Prediction of in-hospital mortality after acute upper gastrointestinal bleeding: cross-validation of several risk scoring systems

• Published in: Journal of international medical research Vol. 50; no. 3; p. 3000605221086442
• Main Authors: Benedeto-Stojanov, Daniela, Bjelaković, Milica, Stojanov, Dragan, Aleksovski, Boris
• Format: Journal Article
• Language: English
• Published: London, England SAGE Publications 01-03-2022; Sage Publications Ltd; SAGE Publishing
• Subjects: Cross-Sectional Studies; Gastrointestinal Hemorrhage - diagnosis; Gastrointestinal Hemorrhage - etiology; Hospital Mortality; Humans; Mortality; Prospective Clinical Research Report; Prospective Studies; Risk Assessment - methods; Risk Factors; Severity of Illness Index; in-hospital mortality risk assessment; Acute upper gastrointestinal bleeding; Glasgow-Blatchford scoring system; Cedars-Sinai medical center predictive index; cross-validation; full Rockall risk scoring system
• ISSN: 0300-0605; 1473-2300
• DOI: 10.1177/03000605221086442

Objective
We aimed to identify the clinical, biochemical, and endoscopic features associated with in-hospital mortality after acute upper gastrointestinal bleeding (AUGIB), focusing on cross-validation of the Glasgow-Blatchford score (GBS), full Rockall score (RS), and Cedars-Sinai Medical Center Predictive Index (CSMCPI) scoring systems.
Methods
Our prospective cross-sectional study included 156 patients with AUGIB. Several statistical approaches were used to assess the predictive accuracy of the scoring systems.
Results
All three scoring systems were able to accurately predict in-hospital mortality (area under the receiver operating characteristic curve [AUC] > 0.9); however, the multiple logistic model separated the presence of hemodynamic instability (state of shock) and the CSMCPI as the only significant predictive risk factors. In compliance with the overall results, the CSMCPI was consistently found to be superior to the other two systems (highest AUC, highest sensitivity and specificity, highest positive and negative predictive values, highest positive likelihood ratio, lowest negative likelihood ratio, and 1-unit increase in CSMCPI associated with 6.3 times higher odds of mortality), outperforming the GBS and full RS.
Conclusions
We suggest consideration of the CSMCPI as a readily available and reliable tool for accurately predicting in-hospital mortality after AUGIB, thus providing an essential backbone in clinical decision-making.