A clinical and biochemical score for mortality prediction in patients with acute dyspnoea: derivation, validation and incorporation into a bedside programme

Background: Risk stratification for patients with acute dyspnoea is a challenging task. No quantitative tool for mortality prediction among patients with acute dyspnoea is available. Methods: 595 dyspnoeic subjects were enrolled in an emergency department. Clinical and biochemical factors independently predictive of death by 1 year were used to develop a mortality risk prediction tool. Results: Seven factors comprised the final tool: age (x0.3), heart rate (x0.2), blood urea nitrogen (x0.3), New York Heart Association class (x5), amino-terminal pro-B-type natriuretic peptide (NT-proBNP) >= 986 pg/ml (18 points), systolic blood pressure < 100 mm Hg (11 points) and presence of a murmur (11 points). A continuous rise in mortality was seen from 1.7% in the lowest score quintile (n = 118; score <= 48.5) to 43.1% in the highest quintile (n = 116, score >= 85.5; p < 0.001 for trend). Receiver operating characteristic curve analysis of the score's accuracy produced an area under the curve (AUC) of 0.82 (95% CI 0.78 to 0.85) with similar AUCs in subjects with acutely destabilised heart failure (AUC = 0.73, 95% CI 0.67 to 0.79) and those without (AUC = 0.83, 95% CI 0.77 to 0.85, p for the comparison = NS). The score was validated in a separate population of dyspnoeic patients (AUC = 0.73, 95% CI 0.64 to 0.82; p < 0.001) and was incorporated into a computer program suitable for near-patient calculation. Conclusion: A new risk stratification tool for acutely dyspnoeic patients has been derived and validated.