Materials and methods

Patients
A retrospective review of patient data and imaging studies was approved by the institutional review board. No patient consent was required for this HIPAA-compliant study.
Between January 22 and February 8, 2020, 88 patients with COVID-19 were identified from six hospitals in Shaanxi Province in China, which is approximately 800 km northwest of Wuhan. The inclusion criteria were as follows: (1) patients with COVID-19 confirmed by at least one positive nucleic acid test performed at the Center for Disease Control and Prevention in Shaanxi Province in China and (2) patients who underwent baseline and at least one follow-up chest CT. According to the WHO, patients are divided into six types: uncomplicated illness, mild pneumonia, severe pneumonia, acute respiratory distress syndrome, sepsis, and septic shock [10]. All the patients were treated based on a standard diagnosis and treatment criteria for COVID-19 which includes initiation of antivirals, interferon, Chinese herbal medications, supplemental oxygen as needed, and hospitalization.

Image acquisition
CT scans of the chest were acquired on 16- to 64-multidector CT scanners (Philips Brilliant 16, Philips Healthcare; GE LightSpeed 16, GE Healthcare; GE VCT LightSpeed 64, GE Healthcare; Somatom Sensation 64, Siemens Healthcare; Somatom AS, Siemens Healthcare; Somatom Spirit, Siemens Healthcare; GE Optima 680, GE Healthcare). The CT parameters were as follows: 120 kVp, current intelligent control (auto mA) of 30–300 mA, and slice thickness reconstructions of 0.6–3.0 mm. All CT examinations were performed without intravenous contrast material.

Data collection and evaluation
The medical records of patients were reviewed to determine the demographic data, medical history, exposure history, clinical symptoms, laboratory findings, and time of disease onset. The date of disease onset was defined as patients’ reported date of symptom onset. The time intervals, in days, between each CT scan and disease onset were determined and categorized into five groups (days 0–3, days 4–7, days 8–10, days 11–14, and days 15–21).
CT images were reviewed independently by two thoracic radiologists each with 10 years of experience. Differences were resolved by discussion to reach consensus. Similar to a previously published report [6], CT findings including the presence and distribution of ground glass opacities (GGO), consolidation, linear opacities, discrete pulmonary nodules, pleural effusion, lymphadenopathy, and cavitation were evaluated. Degree of lobar involvement and overall lung total severity score were recorded. Each of the five lung lobes was assessed for degree or area of involvement and assigned a score of 0 for 0% lobe involvement, 1 for 1–25% lobe involvement, 2 for 26–50% lobe involvement, 3 for 51–75% lobe involvement, or 4 for 76–100% lobe involvement. An overall lung total severity score was reached by summing the five lobe scores (range of possible scores, 0–20) [5].

Statistical analysis
Continuous variables were represented as means and standard deviations, while categorical variables were expressed as counts and percentages. Differences of CT characteristics, lobar involvement, and overall total severity score among different time intervals were statistically compared by the chi-square test or analysis of variance (ANOVA). Linear regression analysis was further used to characterize the linear evolution trends of the variables that were significant at the above analysis.
Statistical analysis was performed by using R software (version 3.6.0; http://www.Rproject.org). The packages in R that were used in this study are “gmodels” packages. All p values were considered statistically significant at p < 0.05.