Ethical Considerations

The Survey and Behavioural Research Ethics Committee of the Chinese University of Hong Kong approved the study (approval no. SBRE-21-0730). All eligible participants provided digital informed consent via electronic means at the beginning of the survey. The survey did not collect personally identifiable information, and we encrypted email addresses collected for survey distribution as anonymized respondent identities. All respondents’ data were recorded in an anonymous manner with a unique internally generated reference key. We assured the respondents of the anonymous, confidential nature of the study, where only aggregate data would be reported. We strictly adhered to the data privacy policy in the approved study protocol.

Survey Instrument

An expert panel consisting of epidemiologists, physicians, public health professionals, professors, and biostatisticians composed and validated the survey. The survey was pilot-tested to optimize coherence and clarity and was subsequently revised according to the pilot findings. During the curation of survey items, we referred to published studies that used validated survey instruments for long COVID symptoms [13,27]. The survey comprised 31 questions and required a median of 13 minutes to complete. The information sheet of the survey provided background information about COVID-19. The survey was delivered in traditional Chinese for participants in Hong Kong and simplified Chinese for respondents in the other 3 cities of mainland China. Before commencement of the survey, prospective participants confirmed that they were aged 18 years or older and had previously been diagnosed with COVID-19.

The survey collected respondents’ sociodemographic information, past medical history, use of chronic medications, history of COVID-19 vaccines received, and severity of COVID-19. The survey contained several parts: First, we collected the respondents’ sociodemographic information, including gender, marital status, employment status, types of jobs, educational level, living status, smoking habits, income level, self-reported health status, and long-term medication. In addition, we asked for the dates of positive and negative COVID-19 tests, the severity of COVID-19, the number of COVID-19 vaccines that were received, and the dates of taking vaccination. To identify the severity of COVID-19, we used the COVID-19 WHO severity classification [28] to assign respondents to 4 categories (mild, moderate, severe, and critical) based on the presence of pneumonia (eg, adults with clinical signs of pneumonia, respiratory rate>30 breaths/minute, severe respiratory distress, or SpO₂<90% on room air; WHO mentioned that a diagnosis can be made on clinical grounds with the assistance of chest imaging, such as radiograph, computed tomography (CT) scan, or ultrasound), hypoxia requiring oxygen, the need for mechanical ventilation, and a history of admission to the intensive care unit (ICU) due to COVID-19.

The respondents’ medical history before and after the diagnosis of COVID-19 was also collected. Respondents were asked whether they had certain types of diseases before diagnosis (yes or no), the year of disease diagnosis, whether they took medications for the diseases (yes or no), and an open-ended field allowing them to document their medications. Medical conditions on the survey were infection, tumor, metabolic diseases, hematological disorders, mental illness, nervous system disease, circulatory system disease, respiratory disease, digestive disease, reproductive and urinary diseases, pregnancy complications, immune system disease, skin and subcutaneous tissue diseases, and musculoskeletal and connective tissue disease, which have been used in previous studies on the association between comorbidities and COVID-19 infection [29].

We also inquired about the presence of COVID-19 symptoms encountered by the respondents that were persistent for at least 3 months after COVID-19 infection. Referring to previous surveys [13,27], we listed 30 common symptoms: fatigue, fever, chills, inability to perform exercise, night sweats, hair loss, headache, dizziness, chest pain, rapid heartbeat, cough, sputum, sore throat, runny nose, dyspnea, arthralgia, myalgia, nausea, vomiting, diarrhea, abdominal pain, stomachache, anosmia, loss of taste, blurred vision, difficulty in concentration, difficulty in fall into asleep, feeling anxious, feeling sad, and memory problems. An open-ended field allowed the respondents to document other symptoms. These symptoms also appeared in “Symptoms of Infection with Coronavirus-19 (SIC),” which is a comprehensive patient-reported outcome measure developed to evaluate vaccines and treatment for COVID-19 [30]. To capture clinically meaningful symptoms, we assigned a 4-point Likert scale to assess the impact of the symptoms on the respondents’ activities of daily living: 1 (no impact), 2 (little impact), 3 (moderate impact), and 4 (high impact). In addition, we classified the symptoms according to the bodily systems involved.

Sampling Strategy

The sample partners of Qualtrics randomly selected eligible respondents for the study, aiming for national representation through routers and a sophisticated application programming interface (API). To enhance the representativeness of the survey, we used a simple random sampling strategy to recruit potential participants by trying to match population demographics to our survey. The study collected data on the population of each city, and eligible participants were randomly selected and invited to complete the survey. The survey was accessed through the website, a smartphone, or other e-devices. Data scrubbing was subsequently performed after receiving all survey responses to remove unfavorable data, optimizing data accuracy and reliability. A more detailed survey methodology has been described elsewhere [26].

Outcome Variables and Sample Size

The primary outcome variable was the prevalence of long COVID symptoms, moderate or severe long COVID symptoms, and severe long COVID symptoms. We defined these 3 outcome variables as having at least 1 symptom with any degree of impact, at least 1 symptom with moderate or severe impact, and at least 1 symptom with severe impact, respectively. The secondary outcomes included the distribution of long COVID symptoms and their patterns. We assumed the proportion of the primary outcome as 50%, which provided a maximum sample size for each group. From the formula precision=1.96 × √[(p) × (1 – p)/N], where "p" refers to proportion of the primary outcome, a sample size of approximately 550 respondents would achieve a precision level of 0.04 for each city, so a total of 2200 respondents were required.

Severity of COVID-19

Most respondents suffered from pneumonia at the time of COVID-19 diagnosis (n=2141, 78.9%), required hospital admission for COVID-19 management (n=1944, 71.7%), needed oxygen due to COVID-19 (n=1427, 52.6%), and received antiviral agents (n=1747, 64.4%); see Multimedia Appendix 1, Table S1. A significant proportion necessitated artificial ventilation (n=986, 36.4%) and admission to the ICU (n=940, 34.7%). Up to 1552 (57.2%) of them had severe or critical COVID-19, and 1160 (42.8%) had mild or moderate severity of COVID-19. Most of the respondents had received vaccination before COVID-19 diagnosis, with the majority having taken 3 or more doses of Coronavac-Sinovac (n=1525, 56.2%), following by those who had taken 2 doses of Coronavac-Sinovac (n=786, 29.0%). There were about 169 (6.2%) respondents who had taken 2 or more doses of Pfizer-BioNTech.

Long COVID Symptoms

Multimedia Appendix 1, Table S2, shows the distribution of long COVID symptoms. The highest proportion of respondents suffered from fatigue (n=914, 33.7%), cough (n=865, 31.9%), sore throat (n=841, 31.0%), difficulty in concentrating (n=828, 30.5%), feeling of anxiety (n=817, 30.2%), myalgia (n=811, 29.9%), arthralgia (n=811, 29.9%), sputum production (n=789, 29.4%), and difficulty in falling asleep (n=792, 29.2%); see Figure 1. The prevalence of having long COVID symptoms of any severity, moderate or high severity, and high severity was 90.4% (n=2452), 62.4% (n=1692), and 31.0% (n=841), respectively. The symptoms were qualified in terms of self-perceived severity, and sensitivity analysis of using 2 or more symptoms showed a better estimate of long COVID syndrome/disorder (Figure 2). These prevalence figures were the highest in Hong Kong (n=541, 97.7%, n=385, 69.5%, and n=194, 35.0%, respectively) and Shanghai (n=662, 92.2%, n=461, 64.2%, and n=287, 40.0%, respectively), followed by Guangzhou (n=630, 88.1%, n=451, 63.1%, and n=212, 29.7%, respectively) and Beijing (n=619, 85.4%, n=395, 54.5%, and n=148, 20.4%, respectively); see Figure 3. The prevalence figures were the highest in the second wave (from July to December 2020; n=67, 91.8%, n=48, 65.8%, and n=16, 21.9%, respectively) and the third wave (from January to June 2021; n=453, 91.7%, n=329, 66.6%, and n=175, 35.4%, respectively), followed by the first wave (from January to June 2020; n=122, 90.4%, n=64, 47.4%, and n=36, 26.7%, respectively), the fifth wave (from January to June 2022; n=1446, 90.1%, n=998, 62.2%, and n=503, 31.3%, respectively), and the fourth wave (from July to December 2021; n=364, 89.9%, n=253, 62.5%, and n=111, 27.4%, respectively); see Multimedia Appendix 1, Figure S1. A slightly negative correlation was found between the timing of COVID-19 and the number of long COVID symptoms (β=–.046, P=.016), supporting that long COVID symptom numbers decline with a longer duration of follow-up.

Open in a separate window

Figure 1

Distribution of long COVID symptoms. The highest proportion of respondents suffered from fatigue (n=914, 33.7%), cough (n=865, 31.9%), sore throat (n=841, 31.0%), difficulty in concentrating (n=828, 30.5%), feeling of anxiety (n=817, 30.2%), myalgia (n=811, 29.9%), arthralgia (n=811, 29.9%), sputum production (n=789, 29.4%), and difficulty in falling asleep (n=792, 29.2%).

Open in a separate window

Figure 2

The prevalence of having long COVID symptoms of any severity, moderate or high severity, and high severity was 90.4% (n=2452), 62.4% (n=1692), and 31.0% (n=841), respectively. The symptoms were qualified in terms of self-perceived severity, and sensitivity analysis of using more than 1 symptom showed a better estimate of long COVID syndrome/disorder.

Open in a separate window

Figure 3

Prevalence of long COVID with severity. These prevalence figures were the highest in Hong Kong (n=541, 97.7%, n=385, 69.5%, and n=194, 35.0%, respectively) and Shanghai (n=662, 92.2%, n=461, 64.2%, and n=287, 40.0%, respectively), followed by Guangzhou (n=630, 88.1%, n=451, 63.1%, and n=212, 29.7%, respectively) and Beijing (n=619, 85.4%, n=395, 54.5%, and n=148, 20.4%, respectively).

Strengths and Limitations

This study was relatively large scale as compared to other surveys, and it adopted a representative sampling strategy involving 4 major cities in China. The response rate was high relative to other similar studies, thus enhancing its generalizability to other settings. Nevertheless, several limitations should be addressed. First, its cross-sectional nature did not allow a cause-and-effect relationship to be established due to the possibility of reverse causality. However, the objective of this study was to identify independent risk factors instead of causes of long COVID. In addition, there existed potential recall bias as some study participants self-reported the presence of long COVID symptoms, especially those with an earlier diagnosis of COVID-19. It should also be noted that the definition of long COVID is different across various authorities, with the National Institute for Health and Care Excellence (NICE) and the US Centers for Disease Control and Prevention (CDC) using 4-12 weeks and 4 weeks, respectively, as the time frame of its persistent symptoms. Furthermore, there could be residual confounders in our multivariate regression analysis, and these could vary according to different cities where the management strategies for COVID-19 might be different. In particular, it is unknown whether the study participants received certain treatments to relieve their long COVID symptoms at the initial stage of the development, thus reducing their risk of long COVID symptoms. Lastly, as in most surveys, not all potential study participants in the sampling frame were eligible before the recruitment process. In addition, the older population who may have a substantially higher burden of both COVID-19 and long COVID symptoms were underrepresented in this study. As the survey had to be completed through e-devices, relatively fewer older people were included as a higher proportion of them did not know how to use e-devices. Future study on the prevalence of long COVID in the older population is warranted.

All authors were responsible for the study concept and design. MC-SW, JH, and Y-YW were responsible for data acquisition and analysis, had full access to all of the data in the study, and take responsibility for data integrity and the accuracy of the data analysis. All authors were responsible for the interpretation of data, drafting, and critical revision of the manuscript for important intellectual content.

This work was supported in part by the Health and Medical Research Fund (HMRF) – Food and Health Bureau Commissioned Research on COVID-19 (project reference number: COVID1903002) to FK-LC and MC-SW. The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.