Comparison of risk scores in upper gastrointestinal bleeding in western India: A prospective analysis.

To study the upper gastrointestinal bleeding (UGIB) characteristics and to validate the Rockall and Glasgow-Blatchford scores (GBS), Progetto Nazionale Emorragica Digestiva (PNED) and albumin, international normalized ratio (INR), mental status, systolic blood pressure, and age > 65 (AIMS65) risk scores in predicting outcomes in patients with UGIB.Three hundred subjects with hematemesis and/or melena were prospectively enrolled and followed up for 30 days. All patients were assessed by hematological investigations, imaging, and endoscopy and risk scores were calculated.The mean age was 43.5 ± 17.2 years, and 207 (69%) were males. Hematemesis was the most common presentation (94%). Variceal bleeding was the most common etiology (47.7%). Thirty patients died (10%) and 50 had rebleeding (16.7%). On univariate analysis, serum albumin ≤ 2.7 gm% (p = 0.008), Glasgow Coma scale ≤ 13.9 (p = 0.001), serum bilirubin > 3 mg/dL (p = 0.004), serum bicarbonate ≤ 15.7 mEq/L (p = 0.001), systolic blood pressure < 90 mmHg (p = 0.004), and arterial pH ≤ 7.3 (p = 0.003) were found to be the predictors of mortality. No variable was found significant on multivariate analysis. All four scores were significant in predicting mortality, but Rockall (area under receiver operating characteristic [AUROC] 0.728) was better than others. Rebleeding was better predicted by PNED (modified) (AUROC 0.705). In predicting the need for transfusion and surgical or radiological intervention, GBS score > 0 was significant while score of < 2 classified patients into low risk for mortality with high negative predictive value.Our study showed that the variceal bleeding was the commonest cause of UGIB. Rockall score was more significant in predicting mortality while PNED for rebleeding. Low risk for mortality, need for blood transfusion, or interventions were accurately predicted by GBS.