A Novel Triage Tool of Artificial Intelligence-Assisted Diagnosis Aid System for Suspected COVID-19 Pneumonia in Fever Clinics
Open Access
- 20 March 2020
- preprint content
- other
- Published by Cold Spring Harbor Laboratory
Abstract
Summary: Background: Currently, the prevention and control of the novel coronavirus disease (COVID-19) outside Hubei province in China, and other countries have become more and more critically serious. We developed and validated a diagnosis aid model without computed tomography (CT) images for early identification of suspected COVID-19 pneumonia (S-COVID-19-P) on admission in adult fever patients and made the validated model available via an online triage calculator.Methods: Patients admitted from Jan 14 to February 26, 2020 with the epidemiological history of exposure to COVID-19 were included [Model development (n = 132) and validation (n = 32)]. Candidate features included clinical symptoms, routine laboratory tests, and other clinical information on admission. Features selection and model development were based on the least absolute shrinkage and selection operator (LASSO) regression. The primary outcome was the development and validation of a diagnostic aid model for S-COVID-19-P early identification on admission.Results: The development cohort contained 26 S-COVID-19-P and 7 confirmed COVID-19 pneumonia cases. The final selected features included 1 variable of demographic information, 4 variables of vital signs, 5 variables of blood routine values, 7 variables of clinical signs and symptoms, and 1 infection-related biomarker. The model performance in the testing set and the validation cohort resulted in the area under the receiver operating characteristic (ROC) curves (AUCs) of 0.841 and 0.938, the F-1 score of 0.571 and 0.667, the recall of 1.000 and 1.000, the specificity of 0.727 and 0.778, and the precision of 0.400 and 0.500. The top 5 most important features were Age, IL-6, SYS_BP, MONO%, and Fever classification. Based on this model, an optimized strategy for S-COVID-19-P early identification in fever clinics has also been designed.Conclusions: S-COVID-19-P could be identified early by a machine-learning model only used collected clinical information without CT images on admission in fever clinics with a 100% recall score. The well-performed and validated model has been deployed as an online triage tool, which is available athttps://intensivecare.shinyapps.io/COVID19/.Keywords
This publication has 21 references indexed in Scilit:
- A novel coronavirus outbreak of global health concernThe Lancet, 2020
- Clinical features of patients infected with 2019 novel coronavirus in Wuhan, ChinaThe Lancet, 2020
- A familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family clusterThe Lancet, 2020
- Machine Learning in MedicineThe New England Journal of Medicine, 2019
- Emergency department triage prediction of clinical outcomes using machine learning modelsCritical Care, 2019
- AKIpredictor, an online prognostic calculator for acute kidney injury in adult critically ill patients: development, validation and comparison to serum neutrophil gelatinase-associated lipocalinIntensive Care Medicine, 2017
- A targeted real-time early warning score (TREWScore) for septic shockScience Translational Medicine, 2015
- Regularization Paths for Conditional Logistic Regression: TheclogitL1PackageJournal of Statistical Software, 2014
- Assessing the Performance of Prediction ModelsEpidemiology, 2010
- Plasma inflammatory cytokines and chemokines in severe acute respiratory syndromeClinical and Experimental Immunology, 2004