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

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.