Chest ImagingE2325. Errors in Lung-RADS Reporting in a Lung Cancer Screening Program
Sakioka J, Guichet P, Prosper A, Liu B, Lee C. Keck School of Medicine of USC, Los Angeles, CA
Address correspondence to J. Sakioka (firstname.lastname@example.org)
Objective: Our objective was to investigate the frequency and types of errors in Lung-Reporting and Data System (RADS) assessment category reporting across the various radiologists at our institution interpreting lung cancer screening low-dose CT (LDCT) examinations.
Materials and Methods: Following interpretation by a dedicated cardiothoracic radiology attending, all screening LDCT examinations were reviewed at a multidisciplinary conference, with attendees including the lead radiologist of the lung cancer screening program as well as faculty from thoracic surgery and pulmonary medicine. Screening LDCT reports found to contain an incorrect Lung-RADS assessment category were subsequently updated with the correct Lung-RADS category by the original interpreting radiologist. Screening LDCT reports and addenda from May 2014 through August 2016 were retrospectively reviewed. LDCT report date, reading radiologist, and assigned Lung-RADS assessment category were recorded. If an LDCT report was updated, the rationale for the addendum and correct Lung-RADS category were also recorded. Radiologists were only included if they read a minimum of 10 screening LDCTs during the study period. Screening LDCT reports interpreted by the lead radiologist attending the multidisciplinary conferences were excluded.
Results: Since the implementation of Lung-RADS in May 2014, 270 screening LDCT examinations were interpreted by the included radiologists. Monthly screening LDCT volume increased during the study time period, growing from 3.5 LDCTs per month in 2014 to 11.5 in 2015, and 26.8 in 2016. Of the 270 LDCT reports, 29 (10.7%) initially contained an incorrect Lung-RADS assessment category and were subsequently updated with the correct Lung-RADS category. Overall, the rate of LDCT report addenda diminished over time, decreasing from 14.3 addenda per 100 LDCTs in 2014, to 13.6 in 2015, and 8.7 in 2016. Of the 29 updated reports, 59% (17/29) altered the patient’s recommended follow-up interval or management recommendations. The rate of LDCT report addenda for individual radiologists ranged from 8 to 15.8 addenda per 100 LDCTs. The most common error in Lung-RADS assessment category selection was inappropriate use of lung nodule size criteria (14/29 = 48%), typically resulting in overestimation of lung nodule malignancy risk (13/14 = 93%). The most common Lung-RADS category correction was downgrading a Lung-RADS 3 to Lung-RADS 2 (12/29 = 41%).
Conclusion: There is an adjustment and learning period following incorporation of Lung-RADS assessment categories into lung cancer screening reports. Meticulous referencing of the published ACR Lung-RADS criteria is beneficial, and accuracy improves with experience. Potential areas for radiologist education include specific nodule size criteria based on nodule type, stability of previous category 3 or 4 nodules resulting in downgrading to Lung-RADS 2, and the addition of the S modifier for significant incidental findings. Review of screening LDCTs in a conference setting is also advantageous.