Purpose or Case Report: Small airways disease (SAD) is a common pathology seen in pediatric patients. In suspected cases of SAD/RSV/bronchiolitis, chest radiographs are often ordered to diagnose disease and to rule out other conditions, such as bacterial pneumonia. While SAD and pneumonia have unique radiographic features, the threshold to diagnose such findings may depend on radiologist experience and training background. The purpose of this study is to investigate intra-radiologist precision and accuracy in diagnosing normal, SAD, and bacterial pneumonia on pediatric chest radiography.
Methods & Materials: An IRB-approved retrospective review was performed of all stat pediatric chest radiographs performed in patients aged 6 months to 6 years over a one-year period. Three fellowship-trained pediatric radiologists at different stages of their careers (early-, mid-, and late-career) assigned a diagnosis of normal, SAD or pneumonia to each of 100 randomly selected qualifying cases. Two to four weeks later, this process was repeated with the 100 cases presented in a different order. Intra-radiologist agreement was then assessed.
Results: Overall intra-radiologist agreement held an average kappa value of 0.64, indicating “good” agreement. Intra-radiologist agreement was 78% for the early-career pediatric radiologist (kappa value of 0.65), 79% for the mid-career pediatric radiologist (kappa value of 0.66), and 76% for the late-career pediatric radiologist (kappa value of 0.61). For all cases, accuracy was 71.5%, 68% and 69% for the early-career, mid-career, and late-career pediatric radiologist, respectively. In the first round, all three pediatric radiologists agreed on the diagnosis of “normal” 71% of the time, “SAD” 66% of the time, and “Pneumonia” 95% of the time.
Conclusions: Our findings demonstrate that there is no statistically significant difference in the precision or accuracy with which pediatric radiologists of different levels of training interpret pediatric chest radiographs. Nevertheless, both overall precision and accuracy were lower than anticipated, which appears to be primarily due to inconsistency in diagnosing SAD. Ongoing studies comparing radiologists of different subspecialties and comparing staff and resident radiologists are currently in process and will allow for more complete characterization of this study’s conclusions.