In an era where digital communication dominates healthcare, the clarity and accessibility of patient education materials (PEMs) are more vital than ever. A recent landmark study conducted at NYU Langone Health reveals how artificial intelligence, particularly large language models (LLMs), can dramatically enhance the readability of these crucial resources. The research addresses a perennial challenge in medical communication: the complexity of information that often surpasses the recommended sixth-grade reading level, rendering it less effective for broad patient populations.

The study meticulously analyzed PEMs sourced from the websites of three leading American health organizations—the American Heart Association (AHA), American Cancer Society (ACS), and American Stroke Association (ASA). These organizations produce patient-directed content designed to inform decision-making and facilitate better health outcomes. Nevertheless, despite their patient-focused intent, the original materials scored an average readability grade level between 9.6 and 10.7, substantially higher than the ideal grade 6 threshold suggested by health literacy experts.

To overcome this barrier, researchers employed three state-of-the-art generative AI models: ChatGPT, Gemini, and Claude. These models operate by leveraging extensive textual datasets from the Internet to predict and generate the next most probable word in a sequence, enabling them to rephrase text in simpler, more digestible terms while maintaining factual accuracy. The application of such LLMs represents a cutting-edge intersection between natural language processing and clinical communication enhancement.

The methodology involved selecting 60 PEMs at random from the specified organizations’ websites. Each text was then fed into the three different LLMs, with prompts instructing the models to reduce the reading complexity to meet or approximate the sixth-grade level. The output was carefully evaluated using established readability formulas to ensure that simplification did not compromise meaning or introduce inaccuracies.

Findings from the study were striking. The three AI tools succeeded in lowering the reading grade levels considerably: ChatGPT brought the average level down to 7.6, Gemini achieved 6.6, and Claude surpassed expectations by reaching an average grade level of 5.6. Moreover, these revisions yielded a noticeable reduction in word counts, enhancing conciseness without sacrificing content quality. This compression translates into easier-to-navigate materials that can better sustain patient attention and comprehension.

Dr. Jonah Feldman, the study’s senior author and medical director of transformation and informatics at NYU Langone, emphasized the transformative potential of AI in healthcare communication. He noted, “Our study shows that widely used large language models have the potential to transform patient education materials into more readable content, which is essential for patient empowerment and better health outcomes.” Feldman further highlighted that even expertly crafted educational resources benefit significantly from AI-based optimization.

The implications of this research extend beyond text simplification. It signals a paradigm shift where healthcare organizations can integrate AI technologies into their communication strategies to bridge the literacy gap among patients. This innovation aligns with broader efforts to promote health equity by ensuring that patients, regardless of educational background, have access to comprehensible information necessary for informed decisions.

Previous studies have documented AI’s utility in generating patient-focused explanations of complex medical data, responding to electronic health queries, and summarizing intricate clinical reports. Building on this foundation, the current study adds empirical evidence supporting the practical application of LLMs for refining patient educational content specifically. The technology’s adaptability and scalability make it a promising candidate for widespread adoption across healthcare systems.

Dr. Paul Testa, chief health informatics officer at NYU Langone and co-author of the study, reflected on the burgeoning role of AI in healthcare. “The breadth of possible AI offerings shows how technology can be leveraged to transform the patient experience across health care systems, and not just in the United States,” he pointed out, underscoring the global relevance of this innovation. Testa also revealed that these AI tools are not merely theoretical; NYU Langone is actively deploying them in clinical trials to assess their impact on patient comprehension post-discharge.

Specifically, the ongoing randomized controlled trial incorporates AI-generated, patient-friendly summaries of hospital discharge instructions. The goal is to evaluate whether such summaries improve patient understanding and satisfaction, ultimately facilitating smoother transitions from hospital to home care. By generating real-world evidence, the team aims to validate the clinical effectiveness and safety of AI-enhanced communication within dynamic healthcare environments.

Dr. Jonah Zaretsky, associate chief of medicine at NYU Langone Hospital—Brooklyn, highlighted the significance of rigorous testing under clinical conditions. “Generating real-world evidence through randomized trials is crucial for validating the effectiveness of AI tools in clinical settings,” he explained. Zaretsky stressed that such research ensures that AI-powered documentation truly serves patients and families without compromising accuracy or safety.

Notably, this important study was self-funded by NYU Langone and involved a dedicated team of researchers including lead author John Will, and co-authors Mahin Gupta and Aliesha Dowlath, alongside Feldman, Testa, and Zaretsky. Their collaborative efforts exemplify the commitment within academic medicine to harness innovative technologies for meaningful improvements in patient care.

As healthcare increasingly embraces digital transformation, the application of large language models to improve the readability and usability of patient education documents marks a significant milestone. It demonstrates how artificial intelligence can serve as a pivotal tool for health literacy, empowering patients with clearer, more concise, and accessible information. Such advancements not only foster better patient engagement but are poised to enhance overall health outcomes by closing the comprehension gap that has long hindered effective communication.

In a world inundated with health information, simplifying and tailoring content to patient needs is paramount. This pioneering work by NYU Langone offers a glimpse into a future where AI-driven solutions are seamlessly integrated into healthcare communication, revolutionizing the way medical knowledge is shared and understood across diverse populations.

Subject of Research:
Artificial intelligence application in patient education for improved readability.

Article Title:
Leveraging Large Language Models to Improve Readability of Online Patient Education Materials: Cross-sectional Study

News Publication Date:
April 10, 2024

Web References:
http://dx.doi.org/10.2196/69955

References:
Published in Journal of Medical Internet Research

Keywords:
Machine learning, Computer science, Patient education, Health literacy, Artificial intelligence, Large language models, Natural language processing, Medical informatics, Readability optimization