Participants and procedure
From 15 March to 15 May 2020, a questionnaire survey was administered to employees in China who were returning to work after the COVID-19 pandemic through an online platform (https://www.wjx.cn/app/survey.aspx). The participants were presented with a description of the study and were informed that participation was voluntary and anonymous. The inclusion criteria were: Older than 18 years old, returning to work between 15 March and 15 May 2020, and able to independently complete the online questionnaire. The exclusion criteria were: Younger than 18 years old (minor), not working or not yet returned to work, having a confirmed or suspected COVID-19 infection after resuming work or during isolation, and being unable to read or write, or to complete the questionnaire independently.
Invitations were sent to all potential participants via WeChat, the most popular social media app in Mainland China with one billion active users daily. At the same time, the research team providing medical services in the quarantine area and in the community offered to scan the QR code on the facility to complete the online survey.
Online informed consent was obtained by asking participants to check the box on their device’s screen with the response of their choice (i.e. ‘I agree to participate in the survey’ or ‘I do not agree to participate in the survey’). If they checked ‘I do not agree’, the computer program terminated automatically. The questionnaire and survey was specifically developed for this study. The full questionnaires are presented in Additional file 1.
The study was approved by the institutional review board of Ningxia Medical University (document number: 2020112).
Part of our online questionnaire consisted of questions concerned with socio-demographic information, including sex, age, marital status, education level, household registration, occupation, years of employment, quarantine status, and source of epidemic-related concerns.
Based on the New Diagnosis and Treatment Scheme for Novel Coronavirus Infected Pneumonia (fifth trial edition)  and the Protocol on Prevention and Control of COVID-19 (sixth edition)  issued by the National Health Commission of China, we developed a definitive questionnaire concerning the epidemiological characteristics, clinical symptoms, and comprehensive prevention and control measures for COVID-19. The scale comprised 10 items, with one point awarded for each correct answer (total score: 0–10). The higher the score, the higher the COVID-19 knowledge level.
Considering the high incidence of depression and anxiety symptoms reported among quarantined subjects in previous studies , depression and anxiety were assessed in the present study, using the Patient Health Questionnaire (PHQ-9) and the Generalized Anxiety Disorder (GAD-7). Both scales have shown good reliability and validity in both foreign and domestic studies [19, 20]. Both scales use a four-point Likert scale (range: 0–3) for each item. For the PHQ-9, which comprises nine items, scores of 0–4, 5–9, 10–14, 15–19, and 20–27 indicate no, mild, moderate, severe, and extremely severe depression, respectively. The GAD-7 comprises seven items, with scores of 0–4, 5–9, 10–14, 15–21 indicating no, mild, moderate, and severe anxiety, respectively.
To evaluate the respondents’ compliance with COVID-19-preventive behaviours when returning to work, items concerning 10 such behaviours were included. The 10 behaviours were 1) wearing masks; 2) covering your mouth and nose when coughing and sneezing; 3) washing your hands frequently; 4) avoiding contact with, buying, or eating the meat of wild animals; 5) being mindful of symptoms such as fever and coughing and performing comprehensive health monitoring; 6) avoiding close contact with people showing symptoms of respiratory disease; 7) avoiding crowded places; 8) keeping rooms clean and opening windows for ventilation; 9) reducing social gatherings and visits to friends and relatives; and 10) maintaining a healthy diet and taking moderate exercise. For these items, a five-point Likert scale was used (1 = ‘completely unnecessary’, 5 = ‘very necessary’). Total scores ranged from 5 to 50 points. Scores of ≥40 indicated higher compliance; scores of < 40 indicated lower compliance; this means that the lower the score is, the lower the behaviour compliance is.
As the questionnaire was electronic, responses were required for all questions before submission; that is, if the questions were not answered completely, they could not be submitted. A ‘skip item’ option was provided to increase the likelihood of obtaining complete and logical responses. To avoid repeat answers, each IP address could only access the questionnaire once. The survey took approximately 8–15 min to complete; a preliminary survey showed that it would take at least 300 s to complete all questions. If the questionnaire is completed in less than 300 s, it indicates that the participants did not read the questionnaire carefully, and that the quality of the responses will be poor; thus, the responses provided in less than 300 s were excluded.
The data were checked and categorised. SPSS version 21.0 statistical software was used for statistical calculations. Two-sided p-values of 0.05 were considered to indicate statistical significance. Categorical variables are presented as frequencies and percentages, while continuous variables are presented as means and standard deviations (±SD). The percentage differences in behavioural compliance across categorical variables were examined using chi-squared tests. An unconditional logistic regression method was adopted to determine the factors influencing respondents’ compliance after controlling for covariates. We used the unconditional logistic analysis method, as the dependent variable is dichotomous and matching design data is not used in this study. The dependent variable was the level (higher/lower) of respondents’ compliance with preventive behaviours (1 = no, 0 = yes). Education level, occupation, COVID-19 knowledge, anxiety, and quarantining were set as independent variables, while gender, age, household registration, and years of employment were set as covariates. Multiple categorical variables (age, occupation, education, and years of employment) were set as dummy variables.
The adjusted odds ratios (OR) for the variables and their 95% confidence intervals (95% CIs) were calculated using the unconditional logistic regression model.