Indicators and prediction models for the severity of Covid-19

Abstract

Objectives: Coronavirus disease 2019 (Covid-19) is outbreaking globally. We aimed to analyse the clinical characteristics, cardiac injury, electrocardiogram and computed tomography (CT) features of patients confirmed Covid-19 and explored the prediction models for the severity of Covid-19.

Methods: A retrospective and single-centre study enrolled 98 laboratory-confirmed Covid-19 patients. Clinical data, electrocardiogram and CT features were collected and analysed using Statistical Package for the Social Sciences software.

Results: There were 46 males and 52 females, with a median age of 44 years, categorised into three groups, including mild, moderate and severe/critical Covid-19. The rate of abnormal electrocardiograms in severe/critical group (79%) was significantly higher than that in the mild group (17%) (P = .027), which (r = 0.392, P = .005) positively related to the severity of Covid-19 (OR: 5.71, 95% CI: 0.45-3.04, P = .008). Age older than 60 years old, comorbidities, whether had symptoms on admission, fatigue, CT features, laboratory test results such as platelet count, lymphocyte cell count, eosinophil cell count, CD3+ cell count, CD4+ cell count, CD8+ cell count, the ratio of albumin/globulin decreased and D-dimer, C-reactive protein (CRP), B-type natriuretic peptide (BNP), cardiac troponin I (cTnI) elevated were the risk factors for the increased severity of Covid-19. The logistic model, adjusted by age, lobular involvement score and lymphocyte cell count, could be applied for assessing the severity of Covid-19 (AUC, 0.903; Sensitivity, 90.9%; Specificity, 78.1%).

Conclusions: Age >60 years old, chronic comorbidities, lymphocytopoenia and lobular involvement score were associated with the Covid-19 severity. The inflammation induced by Covid-19 caused myocardial injury with elevated BNP and cTnI level and abnormal electrocardiograms.

FIGURE 1

Transverse thin‐section CT scan for the moderate patients of Covid‐19. A, a 36‐year‐old male was admitted to the hospital because of cough for four days and fever for half‐day and diagnosed as moderate of Covid‐19. CT demonstrated pure ground‐glass opacity in the left superior lobe. The lobular involvement score was one point, with the lobe involved less than 25%. B, a 36‐year‐old female was admitted to the hospital because of fever for 3 days, and diagnosed as moderate of Covid‐19. CT revealed a subpleural confounding opacity lesion in the lower lobe of the left lung, and a patchy ground‐glass opacity lesion with blurred boundaries in the lower lobe of the right lung (yellow arrows). The lobular score was one each in the left inferior lobe and the right inferior lobe, respectively, because of the involved area less than 25%. C, An 80‐year‐old female was admitted to the hospital because of diarrhoea, dyspnoea, anorexia, fatigue and muscular soreness, no fever or cough and diagnosed as severe of Covid‐19. CT on admission showed diffuse lesions, mainly ground‐glass opacity accompanied by partial consolidation, crazy‐paving pattern in bilateral lungs, which were chiefly distributed under the pleura. The total lobular involvement score was 11, with two in the left superior lobe, two in the left inferior lobe, two in the right superior lobe, two in the right medial lobe and three in the right inferior lobe. The patient was also with atherosclerosis of aorta and coronary artery and a small amount of effusion in bilateral pleura. D, a 36‐year‐old male was admitted to the hospital because of generalised fatigue, muscular soreness for four days, and fever for three days, and diagnosed as severe of Covid‐19. CT on admission indicated diffused irregular mixed patchy and ground‐glass opacity lesions in bilateral lungs. The total lobular involvement score was 14, with three in the left superior lobe, three in the left inferior lobe, three in the right superior lobe, two in the right medial lobe, and three in the right inferior lobe

FIGURE 2

The ROC curves for the three logistic models. Model 1: Logit (P) = −2.942 + 0.311X, (X = Lobular involvement score); ROC curve: AUC, 0.832 (95%CI: 0.725‐0.939, P < .001); cut‐off, 0.43; sensitivity, 68.2%; specificity, 90.6%. Model 2: Logit (P) = −5.905 + 0.059X ₁ + 0.285X ₂, (X ₁ = Age, X ₂ = Lobular involvement score); ROC curve: AUC, 0.876 (95%CI: 0.800‐0.952, P < .001); cut‐off, 0.19; sensitivity, 86.4%; specificity, 78.1%. Model 3: Logit (P) = −3.504 + 0.53 X ₁ + 0.266 X ₂‐1.428X ₃ (X ₁ = Age; X ₂ = Lobular involvement score; and X ₃ = lymphocyte cell count); ROC curve: AUC, 0.903 (95%CI: 0.838‐0.967, P < .001); cut‐off value, 0.18; sensitivity, 90.9%; specificity, 78.1%