Predicting mortality in patients with in-hospital nonvariceal upper GI bleeding: a prospective, multicenter database study

Abstract

Background: Nonvariceal upper GI bleeding (NVUGIB) that occurs in patients already hospitalized for another condition is associated with increased mortality, but outcome predictors have not been consistently identified.

Objective: To assess clinical outcomes of NVUGIB and identify predictors of mortality from NVUGIB in patients with in-hospital bleeding compared with outpatients.

Design: Secondary analysis of prospectively collected data from 2 nationwide multicenter databases. Descriptive, inferential, and multivariate logistic regression models were carried out in 338 inpatients (68.6 ± 16.4 years of age, 68% male patients) and 1979 outpatients (67.8 ± 17 years of age, 66% male patients). A predictive model was constructed using the risk factors identified at multivariate analysis, weighted according to the contribution of each factor.

Settings: A total of 23 Italian community and tertiary care centers.

Patients: Consecutive patients admitted for acute NVUGIB.

Interventions: Early endoscopy, medical and endoscopic treatment as appropriate.

Main outcome measurements: Recurrent bleeding, surgery, and 30-day mortality.

Results: The mortality rate in patients with in-hospital bleeding was significantly higher than that in outpatients (8.9% vs 3.8%; odds ratio [OR] 2.44; 95% confidence interval [CI], 1.57-3.79; P < .0001). Hemodynamic instability on presentation (OR 7.31; 95% CI, 2.71-19.65) and the presence of severe comorbidity (OR 6.72; 95% CI, 1.87-24.0) were the strongest predictors of death for in-hospital bleeders. Other independent predictors of mortality were a history of peptic ulcer disease and failed endoscopic treatment. Rebleeding was a strong predictor of death only for outpatients (OR 5.22; 95% CI, 2.45-11.10). Risk factors had a different prognostic impact on the 2 populations, resulting in a significantly different prognostic accuracy of the model (area under the receiver-operating characteristic curve = 0.83; 95% CI, 0.77-0-93 vs 0.74; 95% CI, 0.68-0.80; P < .02).

Limitations: Study design not experimental, no data on ward specialty, potential referral bias.

Conclusions: In-hospital bleeders have a significantly higher risk of death because they are sicker and more often hemodynamically unstable than outpatients. Predictors of death have a different impact in the 2 populations.