2.1. Study Design and Sampling
The Hong Kong COVID-19 Health Information Survey (CoVHInS) was a population-based dual sampling landline and online survey conducted from 9–23 April 2020. The target population was general Hong Kong residents aged 18 or above. Social Policy Research Limited, a reputable local survey agency, was commissioned to conduct the survey.
We adopted two-stage random sampling in the landline survey using the Web-based Computer-Assisted Telephone Interview system (Web-CATI). Residential telephone directories that covered approximately 76% of Hong Kong residents [21
] were used to generate a random list of telephone numbers for interview. Invalid numbers (e.g., fax line, non-residence line and non-working line), nonresponses, and ineligible households were excluded. After telephone contact had been successfully established with a target household, one eligible person was selected using the “next birthday” rule (i.e., the household member whose birthday was nearest to the interview date was selected). All telephone interviewers completed a half-day training related to COVID-19 knowledge, contents of the questionnaire, sampling procedures and interviewing techniques. Briefing and de-briefing sessions were arranged during data collection, and rigorous quality checks were adopted (17.2% were checked) to ensure research fidelity. Each interview took approximately 25 min. Among the 816 valid telephone number sampled (305 refused, 11 dropout), 500 respondents completed the interview yielding a response rate of 61.3%.
The online survey was conducted on a representative panel of Hong Kong residents. This panel was previously formed by inviting local mobile phone users to join. All mobile phone numbers (prefix starting with 5, 6 or 9), which covered over 90% of Hong Kong residents, received an invitation message. These numbers were generated using the Numbering Plan for Telecommunication Services in Hong Kong provided by the Office of the Communications Authority (OFCA). A total of 100,079 residents covering diverse socio-economic backgrounds joined the panel. Stratified random sampling by sex and age was adopted to select a random list of panel participants, who were then invited to join the online survey by an invitation text message. Participants self-administered the questionnaire via Web-CATI. The response rate was 61.7% (1001 of 1623 eligible panel participants).
Ethics approval was granted by the Institutional Review Board (IRB) of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (UW 20-238). Informed consent was obtained from all respondents before answering the questions.
The adoption of social distancing measures (yes/no) since the first case confirmed in Hong Kong included: (i) avoid going out, (ii) avoid going to crowded places, (iii) avoid going to high-risk places (e.g., bar, wet market, hospital), (iv) avoid social gatherings of more than four people (government regulation, with penalty, issued on 29th March), (v) avoid greetings with physical contact such as handshaking, hugging and kissing, and (vi) keep 1.5 meters from others in public. The total number of social distancing measures was calculated (ranged 0–6). Perceived overall effectiveness of and perceived compliance with the above social distancing measures were measured on a scale ranging from 0–10 with higher scores indicating higher perceived effectiveness and compliance. Stay-at-home, which was voluntary, was measured by the number of days at home except for essential tasks in the past seven days. Personal protection measures including: (i) wearing surgical mask, (ii) washing hands with alcohol-based sanitizers, (iii) using alcohol to clean daily necessities, and (iv) adding water to household drainage system in the past seven days were recorded.
Stress level in the preceding month (from early-March to mid-April 2020) was assessed by the four-item Perceived Stress Scale (PSS-4) [23
]. PSS-4 consists of four items measuring the degree of ability to cope with existing stressors (positive elements, two items, Cronbach’s α = 0.68) and the degree of lack of control and affective reactions (negative elements, two items, Cronbach’s α = 0.69). Each item is rated on a Likert-like scale ranged from 0 (never) to 4 (very often). Higher total scores on the four items (0–16) indicate a higher perceived stress level. The Chinese version of PSS-4 has been validated in our previous study with satisfactory internal consistency (Cronbach’s α = 0.67) [24
Anxiety and depressive symptoms in the past two weeks (from late-March to mid-April 2020) were assessed using the four-item Patient Health Questionnaire (PHQ-4). PHQ-4 consists of the two-item General Anxiety Disorder (GAD-2) and the two-item Patient Health Questionnaire (PHQ-2) [25
]. GAD-2 covered two core criteria for generalized anxiety that screen for social panic and anxiety disorders. PHQ-2 measured depressive symptoms, depressed mood and loss of interest, two core diagnostic criteria for major depression disorder [26
]. Each item scores on a Likert-like scale ranging from 0 (not at all) to 3 (nearly every day). Subscales of the GAD-2 and PHQ-2 scores range from 0–6 with a score of ≥3 indicating anxiety and depression symptoms [27
]. We have previously validated the Chinese version of the PHQ-2 in the Hong Kong population [29
]. The internal consistency of GAD-2 (Cronbach’s α = 0.81) and PHQ-2 (Cronbach’s α = 0.81) were positive in this study.
Sex, age, marital status (never married, married or cohabited, divorced or separated, or widowed), current living arrangement (living alone, co-living with family members, or co-living with other people), and socioeconomic status (SES) including educational attainment (primary or below, secondary, or tertiary), employment status (full-time work, part-time work, student, housemaker, unemployed or retired) and monthly personal income (HK$ ≤ 10,000; 10,001–20,000; 20,001–30,000, 30,001–40,000, or ≥40,001; US$1 = HK$7.8) of the respondents were recorded.
2.3. Statistical Analysis
We used Chi-squared test and t-test to compare sociodemographic characteristics and mental health symptoms between the landline telephone and online self-administrated samples. To improve the representativeness of the sample, all data were weighted according to provisional figures obtained from the Census and Statistics Department on the sex, age, and education attainment distributions of Hong Kong’s general adult population in 2016. Multivariable linear regression was used to examine the associations with sociodemographic characteristics of number of social distancing measures adopted, number of days stayed-at-home and perceived compliance with social distancing measures. The association of mental health symptoms including stress, anxiety and depression with the number of measures adopted, number of days stayed-at-home and perceived effectiveness and compliance were calculated by multivariable linear (for stress) and logistic (for anxiety and depression) regressions. In regression model 1, potential sociodemographic confounders including sex, age, educational attainment, current employment status, and monthly income were adjusted. We additionally adjusted for the four personal protection measures in regression model 2. The association between personal protection measures and mental health symptoms were analyzed by multivariable regression models adjusted for sociodemographic factors and social distancing. Effect modifications by age (18–59, 65+ years) and education attainment (primary or below, secondary, and tertiary) on the associations between mental health symptoms, stay-at-home and compliance with social distancing were assessed using the interaction terms. Analyses were performed using STATA version/MP 15.1 (StataCorp LP, College Station, TX, USA).