Predictors of favourable outcome in non-variceal upper gastrointestinal bleeding: implications for early discharge?

Abstract

Background: There is a lack of validated predictors on which to decide the timing of discharge in patients already hospitalized for upper nonvariceal bleeding.

Aims: Identify factors that appear to protect nonvariceal bleeders from the development of negative outcome (rebleeding, surgery, death).

Methods: Secondary analysis of two prospective multicenter studies. Multivariate analyses for each investigated outcome were performed; a single model was developed including all factors that were statistically significant in each sub-model. A final score was developed to predict favourable outcomes. Prognostic accuracy was tested with ROC curve analysis.

Results: Out of 2398 patients, 211 (8.8%) developed one or more adverse outcomes: 87 (3.63%) had rebleeding, 46 (1.92%) needed surgery and 107 (4.46%) died. Predictors of favourable prognosis were: ASA score 1 or 2, absence of neoplasia, outpatient bleeding, use of low-dose aspirin, no need for transfusions, clean-based ulcer, age <70 years, no haemodynamic instability successful endoscopic diagnosis/therapy, no Dieulafoy's lesion at endoscopy, no hematemesis on presentation and no need for endoscopic treatment. Overall prognostic accuracy of the model was 83%. The final score accurately identified 20-30% of patients that eventually do not develop any negative outcome.

Conclusions: The "good luck score" may be a useful tool in deciding when to discharge a patient already hospitalized for acute non-variceal bleeding.

Keywords: Adverse outcomes; Early discharge; Gastrointestinal bleeding; Prediction rule; Risk assessment.