Center for the Future of the Health Professions April 2024 digest
Posted: April 17, 2024
This month, The Center for the Future of the Health Professions will publish another monthly op-ed column for 2024. Our columns offer strong, well-informed, and focused opinions on issues impacting the future of health professions. The center was established to provide policymakers at the state, local, and national levels, as well as stakeholders in the health system, with accurate, reliable, and comprehensive data and research on the healthcare workforce. This is crucial for effective planning for a sustainable future and optimal utilization of available resources.
This month’s column will focus on artificial intelligence (AI), which is rapidly transforming healthcare delivery and medical education. The integration of AI technologies in healthcare has led to significant advancements in diagnostics, personalized treatment approaches, and patient care coordination in recent years. Furthermore, AI presents cutting-edge opportunities in health professional education, offering tailored tutoring platforms and immersive simulations to enhance and personalize learning experiences.
This article’s author, Ted Wendel, PhD, currently serves as the senior vice president for university planning at A.T. Still University (ATSU). Dr. Wendel has a rich academic background, with experience spanning various institutions and leadership roles in education and research. His contributions to neuropharmacology and hypertension research, as well as his innovative approaches to educating health professionals, have been significant.
In addition to his academic pursuits, Dr. Wendel has been actively involved in humanitarian efforts as a longtime volunteer, serving as a photojournalist for Project HOPE, an international healthcare organization. His dedication to capturing the stories of humanitarian aid initiatives reflects his commitment to making a difference in global health.
We hope you find this article informative and engaging, and we welcome your feedback and comments.
Randy Danielsen, PhD, DHL(h), PA-C Emeritus, DFAAPA
Professor and Director
The Center for the Future of the Health Professions
A.T. Still University
Artificial intelligence in the education of healthcare professionals
Artificial intelligence (AI) is rapidly changing the landscape of healthcare delivery and medical education. In just the last few years, integrating AI technologies into healthcare has propelled advancements in diagnostics, personalized treatment plans, and patient care management. AI offers innovative solutions to adapt and optimize learning experiences within health professions education, from customized tutoring systems to immersive simulations.
This paper explores AI in healthcare and its implications for medical education. Today’s graduates will face future medical practice that will take advantage of AI and be challenged by its flaws. By examining the current state of AI in healthcare and educational settings, this paper aims to elucidate the opportunities, challenges, and ethical considerations surrounding its implementation. This paper seeks to provide a simple understanding of the AI landscape and explain key trends, successes, and areas for further exploration.
To appreciate AI’s future, it is essential to have some understanding of the technology and recent achievements that unlocked potential changes that stagger the imagination. In its simplest form AI strives to create machines that mimic the thinking of humans. The constant evolution of microchip technology over the last 20 years has made it possible for computers to ingest and analyze data at astounding rates. Computer scientists have taken advantage of a computer’s ability to ingest data and establish relationships between and among huge amounts of data organized into datasets.
AI tools such as ChatGPT are based on supercomputers scanning the content of billions of documents posted on the worldwide web and establishing statistical algorithms that define the relationships of the text in the documents. This process, known as training, is used to create a large language model (LLM). These LLMs permit individuals to have seemingly natural conversations with computers that understand the input and respond based on the relationships discovered as part of the training. Computers can be and have been trained on almost any type of data. Data can be derived from words in a document to the arrangement of pixels in and image or the reported speed of a car in rush hour traffic.
Medical research has always been based on collecting and comparing datasets, such as a treated dataset versus a similar dataset of untreated individuals. This comparison type is supervised since a researcher has labeled each patient outcome as either treated or untreated. However, imagine a scenario like an LLM where the data is not labeled and computers discover the statistical relationships existing in a large dataset of patient medical records or millions of images of breast scans. With enough data, and over time, the statistical algorithms become increasingly accurate at defining what patterns may be associated with what pathologies or clinical outcomes in much the same way a healthcare provider becomes an experienced professional.
The rapid evolution of AI in healthcare presents a dynamic and transformative landscape, with innovations emerging regularly. Health professionals should be aware of key trends as illustrated in the following applications of AI that reflect the current and future potential of AI in healthcare. Understanding these trends will help professionals harness AI’s capabilities effectively and ethically. Health professional educators must understand AI to prepare graduates for practice in a world dramatically different from today.
Chatbots
The evolution of LLMs used as chatbots like ChatGPT (as well as other LLMs like Gemini, Perplexity, and others) and the ability to interact using a natural language interface is the most dramatic technological advance in decades. These tools present a transformative potential. By automating administrative tasks such as clinical paperwork, billing, and recordkeeping, LLMs can significantly reduce the burden on healthcare professionals, allowing them to focus more on patient care rather than administrative tasks.
Moreover, the ability of these models to serve as an additional layer of analysis and insight can aid in diagnosis and treatment, acting as a support system for medical professionals. This can lead to more accurate and timely interventions, potentially improving patient outcomes.
The potential for these models to monitor patient compliance and predict clinical interventions is fascinating. By analyzing patient data, AI can identify patterns and may be capable of predicting health events before they occur, enabling proactive rather than reactive healthcare. This could lead to personalized and timely interventions, reducing hospital readmissions and improving overall patient health.
The latest version of Chat GPT released in April 2023 shows significant improvement over the previous version. The advancements in ChatGPT 4.0 encompass significant enhancements in linguistic understanding, context retention, information integration, reasoning, multilingual support, bias mitigation, customizability, interdisciplinary application, and ethical considerations. These improvements underscore the model’s evolution toward more nuanced semantic understanding, comprehensive knowledge representation, and sophisticated problem-solving abilities. Also, these improvements highlight the focus on language equity, adaptability for specific population needs, and the importance of addressing ethical concerns and biases in AI deployment. Chatbots’ sophistication represent technological progress and opens new avenues for academic exploration across computational linguistics, AI ethics, continual learning, and cross-disciplinary applications, emphasizing the broader impact of advanced AI systems in healthcare and society.
Diagnostic imaging
AI has made significant strides in diagnostic imaging, revolutionizing how images are acquired, interpreted, and used for patient care. Automated image analysis of scans such as CT, MRI, X-rays, and ultrasound have improved both the accuracy and speed of diagnosis in several areas such as breast cancer screening1, bone fracture analysis2, and lung disease detection.3
Beyond diagnosis, AI in imaging can predict patient outcomes by analyzing image patterns that correlate with prognosis, helping clinicians make informed decisions about treatment strategies. AI tools can streamline radiology workflows by prioritizing urgent cases, automating routine tasks, and facilitating image triage, ensuring patients requiring immediate attention are identified and treated promptly.
An outstanding example of AI’s impact on diagnostic imaging is EyeArt by Eyenuk, the first FDA-cleared AI technology for autonomous detection of diabetic retinopathy, a condition causing vision loss and blindness that affects millions of Americans annually. EyeArt enables in-clinic, real-time diabetic retinopathy screening for primary care practices, diabetes centers, and optometric offices. This tool provides an inexpensive, rapid, and accurate diagnosis allowing healthcare practitioners to identify referable diabetic retinopathy patients quickly and accurately during a diabetic patient’s regular exam. This AI tool has proven its value in early diagnosis of possible vision loss at a remarkably reduced cost.
Tools like EyeArt illustrate the breadth of AI’s impact on diagnostic imaging, enhancing accuracy, efficiency, and patient care. The field continues to evolve, with ongoing research and development aimed at expanding the capabilities and applications of AI in medical imaging.
Knowledge acquisition
For decades, knowledge has been widely available through the internet using browsers to search for information. Internet searches deliver a list of links to various websites. There is little curation of the search links, and the searcher must scour them to glean the desired information.
SciSpace is an academic search engine that facilitates access to scientific literature. While it primarily serves researchers and academics across various fields, healthcare professionals can significantly benefit from its features and capabilities.
SciSpace is an innovative platform aiming to streamline the process of accessing academic papers, journals, and scholarly articles. It leverages advanced search algorithms to help users find relevant literature quickly and efficiently. The platform offers a user-friendly interface that simplifies the search for high-quality, peer-reviewed scientific information.
AI tools like SciSpace offer healthcare educators access to various scientific articles, including those from medical and healthcare journals. The time saved by using this AI tool cannot be overstated. Anyone who has spent hours searching the literature on a specific topic will be astounded by the time saved using an application that provides immediate access to peer-reviewed, credible literature.
SciSpace and a growing number of similar applications serve as a vital tool for healthcare professionals, enhancing their ability to access scientific literature, stay informed about the latest research, and apply this knowledge to improve patient care, drive research, and contribute to advancing the medical field.
Use with caution
AI tools like those described in this paper must be used and applied with great caution. As with any new and rapidly developing technology, AI tools cannot be relied on for their accuracy. LLMs have been known to hallucinate and provide false or misleading information and then promote and justify these errors. All systems make errors and using tools like ChatGPT demands policies and procedures that check and confirm results. Chatbots are not now, nor will they be in the immediate future, a primary or exclusive source of medical knowledge or care.
AI tools are subject to a training bias. The results provided by these tools are dependent on the populations used for their training. Should the training population be skewed by factors such as age, gender, or race, they will propagate knowledge biased by the distortion of the training population. This echoes the process of drug discovery and testing during the 20th century. Most drugs were tested and evaluated on a population of middle-aged white males, essentially ignoring the critical gender and race differences in drug actions and adverse effects. So, training of AI tools may reflect skewed populations resulting in responses that lack diversity
Moving forward
As healthcare educators move forward it is imperative to establish clear guidelines for using AI tools in the education of healthcare professionals. A July 2023 report from Cornell University’s Center for Teaching and Learning encourages faculty to explicitly set expectations for when and how students can employ generative AI in their work with proper attribution. Future policies and guidelines must be tempered with the dichotomy that recognizes the value of AI tools in practice while still promoting student learning. Ohio University’s Center for Teaching and Learning offers example AI policies and assignments from faculty members at the University to provide a reference point.
In the forthcoming decade, the trajectory of AI tools is anticipated to undergo rapid evolution, a progression aptly represented by the S-Curve model – a graphical illustration traditionally employed to depict the dose-response relationship of pharmaceuticals. Analogous to medications, AI technologies must be subjected to rigorous regulatory frameworks akin to those applied to the discovery and monitoring of new drugs. Such regulation must ascertain the safety and efficacy of AI tools, safeguarding the public from adverse consequences and misuse. Furthermore, regulatory measures must be instituted to ensure the economic implications of AI technologies do not preclude access for socio-economically disadvantaged populations globally.
AI applications are poised to fundamentally transform the domain of healthcare and therefore the education of health professionals. These advancements promise to enhance access to medical services, diminish costs, and alleviate the administrative burdens associated with scheduling, recordkeeping, and billing, among other tasks. As these tools evolve, they are expected to achieve greater accuracy and user-friendliness. Presently, AI technologies serve as an invaluable complement to healthcare practice – a potent assistant capable of swiftly navigating and structuring vast repositories of knowledge.
Author’s note: The author created the original outline and drafts of this paper and used two AI tools. The OpenAI Chatbot V 4.0 was used to clarify some content and improve the readability of several paragraphs. The SciSpace tool noted in the paper was used to identify published resources noted in the publication. These tools were used in support of this paper between November 2023 and March 2024. The author takes full responsibility for the thoughts and ideas expressed in the publication.
References
1 Nisha, Sharma., Jonathan, James., et al. (2023). Comparing Prognostic Factors of Cancers Identified by Artificial Intelligence (AI) and Human Readers in Breast Cancer Screening. Cancers, 2023;15(12):3069-3069. doi: 10.3390/cancers15123069
2 Jonas, Oppenheimer., Sophia, Lüken., Bernd, Hamm., Stefan, M., Niehues. A Prospective Approach to Integration of AI Fracture Detection Software in Radiographs into Clinical Workflow. Reproductive and developmental Biology, 2022; 13(1):223-223. doi: 10.3390/life13010223
3 Yumi, Kuroda., Tomohiro, Kaneko., et al. Artificial intelligence-based point-of-care lung ultrasound for screening COVID-19 pneumoniae: Comparison with CT scans. 2023 PLOS ONE, 18(3):e0281127-e0281127. doi: 10.1371/journal.pone.0281127