Maximum chest CT score predicts progression to severe illness in patients with COVID-19: a retrospective study from Wuhan, China

Background: We investigated the clinical course and imaging ndings of hospitalized patients who were initially diagnosed with moderate COVID-19 symptoms to identify risk factors associated with progression to severe/critical symptoms. Methods: This study was a retrospective single-center study at The Central Hospital of Wuhan. 243 patients with conrmed COVID 19 pneumonia were enrolled in the analysis, of which 40 patients progressed from moderate to severe/critical symptoms during follow up. Demographic, clinical, laboratory and radiological data were extracted from electronic medical records and compared between moderate and severe/critical symptom types. Univariable and multivariable logistic regressions were used to identify the risk factors associated with symptom progression. Results: Patients with severe/critical symptoms were older (p (cid:0) 0.001) and more often male (p=0.046). We found that the combination of chronic obstructive pulmonary disease and high maximum CT scores was associated with disease progression. Maximum CT scores ( ≥ 11) had the greatest predictive value for disease progression. The area under the receiver operating characteristic curve (ROC) was 0.861 (95% CI: 0.811-0.902). Conclusions: Maximum CT scores and COPD are associated with patient deterioration. Maximum CT scores ( ≥ 11) are associated with severe illness.


Background
Worldwide there have been 234,073 con rmed cases and 9,840 deaths resulting from novel coronavirus 2019 (COVID-19) infection as of March 19, 2020. The rate of overall case-fatality was about 2.3% across China. Although few deaths were found in mild COVID-19 patients, the rate of case-fatality was obviously elevated among critical ill patients. Nevertheless, what factors in uence the prognosis of COVID-19 patients has not been clearly clari ed [1,2].
The onset of COVID-19 is associated with symptoms such as fever, cough, and myalgia. The case de nition adopted in China and elsewhere includes further strati cation of cases as mild, moderate, and severe/critical [3]. More than 80% of the laboratory con rmed cases in China (including both nonpneumonia and pneumonia cases) were mild-to-moderate types [4]. A fth of these cases progressed to a severe or critical stage, with the highest case fatality rate of 4.47% in Wuhan before March 2020 [5]. Most previous studies have only focused on the general epidemiological ndings, clinical characteristics, and outcomes of patients with COVID-19 [2,6,7]. However, few studies have investigated the clinical ndings in patients who progress from moderate to severe/critical type symptoms.
In this study, we did a comprehensive analysis of the clinical course and imaging ndings of 243 hospitalized COVID-19 patients who initially presented with moderate type symptoms. These patients were admitted to the isolation ward of the Central Hospital of Wuhan, which is very close to the seafood market and one of the rst hospitals in Wuhan to admit COVID-19 patients. We aimed to identify risk factors associated with progression from moderate to severe/critical type symptoms.

Study design and participants
This was a retrospective single-center study at The Central Hospital of Wuhan. The study was approved by The Central Hospital of Wuhan Ethics Committee (No.2020421), and written informed consent was waived by the Ethics Commission of the designated hospital for emerging infectious diseases.
Patients who were diagnosed with COVID-19 based on WHO interim guidelines were enrolled consecutively from December 25 to February 16, 2020 [8]. Based on the new coronavirus pneumonia diagnosis and treatment protocols (version 6) developed by the National Health Commission of the People's Republic of China (http://www.nhc.gov.cn/), the clinical classi cation of COVID-19 is as follows: moderate types include those with fever, respiratory tract involvement, and other symptoms, as well as imaging ndings of pneumonia; severe types meet any of the following: (1) respiratory distress with a respiratory rate ≥ 30 beats/min; (2) oxygen saturation ≤ 93% at rest; and (3), arterial blood oxygen partial pressure (PaO 2 )/oxygen concentration (FiO 2 ) ≤ 300 mm Hg (1 mm Hg = 0.133 kPa); critical types meet any of the following conditions: (1) respiratory failure requiring mechanical ventilation; (2) occurrence of shock; (3) ICU admission for combined organ failure.

Data Collection
Epidemiological, clinical, laboratory, radiological characteristics, and outcome data were extracted from electronic medical records. The data were reviewed by two physicians, and a third physician adjudicated the difference in interpretation between the two primary researchers. Fever was de ned as axillary temperature higher than 37.3 °C. Epidemiological, clinical, and laboratory data were de ned as the results of the rst consultation or examination in the electronic medical records. The initial chest CT was de ned as the rst chest CT examination of admission.

CT Image Acquisition
Chest CT scans were performed with the patient in a supine position and using a single inspiratory phase on one of the four CT systems (Bright Speed Elite, GE, America; Philips Ingenuity Core128, Philips Medical Systems, Best, the Netherlands; uCT 760, United Imaging, China; SOMATOM De nition AS, Siemens Healthineers, Germany). The following main parameters were used: tube voltage = 120 kVp; automatic tube current modulation (20-130 mAs); matrix = 512 × 512; FOV = 350 mm × 350 mm − 370 mm × 370 mm; pitch = 0.75-1.25 mm; slice thickness = 1-1.5 mm. The reconstructed images were automatically sent to the corresponding post-processing workstation and picture archiving and communication systems (PACS) for multiplanar reconstruction post-processing.

Chest CT Evaluation
All chest CT images were independently analyzed, and the features were scored by a senior thoracic radiologist with more than 20 years of experience who was blinded to clinical and laboratory ndings. A previous semi-quantitative scoring system was performed to estimate the pulmonary involvement of lesions on the basis of the area involved [9]. Each of the 5 lung lobes was visually scored from 0 to 5, as follows: 0, no involvement; 1, < 5% area involvement; 2, 5%-25% area involvement; 3, 26%-49% area involvement; 4, 50%-75% area involvement; and 5, > 75% area involvement. The total CT score was the sum of the scores of the 5 lung lobes (0-25).

Follow-up Chest CTs
After admission, patients were re-examined for chest CT examination. The endpoint of this study was the development of severe/critical illness. We excluded patients transferred from other medical institutions and those transferred out from our hospital. We also excluded fatalities. All patients in our study recovered. From the day of illness onset, follow-up chest CTs were performed anytime from day 3 to day 60.

Statistical Analyses
Continuous variables are presented as mean ± SD or median (IQR), depending on the normality of distribution. Categorical variables are presented as frequencies and percentages. Continuous variables were compared between the moderate type group and the severe/critical type group using a two-sample ttest or Mann-Whitney U test. Categorical variables were compared using a chi-square test or Fisher's exact test as appropriate. Univariable and multivariable logistic regression models were used to explore the risk factors associated with severe/critical symptom type progression from moderate symptom types. Using the new coronavirus pneumonia diagnosis protocols as the standard of reference, diagnostic performance with the highest CT score of severe/critical type progression was determined using analyses of sensitivity, speci city, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC). 95% con dence intervals (CIs) are also reported. Diagnostic ability was illustrated using a receiver operating characteristic curve (ROC). A two-tailed p 0.05 was considered statistically signi cant. All statistical analyses were performed using SPSS version 18 (SPSS, Inc., Chicago, IL).

Results
A total of 457 adult patients diagnosed with COVID-19 were admitted to our hospital from December 25, 2019 to February 16, 2020. There were 191 patients still hospitalized or not con rmed by SARS-CoV-2 RNA detection or death before our nal follow-up, and they were excluded from analysis. We also excluded 7 patients who were diagnosed with severe/critical type symptoms on admission. Sixteen patients were excluded because their medical records were not available. In total, 243 patients with con rmed COVID 19 pneumonia were enrolled in the analysis. Forty patients progressed from having moderate symptoms to severe/critical symptoms during follow-up. The mean age of the 243 patients was 47 years, with a range of 20 to 89 years. One hundred thirty-eight (56.8%) were female ( Table 1). The patients with severe/critical symptoms were older (p 0.001) and more likely male (p = 0.046) than patients with moderate type symptoms. The most common symptoms at onset were fever (83.5%) and cough (63.8%), followed by weakness (45.7%) and sputum (28.8%). Chest tightness was more likely to occur in patients with severe/critical symptoms (42.5%) than in patients with moderate type symptoms (p = 0.002). Patients in the severe/critical symptom type group had signi cantly higher mean levels of Ddimer (2.1 ± 4 vs. 0.8 ± 1.6 mg/L, p 0.001), C-reactive protein (3.6 ± 3.2 vs. 2 ± 3 mg/L, p = 0.038), αhydroxybutyrate dehydrogenase (247 ± 352 vs. 142 ± 57 U/L, p 0.001), lactate dehydrogenase (251 ± 119 vs. 185 ± 79 U/L, p = 0.002), creatine kinase, (273 ± 636 vs. 120 ± 172 U/L, p 0.001), and creatinine (211 ± 449 vs. 67 ± 43 µmol/L, p 0.001) than patients in the moderate type group. The neutrophil-to-lymphocyte ratio was also signi cantly higher in the severe/critical type group (NLR, 10.6 ± 7.7 vs 3.7 ± 3.9, p 0.001).
All patients had abnormal CT imaging features (Fig. 1). There were no signi cant differences in the initial CT scores between the moderate type group and severe/critical type group (p = 0.112). Patients in the severe/critical type group had signi cantly higher median maximum CT scores (14, IQR 3-28 vs. 7, IQR 0-20; p 0.001) than the moderate type group. There was no signi cant difference in the median time from onset of symptoms to hospitalization between the moderate type group and the severe/critical type group (p = 0.733). In comparison with patients in the moderate type group, patients in the severe/critical type group had longer hospital stays (22, IQR 18.5-28.5 vs. 18, IQR 13-24; p 0.001) and a longer time from illness onset to high CT score (13, IQR 8-17.8 vs. 9, IQR 6-12; p 0.001) (Fig. 2).
Using univariable analysis, the calculated odds of progression to the severe/critical type group was found to be higher in patients having any comorbidity, including hypertension and COPD. Age (≥ 60 years), sex (male), chest tightness, levels of D-dimer (≥ 1 mg/L), C-reactive protein (≥ 1 mg/L), α-hydroxybutyrate dehydrogenase (≥ 180 U/L), lactate dehydrogenase (≥ 250 U/L), creatine kinase (≥ 190 U/L), creatinine (≥ 97 µmol/L), NLR, and highest CT score were also associated with disease progression. The length of the hospital stays and time from illness onset to the time of the highest CT score were also associated with disease progression. A collinearity analysis showed no collinearity between variables. We included these variables in the multivariable logistic regression analysis. We found that the combination of COPD (OR = 9.06, 95% CI [1.30-63.36]; p = 0.026) and a higher maximum CT score (OR = 1.39, 95% CI [1.21-1.60]; p 0.001) were associated with disease progression (Table 2). Data are mean (SD), median (IQR), n (%) or n/N (%). NLR, Neutrophil Lymphocyte Ratio.

Discussion
This retrospective study identi ed high CT scores and COPD as risk factors for deterioration in hospitalized patients with COVID-19 in Wuhan, China. Additionally, being older and male, having chest tightness, hypertension, and elevated levels of D-dimer, C-reactive protein, α-hydroxybutyrate dehydrogenase, lactate dehydrogenase, creatine kinase, creatinine, and NLR were associated with progression to severe or critical COVID-19 illness.
Recently, Zhou et al. demonstrated that older age is associated with death in hospitalized COVID-19 patients in a retrospective, multicenter cohort study of 191 patients [10]. Other studies have also shown that older age (> 65 years) is associated with poorer clinical outcome in patients with COVID-19 [6,11,12].
The present study con rmed that more severely or critically ill patients were older ( 60 years) than patients with moderate type COVID-19 symptoms. The age-dependent risk has also been seen in previous studies of SARS and MERS [12,13]. However, in this current study, older age was not shown to be an independent predictor of deterioration in hospitalized patients with COVID-19. The difference between the present and previous studies may be partly due to the different outcomes. The current study was aimed at the progress of moderate type patients with COVID-19 who eventually recovered, rather than died.
Furthermore, we had a relatively small sample size for the severe or critical group.
Chest CT is the routine imaging modality for clinical diagnosis of patients with COVID-19 pneumonia in the Hubei Province. It may help to screen patients with suspected COVID-19 symptoms, especially those with a negative RT-PCR result at the early stages of the disease [14]. In order to comprehensively evaluate the CT features of COVID-19 pneumonia, a semi-quantitative scoring system has been developed to quantitatively estimate the severity of in ammation based on quantifying the extent of pulmonary abnormalities (including ground-glass opacities, consolidations, or other fuzzy interstitial opacities) [9]. Using this method, no signi cant differences of initial CT scores have been found between moderate type and severe/critical type groups. A recent study showed that CT scores in severe-critical type groups were signi cantly higher than those in the moderate type groups [15]. This may be explained by the fact that patients with COVID-19 in the present study were all of the moderate type initially. With the progression of COVID-19, there were signi cant differences in the highest CT scores between moderate type and severe/critical type groups. Moreover, using multivariable analysis, the maximum CT score indices were found to be an important independent predictor of deterioration of patients who progressed from moderate type symptoms to severe/critical type symptoms. Furthermore, ROC analysis showed an optimal cutoff value of a maximum score CT index of 11 (sensitivity of 85% and speci city of 78.3%) to predict deterioration. We hope that this maximum score CT index may be used to identify patients at earlier stages of COVID-19 who may potentially progress to severe/critical stage symptoms from moderate type symptoms. Patients with this maximum score CT index will then receive more aggressive treatment and close monitoring. However, the e cacy of such an approach remains to be validated in multi-center and large sample studies in the future.
Lung complications, especially COPD, are common in patients with pneumonia. According to a national cross-sectional study, the total number of patients with COPD in China approximates 100 million [16].
Most of the recent studies have shown no signi cant differences in COPD between patients with COVID-19 in non-severe types and severe types [1,7,[17][18][19]]. In the current study, COPD was more common in the severe/critical type group than those in the moderate type group, a nding which agreed with results reported by Guan et al [6]. In addition, COPD has been found to be associated with the deterioration of patients with COVID-19. It should be noted that all studies to date, including our study, have used small sample sizes of COVID-19 patients with COPD. The potential impact of COPD on the disease outcomes of patients with COVID-19 requires further observation and research.
Our study has some limitations. First, not all laboratory tests were done in all patients, including the measurement of levels of α-hydroxybutyrate dehydrogenase, lactate dehydrogenase, creatine kinase, procalcitonin and brain natriuretic peptide. Therefore, their roles might be underestimated in predicting disease progression. In addition, we did not analyze the changes in laboratory ndings in the process of the disease progression or patient recovery. Some of these results might also contribute to the deterioration in some patients. Second, this was a retrospective study from a single center with a relatively small sample size and a certain selection bias, as some patients were transferred to other medical institution through government decree. Thus, comparisons of clinical characteristics, laboratory ndings, and imaging features may be skewed. Third, the semi-quantitative methods for measuring COVID-19 pneumonia lesions may be somwhat subjectivie. Last, infants, children and adolescents were not included in the present study, and an effort should be made to include these groups in future studies.

Conclusions
Maximum CT score index and COPD were identi ed as risk factors which may be used to predict COVID-19 disease progression. A maximum CT score ≥ 11 is associated with development of severe illness. As the current manuscript describes the study protocol and no other data, we do not have any raw data to share at the moment.

Abbreviations
Ethics approval and consent to participate The study was approved by The Central Hospital of Wuhan Ethics Committee (No.2020421), and written informed consent was waived by the Ethics Commission of the designated hospital for emerging infectious diseases.

Consent for publication
Not applicable.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.
Authors' contributions JWX and XW conceived and designed the study. YD, SCZ, PY, DD and BL collected the data. JWX and XL wrote the paper. YLX and ZFH analyzed the data, and XW reviewed and edited the manuscript. All authors read and approved the manuscript and agree to be accountable for all aspects of the research. Figure 1 CT images in moderate type group (A-D), severe type group (E-H), and critical type group (I-L).

Figure 2
Scatter diagram of maximum CT score with time from illness onset in two groups.

Figure 3
Receiver operating characteristic curves analyses of the highest CT score for prediction of disease deterioration.