Comparison of Clinicians to Radiologists in Assessment of Abdominal Radiographs for Suspected Intussusception

Objectives: The objectives of this study were to determine agreement of abdominal radiography (AXR) interpretation and to compare test characteristics, between pediatric emergency medicine (PEM) physicians and pediatric radiologists for evaluation of intussusception. Methods: This was a retrospective cohort study of children aged 3 months to 3 years presenting to a pediatric emergency department (ED) between 2007 and 2009. For inclusion, subjects required an ED presentation for suspected intussusception, performance of a 2-view AXR (supine and upright/lateral decubitus views) and abdominal ultrasound performed during the ED visit, and a subsequent criterion-standard measure for intussusception available (contrast enema, operative report, or clinical follow-up). All AXRs were reviewed by 2 blinded PEM physicians and 2 pediatric radiologists. Interrater agreement (kappa) for AXR interpretation and diagnostic test characteristics were calculated for comparison. Results: A total of 286 children were included; intussusception was present in 61 patients (21.3%). Moderate agreement was present between the PEM physicians and radiologists for AXR assessment, with kappa = 0.57 (95% confidence interval [CI], 0.47-0.66). Radiologist AXR interpretation had specificity of 86.7% (95% CI, 81.6-90.5), sensitivity of 62.3% (95% CI, 49.8-73.4), positive predictive value of 55.9% (95% CI, 43.3-67.9), and negative predictive value of 89.4% (95% CI, 84.6-93.2). Pediatric emergency medicine physician AXR interpretation had specificity of 68.9% (95% CI, 62.6-74.6), sensitivity of 78.7% (95% CI, 66.9-87.1), positive predictive value of 40.7% (95% CI, 32.2-49.7), and negative predictive value of 92.3% (95% CI, 87.1-95.5). Conclusions: Agreement between PEM physicians and pediatric radiologists for evaluation of AXR in suspected intussusception is moderate. Sensitivity and negative predictive value of AXR alone are not sufficiently high overall to exclude intussusception; however, PEM physician interpretation of AXR may possess utility in determining need for abdominal ultrasound in low-risk patients given the high negative predictive value.