Links to online surveys and requests to participate in them are now a common sight in WhatsApp groups. Although earlier such messages used to be apparently for small-scale studies intended for conference presentations, off late, the requests often say that the research is part of some thesis or dissertation. Following the recent outbreak of COVID-19, numerous organizations and researchers have chosen the WhatsApp route to conduct surveys exploring its psychosocial impact. However, such use of WhatsApp groups as a data collection avenue for survey research is fraught with the following problems:
- Surveys help us infer the rates in the population from the responses provided by the sample. For the findings to be generalizable, the sample has to be selected from a clearly defined population through random sampling. However, when you are forwarding the link to as many groups as possible and requesting the members to forward the same to as many other groups as possible, your population will not be clearly defined. Consequently, one cannot know the population to which the survey results would be applicable
- Such samples will be convenience ones, with their inherent problems. For example, they would not be suitable for testing hypotheses using inferential statistics. They can, at best, be used for descriptive purposes and generating hypotheses
- The response rate – the number of people who responded to the survey questions, divided by the total number of people approached to participate – is a crucial measure in survey research. A low response rate would suggest that the sample is not representative of the population. However, in these WhatsApp surveys, one can never know the response rate
- Information about the nonresponders is essential because if it is demonstrated that their sociodemographic profile matches that of the responders, then we can conclude that the responders are indeed representative of the population. This, too, is not possible in WhatsApp surveys, and hence, there is no way to assess the magnitude of selection bias. A large sample does not solve this issue
- It is unethical to study a sample larger than what is suggested by appropriate sample size calculations. However, as the links get forwarded from groups to groups, an unnecessarily large number of people may participate in the survey, with no gains from the time and effort they sacrifice
- Considering the privacy issues with open-source software, collecting sensitive data may not be ethical, even after taking consent, as most participants will not be aware of the implications – i.e., it is no longer an informed consent.
One solution would be to randomly select the sample from a list (e.g., membership directory of an organization or a list of the entire staff of a hospital), ring them up to request consent, collect basic sociodemographic information from those who refuse, and then send the link to only those who consented, with a request to not forward it further. Forwarding of the link to inappropriate groups or persons may be prevented to some extent by mentioning the inclusion and exclusion criteria in the request message.
Two other issues, not restricted to the use of WhatsApp groups, too deserve the consideration of the researchers: (i) As in any survey, sufficient attention has to be paid to the reliability and validity of the questionnaire, so as to reduce the measurement error. (ii) Using a different browser or a second SIM card, the same person may participate in a survey more than once; it may not be possible to detect this with Internet Protocol tracking.
The ethics committees that approve these surveys should include experts on data privacy. If the researchers mention, in the message accompanying the link, about such precautions in place, it may improve the response rate. Furthermore, all steps involved should be accurately described in the manuscript (e.g., what instructions were given, any reminders sent, etc.,) to increase the transparency of the findings.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
1. Eysenbach G. Improving the quality of Web surveys: The checklist for reporting results of internet E-surveys (CHERRIES) J Med Internet Res. 2004;6:e34