Accuracy of O-RADS 4 and 5 Classification System in Diverse Patients in an Academic Health System: A Cohort Study [ID 1659]

INTRODUCTION: The Ovarian-Adnexal Reporting and Data System (O-RADS) 4 and 5 lesions correspond with a 10–50% and greater than 50% chance of malignancy, respectively. We examined the accuracy of these classifications across imaging type and patient demographic factors in predicting malignancy. METHODS: The university IRB reviewed and exempted this retrospective cohort study of patients with O-RADS 4 or 5 lesions on ultrasound (US) or magnetic resonance imaging (MRI) from July 1, 2020, to December 31, 2023 (n=425). We examined the positive predictive value (PPV) of O-RADS classifications in predicting malignancy in patients with diagnostic resolution, ie, who received surgery (n=266) or whose lesions resolved on follow-up imaging (n=44). RESULTS: Across all subgroups, the classifications corresponded with the percent chance of malignancy consistent with previously published studies: O-RADS 4 ranged from 0.13 to 1.0; O-RADS 5 ranged from 0.50 to 1.0. Both classifications were more accurate in MRI (0.91 and 0.28) than US (0.57 and 0.18). The PPVs of O-RADS 4 and 5 on US were similar across postmenopausal status: 17% and 56% PPV for premenopausal, 20% and 52% for postmenopausal for O-RADS 4 and 5, respectively. Stratifying by race, PPVs were within published estimates by imaging modality and O-RADS score. Rates of incomplete follow-up were higher for Black than White patients. CONCLUSIONS/IMPLICATIONS: We found that O-RADS scores remained highly predictive of malignancy in a clinical population diverse in age and race. There were modifiable disparities in follow-up by race. Validating diagnostic classification systems in diverse populations is essential for diagnostic and treatment equity.