E2901. Interpretation Patterns of Pediatric MRI for Suspected Appendicitis in Pediatric Radiologists Versus Emergency Department Radiologists
Authors
Valeria Pena-Trujillo;
Harvard Medical School; Massachusetts General Hospital
Sebastian Gallo-Bernal;
Harvard Medical School; Massachusetts General Hospital
Maria Gabriela Figueiro Longo;
Harvard Medical School; Massachusetts General Hospital
Anand M. Prabhakar;
Harvard Medical School; Massachusetts General Hospital
Randheer Shailam;
Harvard Medical School; Massachusetts General Hospital
Michael S. Gee;
Harvard Medical School; Massachusetts General Hospital
Teresa Victoria;
Harvard Medical School; Massachusetts General Hospital
Objective:
To identify and characterize the differences in the interpretation patterns of abdominal MRI for suspected appendicitis in children between pediatric radiologists and emergency-department (ED) radiologists at a tertiary care pediatric emergency department.
Materials and Methods:
This retrospective, IRB-approved, and HIPPA-compliant study examined the imaging reports of all pediatric patients with clinical suspicion of appendicitis who underwent a standard-of-care MRI over five years (1/1/2017 – 3/31/2022). From a total of 243 examinations that met the inclusion criteria, a balanced cohort of 122 patients (51.6% male; median age ±IQR = 11 years ±7 - 13) encompassing 31 appendicitis cases was selected for further analysis. Anonymized MRI images were prospectively and independently interpreted by two fellowship-trained pediatric radiologists and two ED radiologists in a blinded manner. They were asked to categorize each case as positive or negative for appendicitis and assess the visualization of the appendix and the presence of direct and indirect signs of appendicitis using cloud-based software (Pacsbin, Orion Medical Technologies, Baltimore, MD). The results were then compared with the final diagnosis in EPIC (surgical consultation and pathology reports) to evaluate the readers' overall accuracy, sensitivity, and specificity. The total time to interpretation was also registered. Categorical variables were compared using X2-test and ordinal variables using the Wilcoxon rank test. Interobserver agreements were assessed using k statistics. Analyses were performed in R v4.0.5.
Results:
Overall accuracy (true positive + true negative) was was 95.1% for the pediatric radiologists and 86.1% for the ED radiologists (p = 0.037). Moderate agreement was found between the pediatric and ED radiologists for the diagnosis of appendicitis (29/122 vs. 42/122; p = <0.01), and the presence of direct (25/122 vs. 38/122; p = <0.01) and indirect signs of appendicitis (40/122 vs. 34/122 p = <0.01) (k:0.588; 0.599; and 0.575, respectively) and slight agreement (k: 0.220) for the visualization of the appendix (89/122 vs. 71/122 p = 0.010). The pediatric radiologist saw the appendix in 73% of cases vs. 58.2% for the ER radiologists. Sensitivity was higher for the ED radiologists (90.3% vs. 87%), whereas specificity was higher for the pediatric radiologists (97% vs. 84%). A statistically significant difference (p = 0.042) was found in the total time of interpretation, 115 seconds (67-166) for the pediatric radiologists vs. 81 seconds (36 - 190) for the ED radiologists.
Conclusion:
Our findings denote different patterns of imaging interpretation, reflecting the daily life requirements of the contrasting clinical setting: excellent sensitivity for a prompt diagnosis in the emergency department (which could save lives) and exceptional specificity for the pediatric setting to clarify not-so-evident and complex cases.
