Summary

Background:

Chest radiograph is a standard procedure for diagnosis of pneumonia; however, interpretation shows considerable variability among observers.

Objectives:

To assess the extent of agreement between pediatric residents and board-certified radiologists in interpretation of chest radiography for detection of pneumonia. To evaluate the impact of resident experience, patient age, and signs of infection on this phenomenon.

Methods

: The cohort included 935 patients with suspected pneumonia admitted to the pediatric emergency department at a non-tertiary medical center in Israel 2019–2021. All patients had chest radiographs interpreted by a resident and a radiologist. Interobserver agreement was assessed using Κ and prevalence-adjusted bias-adjusted κ (PABAK) with 95% confidence intervals (95%CI). Results were stratified by resident experience (junior or senior), patient age (≤ 3 vs. > 3 years), white blood cells (≤ 15,000 vs. > 15,000 cells/ml), C-reactive protein (≤ 5 vs. > 5.0 mg/dl), and temperature (< 38.0°C vs. ≥ 38.0°C).

Results:

Moderate agreement between pediatric residents and radiologists was demonstrated for diagnosis of pneumonia (κ= 0.45). After adjustment for disease prevalence, the extent of agreement increased to near-substantial (PABAK= 0.59, 95% confidence interval 0.54–0.64). The extent of agreement was higher for children over 3 years of age and in patients without clinical or biochemical features of pneumonia, especially when diagnosis of pneumonia was ruled out.

Conclusions:

A second reading of chest radiographs by an experienced radiologist should be considered, particularly for patients younger than 3 years of age and in those with signs of infection and an initial diagnosis of pneumonia.