Study design
This cross-sectional study was conducted from 18 February to 1 March 2020. Data were collected from five provinces (Hubei, Guangdong, Sichuan, Jiangsu, and Gansu) that were purposely selected to cover different levels of epidemic severity as defined by the numbers of reported COVID-19 cases (as of 5 March 2020: 67,466, 1,351,539, 631, and 102 cases, respectively) and different regions of China (central, southern, western, eastern, and northern). Three to five cities were selected per province; within the selected cities, three to five districts/counties and five to 10 subdistricts/towns were further selected to represent both different outbreak levels and different regions. CDCs workers were investigated at the province, city, and district/county levels, and PHI workers were investigated at the subdistrict/town level. We targeted at least 5,000 public health workers and achieved a ratio of CDC to PHI workers of approximately 1:2.
Participant recruitment
The eligible participants 1) were aged ≥ 18 years, 2) worked at a CDC unit or PHI in the selected locations during the study period, and 3) participated in COVID-19 control and prevention related work. Site investigators (e.g. CDC workers) in each province distributed the survey links through their WeChat/QQ working groups. WeChat and QQ are popular mobile phone communication/social networking applications used ubiquitously in workplace settings in China. All of the participants were informed of the background, aims, anonymous nature, and duration (approximately 8–12 minutes) of the survey. They were also informed that completing the questionnaire signified their informed consent. No compensation was provided to the participants. The study was approved by the ethics committee of the School of Public Health at Sun Yat-sen University (Reference no.: 2020-012).
Measurements
Socio-demographic characteristics.
We collected information on each participant’s age, sex, and job title, and whether they had children younger than 6 years old. Additionally, we collected information about the areas of routine work before the COVID-19 outbreak only from participants in Guangdong province during the pilot trial. We omitted this section from the survey administered to workers in other provinces after feedback from the pilot trial indicated that the questionnaire was too long.
COVID-19 control and prevention work related variables.
We collected information from all of the participants about their work in terms of its content, their readiness, and its start time. Details of the variables are listed below.
1) Work content. The pre-set list of work content included 14 fieldwork questions covering topics such as face-to-face epidemiological investigations of the patients and close contacts, performance of epidemiological investigations by telephone or video calls, medical observation of the close contacts, collection and shipment of specimens, provision of health education, and performance of community-based investigations. It also addressed 11 non-field work topics, such as technical guideline preparation, data analysis and report writing, laboratory testing, comprehensive coordination and publicity, and technical training. The participants were asked to select the work that they performed from the pre-set list and fill in any other work content that was not included on the list.
2) Time spent in training on COVID-19. This was coded as none, 1–4 hours, 5–8 hours, 9–16 hours, or > 16 hours.
3) Knowledge of COVID-19 prevention and control. This was rated from 1, ‘adequate’ to 5, ‘very inadequate’.
4) Date when the participant began COVID-19 prevention and control work. In the data analysis, we selected the cut-off date of 23 January 2020 because this was both the date when Wuhan city was closed down and the day before Chinese New Year.
5) Epidemic severity in the worker’s province. This was rated from 1, ‘very low’, to 5, ‘very high’, according to the number of confirmed cases in each province.
Efforts and sacrifices.
The participants were asked about their efforts and sacrifices during the outbreak. This information included 1) the number of days during which they worked all night, 2) whether they had worked throughout the Chinese New Year holiday, and 3) whether they had made family sacrifices such as not going home or sending their children to their own parents’ homes to avoid infecting family members.
Perceptions related to COVID-19 and work were also evaluated. One item was used to assess the workers’ concerns about being infected at work, and the responses were ranked from 1, ‘not concerned’, to 5, ‘very worried’. One item was used to assess how long the worker was able to maintain their current work intensity, and the responses were coded as 1–2 weeks, 3–4 weeks, 1–3 months, or > 3 months.
Perceived support and perceived troubles at work were measured using self-constructed items that were developed after discussions with CDC and PHI workers and among the research team. The perceived support scale comprised three items intended to measure the perceived support provided by colleagues, family, and society, and were rated on a Likert-type scale ranging from 1, ‘none’, to 5, ‘very much’. The three items showed acceptable internal consistency in this study (Cronbach’s alpha = 0.760). The perceived troubles at work scale comprised five items, which were rated on a 5-point Likert scale ranging from 1, ‘none’, to 5, ‘very much’. For example, the participants were asked how often they had been treated unfairly at work. In this study, the Cronbach’s alpha for perceived troubles was 0.842.
Health outcomes included overall health status, depression, and anxiety. The participants indicated their overall health by ranking their self-rated health status from 1, ‘very poor’, to 5, ‘very good’. This type of scale has been used widely in China and worldwide [28, 29]. The nine-item patient health questionnaire (PHQ-9) was used to assess the presence of depressive symptoms. The Chinese version of the PHQ-9 has been validated for the general population and showed good internal reliability [30]. The participants were asked to rate how often they had experienced depressive symptoms in the past 2 weeks by using a 4-point Likert scale ranging from 0, ‘none’, to 3, ‘nearly every day’. The total scores ranged from 0 to 27, with a higher score reflecting greater severity. A score ≥ 10 was classified as indicating a major depressive disorder. In this study, the Cronbach’s alpha value was 0.922.
The seven-item General Anxiety Disorder scale (GAD) was used to measure anxiety [31]. Each item was rated on a 4-point Likert scale ranging from 0, ‘never’, to 3, ‘often (almost every day)’. Scores at or above the cut-off value of 10 indicated a probable case of moderate anxiety disorder. In this study, the Cronbach’s alpha value was 0.937.
Statistical analysis
Descriptive analysis was conducted to characterise the study. A chi-square test, t-test, and rank-sum test were used to investigate differences between the CDC and PHI workers. To explore the potential factors contributing to the three health outcomes, namely self-rated health, depression, and anxiety, three sets of logistic regression models were performed in parallel. First, bivariate logistic regression analyses were used to examine the associations between all variables of interest and the three outcomes. Next, adjusted logistic regression models were used to identify the associations between COVID-19-related variables (e.g., COVID-19 control and prevention work related variables, efforts and sacrifices during the outbreak, perceptions) and the three outcomes after controlling for potential confounders (sex, age, having children aged < 6 years, and job title). Finally, multivariate forward stepwise logistic regression models were fitted by using all COVID-19-related variables that were found to be significant in the univariate analysis as candidates for selection and by entering sociodemographic variables into the model. Unadjusted odds ratios (ORu) from the univariate logistic regression models, adjusted odds ratios (AOR) from the multiple logistic regression models, multivariate odds ratios (ORm) from the multivariate stepwise logistic regression, and their respective 95% confidence intervals (CIs) are reported. IBM SPSS Statistics 25 was used for the data analysis. A p value < 0.05 was considered to indicate statistical significance.