GHGrace Clarke Hillyer

Grace Clarke Hillyer

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Grace Clarke Hillyer has a total of 236 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Creation of a Tailored Social Media Campaign to Educate Adolescents and Young Adults About Clinical Trials: The Talk to Me Study

    by Grace C. Hillyer, Desiree Walker, Victoria Fedorko, Rhea K. Khurana, Allison Rosen, Maria Frazer, Meital Eisenberger, Corey H. Basch, Edward Bentlyewski, Justine Kahn, et al.

    Purpose: Medical advancements have improved survival of adolescent and young adults (AYAs) for many cancers but not to the same extent as the pediatric and older adult populations. This is at least partly due to this group being systematically underenrolled in cancer clinical trials. To increase AYA knowledge about clinical trials, we developed and tested culturally relevant, developmentally tailored content delivered via social media. Methods: We assembled a multidisciplinary team of AYA cancer care providers and patient advocates to co-develop social media messages to address AYA-reported barriers impacting clinical trial enrollment (CTE). These included concerns about time commitments, costs, toxicities, and general trial information. Messages were strategically disseminated on Instagram, and engagement was evaluated with usage metrics. We also assessed the acceptability and appropriateness of the materials. Results: Over 8 months, we posted 173 messages and acquired 215 followers with 840 profile views; 89.2% of followers were female, and 74.6% of our followers were 18–44 years. Posts reached 5217 unique accounts. The mean monthly follower growth rate was 9.1%, the reach-to-follower ratio was 1.23%, and the engagement rate was 2.53%, supporting strong content visibility beyond our immediate follower base. Conclusion: We demonstrated the feasibility and acceptability of using social media to convey basic information about cancer clinical trials to AYAs. Guided by a multidisciplinary team approach and in close collaboration with patients, we successfully disseminated a social media campaign on Instagram. From this formative work, increasing the reach and uptake and assessing the effect of these messages on CTE decision-making are critical next steps.

  • Open Access

    Promotion of Tanning Behaviors on Instagram: Content Analysis

    by Tailisha Gonzalez, Grace C Hillyer, Bailey Gold, Corey H Basch

    Background: Adolescents and young adults may be particularly susceptible to social media content that promotes tanning and appearance-based beauty standards. Objective: This study examined the characteristics and messaging of pro- and antitanning Instagram posts and the extent to which they promoted health damaging– or health enhancing–behaviors. Methods: A cross-sectional content analysis of 100 Instagram posts using the hashtag #tanning assessed creator characteristics, tanning practices, commercial promotion, tanning-related messaging, beauty standards, and health risks and benefits. χ 2 and Fisher exact tests were used to determine differences between pro vs anti stances for categorical and dichotomous variables, and the Mann-Whitney U test was used for continuous variables. Results: Most videos were protanning (79/100). Protanning videos were significantly greater than antitanning videos to have been posted in 2026 (77.2% vs 47.6%; P =.02), feature teen and young adult creators (87.3% vs 52.4%; P <.001), and include shopping tags promoting tanning products (30.4% vs 0%; P <.001). Overall, messaging focused on promoting tanning (76%) and advocating for tanned skin as a standard of beauty (63%) and downplayed the use of sun protection (11%) and the risks associated with tanning (20%). Messaging varied significantly by tanning stance, with antitanning videos focused on sun protection and health risk. Conclusions: Findings indicate an urgent need for clinicians to lead content creation on popular social media platforms to counter the health-damaging effects of unreliable, incomplete, and potentially misdirected information that is posted under popularized hashtags.

  • Open Access

    Readability and Citation Sources of ChatGPT-Generated Responses to Lymphoma-Related Health Questions: Cross-Sectional Study

    by Corey H Basch, Tailisha González, Grace C Hillyer, Erin T Jacques, Charles E Basch

    Background: Lymphoma is one of the most prevalent groups of cancers among people aged 15 to 29 years. Increasingly, patients with lymphoma and their caregivers use generative AI systems, supported by large language models (LLMs), to seek cancer-related online health information. Despite the need for plain and clear health information for clinical decision-making, there is a dearth of research on the readability and citation source classification of LLM-supported online health information. Objective: The objective of this study was to assess the reading level and reading ease of lymphoma-related health information using OpenAI’s GPT-5.2 model, as well as the citation sourcing of LLM-generated results. Methods: Questions were submitted to the LLM through an automated pipeline. The application code was written with Claude and run and hosted on Vercel. Standardized instructions were submitted to the LLM, asking for evidence-based sources with URLs to enable citation analysis. The final dataset included 392 citations garnered from 96 standardized lymphoma-related questions. Readability was assessed using the Flesch-Kincaid Grade Level and Flesch Reading Ease scores. Citation sources were classified into 9 predefined organization types, and topics included Hodgkin and non-Hodgkin lymphoma and clinical trials. Differences in readability were tested using linear models fitted to all 392 citations, with standard errors clustered by question. Estimated marginal means (EMMs) were used to summarize model-adjusted readability scores by citation source and topic. When the omnibus test was statistically significant, pairwise comparisons were performed using the Tukey-Kramer adjustment for multiple comparisons. Results: The Flesch-Kincaid reading levels of the ChatGPT responses by citation source and topic type were determined to be at the college or university level, well above the recommended sixth-grade level. Grade level did not differ by citation source (P=.58). The reading grade level of responses for Hodgkin lymphoma responses was statistically higher than that of clinical trial responses (mean 16.2, SD 3.4 vs 13.6, SD 2.5; P=.005). Flesch Reading Ease scores across source and topic were found to be very difficult at the university graduate academic level and did not differ by source (P=.27) or topic (P=.17). Conclusions: Although LLMs frequently produce responses that include cited information, the readability of these outputs may limit accessibility for many patients and caregivers. These findings highlight the need for improved readability standards in AI-generated health information and reinforce that LLM outputs should complement, not replace, professional medical guidance.

  • Cancer Clinical Trial Participant Selection: Does Provider Implicit Bias Play a Role?

    by Grace C. Hillyer, Melissa P. Beauchemin, Desiree Walker, Priscila Laforet, Jessica Verp, Hana Amer, Rohit Raghunathan, Edward Bentlyewski, Ana Maria Lopez, John Spivack

  • Portraying Birth in the Digital Age: How YouTube Shapes Perceptions of Waterbirth

    by Carlee Denholtz, Apeksha Mewani, Corey H. Basch, Adi Gutiqi, Grace C. Hillyer

  • Open Access

    Media Framing of Artificial Intelligence in Education: A Content Analysis of US News Coverage

    by Corey H. Basch, Grace C. Hillyer, Victoria H. Wagner, Charles E. Basch

    Purpose: To examine how mainstream US media framed generative artificial intelligence (AI) in education during its early diffusion, assessing whether coverage emphasized problem definition and moral evaluation over pedagogical solutions and institutional innovation. Method: A content analysis was conducted using data from the Dow Jones Factiva database. A total of 2,369 items published between January 2023 and September 2025 were screened, resulting in 542 articles with substantive focus on education after applying exclusion criteria. Coding was performed using Dedoose (Version 10), enabling thematic categorization and iterative comparative analysis. In total, 3,469 coded mentions were analyzed. Findings: Risk-oriented framing was most prominent (21.6%), including themes such as misinformation, hallucinations, privacy, and academic integrity. This was followed by public perception and debate (15.0%) and industry/education technology (13.4%). Coverage consistently highlighted both opportunities (e.g., tutoring, drafting efficiency, productivity gains) and the need for precaution, with greater emphasis on governance and safety than on pedagogy or professional learning. Conclusion: Early media discourse framed AI in education primarily as a policy and risk management issue, with pedagogical transformation remaining secondary. Understanding these framing patterns can support educators and policymakers in balancing precaution with innovation adoption. Originality/Value: This study provides empirical evidence on media framing patterns during a critical phase of generative AI diffusion, contributing to discussions on governance, public perception, and the integration of emerging technologies in education.

  • Open Access

    Readability and Citation Sources of ChatGPT-Generated Responses to Lymphoma-Related Health Questions: Cross-Sectional Study (Preprint)

    by Corey H Basch, Tailisha González, Grace C Hillyer, Erin T Jacques, Charles E Basch

    BACKGROUND Lymphoma is one of the most prevalent groups of cancers among people aged 15 to 29 years. Increasingly, patients with lymphoma and their caregivers use generative AI systems, supported by large language models (LLMs), to seek cancer-related online health information. Despite the need for plain and clear health information for clinical decision-making, there is a dearth of research on the readability and citation source classification of LLM-supported online health information. OBJECTIVE The objective of this study was to assess the reading level and reading ease of lymphoma-related health information using OpenAI’s GPT-5.2 model, as well as the citation sourcing of LLM-generated results. METHODS Questions were submitted to the LLM through an automated pipeline. The application code was written with Claude and run and hosted on Vercel. Standardized instructions were submitted to the LLM, asking for evidence-based sources with URLs to enable citation analysis. The final dataset included 392 citations garnered from 96 standardized lymphoma-related questions. Readability was assessed using the Flesch-Kincaid Grade Level and Flesch Reading Ease scores. Citation sources were classified into 9 predefined organization types, and topics included Hodgkin and non-Hodgkin lymphoma and clinical trials. Differences in readability were tested using linear models fitted to all 392 citations, with standard errors clustered by question. Estimated marginal means (EMMs) were used to summarize model-adjusted readability scores by citation source and topic. When the omnibus test was statistically significant, pairwise comparisons were performed using the Tukey-Kramer adjustment for multiple comparisons. RESULTS The Flesch-Kincaid reading levels of the ChatGPT responses by citation source and topic type were determined to be at the college or university level, well above the recommended sixth-grade level. Grade level did not differ by citation source (P=.58). The reading grade level of responses for Hodgkin lymphoma responses was statistically higher than that of clinical trial responses (mean 16.2, SD 3.4 vs 13.6, SD 2.5; P=.005). Flesch Reading Ease scores across source and topic were found to be very difficult at the university graduate academic level and did not differ by source (P=.27) or topic (P=.17). CONCLUSIONS Although LLMs frequently produce responses that include cited information, the readability of these outputs may limit accessibility for many patients and caregivers. These findings highlight the need for improved readability standards in AI-generated health information and reinforce that LLM outputs should complement, not replace, professional medical guidance. CLINICALTRIAL

  • Analysis of Google News coverage during the 2025 US measles outbreak

    by Corey H. Basch, Erela Datuowei, Apeksha Mewani, Vincent Jones, Grace C. Hillyer


  • Open Access

    Implicit provider bias in cancer clinical trial enrollment: A scoping review

    by Melissa Beauchemin, Jessica Verp, Priscila Laforet, Samrawit Solomon, Desiree Walker, Dawn L. Hershman, Edward Bentlyewski, Ni-Cheng Liang, Grace C. Hillyer

    Background: The lack of racial and ethnic representation in cancer clinical trials further exacerbates existing health disparities and has been linked to inequitable and unequal distribution of resources, social determinants of health, racism, and most recently, provider implicit bias. The purpose of this scoping review is to describe the evidence of implicit bias as it relates to cancer clinical trial enrollment. Methods: Systematic methods were followed to search online databases for publications that assessed provider bias and cancer clinical trial offers through August 1, 2023. Variables of interest included how implicit bias was defined or measured, what type of bias was examined, and if provider implicit bias was identified or described in the cancer clinical trial enrollment process. Results: We identified 10 publications that met the inclusion criteria. Nine of ten studies were conducted in the United States, and all studies utilized observational or qualitative study design. Most assessed the provider perspective, and one study used a validated measure to identify implicit bias. Some studies identified evidence of implicit provider bias in clinical trial discussions and offers; however, it was observed in situations where system constraints, such as lack of resources or time were present that may be common in oncological settings. Conclusions: Multilevel support and dedication to improving racial, ethnic, and other patient-level representativeness are required to mitigate the influence of implicit bias in resource-poor and stressful clinical settings to ensure equitable access to and enrollment in clinical trials for all patients with cancer. Registration: The protocol for this scoping review was registered to Open Science Framework (1), https://osf.io/a6kbt/?view_only=0a857280364a49b9b3f8363b91b00994 .


  • Open Access

    Artificial Intelligence in Higher Education: Student Knowledge, Attitudes, and Ethical Perceptions in the United States

    by Corey Basch, Grace Hillyer, Bailey Gold, Helen Yousaf

    Purpose: This study examines undergraduate college students’ knowledge, attitudes, and behaviors regarding artificial intelligence (AI), with a particular focus on its use in classroom, coursework, and personal applications. The objective is to provide empirical insights into how students perceive the benefits, risks, and ethical implications of AI adoption in higher education. Methodology: A cross-sectional survey was conducted among 319 students enrolled in a personal wellness elective course in the United States, of whom 258 (81%) completed the questionnaire. Data collection focused on familiarity with AI, frequency and context of use, perceived benefits, ethical concerns, and trust in AI systems. Statistical analyses included chi-square tests, Fisher’s exact tests, and ANOVA to examine associations between frequency of AI use and student perceptions. Findings: More than half of the students reported familiarity with AI, most frequently learning about it through the internet. Frequent AI users expressed higher levels of trust in AI, were more likely to believe that AI makes learning easier, and associated it with improvements in writing, critical thinking, and communication skills. In contrast, infrequent users were more likely to perceive AI use in class as cheating, express distrust in AI safety, and emphasize the need for stricter regulation. Overall, results highlight a “digital divide” between frequent and infrequent users, with implications for teaching practices and institutional policies. Originality/Contribution: This study contributes to the emerging literature on AI in higher education by offering one of the few empirical analyses of U.S. college students’ perspectives. It advances understanding of the ethical and pedagogical challenges posed by AI integration and provides evidence that can inform the design of institutional policies, curricula, and training programs. Practical Implications: The findings underscore the urgent need for universities to develop clear policies and training programs on the responsible use of AI. Institutions should promote AI literacy among both students and faculty to address ethical concerns, reduce academic integrity risks, and prepare students for future workforce demands shaped by technological innovation. Limitations: The study is limited by its cross-sectional design, single-institution sampling frame, and reliance on self-reported data. Future research should expand to multiple universities, adopt longitudinal designs, and explore the perspectives of both students and faculty in greater depth. Alignment with the SDGs: This research contributes to the advancement of the United Nations’ Sustainable Development Goals, particularly SDG 4 (Quality Education), by addressing the role of AI in shaping educational practices, and SDG 9 (Industry, Innovation, and Infrastructure), by analyzing how emerging technologies influence higher education. It also supports SDG 16 (Peace, Justice, and Strong Institutions) by engaging with ethical and regulatory considerations surrounding AI governance in educational contexts.