Incremental value of upper endoscopy for triage of patients with acute non-variceal upper-GI hemorrhage

Background: Risk scores for triage of patients with acute upper-GI hemorrhage that incorporate endoscopic variables (e.g., the complete Rockall Score) may have better test characteristics for identification of "low-risk" bleeding episodes than those (e.g., Blatchford Score, clinical Rockall Score) that rely solely on clinical variables. An endoscopy-based risk score was compared with two clinically based risk scores in a large cohort of patients hospitalized for treatment of acute upper-GI hemorrhage to quantify the incremental value of endoscopy in the identification of low-risk bleeding. Methods: ICD-9-CMcodes for discharge diagnosis were used to identify a cohort of patients(n=175) hospitalized at a university medical center with acute non-variceal upper-GI hemorrhage. Medical record data were abstracted by two data abstractors blinded to the study intent by using a standardized data collection instrument. Blatchford and Rockall Scores were generated for each case. Low risk was defined as a Blatchford Score of 0, a clinical Rockall Score of 0, or complete Rockall Score of 2 or less. Results: The Blatchford Score risk stratified only 14 of 175 (8%) patients with acute, non-variceal upper-GI hemorrhage as "low risk," while the clinical Rockall Score identified 12%. However, the complete Rockall Score identified the greatest number of low-risk cases, 53/175 (30%) (p<0.0001), when compared with either the Blatchford or clinical Rockall Score. Conclusions: The complete Rockall Score identified significantly more low-risk patients with acute upper-GI hemorrhage than either the clinical Rockall Score or the Blatchford Score. Identification of additional low-risk patients via this endoscopy-based score could lead to decreases in the use of hospital-based services, iatrogenic complications, and time lost from work or usual activity, while improving quality of care. Use of the clinical and complete Rockall Scores sequentially, with consideration of outpatient care for patients at identified as low risk, is recommended.