Discrepancy between Reviewing Clinicians and Radiologists in Reporting of Chest Radiograph Signs of Coronavirus Disease 2019 (COVID-19)

Author(s): Keir Ovington, Rhys Metters

Introduction: Chest radiographs form an important aid to COVID-19 diagnosis, however their utility is limited by the reviewers ability to accurately assess for its radiological features, This study seeks to assess for any difference in reporting between radiologists and clinicians.

Methods: 135 admission chest radiographs of patients without a known COVID-19 diagnosis were gathered opportunistically. Radiologist reports, reviewing clinicians and clinicians with no knowledge of the patient categorised radiographs as having either “no covid signs” (category 0), “indeterminate covid signs” (cat. 1) or “classic/probable covid signs” (cat. 2). Cohen’s Kappa was used to evaluate the inter-reporter reliability between these groups.

Results: Radiologists identified 69% as cat. 0, 29% as cat. 1, 5% as cat. 2. Reviewing clinicians agreed with 73% of these reports achieving a Kappa of 0.43 (95% CI 0.32 - 0.54). Consultants performed best with a kappa of 0.77 (0.56-0.98). Clinicians without knowledge of the patient agreed with 54% of reports, Kappa 0.17 (-0.16 -0.50).

Conclusion: There is a significant discrepancy between radiologist and non-radiologist reporting of chest radiographs in COVID-19 supporting the use of rapid radiologist reporting of chest radiographs to aid with diagnosis.

Advances in Knowledge: This is the first paper to our knowledge to assess the difference in reporting of COVID-19 between radiologists and reviewing clinicians, indicating that radiologist reporting of chest x-rays has a measurable advantage in detecting COVID-19 signs, compared to clinician reports alone.