Accuracy and reliability of physical signs in the diagnosis of pleural effusion.

Although pleural effusion is a common disorder among patients presenting with respiratory symptoms, there is limited evidence on the accuracy and reliability of symptoms and signs for the diagnosis of pleural effusion. In our study, conducted at a rural hospital in India, two physicians, blind to history and chest radiograph findings, and to each other's results, independently evaluated 278 patients (196 men), aged 12 and older, admitted with respiratory symptoms. We did a blind and independent comparison of physical signs (asymmetric chest expansion, vocal fremitus, percussion note, breath sounds, crackles, vocal resonance and auscultatory percussion) with the reference standard (chest radiograph). We measured diagnostic accuracy by computing sensitivity, specificity, and likelihood ratios (LRs), and inter-observer reliability by using kappa (kappa) statistic. We performed multivariate analysis to identify the clinical signs that independently predict pleural effusion. The prevalence of pleural effusion was 21% (57/278). The LRs of positive signs ranged from 1.48 to 8.14 and their 95% confidence intervals (CIs) excluded 1. Except for pleural rub, the LRs for negative signs ranged between 0.13 and 0.71. The interobserver agreement was excellent for chest expansion, vocal fremitus, percussion and breath sounds (kappa 0.84-0.89) and good for vocal resonance, crackles and auscultatory percussion (kappa 0.68-0.78). The independent predictors of pleural effusion were asymmetric chest expansion (odds ratio [OR] 5.22, 95% CI 2.06-13.23), and dull percussion note (OR 12.80, 95% CI 4.23-38.70). For the final multivariate model, the area under receiver operating characteristic curve (ROC curve) was 0.88. In conclusion, our data suggest that physical signs may be helpful to rule out but not rule in pleural effusion.