Primary Menu

Education, Events, Publication

Funding & Recognition

Predictors of Misinformation and Harmful Information on Social Media

Semester: Summer 2023


Presentation description

Background: There is little scientific evidence on identification of features associated with misinformation or harmful cancer treatment information on social media. Here, we evaluate predictors of inaccurate and harmful cancer treatment information on social media, as well as predictors of accurate and non-harmful information.
Method: An observational study examined the most popular cancer treatment articles shared on social media in 2018 and 2019. These English-language cancer treatment articles were selected across four platforms (Facebook, Twitter, Pinterest, and Reddit). For the four most common cancer sites (breast, prostate, colorectal, and lung), 50 articles each were selected for analysis. Two cancer experts reviewed these 200 articles, assessing the primary cancer claim's information accuracy and potential harm on a scale. Logistic regression analysis identified predictive factors associated with misinformation and harmful information, as evaluated by the experts.
Results: Out of the 200 articles analyzed, 103 (51.5%) were accurate, while 48 (24%) were categorized as misinformation. Additionally, 96 (48%) articles were non-harmful, while 61 (30.5%) were flagged as harmful. The multivariable analysis revealed that articles containing claims about herbal or natural remedies were independently associated with misinformation categorization. Attributions to authors in the medical field, including quotes from doctors, and the absence of specific claims were associated with a true categorization by both reviewers. Articles related to prostate cancer and claims concerning herbal and natural remedies were independently associated with a non-harmful categorization. Only the absence of specific claims was independently associated with a non-harmful categorization by both reviewers.
Conclusion: Factors linked to misinformation and potentially harmful cancer information on social media were identified. These associations can aid cancer information consumers to better evaluate sources and guide social media users toward safer and more accurate information.

Presenter Name: Angelica Uzoigwe

Presentation Type: Poster
Presentation Format: In Person
Presentation #115
College: Medicine
School / Department: Oncological Sciences
Research Mentor: Skyler Johnson
Date | Time: Thursday, Aug 3rd | 10:30 AM