Interobserver Variability in Ultrasound Reporting - Tertiary Hospital Radiologists Do Better

Abstract

Introduction: Thyroid Imaging Reporting and Data System (TI-RADS) was developed to provide a standardized risk-stratification system for patients with thyroid nodules. Single-center studies have demonstrated an acceptable level of interobserver agreement in applying TI-RADS in clinical practice, however data regarding consistency among different centers is limited. In Israel, thyroid nodules are initially evaluated by ultrasound performed by radiologists at the health maintenance organization (HMO) and then patients are referred to tertiary hospitals for ultrasound-guided fine needle aspiration (FNA) biopsy when indicated. Objective: To evaluate the interobserver concordance in TI-RADS classification system reporting between the HMO and a tertiary hospital. Methods: We performed a retrospective analysis of the sonographic features of 370 thyroid nodules TI-RADS category 2 or higher, from 350 patients evaluated by ultrasound at the HMO and at Hadassah Medical Center from January 1, 2018 to December 31, 2019. The primary outcome was concordance between the TI-RADS classification at the HMO compared to the hospital. Additional endpoints included correlation of TI-RADS to the Bethesda category following FNA, and correlation of TI-RADS with malignancy on final pathology. Results: Of 370 nodules, only 73 (19.8%) demonstrated concordance between the HMO and the hospital. The level of agreement was poor, with 277 (74.8%) nodules demonstrating higher TI-RADS at the HMO compared to the hospital, and 20 (5.4%) with lower TI-RADS at the HMO compared to the hospital (p<0.001, weighted Kappa = 0.120). Of the nodules referred to the hospital, 241 (65.1%) were selected for FNA. A strong correlation between the hospital TI-RADS and Bethesda category was demonstrated (p<0.001). Furthermore, 60 (16.2%) nodules were surgically removed. A strong correlation was identified between the hospital TI-RADS and malignancy on final pathology (p<0.001), yet there was no correlation with the TI-RADS of the HMO (p=0.346). Conclusions: There is poor concordance between TI-RADS classification on ultrasound performed in the HMO compared to a tertiary hospital. In patients who underwent FNA and eventually surgery, the hospital TI-RADS strongly correlated with Bethesda category and final risk of malignancy. Standardization of thyroid ultrasound terminology and dedicated training in thyroid imaging are needed to improve the interobserver concordance in clinical practice.