A novel artificial intelligence-assisted triage tool to aid in the diagnosis of suspected COVID-19 pneumonia cases in fever clinics

摘要: Background Currently, the need to prevent and control spread of 2019 novel coronavirus disease (COVID-19) outside Hubei province in China internationally has become increasingly critical. We developed validated a diagnostic model that does not rely on computed tomography (CT) images aid early identification suspected COVID-19 pneumonia (S-COVID-19-P) patients admitted adult fever clinics made available via an online triage calculator. Methods Patients from January 14 February 26, 2020 with epidemiological history exposure were included study [model development group (n=132) validation (n=32)]. Candidate features clinical symptoms, routine laboratory tests, other information admission. The selection based least absolute shrinkage operator (LASSO) regression. primary outcome was for S-COVID-19-P Results cohort contained 26 cases seven confirmed (C-COVID-19-P). final selected one demographic variable, four vital signs, five blood values, signs infection-related biomarker. model's performance testing set resulted area under receiver operating characteristic (ROC) curves (AUCs) 0.841 0.938, F1 scores 0.571 0.667, recall 1.000 1.000, specificity 0.727 0.778, precision 0.400 0.500, respectively. top most important age, interleukin-6 (IL-6), systolic pressure (SYS_BP), monocyte ratio (MONO%), classification (FC). Based this model, optimized strategy also been designed. Conclusions A machine-learning solely CT able perform admission 100% score. This high-performing deployed as tool, which is at https://intensivecare.shinyapps.io/COVID19/.