Users of the social media and messaging platforms Instagram, Twitter, and WhatsApp, who were aged ≥18 years, were recruited to participate in the survey via a web link. The link was posted on the public social media pages of the following authors KFA, MHJ, and MA. Adult users who viewed the authors’ public pages were asked to voluntarily fill out the survey. Resharing of the survey link by users was permitted on all three platforms for snowball sampling. The survey link was active from March 28 to April 4, 2020. Users with the highest visibility of the authors’ accounts resided predominantly in the Arabian Gulf countries: Bahrain, Kuwait, Saudi Arabia, and United Arab Emirates. The survey was administered in both English and Arabic, the two predominantly spoken languages in these countries.

Our survey was composed of 13 questions (see Multimedia Appendix 1 for questions). The first 7 questions addressed topics such as source(s) of information on COVID-19, viral preventative behavior(s), and knowledge of symptoms of COVID-19. We also asked whether the participant had acquired the infection; if they answered “no,” they were asked to state their perceived likelihood of acquiring it in the next 3 months. Finally, we asked participants whether they would vaccinate against COVID-19. The other 6 questions pertained to subject demographics: age group, gender, educational level, country of current residence, underlying illness(es), and if the participant works or studies in the health care sector. The survey questions were partially adapted from a questionnaire designed and published by the Understanding America Study, which is maintained by the Center for Economic and Social Research at the University of Southern California [13]. The survey was designed on the online platform Zoho Creator (Zoho Corp). Informed consent in the form of agreement to an information leaflet was obtained from all subjects prior to survey initiation. The survey overall took less than 5 minutes to complete. The study was approved by the Research Ethics Committee at the Royal College of Surgeons in Ireland-Medical University of Bahrain.

All data on personal demographics (age groups, gender, highest educational level, working or studying in the health care sector, presence of medical comorbidities, country of current residence, and status of COVID-19 infection) were categorically expressed as counts and percentages of total respondents. Data on source(s) of information during COVID-19, viral preventative behavior(s), and knowledge of symptoms of COVID-19 were also expressed as counts and percentages of total respondents.

Subgroup categorical analyses were conducted to examine the effects of age (≥35 years versus <35 years), gender (female versus male), underlying illness(es) (yes versus no), and status of working or studying in the health care industry (yes versus no) on the perceived likelihood of being infected with COVID-19 in the next 3 months, as well as the likelihood of vaccinating against COVID-19. Results were categorically expressed as counts and percentages of total respondents in each of the subgroups.

The development of the research questions and outcomes was primarily executed by the study investigators and guided by current literature. Patients were not involved in the design of this study. The involvement of patients and the public, however, was crucial during the recruitment phase of the study (ie, dissemination of the survey link, via snowballing effect, through users’ social media accounts to others in the digital community). After publication, the results will be disseminated to participants on the same platforms that hosted the survey link, through a short, interactive video recorded and posted by the principle investigator of the study.

In the midst of an unprecedented global health crisis, it is critical to know if the educational messages for the prevention of COVID-19 are being delivered, understood, and implemented by the public. Our survey, conducted through three social media outlets, facilitated a large and rapid collection of 5677 responses within 7 days and showed that the majority of respondents were well aware of the symptoms of COVID-19 and the measures necessary to prevent it, as per the WHO guidelines [14]. Additionally, a large portion of subjects (around 50%) utilized reliable sources of information such as the WHO. Such collective responses are indicative of a well-informed cohort, which may be explained by the younger, more educated (75% had higher education) users of social media. It is of note that a large number of these responses (56.0%) were from people in Bahrain. The spread of COVID-19 in Bahrain was well contained at the time of survey distribution (998 cases as of April 11, 2020) [2]. The number of fatalities (7 as of April 11, 2020) [2] has been very low considering it is the third most densely populated country in the world. Such statistics can be due to the early and extensive communication with the public, particularly through social media outlets [15,16]; this highlights the importance of conveying general knowledge about the virus to residents to control spread.

Our subgroup analysis examining the effects of age (<35 versus ≥35 years old), gender, presence of underlying illnesses, and work or study in health care revealed some intriguing findings. Younger subjects (<35 years old) reported a higher likelihood of vaccinating against COVID-19 compared to older participants. This may be a reflection of their higher perceived likelihood of acquiring COVID-19, as demonstrated in our subgroup analysis. A higher perceived likelihood of infection may be explained by higher chances of coming into contact with perceived high-risk groups as mandated by work, social, or study environments. Similarly, male subjects reported a higher likelihood of acquiring the infection and vaccination than their female counterparts. This may have been triggered by late reports from China, Italy, France, Germany, and South Korea indicating higher rates of mortality, as high as 89%, in males compared to females [17]. Infection rates, however, have not widely differed between the sexes [17]. It has been hypothesized that the differences in mortality rate may be due to higher rates of cigarette smoking, alcohol consumption, and the number of pre-existing comorbidities among men compared to women [18]. Lastly, workers or students in the health care sector reported a higher likelihood of acquiring the infection, as well as a higher likelihood of vaccinating, as one might expect. No major differences were noted amongst those with and without underlying illnesses.

One major limitation of this study is its sampling technique. Convenience sampling, typically utilized in cross-sectional studies, is a type of nonprobability sampling that allows for data collection from a group of people who are easy to contact or reach [19]. This may have introduced a sampling bias in our subject cohort. For instance, female and younger subjects are more likely to be represented in social media compared to their male and older counterparts [11], as seen in our study population. Another major limitation is the susceptibility of the study to a nonresponse bias. Our sample also had an unusually high prevalence (22%) of subjects either working or studying in the health care sector. This may be explained by the fact that the survey link was posted on platforms managed by physicians. Such platforms would typically attract users from the same profession. Additionally, the majority of respondents resided in a confined geographical area located within the Arabian Gulf Peninsula. Thus, the data may not be applicable to subjects residing elsewhere. We plan to address such limitations with future studies targeting a wider and more diverse sample of the population.

Our results are indicative of a well-informed cohort, implementing appropriate protective measures, with the majority reporting social media as their main source of COVID-19 information. This demonstrates that social media is an impactful way of collecting data and delivering information, even though there may be inherent limitations to this study design. We plan to use this platform again to determine if such changes are maintained with the continuation of the pandemic.