University of Maryland: Gender Bias Plagues Online Doctor Reviews, Study Finds
Online reviews are a go-to source for many people looking for a new doctor, but how can health care consumers judge the quality of the reviews themselves—and particularly, whether reviewer bias might be slanting the results?
While previous studies have confirmed that men have a ratings edge, new research including University of Maryland faculty and alums, is the first to set an objective bar: starting with doctors who’d all been sanctioned for negligence, malpractice or other improper behavior by state medical boards, the study analyzed numerical ratings and language used in doctor reviews to demonstrate that women are systematically assessed differently online than men. The team found that female doctors who were sanctioned get worse ratings and have more negative language in their reviews than male doctors who face being similarly penalized.
“Online ratings are biased against females, not just for health care, across service industries,” said Margrét Bjarnadóttir, an associate professor of management science and statistics and a co-author of the study published recently in JMIR Formative Research. “What we were interested in was to be able to demonstrate that these biases hold, even when controlling for quality. It highlights the need to create a better structure around online reviews to make them more neutral and unbiased.”
Her co-authors included two Robert H. Smith School of Business alums, Villanova University Assistant Professor David Anderson Ph.D. ’13 and Chong Chen M.S. ’19, now at Amazon.
The team used natural language processing, a subfield of machine learning, to analyze reviews to find which words are associated with which type of physician. They found that the words most associated with sanctioned female doctors have much more negative connotations than those applied to men, including the terms “unprofessional,” “cold” and “dismiss.”
But in web-based reviews of male doctors, the team found that it’s hard to tell the difference between those who have and haven’t been disciplined. Words like “save,” “compassion” and “superb” are exclusively associated with reviews of male-sanctioned physicians; they never appear in a sanctioned female doctor’s review.
This begs the question: Are men or women more biased against female doctors who are being penalized? Bjarnadóttir said that while the reviews used in the research are anonymous, “we did however look specifically at ob-gyn doctors, where we can argue that most of the reviewers must be female, and the same biases carried through, even when everybody rating was female.”
The study concludes that websites and apps featuring physician reviews should follow best practices for reducing gender and racial bias. For instance, as previous research has shown, asking specific questions rather than providing open-ended boxes for reviews can cut down on bias.
Bjarnadóttir said, “Because these platforms and ratings matter in patient choice,” this research “is a call on platforms to redesign their evaluations in the online reviews to make sure they are as neutral and objective as possible.”