Related Experiment Videos
Variability in the classification of radiographs using the 1980 International Labor Organization Classification for
L S Welch1, K L Hunting, J Balmes
1Department of Occupational/Environmental Medicine, Washington Hospital Center, Washington, DC 20010-2975, USA.
Chest
|January 1, 1999
Summary
Interreader agreement for classifying chest radiographs using the International Labor Organization (ILO) classification was assessed. Five of six readers showed fair to good agreement for profusion and pleural findings, indicating reliable classification for pneumoconiosis detection.
Area of Science:
- Radiology
- Occupational Health
- Medical Imaging Analysis
Background:
- The International Labor Organization (ILO) classification is crucial for standardizing the assessment of pneumoconiosis from chest radiographs.
- Evaluating interreader agreement is essential for ensuring the reliability and consistency of diagnostic interpretations in occupational lung disease screening.
Purpose of the Study:
- To quantify the extent of agreement among multiple readers in classifying chest radiographs for pneumoconiosis using the ILO classification system.
- To assess agreement on profusion of small opacities and the presence of pleural abnormalities.
Main Methods:
- Six readers (three from Canada, three from the US) evaluated 119 chest radiographs.
- Radiographs were classified for profusion (0/- to 3/+) and pleural abnormalities (yes/no) according to the ILO system.
- Statistical methods including concordance, observed agreement, and kappa statistic were employed to evaluate interreader agreement.
Main Results:
- Five of six readers demonstrated fair to good interreader agreement for both pleural findings and profusion (dichotomous: >=1/0 vs <=0/1), as measured by the kappa statistic.
- One reader exhibited poor agreement, highlighting potential variability in interpretation.
- Concordance was higher for normal radiographs compared to those with disease, and the kappa statistic was used to account for this bias.
Conclusions:
- The study indicates generally good interreader reliability for ILO classification of pneumoconiosis among experienced readers, with some individual variability.
- The kappa statistic is a valuable tool for assessing agreement, controlling for chance, and identifying potential biases like 'underreading' or 'overreading'.