Diagnostic accuracy of screening tests for patients suspected of COVID-19, a retrospective cohort study.

Moretti, Marco; Van Laethem, Johan; De Geyter, Deborah; Cools, Wilfried; Ilsen, Bart; Allard, Sabine Danielle; Pierard, Denis

BACKGROUND The diagnostic gold standard for Coronavirus-2019 disease (CoViD-19) is reverse transcriptase-polymerase chain reaction (RT-PCR). However, its sensitivity might be suboptimal. The current study aims to investigate predictive factors for false-negative nasopharyngeal RT-PCR in CoViD-19 patients. Additionally, the specificity and sensitivity of RT-PCR on the nasopharyngeal swab, serology and chest computerized-tomography (CCT) as a screening tool for the diagnosis of CoViD-19 were investigated. METHODS Medical records of patients admitted at the university hospital UZ Brussel during the CoViD-19 epidemic were reviewed. A group of CoViD-19 patients with false-negative RT-PCR was identified through scrupulous examination of medical records. Serological testing was performed through chemiluminescent microparticle assay. RESULTS Eighteen CoViD-19 patients with 'false negative' RT-PCR were identified and compared to 51 'true positives'. Logistic regression for prediction of 'false negative' RT-PCR found significantly higher serology results at hospitalization and more intensive care unit admission in the group with false-negative testing. In a cohort of 228 patients, the sensitivity of RT-PCR for the diagnosis of CoViD-19 was 85%. The sensitivity of serology was 86% and its specificity 92%. Chest computerized-tomography (CCT) showed a sensitivity of 93%, its specificity was 62%. By combining RT-PCR and serology results any 'false negative' could be excluded. CONCLUSIONS In this cohort, the sensitivity and specificity of RT-PCR and serology for the diagnosis of CoViD-19 were high and comparable. CCT had the highest sensitivity and confirmed its efficacy as a screening tool. CoViD-19 patients, who have a more severe presentation, might have negative RT-PCR and positive serology results.