Abstract
PURPOSE: Breast cancer remains a global public health burden. This study aimed to evaluate the readability of breast cancer articles shared on X (formerly Twitter) during Breast Cancer Awareness Month (October 2024), and it explores the possibility of using artificial intelligence (AI) to improve readability. METHODS: We identified the top articles (n = 377) from posts containing #breastcancer on X during October 2024. Each article was analyzed using 9 established readability tests: Automated Readability Index (ARI), Coleman-Liau, Flesch-Kincaid, Flesch Reading Ease, FORCAST Readability Formula, Fry Graph, Gunning Fog Index, Raygor Readability Estimate, and Simple Measure of Gobbledygook (SMOG) Readability Formula. The study categorized sharing entities into five groups: academic medical centers, healthcare providers, government institutions, scientific journals, and all others. This comprehensive approach aimed to evaluate the readability of breast cancer articles across various sources during a critical awareness period of peak public engagement. A pilot study was simultaneously conducted using AI to improve readability. Statistical analysis was performed using SPSS. RESULTS: A total of 377 articles shared by the following entities were analyzed: academic medical centers (35, 9.3%), healthcare providers (57, 15.2%), government institutions (21, 5.6%), scientific journals (63, 16.8%), and all others (199, 53.1%). Government institutions shared articles with the lowest average readability grade level (12.71 ± 0.79). Scientific journals (16.57 ± 0.09), healthcare providers (15.49 ± 0.32), all others (14.89 ± 0.17), and academic medical centers (13.56 ± 0.39) had higher average readability grade levels. Article types were also split into different categories: patient education (222, 58.9%), open-access journal (119, 31.5%), and full journal (37, 9.6%). Patient education articles (15.19 ± 0.17) had the lowest average readability grade level. Open-access and full journals had an average readability grade level of 16.65 ± 0.06 and 16.53 ± 0.10, respectively. The mean values for Flesch Reading Ease Score are patient education 38.14 ± 1.2, open-access journals 16.14 ± 0.89, full journals 17.69 ± 2.14. Of note, lower readability grade levels indicate easier-to-read text, while higher Flesch Reading Ease scores indicate more ease of reading. In a demonstration using AI to improve readability grade level of 5 sample articles, AI successfully lowered the average readability grade level from 12.58 ± 0.83 to 6.56 ± 0.28 (p < 0.001). CONCLUSIONS: Our findings highlight a critical gap between the recommended and actual readability levels of breast cancer information shared on a popular social media platform. While some institutions are producing more accessible content, there is a pressing need for standardization and improvement across all sources. To address this issue, sources may consider leveraging AI technology as a potential tool for creating patient resources with appropriate readability levels.