[A scoring model for a differential diagnosis of tuberculous and non-tuberculous pleurisy]

To evaluate various clinical features and laboratory biochemical markers so as to develop a predictive model for differentiating tuberculous pleurisy (TBP) from non-tuberculous pleurisy (non-TBP).

A total of 241 TBP patients and 212 non-TBP patients who were hospitalized between January 2007 and December 2009 at our hospital were studied retrospectively. Their symptoms and laboratory parameters were recorded. The statistically different variables were selected to undergo binary logistic regression to calculate a scoring system (range: 0 - 10) according to their β coefficients. A receiver operating characteristic (ROC) curve was used to calculate the best cut-off value. The performance of the model was tested in a sample of 82 new cases with pleural exudates.

Seven variables were selected in the present scoring model: temperature > 38°C (1.0 point), purified protein derivative testing positive (1.0 point), serum C-reactive protein ≥ 26 mg/L (1.5 points), pleural fluid lymphocyte percentage ≥ 85% (1.0 point), pleural fluid protein ≥ 49 g/L (1.0 point), pleural fluid adenosine deaminase ≥ 43 U/L (2.5 points) and serum and/or pleural fluid mycobacterium tuberculosis antibody positive (2.0 points). With a cut-off value of 6.0 points, the sensitivity, specificity and accuracy of differentiating TBP from non-TBP was 90.1%, 94.3% and 92.1% respectively. With the application of this model, 82 new pleural effusions showed a sensitivity of 94.1%, a specificity of 93.8% and an accuracy of 93.9%.

The scoring model provided a simple and feasible way of facilitating a differential diagnosis of TBP and non-TBP patients.