Objectives: The identification of prognostic factors of easy application in clinical practice can improve the diagnostic and therapeutic decision making process in upper gastrointestinal bleeding (UGB). The present study estimates the capacity to predict an unfavorable clinical course (mortality, unstable bleeding, and/or interventional therapy) on the basis of the preendoscopic and endoscopic clinical analytical findings in patients with UGB owing to peptic ulcer.
Method: A retrospective cohort study was made of 473 adult patients seen in the Emergency Service of a District Hospital, and diagnosed with UGB secondary to gastroduodenal ulcer. Logistic regression analysis was used to construct different models, with the evaluation of their predictive capacity based on calculation of the area under the receiver operating curve (ROC). The final model was used to calculate the probabilities of an unfavorable clinical course for different profiles, with the purpose of constructing an algorithm of help in the decision making process applied to patients initially considered to be at low risk (Forrest classification IIb and III).
Results: The model with the Forrest variable showed a high predictive capacity: ROCa=0.81 (95% confidence interval, 0.76-0.85). Incorporation to the model of clinical and preendoscopic factors (type of UGB, hematocrit, kidney failure, and liver disease) significantly increased its predictive capacity: ROCa=0.87 (95% confidence interval, 0.83-0.91). This model allows the differentiation of different complication risk levels in patients initially at low risk according to the Forrest classification (IIb and III).
Conclusions: The Forrest classification is the principal predictive factor for an unfavorable course in patients with gastrointestinal bleeding owing to peptic ulcer, though clinical factors are also important and should complement the decision taking process.