RESUMO
OBJECTIVE: The diagnosis of SARS-CoV-2 infection presents some limitations. RT-PCR in nasopharyngeal swabs is considered the gold standard for the diagnosis, although it can have false negative results. We aimed to analyze the accuracy of repeating nasopharyngeal swabs based on different clinical probabilities. METHODS: Retrospective observational study of the first patients admitted to a two COVID Internal Medicine wards at the University Hospital Marqués de Valdecilla, Santander, from March to April 2020. RT-PCR targering E, N, RdRP and ORFab1 genes and antibody tests detecting IgG. RESULTS: A total of 145 hospitalized patients with suspected SARS-Cov2 infection were admitted and in 98 (67.5%) diagnosis was confirmed. The independent predictive variables for SARS-CoV-2 infection were: epidemiological contact, clinical presentation as pneumonia, absence of pneumonia in the last year, onset of symptoms > 7 days, two or more of the following symptoms -dyspnea, cough or fever- and serum lactate dehydrogenase levels >350 U/L (p<0.05). A score based on these variables yielded an AUC-ROC of 0.89 (CI95%, 0.831-0.946; p<0.001). The accuracy of the first nasopharyngeal swabs was 54.9%. Repeating nasopharyngeal swabs two or three times allows to detect an additional 16% of positive cases. The overall accuracy of successive RT-PCR tests in patients with low pre-test probability was <5%. CONCLUSIONS: We have defined a pre-test probability score based on epidemiological and clinical data with a high accuracy for diagnosis of SARS-CoV-2. Repeating nasopharyngeal swabs avoids sampling errors, but only in medium of high probability pre-test clinical scenarios.