This was a retrospective longitudinal cohort study involving COVID-19 patients that were confirmed by reverse transcriptase polymerase chain reaction (RT-PCR), based on the Chinese diagnosis and treatment protocol for COVID-19 (Trial Version 7) (10) admitted at the Jinyintan Hospital. Epidemiological, clinical, and laboratory data for patients were collected between January 24, 2020 and March 31, 2020, at which point patients were admitted and no more patients were admitted. Individuals who were 18 years of age or older, and who tested positive for COVID-19 by real-time RT-PCR were included in this study. No exclusion criteria were applied.
Nose and throat swab specimens or nasopharyngeal aspirates collected from patients were tested for COVID-19 using RT-PCR. Epidemiological information, medical history, and clinical symptoms were recorded upon admission. Imaging and laboratory tests were routinely performed for all patients. Data were extracted from hospital information systems (HISs) using a standardized data collection form. All data were recorded daily. A group of experienced physicians determined the frequency of examinations and then reviewed and cross-checked the data. Implausible values for variables were identified with the data entry department and confirmed by physicians.
An axillary temperature of at least 37.3 °C was considered as fever. The six-point ordinal scale of clinical status for respiratory support was categorized as follows: 1) 6 points, death; 2) 5 points, need for invasive ventilator; 3) 4 points, high flow humidification oxygen/high frequency oxygen/non-invasive ventilator are needed; 4) 3 points, low flow (1 liters – 5 liters per minute) face mask/nasal cannula for oxygen inhalation; 5) 2 points, oxygen saturation of above 93 without oxygen inhalation; 6) 1 point, discharge (11). Mild illness was defined as 2 points, while severe illness was defined as 3 points or above. Escalation of respiratory support was defined as a transition from 2 or 3 points to 4 points or higher.
Chest x-ray and chest computed tomography abnormality (CXR): 0, bilateral pneumonia with no abnormal lesions (unilateral or bilateral pneumonia); 1, unilateral pneumonia with multiple mottling and ground-glass opacity; 2, bilateral pneumonia with multiple mottling and ground-glass opacity.
The practical definition of liver dysfunction was modified from studies on the Child-Pugh score, and other studies that reported that liver dysfunction is highly correlated with COVID-19 progression and associated mortality (12-14). The MCP was defined as follows (Table 1): Aspartate aminotransferase (AST) (U/L): <35, 1 point; 35–70, 2 points; >70, 3 points (35 U/L is the normal standard in Jinyintan Hospital). Total bilirubin (TBIL) (μmol/L): <34, 1 point; 34–51, 2 points; >51, 3 points. Albumin (ALB)(g/L): >35, 1 point; 28–35, 2 point; <28, 3 points. Prothrombin time (PT) (s): prolong <4, 1 point; 4–6, 2 points; >6, 3 points. It was graded as 5 points or below for Child-Pugh-A; 6–9 points for Child-Pugh-B; and 10–15 points for Child-Pugh-C. The AST cut-off values were obtained from literature reviews (8,15-16) and extensive discussion with hepatology experts and infectious disease specialists.
Item Score 1 2 3 Total bilirubin, mmol/L <34 34−51 >51 Albumin, g/L >35 28−35 <28 Aspartate aminotransferase, U/L <35 35−70 >70 Prothrombin time prolongation, s <4 4−6 >6
Table 1. Modified Child-Pugh scoring system.
Categorical and continuous variables were presented as percentages (%) and as means [95% confidence interval (CI)], respectively, while normally distributed data were presented by their medians [interquartile range (IQR)]. For normally distributed data, independent t tests were performed to compare means between two groups for continuous variables; otherwise, the Mann-Whitney test was used. For categorical variables, we used Pearson’s chi-squared test or Fisher’s exact test to determine significant differences.
Probability of escalation was calculated using Kaplan-Meier curves with death or escalation up to 2 points or above as events, and surviving patients recorded in 30 days. Longitudinal analysis was performed using scatterplots of daily results against time, which was visualized by the estimated survival time distribution plot. The Cox regression model was used to simultaneously evaluate the relationship between various risk factors and survival time of patients while the log-rank test was used to determine whether differences in survival rates were statistically significant. Missing data were carried by results from the previous day and only estimated for the repeated measures multivariate model. If more than one result was recorded on a certain day, the most deviated result was chosen to reveal the reality of availability of given information to a clinician at a random time. A locally estimated scatterplot smoothing (LOESS) curve showing 95% CIs for the daily mean was used to present dynamic changes in liver function markers based on the severity of COVID-19 (default settings with a polynomial degree of 2, Gaussian kernel, and interpolation on cell size of 0.2). The mixed-effect Cox proportional hazards regression models were used to establish the factors that were associated with escalation of respiratory support or mortality. A two-sided α of <0.05 was considered statistically significant. All analyses were performed using R statistical software (version 3.6.2; The R Foundation for Statistical Computing, Vienna, Austria)
Study Design and Participants