Clinical characteristics of inpatients with coronavirus disease 2019 (COVID-19) in Sichuan province

We conducted a multi-center retrospective study to investigate the clinical characteristics and outcomes of inpatients with confirmed and suspected COVID-19 in Sichuan province. Sichuan locates in West China and has about 80 million populations. It governs 18 municipal cities and 3 autonomous regions with Yi, Qiang and Tibetan minorities. The first case in Sichuan was confirmed on January 21 [19]. We collected medical records of patients who were admitted to 15 hospitals for COVID-19 from January 21 to February 11. These 15 hospitals covered ten cities and one autonomous region (i.e., Ganzi Tibetan autonomous region), and has treated more than one third patients with COVID-19 in Sichuan. To ensure effective implementation of the study, we developed a multidisciplinary research team, including experts in respiratory medicine and intensive care medicine, epidemiologists, statisticians, and informatics. The Institutional Review Board of West China Hospital approved the study on 10 February 2020 (WCH2020–129), and waived patient consent.

Confirmed and suspected patients were diagnosed according to the New Coronavirus Pneumonia Prevention and Control Program issued by the National Health Commission of China [20]. Confirmed cases were defined as patients who had a positive result of high-throughput sequencing or real-time reverse-transcriptase polymerase-chain-reaction (RT-PCR) assay for respiratory tract or blood specimens. Suspected cases were identified based on exposure history and clinical features. Patients who resided or traveled to Wuhan, or had close contact with confirmed cases or patients with fever or respiratory symptoms within 14 days were considered as individuals with exposure risk.

For all suspected patients, nucleic acid detections of COVID-19 were performed by the local Center for Disease Control and Prevention, consistent with the WHO protocol [21]. All provinces in China adopted the uniform laboratory testing procedures since January 24. The sequences were detected as following: for Open reading frame 1 ab fragment, the forward primer was 5′-CCCTGTGGGTTTTACACTTAA-3′, reverse primer sequence was 5′-ACGATTGTGCATCAGCTGA-3′, and probe was 5′-FAM-CCGTCTGCGGTATGTGGAAAGGTTATGG-BHQ1–3′; for the N region of the viral sequence, the forward primer was 5′-GGGGAACTTCTCCTGCTAGAAT-3′, reverse primer was 5′-CAGACATTTTGCTCTCAAGCTG-3′, and the probe was5′-FAM-TTGCTGCTGCTTGACAGATT-TAMRA-3’ [18].

Severe COVID-19 was defined in patients meeting any of the following criteria: presence of respiratory distress with an oxygen saturation of blood ≤93%; or oxygenation index ≤300 mmHg. Patients who required care at intensive care unit or mechanical ventilation, or developed shock were defined as critically severe cases [20].

Medical records of patients with confirmed and suspected COVID-19 and those excluded from the infection were photocopied and sent to the data coordination center at West China hospital in Chengdu, Sichuan. A team of experienced respiratory clinicians then reviewed and abstracted data according to a pre-defined, pilot-tested questionnaire, modified from the WHO Case Report Form (CRF). The data coordination center conducted training on data abstractors, and consensus was achieved regarding the rules of data abstraction. The data abstractors collected data by using the Epi-Data software, version 3·1 (EpiData Association), and all abstracted data were checked by a second abstractor.

The CRF included information regarding demographic characteristic (e.g., gender, age), exposure history (e.g., Wuhan exposure and special occupational exposure), symptoms, or signs (e.g., constitutional, respiratory, gastrointestinal symptoms), laboratory and radiologic findings (e.g., routine blood tests, serum creatinine, transaminases, chest X-ray, or computed tomography (CT)), co-morbidities (e.g. diabetes, hypertension, chronic obstructive pulmonary disease (COPD), cerebrovascular disease (CVD)), complications (e.g., acute respiratory distress syndrome (ARDS), shock, or sepsis), treatment pattern (e.g., antiviral, antimicrobial, or supportive treatment), and outcomes (death and admission to ICU). Information regarding symptoms and signs before and after admission was collected separately. We recorded the first laboratory and radiologic findings after admission and the laboratory tests with highest and lowest values during hospitalization.

We summarized clinical features, radiographic and laboratory findings, and treatment patterns for confirmed and suspected patients. Continuous data were summarized as the means and standard deviations or median and interquartile range (IQR). Categorical data were expressed as numbers and percentages.

For confirmed cases, the median age was 44 (IQR, 33–50) years; 57 (38·8%) cases were females; 82 (61·2%) either resided or ever traveled to Wuhan, 23 (19·3%) were infected by imported cases, and two (1·5%) were healthcare workers (Table 1). The most common symptoms of confirmed cases from onset to admission were cough (70·7%), fever (70·5%), and sputum (33·3%), while fatigue (21·8%) and diarrhea (10·2%) were less frequent (Table 2). Almost one-fifth (19·0%) of confirmed patients developed no fever, and 12·2% had no respiratory symptoms throughout the course of disease.

Among confirmed patients, lymphopenia and eosinopenia was reported in 71 (49·7%) and 81 (60·9%), respectively, 113 (87·6%) had abnormal chest CT findings in the first examination, and 118 (89·4%) had abnormal CT findings during hospitalization (Table 3). The median time from illness onset to abnormalities on CT was 3 days. The most common patterns on chest CT during hospitalization were patchy or stripes shadowing (78·0%), ground-glass opacity (74·2%), and most were bilateral pneumonia (76·5%, Table 3 and Fig. 1).

The median time from disease onset to first medical visit was 1 day (IQR, 0–5). Nine (6·2%) confirmed patients developed respiratory failure, and four (2·7%) developed secondary bacterial pneumonia. A total of 144 (98·6%) patients received antiviral therapy, 81 (56·3%) received antibiotics, and 34 (27·0%) received glucocorticoids. Twelve (8·2%) patients required mechanical ventilation, of which 11 (7·5%) received non-invasive ventilation, and two (1·4%) received invasive mechanical ventilation. Four (2·7%) confirmed patients were transferred to intensive care unit, and no patients died (Table 4).

Among 22 suspected patients, seven (31·8%) were females, and the median age was 51 (IQR, 34–56) years. Four (30·8%) either resided or ever traveled to Wuhan, and four (33·3%) contacted with people from Wuhan. The most common symptoms from onset to admission were cough (86·4%), fever (76·2%), and sputum (45·5%) (Table 2). Lymphopenia and eosinopenia occurred in 6 (54·5%) of suspected patients. Thirteen (92·9%) suspected patients had abnormal CT findings, and the most common patterns were patchy or stripes shadowing (85·7%) during hospitalization (Table 3). One (5·3%) patients developed respiratory failure, and two (9·1%) received non-invasive mechanical ventilation (Table 4).

The median age of severe and non-severe cases was 50 and 43 years (P = 0·73). Compared to non-severe cases, severe ones were more likely to have underlying comorbidities (62·5% vs 26·2%, P = 0·0010), including hypertension (33·3% vs 9·0%, P = 0·0040) and pulmonary diseases (25·0% vs 4·1%, P = 0·0020) (Table 1). The temperature was higher in severe cases both from onset to admission (38·5 °C vs 37·8 °C, P = 0·004) and throughout the course of disease (38·5 °C vs 38 °C, P = 0·0080). Respiratory symptoms during onset to admission were more commonly presented in severe cases than non-severe ones, including cough (92·0% vs 66·4%, P = 0·020), sputum (60·0% vs 27·9%, P = 0·0040) and shortness of breath (40·0% vs 8·2%, P <  0·0001). The heart rates at admission were higher in severe cases (104 times per minute [IQR, 91–109] vs 90 times per minute [IQR, 80–99], P = 0·0050, Table 2). No significant differences were found in gastrointestinal symptoms between the two populations, although appearing higher in non-severe cases both from onset to admission (8·0% vs 12·3%) and throughout the course of disease (20·0% vs 40·2%).

Severe cases had more abnormalities on the first laboratory tests, including lower lymphocyte counts (0·8 × 10⁹/L [IQR, 0·5–1·0]) vs 1·2 × 10⁹/L [IQR, 1·0–1·7], P <  0·0001) and eosinophils count (0·00 × 10⁹/L [IQR, 0·00–0·01] vs. 0·02 × 10⁹/L [IQR, 0·00–0·05], P = 0·0030), and higher proportion of lymphopenia (79·2% vs 43·7%, P = 0·0030) and eosinopenia (84·2% vs 57·0%, P = 0·046) (Table 3). Severe patients also had a higher level of D-dimer (414 μmol/L [IQR, 163–930] vs 175 μmol/L [IQR, 90–368], P = 0·025) and C-reactive protein (31·7 mg/L [IQR, 14·2–54·2] vs 5·7 mg/L [IQR, 1·9–15·6], P <  0·0001). The finding during hospitalization were similar to the first laboratory tests (Supplementary Table 1).

During hospitalization, severe cases were more likely to have bilateral pneumonia (100·0% vs. 72·3%, P = 0·016) and pleural effusion (15·0% vs. 2·7%, P = 0·045, Table 3), and were more likely to receive antibiotics (91·7% vs 49·2%, P <  0·0001), glucocorticoids (72·0% vs 15·8%, P <  0·0001), intravenous immunoglobulins (16·0% vs 3·3%, P = 0·010) and oxygen therapies (72·0% vs 33·6%, P <  0·0001) (Table 4).

Univariate logistic analysis showed that patients with pulmonary diseases (OR 7·80, 2·14–29·72), hypertension (OR 5.05, 1.73–14.48), white blood cell count > 10× 10⁹/L.

(OR 7.38, 1.81–32.10), lymphocyte count< 1·1× 10⁹/L (OR 4·26, 1·67–12·39), bilateral pneumonia (OR 11·46, 2·24–209·65), and pleural effusion (OR 9·44, 1·47–75·78) were more likely to develop into severe cases. Higher temperature (OR 2.52, 1.37–5.08)) and higher heart rate (OR 1.04, 1.01–1.07) were associated with increased risk of developing severe cases (Table 5).

In this study, we found that most confirmed COVID-19 cases were adults, particularly males. The most common symptoms of confirmed patients were cough, fever and sputum, while gastrointestinal symptoms were less frequent. Presence of lymphopenia and eosinopenia was also frequent. A typical finding of CT scan for COVID-19 was bilateral ground-glass opacity, occurring in two-third of patients. Nearly all patients received antiviral treatments, with lopinavir/litonavir being the most often used.

Generally, the symptoms of patients in Sichuan were relatively mild, and the clinical outcomes were better than those in Wuhan. We found that about 19·0% of confirmed cases did not have fever, and 10·6% had no radiologic abnormality throughout the course of disease. In contrast, the proportion of absence of fever and radiologic abnormality were reported less than 5% in Wuhan [3, 14, 15]. Only two (1·5%) healthcare workers were infected, the proportion of which was much lower than that in Wuhan [14, 16]. Among the 147 confirmed cases in Sichuan, 12 (8·2%) received mechanical ventilations and no death occurred, which contrasted the reported mortality ranging from 4·3% to 15·0% in Wuhan [3, 14, 15, 22]. Similar to our findings, Xu et al also suggested that patient symptoms in Zhejiang province were mild [17]. The differences in the clinical features and outcomes between Wuhan versus other regions may be due to the facts that limited healthcare and human resources were available in Wuhan, particularly at the early stage of the outbreak. Indeed, the median time from disease onset to the first medical visit was 1 day in Sichuan as opposed to 7 days in Wuhan [14].

The COVID-19 had both similar and distinct characteristics in comparison with severe acute respiratory syndrome coronavirus (SARS-CoV) [23]. Although symptoms were similar [3, 15, 16], a fever-free condition was observed in 19% of confirmed cases in our study, much higher than that in SARS-CoV (less than 1%) [24, 25]. SARS-CoV was clearly a more serious condition, with about 17·0% of patients receiving invasive mechanical ventilations and 9·6% died [24]. In contrast, our study showed that 2·7% of COVID-19 patients were transferred to ICU, and 1·4% received invasive mechanical ventilation. Until now, the human has not developed specific antiviral drugs for both coronaviruses.

Our study found that older patients and those with underling comorbidities were more likely to develop severe illness, consistent with other reported findings [3, 14, 15]. We also found that severe cases were more likely to have fever, respiratory symptoms, abnormal laboratory and radiologic findings, and to develop complications including respiratory failure, coagulation disorders, and sepsis than non-severe cases. These are all consistent with published studies [3, 14, 22]. In particular, we found that eosinopenia appeared in 57·0% and 84·2% of non-severe cases and severe cases, suggesting a potential of use for differentiating disease severity. Indeed, eosinopenia has been identified as a good diagnostic marker for severe infections, such as sepsis and bloodstream infection [26, 27]. A study including 177 ICU patients suggested that eosinopenia may be a good diagnostic marker in distinguishing between non-infection and infection with an area under receiver operating characteristic curve of 0·89 [26]. A possible explanation was that eosinophils migrated to the inflammatory site due to chemotactic substances released during acute inflammation [26].

Our study has several strengths. To our best knowledge, this is the first study investigating clinical characteristics of COVID-19 patients from west China. To ensure representativeness, we collected data from 15 hospitals specifically responsible for treating COVID-19 patients. We also implemented rigorous approaches to collect clinical data to ensure the quality of data.

Meanwhile, our study was a retrospective study, and the data accuracy and completeness were not optimal. Nevertheless, we implemented a strong data collection strategy to minimize potential bias. Secondly, most patients were hospitalized at the time of data collection. Thus, we were unable to investigate outcomes of those infected patients. Thirdly, we included a limited number of patients. As such, we were unable to conduct more sophisticated analyses to control for potential confounding effect. Fourthly, the total number of patients visiting the studied hospitals is not available, and hence the proportions of suspected and confirmed patients with COVID-19 among all hospital visits were unclear.