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

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

Analysis of clinical characteristics of critically ill patients during prevention and control of new coronavirus pneumonia

OBJECTIVE: To analyze the clinical characteristics of critically ill patients during the prevention and control of new coronavirus pneumonia and to improve the diagnosis and treatment of critically ill patients with fever. METHODS: A retrospective case analysis method was used to collect clinical data of 80 patients admitted to the critically ill area of fever clinic in the first medical center of PLA general hospital from February to March,2020. Pharyngeal swab specimens were collected from all patients at 24 h intervals and tested for new coronavirus nucleic acid twice.At the same time, a combination of influenza A and B virus nucleic acid were determined, and lung computer tomography scan examination were performed. The differences in clinical data of patients with different blood leukocyte count groups were compared. RESULTS: All patients had negative results for new coronavirus nucleic acid tests, and 2 patients had a positive test for influnenza A virus nucleic acid. One patient was tested positive for hepatitis B virus nucleic acid. There were 60 patients(75.00%) with age over 65 years, 47 patients(58.75%) with more than two basic medical diseases, 70 patients(87.50%) with lung CT indicating lesions, 30 patients(37.50%) with acute kidney injury. Compared with patients with while blood cell count <1010~9/L, patients with WBC>=1010~9/L had higher incidence of diabetes mellitus WBC(P=0.024). There was no significant difference in retention time, body temperature, oxygenation index, and incidence of acute kidney injury between two groups of patients. CONCLUSION: The critically ill patients diagnosed in hospital had the characteristics of old age, many chronic diseases, widespread lung disease, and prone to acute kidney injury during screening. Nucleic acid detection is the most important differential diagnosis basis for the screening of new coronavirus infection.

Prediction of fatal adverse prognosis in patients with fever-related diseases based on machine learning: A retrospective study.

BACKGROUND: Fever is the most common chief complaint of emergency patients. Early identification of patients at an increasing risk of death may avert adverse outcomes. The aim of this study was to establish an early prediction model of fatal adverse prognosis of fever patients by extracting key indicators using big data technology. METHODS: A retrospective study of patients' data was conducted using the Emergency Rescue Database of Chinese People's Liberation Army General Hospital. Patients were divided into the fatal adverse prognosis group and the good prognosis group. The commonly used clinical indicators were compared. Recursive feature elimination (RFE) method was used to determine the optimal number of the included variables. In the training model, logistic regression, random forest, adaboost and bagging were selected. We also collected the emergency room data from December 2018 to December 2019 with the same inclusion and exclusion criterion. The performance of the model was evaluated by accuracy, F1-score, precision, sensitivity and the areas under receiver operator characteristic curves (ROC-AUC). RESULTS: The accuracy of logistic regression, decision tree, adaboost and bagging was 0.951, 0.928, 0.924, and 0.924, F1-scores were 0.938, 0.933, 0.930, and 0.930, the precision was 0.943, 0.938, 0.937, and 0.937, ROC-AUC were 0.808, 0.738, 0.736, and 0.885, respectively. ROC-AUC of ten-fold cross-validation in logistic and bagging models were 0.80 and 0.87, respectively. The top six coefficients and odds ratio (OR) values of the variables in the Logistic regression were cardiac troponin T (CTnT) (coefficient=0.346, OR = 1.413), temperature (T) (coefficient=0.235, OR = 1.265), respiratory rate (RR) (coefficient= -0.206,OR = 0.814), serum kalium (K) (coefficient=0.137, OR = 1.146), pulse oxygen saturation (SPO2) (coefficient= -0.101, OR = 0.904), and albumin (ALB) (coefficient= -0.043, OR = 0.958). The weights of the top six variables in the bagging model were: CTnT, RR, lactate dehydrogenase, serum amylase, heartrate, and systolic blood pressure. CONCLUSIONS: The main clinical indicators of concern included CTnT, RR, SPO2, T, ALB and K. The bagging model and logistic regression model had better diagnostic performance comprehesively. Those may be conducive to the early identification of critical patients with fever by physicians.