Observer variation in classifying chest radiographs for small lung opacities and pleural abnormalities in a population sample.

Abstract

The purpose of this study was to assess inter-and intraobserver variation in the radiographic categories of small lung opacities (profusion) and pleural abnormalities classified according to the ILO classification of pneumoconioses with some modifications. Chest radiographs derived from a representative adult population sample (n = 7,095) were classified by two radiologists. Observer variation was assessed on the basis of kappa (kappa)-type statistics. The observers agreed on profusion categories in 69% of cases of the total material. Up to 98% of the classifications fell into the same category or deviated by no more than one category. The corresponding kappa (kappa) coefficient was 0.48 (95% CI = 0.46=0.49) and the weighted kappa 0.72. When a selected subsample was reclassified by the observers, the proportions of crude agreement on profusion of small opacities ranged from 42% to 47% (weighted kappa 0.52-0.55). The proportions of agreement on the main pleural abnormalities were 92% or over, and the corresponding kappa coefficients at least 0.73. The classification of lung opacities was subject to considerable observer variation, which calls for caution when results from different studies are compared. This variation, however, rarely exceeded one category, and thus appears to be small enough for meaningful comparisons between groups, at least within a single study.