Symptomatic Features of 932 Hospitalized Patients with COVID-19 in Wuhan

This study reported detailed symptomatic features in severe and non-severe patients with COVID-19 in China. The prevalence and duration of COVID-19-related symptoms have been documented. The analysis of the relationship between clinical symptoms and illness severity revealed a cluster of predictors for COVID-19. Our study also showed an optimistic prognosis in most patients through three months of follow-up after discharge.


Introduction
Coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, has spread unprecedentedly, causing an ongoing worldwide pandemic [1,2]. This situation is still not optimistic, even though this illness's spread is basically under control in many countries [3]. Globally, as of 31 December 2020, there have been 81,475,053 con rmed cases of COVID-19, including 1,798,050 deaths, according to the WHO website data [4]. The outbreak and rapid spread of COVID-19 have caused tremendous pressure on medical resources and clinical treatment capacity worldwide [5].

Relevant studies have been published involving the clinical characteristics of patients with COVID-19
during the past few months. Although large case series have been described [6][7][8], most reports focused on the onset and early stages of the illness. Follow-up data regarding the illness course and the outcome after discharge from the hospital remain limited. Moreover, the spectrum of clinical manifestations of COVID-19 varies with diverse population demographics and prevalence of comorbidities. Besides, the relationship between symptomatic features and illness severity has not been fully elucidated. Therefore, it is essential to characterize symptomatic features and the progression of patients with COVID-19.
This study performed an updated analysis of detailed clinical characteristics and a 3-month follow-up in hospitalized COVID-19 patients in Wuhan, China. Our ndings sought to disclose the symptomatic features of COVID-19 and help physicians make correct judgments of this disease's severity and prognosis.

Study Design and Patients
In this cohort study, we recruited hospitalized COVID-19 patients at the Optical Valley Branch of Maternal and Child Hospital of Hubei Province, Wuhan, which is a designated hospital to treat adult COVID-19 patients (aged ≥ 14 years).
Between Feb 20 and Mar 31, 2020, this hospital admitted 1767 suspected COVID-19 patients.  was diagnosed based on the National Health Commission of China (version 7) [9]. Only laboratorycon rmed cases were included in this study. As shown in Figure 1, seven patients were excluded due to a lack of medical records. Fifty-eight patients were excluded due to a lack of contact information. A total of 277 patients were excluded due to negative results on nucleic acid and speci c serum antibody detection. A total of 1425 cases were included in the list for the second round of medical history review and threemonth follow-up after discharge, of which 281 cases were lost to follow-up, and 212 cases were not completed. Finally, We included 932 laboratory-con rmed COVID-19 patients with complete follow-up data in our cohort for further analysis, consisting of 880 non-severe cases and 52 severe cases. treatment, and outcome data were extracted from electronic medical records. All needed variables with necessary explanations were de ned rst by the research team. An experienced respiratory clinician team reviewed the copies of the medical records and abstracted the data. The abstracted data were entered into a computerized database, and double-entry was required for all variables. The data-entry team ensured that all data needed were collected. If the core data were missing, requests for clari cation were immediately sent to the coordinators, who subsequently contacted the attending clinicians.
We designed a standardized electronic follow-up questionnaire form to supplement detailed symptomatic information and three-month follow-up after discharge, including onset symptoms, duration, dates, times and frequency of the symptoms, remaining symptoms after discharge, etc. The intact follow-up questionnaire form is shown in the Appendix. To ensure the validity of the collected data, patients who died, who were unable to cooperate or had unreliable narrators (i.e., elderly patients, dementia patients, and those with communication problems), were excluded from the cohort. The nal date of the follow-up was Aug 02, 2020. All the authors checked the accuracy and completeness of the data.
Data cleaning, including logical checks, outlier checks, and variable engineering, was performed by experienced programmers. An experienced clinician assisted the variable engineering. The original variables were transformed if needed, including but were not restricted to converting continuous variables into categorical variables and combining multiple variables into single variables for information integration.

De nitions
Laboratory-con rmed COVID-19 was de ned as a positive result on reverse transcriptase-polymerase chain reaction (RT-PCR) assay of nasal and pharyngeal swab specimens or speci c serum antibodies for SARS-CoV-2. The severity of COVID-19 was evaluated following guidelines issued by the National Health Commission of China (version 7.0), as mentioned above [9]. Patients with at least one of the following conditions were de ned as severe cases: 1) dyspnea (respiratory rate ≥ 30 times/min); 2) nger oxygen saturation ≤ 93% in the resting state; 3) arterial oxygen partial pressure/oxygen absorption concentration ≤ 300 mmHg; and 4) pulmonary imaging showing that foci that progressed by > 50% in 24-48 hours.
The onset date was de ned as the earliest day when symptoms were noticed by patients themselves or physicians. Fever was de ned as 37.3°C or higher on axillary temperature. Septic shock was de ned according to the 2016 Third International Consensus De nition for Sepsis and Septic Shock [10]. Acute kidney injury was de ned according to the KDIGO clinical practice guidelines [11]. The de nition of acute respiratory distress syndrome referred to the Berlin diagnostic criteria [12]. Acute myocardial injury was de ned as serum levels of ultrasensitive cardiac troponin I above the 99th percentile upper reference limit [13].
The Modi ed Medical Research Council (mMRC) Dyspnea Scale was de ned as follows: level 0 for no breathlessness, level 1 for breathless when hurrying or walking up a hill, level 2 for breathless when walking slower than people of same age or requirement to stop when walking, level 3 for breathlessness or stopping walking after ∼100 m or a few minutes, and level 4 for breathless when dressing or not able to leave the house [14].

Statistical analysis
Categorical variables were described as frequencies and percentages. Continuous variables were described with mean, median, and interquartile range (IQR) values. Independent group t test or Mann-Whitney test was used to compare means for continuous variables depending on the data's distribution. The χ2 test or Fisher's exact test was used for the comparison of frequencies for categorical variables.
Univariate and multivariate logistic regression models were used to explore the symptomatic risk factors for severe COVID-19. Based on previous ndings and clinical considerations, a total of 11 variables, including signi cant basic demographic data and symptomatic characteristics in univariate analysis, were included for multivariate analysis. If the difference between groups was not signi cant, the number of events was too small, odds ratios could not be calculated, or collinear with other variables, we excluded this variable from univariate analysis.
All analyses were performed using SAS JMP Pro 15.0. The tests with a P value < 0.05 were considered statistically signi cant.

Fever
As previously reported, fever is the most common symptom. Still, most of the patients had a persistent low fever (37.3°C -38°C). Moderate fever (38°C -39°C) was more common in severe patients (Table 2). Severe COVID-19 patients had a higher proportion of fever than non-severe patients (78.9% vs 58.9%, P = 0.0043). The median duration of fever was 7 days (IQR 4-11 days) in the non-severe group and 9 days (IQR 6-12 days) in the severe group ( Figure 2, Table S1).

Cough
Most patients complained of dry cough, and white sputum was common in patients with sputum cough ( Table 2). The median duration of cough was 17 days (IQR 9-35 days) for the whole cohort, while the median duration of cough in the severe patients was as long as 30 days (12.5-51.5 days, P < 0.05), almost double that of non-severe patients ( Figure 2, Table S1).

Dyspnea
Dyspnea is a characteristic symptom of severe COVID-19. In our cohort, dyspnea in severe patients occurred later, lasted for a longer time, and reached higher mMRC levels ( Figure 3). As shown in Table S1, the median time from illness onset to dyspnea was 4 days (IQR 1-7 days) in all patients, with 3 days (1-7 days) in non-severe patients and 7 days (1-12 days) in severe patients. The most severe dyspnea was observed on the 8th day from the beginning in non-severe patients and on the 15th day in severe patients with the disease's progression. The median duration of dyspnea in severe patients was as long as 28 days (IQR 14-51 days), which was signi cantly longer than that in non-severe patients (15 days, IQR 8-31 days). As shown in Table 2, severe patients showed a higher degree of dyspnea on disease onset, as quanti ed by mMRC levels. At the peak stage, 42.3% of the severe patients' mMRC levels reached grade 4. Even when dyspnea was relieved considerably on the 43rd day from the onset, 26.9% of the whole cohort's mMRC level was still at grade 2.

Fatigue and anorexia
Fatigue and anorexia often coincided with fever and lasted longer than dyspnea ( Figure 3). The median duration of fatigue was 14 days (8-27 days) in the non-severe group and 32 days (21-44 days) in the severe group. The median duration of anorexia was 12 days (8-24 days) in the non-severe group and 27 days (17-40 days) in the severe group (P < 0.001, Table S1). 14.5% of the patients had anorexia at the onset, with a larger proportion in the severe group than in the non-severe group (30.8% vs 13.5%, P = 0.0006, Table S2). Nausea or vomiting, and diarrhea Nausea or vomiting and diarrhea were not common. As shown in Table 2, only 32/932 (3.4%) patients experienced nausea or vomiting. A total of 8.6% of patients complained of diarrhea, most of which began on the third day after illness onset, and the median duration was 5 days (IQR 4-12 days). There was no signi cant difference in the incidence of these symptoms between the severe and non-severe groups (Table S1).
Taste loss and olfactory disturbance Unexpectedly, the incidence of taste loss and smell loss in our cohort was only 6.2% and 3.1%, respectively ( Table 2). The median duration of both symptoms was 15 days. However, the median duration of these two symptoms in severe patients (22 days and 27 days, respectively) were longer than those of non-severe patients (both were 14 days) (Figure 2, Table S1).

Symptom clusters
Cluster analysis of symptom correlation showed that speci c symptoms were positively related and formed several groups ( Figure 4). Anorexia, nausea or vomiting, and diarrhea formed a gastrointestinal symptom cluster. Headache and dizziness formed a neuropsychiatric symptom cluster. Loss of taste and smell were linked and often co-occurred, which formed an otorhinolaryngological symptom cluster. Fatigue and myalgia or arthralgia were highly correlated. In addition to nausea or vomiting and diarrhea, symptoms associated with anorexia included loss of taste, fatigue, and dyspnea. The only sign highly related to dyspnea was anorexia.

Onset symptoms
As shown in Table S2, fever and cough are the two most common symptoms. Only 17.3% of severe patients had dyspnea at onset, while 30.8% had anorexia. As shown in Table S3, a total of 82.7% of severe patients and 63.3% of non-severe patients started with systemic symptoms, including fever, fatigue, myalgia or joint pain. 32.7% of severe patients started with gastrointestinal symptoms, including anorexia, nausea, vomiting or diarrhea, which was signi cantly higher than that of non-severe patients.
Rehabilitation symptoms at 3 months after discharge All patients were followed up for three months after discharge ( Table 4). The results showed that 6.2% of the patients still had cough, mostly dry cough. A total of 1.8% of patients had fatigue, 0.8% of patients had taste loss, and 0.7% of patients had olfactory disturbance. After discharge from hospital, a total of 7.2% patients still had dyspnea, most with mMRC level 1. As a remaining symptom, the incidence of dyspnea was higher in the severe group than in the non-severe group (P < 0.0001), and severe patients had a higher degree of mMRC levels (P < 0.01).

Discussion
While numerous studies have reported the clinical characteristics of COVID-19 patients, this disease's symptomatic features are still not thoroughly investigated. With this study, we deeply examined the clinical data of COVID-19 patients. We uncovered more detailed information on the symptomatic features, which is essential for physicians to grasp the characteristics of COVID-19 and make a correct judgment of the illness. We found that the symptoms of COVID-19 are usually not typical but diverse. The most common presenting symptoms were fever (60%), cough (50.8%), and fatigue (36.4%), which is consistent with previous reports [15,16]. However, the prevalence of olfactory and taste disorders (OTDs) is much lower in our cohort than in other reports [17][18][19][20]. It should be noted that 11.6% of the COVID-19 patients, including two severe cases, were asymptomatic, suggesting that the screening of COVID-19 based on symptoms is not reliable.
Although fever is the most common symptom, most patients present with only persistent low fever; of note, nearly 40% of the patients did not have a fever during the illness course. Additionally, high fever (> 39 °C) was rare in these patients. However, it should be mentioned that severe COVID-19 patients were more prone to suffer fever with a longer duration. Dry cough was also one of our patients' most common symptoms, which usually lasted a much more extended period than fever. As expected, the time of cough in severe patients was much longer than that in non-severe patients.
We know that dyspnea is a kind of subjective experience common in severe pneumonia patients. The occurrence of dyspnea appeared late and atypically, not always following a decline in blood oxygen saturation. Our study showed that the interval time from illness onset to dyspnea occurrence in severe patients was longer than that in non-severe patients (7 days vs. three days), presumedly caused by dyspnea's insensitivity in critical patients. In con ict with the illness severity, 40.4% of severe COVID-19 patients did not claim dyspnea. This phenomenon is called "happy hypoxia " [21] or "silent hypoxia" [22], which means people are unaware they are being deprived of oxygen. A hypothesis was proposed that SARS-CoV-2 infection-mediated in ammation of the nucleus tractus solitarius affected the brain's hypoxia perception ability [23]. Therefore, it seems that the value of dyspnea as an alert sign for severe COVID-19 was limited.
Fatigue and anorexia also had relatively high incidence rates (36.4% and 21.8%, respectively). Onset symptoms with anorexia were associated with severe illness. This was probably because insu cient food intake in patients with anorexia resulted in malnutrition and hypoproteinemia [24], which are potential risk factors for severe pneumonia. The signi cance of anorexia varied in different age groups of COVID-19 patients. It has been reported that anorexia could be the unique onset symptom in one-third of SARS-COV-2-infected children in the UK [25].
Initially, loss of smell and taste were not considered essential symptoms for COVID-19 patients. Nevertheless, they were de ned as commonly recognized features in later reports from Western countries. In a cross-sectional survey of the prevalence of OTDs, 33.9% (20/59) of SARS-CoV-2-positive hospitalized patients presented at least taste or olfactory disorder, and 18.6% of cases had both disorders [17]. Lechien et al. found that hospitals reported olfactory and gustatory dysfunctions, respectively [18]. However, OTDs reported in COVID-19 patients were much rarer in Asian people, ranging from 5.1-15.7% [26,27]. In this report, the incidence rates of dysgeusia and dysosmia were only 6.2% and 3.1%, respectively, which were markedly lower than those from Western countries. There are some possible reasons for this issue. The clinical spectrum of COVID-19 is broad and varies according to ethnicity, which can be perfectly de ned by the remarkable difference in the incidence rate of OTDs in different regions. It has been reported that SARS-CoV-2 has two major lineages, designated L and S. The prevalent lineage of SARS-CoV-2 in China was different from that in Europe. The L lineage was more prevalent than the S lineage in the early outbreak of COVID − 19 in Wuhan, China [28]. Various SARS-CoV-2 lineages with large molecular divergence might cause distinct clinical symptoms.
It has been reported that older age, high SOFA score, and d-dimer greater than 1 µg/mL are risk factors for COVID-19 patients with poor prognosis at an early stage [16]. In this study, we also found that the risk factors for severe illness included age over 65, male sex, fever lasting for more than 5 days, anorexia at the onset, and mMRC level above grade 2. These data implied an underlying predictive value of speci c clinical symptoms for the severity degree classi cation of COVID-19 patients. Further exploration of the symptom-based prediction model on the illness classi cation of COVID-19, if possible, would help health workers triage patients and recognize patients with COVID-19 requiring hospital admission. It will also be useful for doctors to screen patients inclined to develop the severe illness at an early stage and adopt intensive treatment before it is too late.
In this study, we also followed up with the patients for three months after discharge. Most non-severe and severe patients showed favorable prognoses. Cough, fatigue, and dyspnea were the most common remaining symptoms in convalescent COVID-19 patients. Some patients still did not recover from taste and olfactory disturbance. Compared with non-severe patients, severe patients were more likely to have remaining symptoms after discharge, especially dyspnea.
Our study has some limitations. As a retrospective study, some cases had incomplete documentation of the medical history and clinical manifestation. We had limited time to extract the prospective data in emergent situations. Furthermore, in our cohort, most patients (94.4%) were non-severe cases. All con rmed COVID-19 cases were required to be admitted to designated hospitals or temporary cabin hospitals in China. Thus, the proportion of mild and moderate COVID-19 patients in Chinese hospitals was higher than that in other countries.

Conclusions
This study reported detailed symptomatic features in severe and non-severe patients with COVID-19 in Wuhan, China. The prevalence and duration of COVID-19-related symptoms have been documented. The analysis of the relationship between clinical symptoms and illness severity revealed a cluster of predictors for COVID-19. Our study also showed an optimistic prognosis in most patients through three months of follow-up after discharge. We thank Prof. Xiao-Fei Ye, Department of Health Statistics, Second Military Medical University, for his advice regarding the clinical study, data analysis, and manuscript preparation. We thank all the patients who consented to donate their data for analysis and all the medical staff members on the front line caring for patients.

Authors' contributions
All authors had full access to all the data in the study and took responsibility for the data integrity and data analysis accuracy. WX and HH designed the study and revised the manuscript. LS and JW drafted the manuscript. LS and YW performed the analysis. LS, YW, ZZ, WC, BN, CY, JW, QW, CW, YL, CC, and JY collected the data. LS, PH, and YH designed the electronic follow-up questionnaire form.

Funding
There was no funding source for this study. The corresponding author had the nal decision to submit for publication.

Availability of data and materials
All data is available by application to the study authors.

Ethics approval and consent to participate
The study protocol was approved by the Optical Valley Branch of Maternal and Child Hospital and Shanghai Changzheng Hospital (2020SL007). Written informed consent was waived by the Ethics  Univariate and multivariate logistic stepwise regression models were used to explore the risk factors for severe COVID-19. Based on previous ndings and clinical considerations, a total of 11 variables, including signi cant basic demographic data and symptomatic characteristics in univariate analysis, were included for multivariate analysis. If the difference between groups was not signi cant (e.g., the onset of fever) or the number of events was too small (e.g., the onset of nausea and vomiting), odds ratios could not be calculated, or they were collinear with other variables (e.g., cancer is one of the comorbidities), we excluded this variable from univariate analysis.  Figure 3 Timeline of patients hospitalized with COVID-19 who were dyspnea.