You may not be sympathetic for what I described last week. Can I not predict the severity of gastrointestinal bleeding based on the clinical data provided, and thus reject some cases of urgent endoscopy?
Yes and no. There is indeed no shortage of prediction scores (of anything) in the literature. How they may help depends on the setting and application.
In a study published earlier this year, the Glasgow-Blatchford bleeding score (GBS) was validated in 4 UK hospitals.[Stanley AJ et al. Lancet 2009;373:42-7] The overall accuracy of GBS to predict the need for endoscopic intervention and death among patients with symptoms of gastrointestinal bleeding was good, with an area under ROC curve of 0.90. The investigators also prospectively implemented the score in real clinical practice and demonstrated that the need for hospitalization decreased.
Can I apply it? When one looks at GBS carefully, the low risk criteria include normal serum urea, normal hemoglobin level, normal blood pressure and heart rate, and absence of melena, syncope, cardiac failure, or liver disease. If a patient has none of these, you really would wonder if he/she suffers from anything at all. In fact, in the validation study, 56% of the patients fulfilling the low risk criteria (GBS=0) had no lesion or just hiatus hernia on endoscopy. It looks as if the score is most useful in helping those who cannot clerk a proper history and make the diagnosis of upper gastrointestinal bleeding. Sadly, whenever I am consulted, the patients would have at least several of the risk factors. According to evidence-based medicine, I should have no choice but come back to the hospital every time.
Another factor we cannot ignore is that the negative predictive value of any test depends on the prevalence of disease. At the primary care clinic or emergency department where most patients with symptoms of gastrointestinal bleeding actually would not die or require endoscopic therapy, using GBS to rule out significant bleeding is safe and accurate. Managing old and frail in-patients is obviously another ball game.