This study was part of the COVID-19 Health Information Survey (CoVHIns)—which is a cross-sectional survey among adults in Hong Kong who are aged ≥18 years—which investigated COVID-19–related information use, preventive behaviors, and well-being. The survey was conducted on April 9-23, 2020, after the peak of the second wave of the outbreak, and when social distancing measures were implemented. Data were collected using telephonic and web-based surveys. All interviews were conducted by trained interviewers of Social Policy Research Limited through the Web-based Computer Assisted Telephone Interview system.

The details of CoVHIns have been reported previously [22,23]. Briefly, a 2-stage sampling method was adopted for the telephonic survey. First, telephone numbers were retrieved from the residential telephone directories and randomly listed for interview. Invalid numbers, lack of responses (after being called for a maximum of 5 times), and ineligible households (including individuals aged <18 years or those who are unable to communicate in Cantonese or Mandarin) were excluded. Second, once a household was successfully contacted, the eligible family member whose birth date was closest to the interview date was invited to complete the interview. Each interview lasted approximately 20 minutes. A total of 816 landline telephone numbers were successfully sampled, and 500 participants consented to and completed the interview (response rate=61.3%).

In addition, web-based surveys randomly sampled participants from a representative panel of >100,000 mobile phone users, which was generated by sending text messages to a random list of mobile phone numbers provided by the Numbering Plan for Telecommunication Services (prefixes 5, 6, 9). Stratified random sampling by sex and age was adopted. Text messages with an invitation were sent to the randomly selected members in the panel. Among 1623 eligible individuals contacted, 1001 participants consented to and completed the questionnaire on the internet (response rate=61.7%). Ethics approval was granted by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (approval# UW-20-238).

Seeking of web-based information on COVID-19 was self-reported (sought or no-sought). eHealth literacy was assessed among individuals who had sought web-based information on COVID-19, considering that eHealth literacy is based on the experience of access to web-based information [24]. We used the Chinese version of the eHealth literacy scale (eHEALS) to measure eHealth literacy levels by asking participants’ about their past experience in seeking web-based information on COVID-19 (Multimedia Appendix 1). The eHEALS contains 8 items scored with a 5-point Likert scale (ranging from “strongly disagree” to “strongly agree”). The total score ranged from 8 to 40, with a higher score indicating higher eHealth literacy [25]. A Cronbach α of .95 was used in our study. Consistent with the Chinese version of eHEALS [25], we observed a unidimensional structure of the Chinese eHEALS with adequate model fitness (comparative fit index 0.974 [>0.95, acceptable], root mean square error of approximation 0.097 [close to 0.06, acceptable], and Tucker-Lewis index 0.964 [>0.95 acceptable]) [26]. We divided the eHEALS score into 4 categories (Q1-Q4) based on quartile values (median 28, IQR 22-32) in accordance with previous studies, using the median as the cutoff [7,27]. Specifically, Q1 was the interval of the overall eHEALS score ranging from the lowest value (ie, 8) to a score of 22; Q2, from 22 to 28; Q3, from 28 to 32; and Q4, from 32 to the highest value (ie, 40).

Based on World Health Organization guidelines for COVID-19 prevention [28], we assessed personal preventive behaviors in the past 7 days, including the following: “wearing surgical masks when going out,” “wearing fabric masks when going out,” “washing hands with alcohol-based sanitizer,” “adding water/bleach to the household drainage system,” and “keeping a social distance from people in public areas (eg, 1.5 meters),” with responses of “never,” “occasionally,” “sometimes,” and “often” (Multimedia Appendix 2). Adherence to personal preventive behavior was dichotomized (low or high adherence) on the basis of previous studies on the association between eHealth literacy and health behaviors [14,24]. Responses of “never,” “occasionally,” and “sometimes” were together considered as low adherence and “often” was considered as high adherence.

Education levels and income were considered as indicators of SES. Education levels were measured as categorical variables (“primary or below,” “secondary,” and “tertiary or above”) on the basis of the highest education level attained. We measured monthly personal income in accordance with 6 predefined categories (from ≤HK $10,000-$50,001 [US $1=HK $7.8]). Since few participants had an income of HK $40,001-$50,000 and ≥HK $50,001, the data were recoded in 4 categories including ≤HK $10,000, HK $10,001-$20,000, HK $20,001-$30,000, and >HK $30,000 to obtain robust outcomes on regression analyses.

Other demographic data included sex, age, and marital status (never been married, married or cohabitating, and divorced or separated or widowed). Employment status was categorized as economically active (full-time work or part-time work) and economically inactive (student, homemaker, unemployed, and retiree) [29]. Any chronic diseases were self-reported (any or none).

All data were weighted by sex, age, and education levels in accordance with the 2016 population by census to improve the representativeness of the sample.

First, disparities in seeking web-based information on COVID-19 (dichotomized variable) by SES were assessed through multivariable logistic regression, which yielded adjusted odds ratios (aOR) for seeking web-based information on COVID-19. Second, socioeconomic disparities in eHealth literacy, being a continuous variable, were assessed through linear regression, which yielded unstandardized regression coefficients to reflect the change in the eHEALS score for a unit change in the independent variable. Third, we used the slope index of inequality (SII) to estimate the absolute difference in the eHEALS score between the most advantaged and most disadvantaged groups. The SII has been recommended by the World Health Organization, and a high SII indicates severe inequality [30]. Income categories were first ranked from the lowest to highest, and the cumulative proportion of participants were assigned to each category on the basis of the midpoint of range as the code for each category. The eHEALS score was then regressed against the cumulative proportion of each income category [30]. A similar analysis was performed for education-related SII. As each personal preventive behavior was dichotomized as low and high adherence, the associations (determined with aOR and 95% CI values) of seeking web-based information on COVID-19 and the eHEALS score with each personal preventive behavior were estimated through multivariable logistic regression adjusted for demographic variables, SES, and chronic disease. All analyses were performed using Stata (version 15.1, Stata Corp).

This study is the first to report socioeconomic disparities in seeking web-based information on COVID-19 and eHealth literacy during the COVID-19 pandemic and their association with a high adherence to COVID-19–related preventive behaviors, including wearing surgical masks, washing hands, adding water or bleach to the household drainage system, and social distancing.

Seeking of web-based information on COVID-19 was observed among younger participants in our study, concurrent with previous studies on web-based health information seeking behaviors [31]. A recent study also indicated that younger family members sought web-based information for the elderly during the pandemic [32]. Such an age disparity in information seeking can be attributed to the higher penetration rate of internet devices such as personal computers and smartphones among younger rather than older individuals [18]. Small font sizes, crowded visual presentations, and distracting flashes on most web-based information sources could be barriers to web-based information seeking among the elderly [33]. More frequent health information seeking from traditional media such as the radio and newspapers were observed among the elderly in our previous population-based study [20]. Our finding that higher SES including education levels and income is associated with seeking web-based information on COVID-19 is consistent with that of previous studies on seeking of web-based health information conducted before the COVID-19 pandemic [13,31]. Compared to our previous study, which measured SES disparities in seeking of web-based health information in 2009-2012, the ORs of web-based information seeking were found to decrease (eg, tertiary or above education: 2.98 in 2020 vs 8.00 in 2009-2012) [20]. Such a reduction in the effect size could result from the increased popularity of internet-accessible devices in the general population in Hong Kong [34]. Alternatively, the decreased ORs could be attributed to increased information seeking behaviors in crisis events, which were suggested as a means of reducing situation uncertainty and controlling risk [35].

Furthermore, we found age and SES disparities in eHealth literacy, thus revealing disparities in locating, understanding, and the utility of web-based information among those who sought web-based information on COVID-19. The associations between sociodemographic characteristics (eg, age and SES) and eHealth literacy observed in our study were concurrent with previous findings on health literacy [36,37]. Our study focused on eHealth literacy because the Internet has been the major platform for disseminating health information during the COVID-19 pandemic, since it is easily available and can instantly update information on, for example, preventive behaviors and access to social and health services. Considering that web-based information is complex and misinformation on the internet has led to inappropriate behaviors [38], those who used the internet for health information but with limited eHealth literacy skills to discern the quality of different information sources were the potential at-risk populations and require more attention. We observed stronger associations between eHealth literacy and education rather than income, which probably reflects the notion that knowledge and skills are more affected by cognitive function than by the available material. The education-related disparity in eHealth literacy was larger than the income-related disparity in our study, which suggest that education plays a more crucial role than income in affecting eHealth literacy. Other studies have also suggested that disparities in eHealth literacy were due to knowledge gaps rather than merely physical barriers to internet access [39]. Furthermore, we noticed that eHealth literacy and seeking of web-based information on COVID-19 have similar risk factors including an older age and a lower SES [13]. eHealth literacy enables seeking of web-based information [8]; further studies can explore the extent to which low eHealth literacy hinders seeking of web-based information among older and low-SES individuals.

Successful control of the COVID-19 pandemic would need universal adherence to preventive behaviors, which have proven very effective in reducing disease spread [40]. Seeking of web-based information on COVID-19 is associated with the adherence to preventive behaviors, suggesting the need to understand the barriers to the low SES group, including low eHealth literacy to use the internet to obtain health information. Our participants with higher eHealth literacy showed high adherence to personal preventive behaviors, which was consistent with the results of previous non–COVID-19 studies that eHealth literacy correlated with health behaviors such as regular physical exercise and balanced diets [24,41,42]. Our study extended those findings to COVID-19 preventive behaviors in the specific context of the COVID-19 pandemic, in which increasing misinformation has been disseminated through the internet. Low eHealth literacy could lead to difficulties in fact-checking and mistrust in peoples’ beliefs in coronavirus conspiracies, which would impede the performance of preventive behaviors [43]. Such disparities in eHealth literacy have led to disparities in guidelines on preventive behaviors, their profound consequence being health inequality [44]. Web-based information should be better designed to address the eHealth literacy levels of target users, particularly those with a low SES, to bridge the existing knowledge gap. Further studies are needed to explore how to improve eHealth literacy effectively and the approach involving the use of eHealth literacy to facilitate better health behaviors.

Our study has some limitations. First, the cross-sectional data cannot confirm the causal association, although it is unlikely that higher eHealth literacy or seeking of web-based information on COVID-19 would lead to higher education and income. Second, we measured the perceived eHealth literacy instead of actual performance on the internet. Some studies have measured performed eHealth literacy and reported a weak or moderate correlation between perceived and performed eHealth literacy [15,45]. Third, eHEALS, the most commonly used validated scale, was developed at the early stage of internet technology; its fit with Web 2.0-related technologies (social media) was not clear because of the considerable changes on the internet (more participatory and interactive web) [46]. Future studies are needed to improve the model of eHealth literacy with the evolution of the internet and the COVID-19 pandemic [46,47]. Fourth, we did not collect data on channels of web-based information on COVID-19; hence, further studies should include details of the frequency and channels on the internet for seeking information on COVID-19.

This study provides the first evidence that Chinese adults with a higher SES had higher eHealth literacy and sought web-based information on COVID-19, and that both these factors are associated with high adherence to the guidelines on preventive behaviors during the COVID-19 pandemic. Effective interventions are needed to enhance the low eHealth literacy skills of low-SES individuals to combat the COVID-19 pandemic.