An evaluation of endoscopic indications and findings related to nonvariceal upper-GI hemorrhage in a large multicenter consortium

Background: The majority of studies on upper-GI hemorrhage (UGIH) to date have been based on clinical experience at academic centers. There is limited information about patients with UGIH who are evaluated by endoscopy at nonacademic settings, which comprise the majority of endoscopic practices in the United States. Objective: The aims of this retrospective study were to use a large national endoscopy database to describe the population of patients who received an EGD for nonvariceal upper-GI hemorrhage at diverse clinical practice sites and to characterize their endoscopic diagnoses. Methods: Between 2000 and 2004, the Clinical Outcomes Research Initiative (CORI) database received endoscopic reports from 72 diverse practice sites in the United States. All adult patients who under-went an EGD during this period at a CORI site for the indications of hematemesis, melena, or "suspected upper-GI bleed" were identified. Variceal bleeding was excluded. The nonvariceal UGIH cohort was described, and this group's endoscopic findings were characterized by the prevalence, indication, and location of an EGD (inpatient vs outpatient). A repeat EGD within a 2-week period was also evaluated. Results: A total of 243,427 EGDs were performed during the study period; 12,392 (4.9% of all EGDs) were performed for evaluation of hematemesis, melena, or "suspected" upper-GI bleed. Patients in this cohort were more likely to be older, men, and nonwhite compared with their nonbleeding counterparts. Melena was the most common indication for an EGD. The most common endoscopic finding was an ulcer (32.7%), followed by erosion (18.8%). A "normal" EGD was reported on 17.2% of EGDs. Among patients with ulcers, gastric ulcers were more common that duodenal ulcers (54.4% vs 37.1%), whereas clean-based ulcers comprised 52% of all ulcers. Ulcers and Mallory-Weiss tears were significantly more common on inpatient procedures. Ulcers identified on outpatient EGDs were more likely to be clean based. A repeat endoscopy was performed on 4% of the cohort within a 2-week follow-up period, hematemesis was the most common indication for a repeat EGD, and 24.3% of the cohort had a documented therapeutic intervention on their initial EGD. Among inpatients, 6.0% had a repeat EGD within two weeks for an UGIH indication compared with only 1.4% of outpatients (P < .0001). Conclusions: These results allow association of EGD findings with patient presentation. Furthermore, it allows us I to characterize endoscopic findings in a particular patient population (those patients with UGIH) in clinical practice, outside of traditional academic medical centers. The CORI database is a valuable resource for characterizing the epidemiology of endoscopic findings and, in particular, patient populations and across diverse practice settings, thus, enabling the development of hypotheses for future studies.