Classical and Latent Class Analysis Evaluation of Sputum Polymerase Chain Reaction and Urine Antigen Testing for Diagnosis of Pneumococcal Pneumonia in Adults

Diagnosis of pneumococcal pneumonia is complicated by the lack of a diagnostic reference standard that is highly sensitive and specific. Latent class analysis (LCA) is a mathematical technique that relates an unobserved ("latent") infection to multiple diagnostic test results by use of a statistical model. We used classical analysis and LCA to evaluate the sensitivity and specificity of blood culture, sputum Gram stain, sputum polymerase chain reaction (PCR), and urine antigen testing for diagnosing pneumococcal pneumonia among 149 adults with community-acquired pneumonia. On the basis of LCA models, sensitivity of autolysin PCR and pneumolysin PCR was 82% and 89%, respectively, but specificity was low, 38% and 27%, respectively. For urine antigen testing, sensitivity was 77%-78%, and specificity was 67%-71%. Results of the LCA models were comparable with those obtained from classical analysis. LCA may be useful for diagnostic test evaluation and for determining the prevalence of pneumococcal infection in epidemiological studies of community-acquired pneumonia and in vaccine efficacy trials.