Quantitative assessment of clinical signs for the detection of classical swine fever outbreaks during an epidemic.

The performance of clinical signs as a diagnostic test for the detection of classical swine fever (CSF) outbreaks during the 1997-1998 CSF epidemic in The Netherlands was evaluated by constructing and analysing a receiver operating characteristic (ROC) curve. This curve assesses the discriminating ability of a diagnostic test over a range of test signals. The cut-off values for a defined diagnostic test to detect CSF outbreaks were set by different combinations of clinical signs observed. The area under the ROC curve, which is a quantitative measure of test performance, was significantly (P<0.001) larger than the area under the random ROC curve. This indicates that clinical signs have a significantly higher performance as a diagnostic test for the detection of CSF than for flipping a coin. However, the gain in diagnostic performance compared to a random process is not as much as we would wish it to be. The optimal efficient diagnostic test combined a sensitivity of 72.7% with a specificity of 52.7%, with a combination of the following clinical signs: unsteady gait/ataxia, not eating, not reacting to antibiotic treatment, conjunctivitis, hard faecal pellets.