Introduction and Objectives BTS guidelines suggest the diagnostic test for pneumonia (in hospital) is a CXR. Since management should be commenced within 4 h of arrival the initial interpretation of the CXR is most often performed by junior doctors. Consultants review patients and their CXRs within 24 h of admission and a radiologist's report is issued at a later time point but diagnosis dependent decisions—whether to admit, risk stratification and whether to give antibiotics—fall to junior doctors. We measured the inter-observer agreement within groups of doctors involved at key points in CAP diagnosis and management.
Methods 24 admission CXRs from patients attending a large teaching hospital between February and April 2011 with a suspected diagnosis of CAP were collected on a radiology computer workstation. Eleven reporting radiologists, eight attending consultants and twenty-two junior doctors from a range of acute medical specialties were recruited. All 41 recruits independently interpreted the CXRs, answering “yes” or “no” for presence of pneumonic infiltrate. Agreement within groups was tested using the Fleiss-Cuzick extension of Cohen's κ statistic (abbreviated here as κ).
Results Agreement (κ scores) for each group of doctors are shown in Abstract P251 figure 1. Radiologists had moderate agreement (κ 0.52, CI 0.49 to 0.55), junior doctors had moderate agreement (κ 0.47, CI 0.45 to 0.48) and attending consultants fair agreement (κ 0.34, CI 0.30 to 0.38).
Conclusions In the context of possible pneumonia, the CXR was not consistently reported by any group. Junior doctors were more consistent than attending consultants and radiologists were most consistent. A possible explanation for these differences is that junior doctors, by necessity, have developed similar ideas of what they will consider pneumonic where as consultants who less frequently make the initial diagnostic decision vary in their criteria for diagnosing pneumonia. This study does not present a “gold standard” interpretation and therefore does not address the issue of accuracy but it does raise questions about to what extent the CXR ever be regarded as a “reliable” diagnostic test.