Compare the diagnostic properties of five case definitions of suspected COVID-19 that were used or proposed in Chile during the first eight months of the pandemic.
An analysis was done of the diagnostic properties (sensitivity, specificity, and positive and negative predictive values) of three case definitions of suspected COVID-19 used in Chile between March and October 2020, as well as two alternative proposed definitions. The sample was 2,019 people with known results for the polymerase chain reaction (PCR) test for SARS-CoV-2. Stepwise logistic regression was used to develop criterion 5, optimizing sensitivity and specificity values. Multifactor logistic regression was used to explore the association between demographic variables, symptoms and signs, and PCR positivity. Different positivity scenarios were analyzed and ROC curves were compared.
The presence of anosmia (OR = 8.00; CI95%: 5.34–11.99), fever (OR = 2.15; CI95%: 1.28–3.59), and having been in close contact with a person sick with COVID-19 (OR = 2.89; CI95%: 2.16–3.87) were associated with a positive PCR result. According to the analysis of the ROC curve, criterion 5 had the highest capacity for discrimination, although there were no significant differences with the other four criteria.
Criterion 5—based on anosmia, close contact with people with COVID-19, and fever as sufficient unique elements—was the most sensitive in identifying suspected cases of COVID-19, a key aspect in controlling the spread of the pandemic.