1. Home>
2. Specialties>
3. Gastroenterology>
4. Summary and Comment

Risk Prediction for Acute Upper GI Hemorrhage

A new scoring system identifies patients who can be treated without hospital admission.

Acute upper gastrointestinal bleeding (AUGIB) results in 250,000 to 300,000 hospitalizations annually in the U.S., but several studies have shown that many AUGIB patients have low risk for rebleeding and could be treated as outpatients. Multiple risk-prediction scales have been developed, but few have been validated prospectively, and none has been applied in general clinical practice.

Investigators at three Veterans Affairs hospitals prospectively collected data on 391 patients who were admitted with diagnoses of AUGIB. Clinical and endoscopic data were collected using standard forms at the time of admission, throughout the hospital stay, and for 30 days postdischarge. The researchers identified predictors of two adverse outcomes: GI bleeding (rebleeding, surgery, hospital mortality) and global outcomes (GI bleeding plus worsening of underlying comorbidities or new comorbid conditions that occurred during the hospitalization). Data from two thirds of the patients (the derivation set) were used to create a prediction model that was validated using data from the other one third of the patients (the validation set).

Statistical analysis of the derivation set showed that three variables correlated with worse outcomes for bleeding: APACHE II score 11, endoscopic stigmata for rebleeding, and esophageal varices. This model produced good discrimination (C=0.81) and calibration (P=0.80). This model, plus the presence of unstable comorbidity on admission, correlated with worse global outcomes (C=0.78; P=0.69). Results in the validation set were similar to those in the derivation set.

The researchers developed a scoring system, based on the presence or absence of each risk factor. Low risk was defined as a score of 0 (no risk factors), intermediate risk as 1, and high risk as 2. Among patients with low, intermediate, and high risk, adverse bleeding outcomes occurred in 1.1%, 5.0%, and 25.5%, respectively; poor global outcomes occurred in 4.8%, 16.7%, and 46.5%, respectively. Dichotomizing the scale, presence of any risk factor (score, 1) identified 25 of 27 patients with poor bleeding outcomes and 71 of 78 patients with poor global outcomes. A score of 0 identified patients who could be discharged safely; this included 136 (37.8%) patients when bleeding-related factors were considered and 116 (32.2%) when unstable comorbidity was added. The authors concluded that these simple clinical prediction rules can identify patients who can be treated safely as outpatients.

Comment: As noted by the authors, this study is limited by the fact that VA hospitals serve an overwhelmingly male population (99%) and because the distribution of comorbid conditions could be different from that seen in community hospitals. Although these rules performed better than the Rockall score, they rely on the APACHE II score, which often is not available at the time of admission. Patients with varices have uniformly worse outcomes and should be excluded from any scoring system. Ultimately, all scoring systems rely on quantifying the same clinical data: the hemodynamic effect of bleeding; the endoscopic appearance of the lesion; and the presence of comorbid conditions, including liver disease. When logically applied, this information identifies the risk level for a poor outcome and can guide decisions about admission and hospital care.

— David J. Bjorkman, MD, MSPH (HSA), SM (Epid.)

Published in Journal Watch Gastroenterology June 25, 2007

Citation(s):

Imperiale TF et al. Predicting poor outcome from acute upper gastrointestinal hemorrhage. Arch Intern Med 2007 Jun 25; 167:1291-6.

• Original article (Subscription may be required)
• Medline abstract (Free)