Can history and exam alone reliably predict pneumonia?

Background: Prediction rules based on clinical information have been developed to support the diagnosis of pneumonia and help limit the use of expensive diagnostic tests. However, these prediction rules need to be validated in the primary care setting. Methods: Adults who met our definition of lower respiratory tract infection (LRTI) were recruited for a prospective study on the causes of LRTI, between November 15, 1998 and June 1, 2001 in the Leiden region of the Netherlands. Clinical information was collected and chest radiography was performed. A literature search was also done to find prediction rules for pneumonia. Results: 129 patients - 26 with pneumonia and 103 without - were included, and 6 prediction rules were applied. Only the model with the addition of a test for C-reactive protein had a significant area under the curve of 0.69 (95% confidence interval [Cl], 0.58-0.80), with a positive predictive value of 47% (95% Cl, 23-71) and a negative predictive value of 84% (95% Cl, 77-91). The pretest probabilities for the presence and absence of pneumonia were 20% and 80%, respectively. Conclusions: Models based only on clinical information do not reliably predict the presence of pneumonia. The addition of an elevated C-reactive protein level seems of little value.