Classification of Diverticular Disease (CDD) – assessment of the intra- and interobserver agreement in abdominal CT scans

Abstract

Purpose Along with ultrasound, computed tomography (CT) is one of the imaging modalities of choice in patients with suspected diverticular disease (DD). Recently, a newer Classification of Diverticular Disease (CDD) has been proposed. However, its reliability in daily radiological practice has never been proven. Therefore, our aim was to evaluate the intra- and interobserver agreement of the CDD in abdominal CT scans. Methods In this retrospective study, 481 CT scans of patients with suspected DD were included. Two readers (one board-certified radiologist with 6 years of experience, one 3 rd year radiology resident) individually evaluated all CTs in two reading sessions using the CDD. A composite endpoint of a prior consensus reading, follow-up, and intraoperative findings served as the reference. Intra- and interobserver agreement were calculated using Cohen-k statistic. Results DD was present in 317 cases (66 %), mostly classified as CDD stage 0, 1b, and 2a (28 %, 30 %, und 14 %). Intraobserver agreement was almost perfect for both readers (kappa 0.93 and 0.88). Interobserver agreement was high and improved from substantial (kappa 0.77) in the first reading session to almost perfect (kappa 0.84) in the second reading session. The interobserver agreement was best for CDD types 0 (diverticulosis) and 2c (free perforated diverticulitis) (mean kappa 0.83 and 0.86) and poorest for CDD types 1a (diverticulitis without phlegmon) and 2b (covered diverticulitis with macroabscess) (mean kappa 0.17 and 0.38). Intra- and interobserver agreement of acute uncomplicated (CDD type 1) and acute complicated diverticulitis (CDD type 2) were substantial to almost perfect (mean kappa 0.63–0.86). Agreement with the reference was almost perfect for both observers (mean kappa 0.86 and 0.82). Administration of rectal contrast did not significantly improve the diagnosis. Conclusion The CDD is a classification based on relatively clear imaging characteristics, which can be readily applied by radiologists with different expertise. In our study, the CDD had a high intra- and interobserver agreement, enabling a reliable therapy-related categorization of DD. Key Points: