Nonvariceal upper gastrointestinal bleeding: differences in outcome for patients admitted to internal medicine and gastroenterological services.

OBJECTIVE: It has been suggested that admission to a gastroenterology service (GAS) is associated with a better prognosis and lower cost for treatment of gastrointestinal (GI) diseases, such as upper GI bleeding (UGB). However, a large potential bias by higher comorbidity on internal medicine services (MED) could not be excluded from these studies. We therefore compared patients with upper GI bleeding admitted to a gastroenterology or internal medicine department, with special emphasis on prognostic factors, such as comorbidity, and outcome.

METHODS: Between 1991 and 1995, 322 patients were admitted to our hospital for UGB. Forty-five patients had variceal and 277 patients had nonvariceal upper GI bleeding (NUGB). Of 232 patients with primary NUGB, 125 were admitted to GAS and 93 to MED. The charts of these patients were revised, comorbidity was carefully recorded, and the Rockall risk score was calculated. All deaths were individually classified as unavoidable, mostly due to severe underlying illness, or potentially avoidable.

RESULTS: No differences in delay for endoscopy or treatment were observed between GAS and MED. The rebleeding, surgery, and mortality rates in GAS and MED patients were 11.6% versus 11.5% (NS), 7.8% versus 7.3% (NS), and 2.4% versus 10.8% (p = 0.02), respectively. Rockall scores differed between GAS and MED patients (3.1 +/- 1.8 vs 3.7 +/- 1.7, p = 0.02). The mortality rate stratified by Rockall score was lower for the GAS patients. However, individual analysis revealed that only three of 13 deaths were potentially avoidable: two of 10 at the MED and one of three at the GAS.

CONCLUSION: The lower mortality among nonvariceal upper GI bleeding patients admitted to a gastroenterological service compared to an internal medicine service was mainly due to lesser comorbidity. This effect was not detected by stratification according to Rockall, but shown with analysis of individual patient charts only. The latter underscores the potential pitfalls when comparing outcome or cost of treatment between different medical services.