Is the Blatchford score more reliable than the Rockall score in predicting which patients with nonvariceal upper gastrointestinal bleeding will require clinical interventions?

The Blatchford score identified more patients who needed transfusion or interventions to stop bleeding among those with nonvariceal upper gastrointestinal (GI) bleeding. (LOE = 2b)

Chen IC, Hung MS, Chiu TF, Chen JC, Hsiao CT. Risk scoring systems to predict need for clinical intervention for patients with nonvariceal upper gastrointestinal tract bleeding. Am J Emerg Med 2007;25(7):774-779.  [PMID:17870480]

Cohort (retrospective)

Unknown/not stated

Inpatient (any location)

These authors were interested in comparing different clinical scoring systems to guide the management of patients with upper GI bleeding. The Blatchford score uses clinical and laboratory data; the Rockall score also uses endoscopic data (both of these clinical scoring systems can be found in InfoRetriever). The advantage of a scoring system that doesn't rely on endoscopy is the potential for wider application. The authors retrospectively reviewed 354 hospital records of patients admitted for upper GI bleeding who underwent gastroscopy. They excluded patients with bleeding esophageal varices. Although the data collection was performed by a single research assistant masked to the study purpose and to the scores, the authors don't say much about the data extraction methods, a common problem with chart audits. This is unfortunate since good guidelines for conducting chart audits exist.* Additionally, the authors don't mention if any of these scores had been used to guide physicians' decisions. From these audits, they determined if the patient needed clinical intervention (blood transfusion or operative/endoscopic intervention to control bleeding) and calculated the Blatchford score, the pre-endoscopy Rockall score, and the postendoscopy Rockall score. A Blatchford score and a pre-endoscopy Rockall score greater than 0 and a postendoscopy score greater than 2 was considered high risk. A total of 246 patients received transfusions or underwent procedures to stop bleeding. Although all the scores identified more than 90% of these patients, the Blatchford score identified all but 1. Overall, the scores have limited value because of their low specificity. Remember that given the same specificity, increasing sensitivity gives higher rates of false-positive results. The positive likelihood ratios all hover around 1 (except for the postendoscopy Rockall score). This means that they are nearly useless in accurately distinguishing high-risk patients. On the other hand, the negative likelihood ratio for the Blatchford score was much better than the other scores: 0.02 (95% CI, 0.003-0.13). This means that the Blatchford score was much better at ruling out high-risk patients. Now I'd like to see a prospective trial comparing these scoring systems directly. *Jansen AC, et al. J Clin Epidemiol 2005;58(3):269-74.