R/DoctorsUK (Sampling Bias in Action): How English Junior Doctors Pay Referendum Results Compared to Those Predicted on the Doctors UK Reddit.

Reddit is a popular social media platform that is made up of subreddits, a kind of special interest page. One of these is DoctorsUK, which has over 45,000 members and claims to be a community for UK-based doctors. There is, however, no way of verifying who uses the page, as Reddit is essentially anonymous. Following a period of industrial action in England, a pay deal was put forward and voted on by over 45,000 resident doctors in England. During the voting window, many posts on the subreddit DoctorsUK referenced the pay deal with various comments stating that members were either voting for or against the deal. I wanted to compare the numbers who stated their views on the page to the results obtained in the official ballot. I analyzed the four most commented - on posts relating to the pay deal and read a total of 1,297 comments on these posts. Only a user's first comment on a post was included in the analysis, and deleted comments were not retrieved. This left me with 548 comments to include in the study. I then grouped the comments based on whether they clearly stated their voting intention on the ballot. The voting intention was clearly stated in 230 comments. Of these 230 comments, 60 comments (26.1%) said they would be accepting the deal, and 170 (73.9%) said they would be rejecting it. In the formal ballot, 66% of resident doctors voted to accept the deal. A sentiment analysis was also performed, which found an overwhelming negative sentiment in three of the four posts. A chi-squared test found a P-value of <0.001, suggesting a significant difference between predicted and actual values. While this paper looked at a single question, it found a significant amount of bias in the subreddit. There are few surveys of doctors at such volume that other issues could be assessed in this way. This study should lead to a questioning of the veracity of views on a page that is anonymous and a realisation that we can, not in good faith, assume that all posters on DoctorsUK are UK-based doctors. With this in mind, we may need to question the possible alternative motives for the propagation of such ideas. It is also worth considering how these ideas being shared in a public forum associated with the medical profession may affect large language models and AI, given that Reddit has allowed AI companies access to its data.