Observer variation in detecting lymphadenopathy on chest radiography.

OBJECTIVE To assess inter- and intra-observer agreement in the detection of lymphadenopathy on chest radiography in children at risk for tuberculosis. PATIENTS AND METHODS Retrospective examination of the antero-posterior and lateral chest radiographs of children aged 1 month to 11 years discharged from the short-stay ward of the Red Cross Children's Hospital, Cape Town, with a diagnosis of tuberculosis or pneumonia. Four paediatric pulmonologists viewed the radiographs independently. The main outcome measures were inter- and intra-observer agreement on the presence or absence of lymphadenopathy, reported as present, absent or equivocal, and expressed as weighted kappa statistics. RESULTS Weighted kappa for the six pairs of observers ranged from 0.14 (95%CI 0.02-0.30) to 0.52 (95%CI 0.35-0.69). After a 3-month interval, intra-observer agreement ranged from 0.44 (95%CI 0.25-0.62) to 0.71 (95%CI 0.56-0.87). The average weighted kappa for inter-observer agreement was 0.33, and the average intra-observer kappa was 0.55. CONCLUSIONS There was 'fair' inter- and 'moderate' intra-observer agreement among paediatric pulmonologists in detecting lymphadenopathy on chest radiography in children. Caution is necessary when basing clinical decisions on the presence of lymphadenopathy on chest radiography.