Chest radiograph abnormalities associated with tuberculosis: reproducibility and yield of active cases.

SETTING: Tertiary care referral centre specialising in respiratory diseases.
OBJECTIVES: Chest radiography is a major screening and diagnostic tool for tuberculosis (TB). We evaluated the reproducibility of a radiographic classification system for screening for active TB of immigration applicants to Canada. We also evaluated the validity of this classification system for detection of prevalent active TB among the screened applicants, as well as tuberculin-positive close contacts and symptomatic patients.
METHODS: Reproducibility was assessed by re-reading a randomly selected 10% sample of screening chest films. Validity was estimated from the final clinical and microbiologic diagnosis of patients undergoing detailed clinical evaluation.
RESULTS: Inter-reader agreement using five broad categories was moderate (kappas of 0.44-0.56), while intra-reader agreement was substantial (kappas of 0.59-0.72). After adjustment for age and patient group, the adjusted odds of active tuberculosis, relative to normal or minor findings or granulomas, for fibronodular changes was 10.2 (95% confidence interval [CI] 3.2-33), for mass or pleural effusion it was 11.6 (95%CI 3.6-37), and for parenchymal infiltrate it was 46.1 (95%CI 18-117). Among tuberculin-positive close contacts, the probability of active tuberculosis was more than 50% if the radiographs showed any mass, pleural disease, or parenchymal infiltrates.
CONCLUSION: A simple classification of TB-related chest radiographic abnormalities into five broad categories had moderate to substantial reproducibility of readings, with reasonable validity.