Health Education in the Era of AI: Curating
Patient-Centric Learning Material with LLMs
Executive Summary
- 🌟 Vision: We strive to transform healthcare education by harnessing Large Language Models (LLMs). Our vision is to ensure authenticity, break barriers, and empower global communities with accessible, reliable health information in their native languages for informed decision-making.
- 🚀 Mission: Our mission revolves around crafting accurate patient education content using refined LLMs. We prioritize linguistic accuracy, medical reliability, and global accessibility, aiming to empower informed health decisions, break language barriers, and foster universal health equity.
- 🤝 The Driving Force: The Institute of Health Innovation and Education, a registered non-profit organization in the Texas, United States.
- ⏰ Future Plans: We are ready to start as soon as funding is received.
- 📍 Rollout Strategy: We will disseminate the health education material at a global level through digital platforms.
Why Fund?
There is an absolute need for advancing health education globally. The era of AI will produce more misinformation and disinformation. IHIE, in collaboration with NeuroCare.AI, has both the data and human resources to accomplish this goal.
This project will impact 5 billion people globally that is less than $0.00008 per person for one year!
Project Outcomes
- LLM Fine-Tuning for Reliable Content: The project’s focus on fine-tuning LLMs ensures the production of highly dependable health education materials. This meticulous refinement guarantees content accuracy, fostering trust in the information provided.
- Real-Time Accuracy Checks by Expert Physicians: Our process integrates real-time evaluations by expert physicians, ensuring content precision and relevance. This hands-on review system enhances the material’s quality, aligning it with established medical standards.
- Multilingual Translation for Universal Accessibility: Through comprehensive multilingual translations, we ensure health education materials are easily understandable across diverse linguistic backgrounds. This inclusivity breaks language barriers, empowering global communities with vital health insights.
Sponsorship Challenge: Patient Education in the Age of Generative AI
The advent of Generative AI and Large Language Models (LLMs) has introduced a new paradigm in accessing information on the internet. This technology has the potential to revolutionize patient education and healthcare information dissemination, but it also comes with its own set of challenges that must be addressed.
- Lack of Authenticated Source of Medical Information on the Internet: The internet is saturated with medical information, but much of it is unverified, unreliable, and sometimes even misleading. Patients often struggle to distinguish between credible and dubious sources, leading to incorrect self-diagnoses and potentially harmful treatment choices (Yaacoub JPA et al, 2020).
- Generative AI is Trained on a Massive Dataset, Lacking Authenticity: Generative AI (LLMs) learns from the entirety of the internet, which includes both accurate and misleading information. This wide-ranging dataset can result in generated content that is inconsistent in quality and accuracy, posing a significant challenge to reliable patient education (De Varies K, 2020).
- Helpfulness vs. Harmlessness: While Generative AI can generate information, it may inadvertently produce content that is not only inaccurate but also potentially harmful. Patients who rely on AI-generated content may receive misguided advice that could have serious consequences for their health (Spallek S et al., 2023; Tan X et al., 2023).
- Authenticated Material is Behind Paywall: Many authenticated medical resources are often locked behind paywalls, making them inaccessible to the general public. This lack of free, accessible, and trustworthy medical information leaves patients with limited options for educating themselves (Day S et al., 2020).
- Authenticated Material is Technical and Hard to Understand: Authenticated medical sources often use complex language and terminology, making it challenging for the average person to understand. This language barrier can hinder effective patient education (Kaufman DM, 2018).
Why Prioritize this Challenge?
- The Digital Age is Expanding: The world is rapidly transitioning to a digital age, with a significant increase in mobile internet use. As more people rely on the internet for information, it is essential to ensure that they have access to accurate and reliable medical information.
- LLMs Will Take Over as the Primary Source of Information: Large Language Models are becoming an integral part of information retrieval and generation. With their increasing influence, there is a growing need to ensure that these models provide information that is both accurate and safe.
- The Urgent Need for an Authenticated Source of Information: With the proliferation of LLMs, there is a pressing need to establish authenticated sources of medical information that can be used to train and fine-tune these models. This would help ensure that the information generated by these models is accurate, up-to-date, and reliable.
Language Barriers in Health Equity
While English has served as the “universal language of science,” enabling unimpeded idea exchange and facilitating access to worldwide scientific literature, its predominant use has presented obstacles for non-native English speakers (Marquez MC et al., 2020).
Despite English being the native language of just 7.3% of the global population and only a fraction, less than 20%, being proficient in it, almost 75% of all medical publications are in English. Approximately, 17% of the world’s population is capable of conversing in English, mostly as a secondary language (Figure 1a). Consequently, the prevalence of English in the global medical community results in the exclusion of 80% of the world’s inhabitants from engaging with and contributing to the majority of the world’s published scholarly works (Bahji A et al., 2023) (Figure 1b).
These statistics demonstrate a substantial discrepancy between language accessibility and the dissemination of global scientific knowledge.
Figure 1: (A) – English language native and proficient population in the world versus (B) – Percentage of total scientific literature available in English language
Proposed Solution: Leveraging LLMs for Authentic Patient Education Material
Our proposed solution aims to revolutionize patient education by harnessing the power of advanced LLMs, to create an authentic, reliable, and accessible source of patient education material. Here’s how we plan to implement this solution:
- Creating Authentic Patient Education Material Using LLMs: Leveraging state-of-the-art natural language processing capabilities, LLMs will be instrumental in generating patient education material. This technology assists in creating highly informative, readable, and comprehensible content tailored to the average patient’s needs. Employing LLMs ensures linguistic accuracy and audience relevance in the produced content.
- Ensuring Accuracy and Reliability: Acknowledging the importance of medical accuracy in patient education, our content generated through LLMs will undergo a rigorous review by medical professionals and subject matter experts. This review ensures adherence to clinical guidelines and verifies medical accuracy, maintaining the highest standards to foster trust among patients and healthcare providers.
- Fine-Tuning LLMs for Patient Education: To refine the precision of patient education content, we will conduct a tailored fine-tuning process for the selected LLMs. This involves training the model on a verified medical content corpus and patient education resources. Fine-tuning LLMs ensures alignment with the latest medical advancements and guidelines. Upon completion, we aim to open source this fine-tuned LLM version, enabling wider accessibility and empowering diverse entities to create accurate patient education materials.
- Multilingual Accessibility Using SeamlessM4T: Access to healthcare information should be universal, transcending language barriers. To achieve this, we will employ advanced machine translation technology: SeamlessM4T. This tool will facilitate the translation of our patient education content into different languages, ensuring that individuals worldwide can benefit from the information. SeamlessM4T excels in accurately translating complex medical terminology, maintaining the original content’s accuracy and integrity in multiple languages. This approach guarantees that our patient education material is accessible and beneficial to diverse global populations.
Project Impact
The impact of our project can be summarized in the following points:
Promoting Health Equity: This project’s multilingual health education materials cater to diverse audiences, promoting health equity and breaking down language barriers in healthcare. By providing information in multiple languages, we will ensure equitable access to essential health knowledge for patients and caregivers from varied linguistic backgrounds.
Overcoming language barriers fosters inclusive communication, trust, and empowers individuals to make informed healthcare decisions. This inclusive approach transforms healthcare interactions, fostering a more equitable and accessible healthcare environment for all.
Enhanced Understanding: By providing valuable information, our project enhances the understanding of patients and caregivers regarding various disorders. This increased knowledge empowers them to make informed decisions and better manage the challenges they may encounter.
Improved Management: Equipped with relevant and reliable information, patients and caregivers are better able to manage disease conditions. They gain the necessary tools and understanding to navigate their healthcare journey effectively, leading to improved outcomes and quality of life.
Expertise and Reliability: The involvement of experienced faculty members ensures the highest quality of content. Their expertise and affiliation with our institution guarantee that the course material is reliable, up-to-date, and tailored to meet the specific needs of our target audience.
Wide Accessibility: By utilizing online platforms and social media channels, we ensure easy accessibility and convenience for patients and caregivers. This broad dissemination strategy allows us to reach a wider audience, including individuals who may have limited access to traditional educational resources.
Empowerment: Our project aims to empower patients and caregivers by equipping them with the knowledge and understanding necessary to navigate disease conditions more effectively. This empowerment enables them to actively participate in their care, make informed decisions, and advocate for themselves or their loved ones.
The impact of our project extends beyond the immediate educational resources. By empowering patients and caregivers with knowledge and understanding, we contribute to their long-term well-being, enabling them to make informed choices, advocate for their needs, and lead healthier lives (Figure 1).
Figure 2: Impact and benefits of spreading health education among patients and caregivers
References
Bahji, A., Acion, L., Laslett, A. M., & Adinoff, B. (2023). Exclusion of the non-English-speaking world from the scientific literature: Recommendations for change for addiction journals and publishers. Nordic Studies on Alcohol and Drugs, 40(1), 6-13.
Day, S., Rennie, S., Luo, D., & Tucker, J. D. (2020). Open to the public: paywalls and the public rationale for open access medical research publishing. Research involvement and engagement, 6(1), 1-7.
De Vries, K. (2020). You never fake alone. Creative AI in action. Information, Communication & Society, 23(14), 2110-2127.
Kaufman, D. M. (2018). Teaching and learning in medical education: how theory can inform practice. Understanding medical education: evidence, theory, and practice, 37-69.
Márquez, M. C., & Porras, A. M. (2020). Science communication in multiple languages is critical to its effectiveness. Frontiers in Communication, 31.
Spallek, S., Birrell, L., Kershaw, S., Devine, E. K., & Thornton, L. (2023). Can we use ChatGPT for Mental Health and Substance Use Education? Examining Its Quality and Potential Harms. JMIR Medical Education, 9(1), e51243.
Tan, X., Shi, S., Qiu, X., Qu, C., Qi, Z., Xu, Y., & Qi, Y. (2023, December). Self-Criticism: Aligning Large Language Models with their Understanding of Helpfulness, Honesty, and Harmlessness. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track (pp. 650-662).
Yaacoub, J. P. A., Noura, M., Noura, H. N., Salman, O., Yaacoub, E., Couturier, R., & Chehab, A. (2020). Securing internet of medical things systems: Limitations, issues and recommendations. Future Generation Computer Systems, 105, 581-606.
The need for modern care and education system has never been greater. Sign up for our newsletter.