Accuracy and interobserver agreement between MR-non-expert radiologists and MR-experts in reading MRI for suspected appendicitis

Abstract

Objective: To compare accuracy and interobserver agreement between radiologists with limited experience in the evaluation of abdominal MRI (non-experts), and radiologists with longer MR reading experience (experts), in reading MRI in patients with suspected appendicitis.

Methods: MR imaging was performed in 223 adult patients with suspected appendicitis and read independently by two members of a team of eight MR-inexperienced radiologists, who were trained with 100 MR examinations previous to this study (non-expert reading). Expert reading was performed by two radiologists with a larger abdominal MR experience (>500 examinations) in consensus. A final diagnosis was assigned after three months based on all available information, except MRI findings. We estimated MRI sensitivity and specificity for appendicitis and for all urgent diagnoses separately. Interobserver agreement was evaluated using kappa statistics.

Results: Urgent diagnoses were assigned to 147 of 223 patients; 117 had appendicitis. Sensitivity for appendicitis was 0.89 by MR-non-expert radiologists and 0.97 in MR-expert reading (p=0.01). Specificity was 0.83 for MR-non-experts versus 0.93 for MR-expert reading (p=0.002). MR-experts and MR-non-experts agreed on appendicitis in 89% of cases (kappa 0.78). Accuracy in detecting urgent diagnoses was significantly lower in MR-non-experts compared to MR-expert reading: sensitivity 0.84 versus 0.95 (p<0.001) and specificity 0.71 versus 0.82 (p=0.03), respectively. Agreement on urgent diagnoses was 83% (kappa 0.63).

Conclusion: MR-non-experts have sufficient sensitivity in reading MRI in patients with suspected appendicitis, with good agreement with MR-expert reading, but accuracy of MR-expert reading was higher.

Keywords: Abdominal pain; Magnetic resonance imaging appendicitis observer variation sensitivity and specificity.