Background: Diagnosing heart failure in an on-call setting can be difficult, and international studies report diagnostic accuracy among duty doctors, as measured using area under the ROC curve (AUC), to be 0.76-0.90. This study has examined the accuracy with which doctors in the internal medicine out-of-hours service in a Norwegian university hospital distinguish heart failure from no heart failure in patients with dyspnoea.
Material and method: Information was gathered on 468 patients admitted to Akershus University Hospital with dyspnoea between June 2009 and November 2010, and 314 patients were included in the study. The duty doctors estimated the probability of heart failure (0-100%) before N-terminal pro-B-type natriuretic peptide (NTproBNP) concentrations were known. The final diagnosis for the hospital admission was made retrospectively by two independent doctors after review of the medical records, including supplementary tests and the patient outcome.
Results: Heart failure was considered the cause of hospitalisation in 143 patients (46%). Patients with heart failure were older, more often men, had a higher prevalence of heart disease, reduced/impaired renal function, and higher NTproBNP concentrations than patients with non-heart failure dyspnoea. The diagnostic accuracy among duty doctors for heart failure (AUC) was 0.86 (95% confidence interval 0.82-0.90). The doctors' diagnostic accuracy was lower when the patient had heart failure with left ventricular ejection fraction [LVEF] ≥ 50% (n=52): AUC 0.83 (0.77-0.87).
Interpretation: The duty doctors at Akershus University Hospital from 2009-2010 demonstrated similar diagnostic accuracy for heart failure as previously reported from international centres. Diagnostic accuracy was lower for heart failure patients with LVEF ≥ 50%.