Background: To better understand the ability of physicians to predict the need for endoscopic therapy and to accurately predict specific endoscopic lesions, we performed a prospective, nonrandomized, observational cohort study in patients presenting with upper gastrointestinal hemorrhage (UGIH) who were undergoing endoscopy. Aim: We aimed to evaluate the pre-endoscopy diagnostic accuracy and the correct prediction of high-risk lesions in patients with UGIH according to the level of clinical expertise. Methods: One hundred twenty-one patients presenting with hematemesis and/or melena within 48 hours were studied. A questionnaire was given to primary physicians, gastroenterology fellows, and gastroenterology faculty, asking them to predict the need for endoscopic therapy and the cause of the bleed. Results: The need for endoscopic therapy was predicted accurately by 68% of the primary physicians, 70% of the fellows, and 74% of the faculty physicians (P = 0.61). The faculty were able to predict which patients did not need therapy more accurately than the fellows and the residents: 85%, 78%, and 68%, respectively (P = 0.03). The diagnostic accuracy of the clinicians - that is, the ability to accurately predict the bleeding lesion among the primary physicians, fellows, and faculty physicians, was similar at 46%, 52%, and 48%, respectively (P = 0.65). Conclusions: The accuracy of predicting the need for endoscopic therapy and the culprit cause of UGIH, based on clinical evaluation, was similar across levels of expertise. However, the faculty gastroenterologists were better than the gastroenterology fellows and the primary providers in predicting which patients do not require endoscopic treatment. We conclude that the relative inability of any group of physicians to accurately predict the presence of high-risk lesions requiring endoscopic therapy suggests that most patients with UGIH should undergo upper endoscopy for diagnosis and possible therapy.
ASJC Scopus subject areas
- Biochemistry, Genetics and Molecular Biology(all)