

The aims of this study are:

1. to describe the characteristics of patients hospitalized for UGIB after their visit to the ED
2. to assess the predictive factors of death or therapeutic intervention (i.e, blood transfusion, endoscopic hemostasis, surgery or interventional radiology)
3. to assess the accuracy of existing prognostic scores to classify patients according their risk of death or need for therapeutic intervention (external validation) and to identify low-risk patients not requiring intervention.
4. Depending on the results, a new score could be derived to identify patients at low risk for intervention or death.

Inclusion Criteria:

• patient ≥ 18 years old
• with a diagnosis of upper gastrointestinal bleeding (defined by ICD10 codes)
• visited a hospital with ED data available in the data warehouse

Exclusion Criteria:

• reason for consultation not related to upper gastrointestinal bleeding (identified by manual reading of hospital notes)

Patient ≥ 18 years old with a diagnosis of upper gastrointestinal bleeding hospitalized after an ED visit in hospital with data available in the Assistance Publique-Hôpitaux de Paris data warehouse