Identifying COVID-19 cases in primary care settings

Mao, Y.; Tan, Y.-R.; Thein, T.-L.; Chai, L.; Cook, A.; Dickens, B.; Lew, Y.-J.; Lim, F. S.; Lim, J. T.; Sun, Y.; Sundaram, M.; Soh, A.; Tan, G.; Wong, F.; Young, B.; Zeng, K.; Chen, M.; Ong, D.

Case identification is an ongoing issue for the COVID-19 epidemic, in particular for outpatient care where physicians must decide which patients to prioritise for further testing. This paper reports tools to classify patients based on symptom profiles based on 236 SARS-CoV-2 positive cases and 564 controls, accounting for the time course of illness at point of assessment. Clinical differentiators of cases and controls were used to derive model-based risk scores. Significant symptoms included abdominal pain, cough, diarrhea, fever, headache, muscle ache, runny nose, sore throat, temperature between 37.5{degrees}C and 37.9{degrees}C, and temperature above 38{degrees}C, but their importance varied by day of illness at assessment. With a high percentile threshold for specificity at 0.95, the baseline model had reasonable sensitivity at 0.67. To further evaluate accuracy of model predictions, we firstly used leave-one-out cross-validation, which confirmed high classification accuracy with an area under the receiver operating characteristic curve of 0.92. For the baseline model, sensitivity decreased to 0.56. Secondly, in a separate ongoing prospective study of 237 COVID-19 and 346 primary care patients presenting with symptoms of acute respiratory infection, the baseline model had a sensitivity of 0.57 and specificity of 0.89, and in retrospective notes review of 100 COVID-19 cases diagnosed in primary care, sensitivity was 0.56. A web-app based tool has been developed for easy implementation as an adjunct to laboratory testing to differentiate COVID-19 positive cases among patients presenting in outpatient settings.