Scoring systems and risk assessment for upper gastrointestinal bleeding

Abstract

Mortality associated with acute upper gastrointestinal bleeding remains high despite advances in diagnosis and therapy. This was emphasized by the findings of the seminal English National Audit of acute gastrointestinal haemorrhage undertaken by Rockall and associates in the mid-1990s. The apparent lack of progress is largely due to less selective reporting in an ageing population with greater co-morbidity. Thus some deaths will be unavoidable even with exemplary treatment. Managing high risk patients in a dedicated area with close cooperation between medical and surgical gastroenterologists has been shown to improve outcome. The challenge is to select those patients who have most to gain from such a scarce and expensive resource so that their treatment can be optimized. Various risk factors have been identified to help achieve this end. Rockall's national audit data suggest that avoidable deaths remain a problem in most district general hospitals. A simple numerical score was derived from these audit data (Rockall score) to predict rebleeding and mortality. The score is based on five variables: age, shock, co-morbidity, endoscopic diagnosis and stigmata of recent haemorrhage. It has the advantage that pre-endoscopic assessment can be made by inexperienced medical or nursing staff. The system was validated internally in a second audit by Rockall and co-workers, and subsequent external validation has come from New Zealand and the Netherlands. The score is less reliable at predicting rebleeding than death and so is, as yet, an imperfect instrument. The scoring system has also proven valuable in selecting low risk patients for early discharge (resulting in health care economies) and for comparing outcome data from different hospitals or populations. Endoscopic treatment has recently been shown to reduce rebleeding rates and perhaps mortality. These advances in therapy are becoming more widely adopted and may influence the predictive ability of the Rockall score. The study from Edinburgh, in this issue, although small and with wide confidence intervals, supports the ability of the Rockall score to identify high risk cases amongst those given endoscopic treatment. It also suggests that an adjustment of the score may be required in these circumstances to prevent overcalling the risk of rebleeding and death.