Abstract
The most obvious concern is the 16% response rate. With such a poor response rate and with no knowledge of how respondents differed from non-respondents, it is impossible to have confidence in the population estimates for behaviors such as participation in online social networking (OSN). For example, suppose the 84% of the sample who did not respond to the survey are different from the 16% of the people who did respond, perhaps because use of OSNs is correlated with the decision to respond. For practicing physicians, the survey reports that about 40% use OSNs. If the rate of use among the 84% of the sample who did not respond to the survey were really 20%, then the overall population estimate for use among practicing physicians would be 23.2%. This would be about half of the reported 40% and could lead to a totally different perspective on the results.
Second, the data are not weighted to adjust for differences in response rates and the different probabilities of selection within the sample. Development and application of the survey weights are essential to appropriately develop strata level and population estimates. Failing to conduct weighted analyses that adjust for both non-response and survey design effects calls into question all of the population-level estimates presented in this paper.
Third, the study significantly over-represents female physicians. This over-representation is problematic since female gender is also strongly associated with choice of specialty, both personal age and professional age—factors that could be related to participation in OSN and other forms of social behaviors.
Generally, fatal flaws of this nature are easily detectable upon peer-review. We are surprised that such a paper was accepted for publication in a journal of JGIM’s quality. It is our opinion that any basic survey text or any survey professional could have helped the authors to take up this interesting research question and provide data to inform it.
