PubMed1987
We analysed 103 episodes of upper gastrointestinal bleeding in 88 elderly patients (age 76 +/- 7.7 years) to determine which of a group of 52 clinical and laboratory variables, measured on admission, best predicted continued or rebleeding, and death in these patients. Variables which related directly to the size of the bleed (blood urea, haemoglobin, pulse rate, systolic blood pressure) were all strongly predictive of both outcomes (P less than 0.001). Of the variables unrelated to the size of the bleed, prolonged prothrombin time and elevated serum creatinine were most strongly predictive of a poor outcome, suggesting that haemostatic dysfunction may be a major contributor to death from upper gastrointestinal haemorrhage in elderly patients. Other variables with strong predictive potential were age (P less than 0.001), the presence of multiple disease states (P less than 0.01), therapy with multiple drugs (P less than 0.01) and acute stroke or obtundation on admission (P less than 0.01). In general terms the size of the bleed was as significant as the premorbid condition of the patient in predicting the outcome. This, together with the fact that half the patients died of hypovolaemia, suggests that death from upper gastrointestinal bleeding in the elderly is not inevitable and that further reduction in mortality from this cause is attainable.