Early prediction of mortality risk among patients with severe COVID-19, using machine learning

Abstract

Background: Coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 infection, has been spreading globally. We aimed to develop a clinical model to predict the outcome of patients with severe COVID-19 infection early.

Methods: Demographic, clinical and first laboratory findings after admission of 183 patients with severe COVID-19 infection (115 survivors and 68 non-survivors from the Sino-French New City Branch of Tongji Hospital, Wuhan) were used to develop the predictive models. Machine learning approaches were used to select the features and predict the patients' outcomes. The area under the receiver operating characteristic curve (AUROC) was applied to compare the models' performance. A total of 64 with severe COVID-19 infection from the Optical Valley Branch of Tongji Hospital, Wuhan, were used to externally validate the final predictive model.

Results: The baseline characteristics and laboratory tests were significantly different between the survivors and non-survivors. Four variables (age, high-sensitivity C-reactive protein level, lymphocyte count and d-dimer level) were selected by all five models. Given the similar performance among the models, the logistic regression model was selected as the final predictive model because of its simplicity and interpretability. The AUROCs of the external validation sets were 0.881. The sensitivity and specificity were 0.839 and 0.794 for the validation set, when using a probability of death of 50% as the cutoff. Risk score based on the selected variables can be used to assess the mortality risk. The predictive model is available at [https://phenomics.fudan.edu.cn/risk_scores/].

Conclusions: Age, high-sensitivity C-reactive protein level, lymphocyte count and d-dimer level of COVID-19 patients at admission are informative for the patients' outcomes.

Keywords: COVID-19; death; fatality rate; machine learning; predictive model.

Figure 1

The study flow chart. bagFDA, bagged flexible discriminant analysis; PLS, partial least squares; AUROC, area under the receiver operating characteristic curve

Figure 2

The top 20 important variables selected by five machine learning models (A-E) and the model performance based on the selected variables (F). NT-proBNP, N-terminal pro-brain natriuretic peptide; hsCRP, high-sensitivity C-reactive protein; CRE, creatinine; ALT, alanine aminotransferase; IL8, interleukin 8; LYM, lymphocyte count; IL6, interleukin 6; WC, white cell count; eGFR, estimated glomerular filtration rate; ALP, alkaline phosphatase; RC, red cell count; TCH, total cholesterol; FIB, fibrinogen; PLT, platelet count; TB, total bilirubin; ALB, albumin; PT, prothrombin time; IL10, interleukin 10; IL2R, interleukin-2 receptor; FER, ferritin; LDH, lactate dehydrogenase; hscTnI, high-sensitivity cardiac troponin I; Mono, monocyte count; UA, uric acid; ESR, erythrocyte sedimentation rate; No.Com., number of basic conditions

Figure 3

The area under the receiver operating characteristic curve (AUROC) of the logistic regression model based on selected variables in the derivation set (A) and the external validation set (B)

Figure 4

The relationship between probability of death and the normalized probability (A) and the cost curves of the predictive models using different cutoffs (B)

Figure 5

The formula to calculate the risk scores (A) and their distributions among survivors and non-survivors (B) and the corresponding probability of death (C)