P17 The Accuracy of a Diagnosis of Pneumonia in a UK Teaching Hospital

Abstract

Background Obtaining an accurate diagnosis of pneumonia is an essential part of optimal patient care. Analysis of patients' hospital records allows clinical coding (ICD-10) of admission events which assist development of clinical decision algorithms, assessment of quality of care and public health evaluation. We sought to evaluate the reliability of applied clinical codes and the accuracy of a diagnosis of pneumonia in our institution. Methods A retrospective case note review of all patients admitted to University Hospital Llandough in 2011 with a final clinical code diagnosis of pneumonia. Pneumonia was defined as the presence of new radiographic infiltrate in patients with symptoms consistent with an acute lower respiratory tract infection.1 The chest radiographs of each patient were reviewed by a respiratory physician (KP, HED) and the formal radiology report was independently scrutinised (IM). Results 710 patient episodes of ICD-10 coded pneumonia were identified in a 1 year period at our hospital. Ten patients had no chest x-ray performed and one x-ray had no report. Radiological confirmation of pneumonia (by radiology reporting) occurred in 69.8% (488/699); a radiological diagnosis of pneumonia was made by a respiratory physician (KP, HED) in 71.8% (502/699) of patients. There was 85% agreement between the Respiratory and Radiology reports (592/699 cases) with a kappa of 0.66 (95% CI 0.57 to 0.69). The accuracy of a pneumonia diagnosis differed little between patients cared for by a respiratory physician (72.3% agreement with radiology report) and those admitted to a non-respiratory ward (68.1%). In 27.0% and 31.9% of patients respectively there was no radiological evidence of pneumonia. Conclusions A clinical coding diagnosis of pneumonia is unreliable with 30.2% of patients not having compatible radiograph change. This has implications for the validity of any research performed on data selected on the basis of clinically coded information. Misdiagnosis by clinicians is the most likely reason for this discrepancy.