What’s new in Academic International Medicine: Artificial intelligence in medical education – A once-in-a-century opportunity to achieve rapid global parity

Abstract

The rapid evolution and mainstream adoption of artificial intelligence (AI) and machine learning (ML) capabilities have the potential to transform the health-care environment in many, often unexpected ways.[1,2] While high-income economies are well positioned to benefit from the improved efficiencies and greater productivity of an AI-enabled workforce,[3,4] there are also potential challenges related to rapid and not always orderly implementations of AI.[3,5] Due to inherent structural differences between high-income regions (HIRs) and low- and middle-income regions (LMIRs) around the globe, considerations specific to LMIR-centric implementations of AI in health-care are not only unique and different but also rich with once-in-a-generation transformational opportunities. When it comes to AI/ML, medical education (ME) is likely to be one of the most impacted areas of broadly defined “health-care” across the planet and specifically within LMIRs.[6-8] Potential impacts may be both positive and negative, depending on the aspect of ME affected. The AI revolution in ME across LMIRs is likely to result in what one might call a “generational leap” – not dissimilar to the very rapid (and highly transformational) adoption of cellular telephony across regions of the world where traditional “cable-based” telephony and the Internet were not feasible, too costly, or restricted to high-income residents.[8-10] Within a very short period of time, large segments of the population gained access not only to telephony in general but also to the Internet and all of the downstream benefits thereof.[11-13] Coincidentally, much greater access to the Internet around the world also forms an important element of future development, including the ability to rapidly deploy high-quality, AI-based educational platforms, with a focus on creating synergies and closing resource gaps that were previously considered insurmountable.[14] Immediate questions arise in the context of this potential transformational change, including “why should we do this?”; “who stands to benefit and who stands to lose?”; and “how does the AI-enabled ecosystem affect traditional components of ME, including current stakeholders?” There are, of course, many other potential questions around the actual implementation, resources required, and other highly granular aspects of any future deployments of AI-enabled/aided/facilitated ME in LMIRs. That said, let us focus on the three main concerns stated above in a sequential manner. The first question, “why should we do this?,” is answered relatively easily. From purely pragmatic standpoint, the availability of high-quality, AI-enabled/enhanced ME in world regions where such education is currently not available would be truly transformational.[15-17] Where opportunities simply did not exist, the availability of high-quality ME would drastically improve the public health equation; the most important considerations here include workforce development, patient access to care, and future provider retention and availability.[18,19] This, in turn, could translate into sustainable quality-of-life improvements for billions of people across the planet. Beyond these basic considerations, future downstream benefits of AI-enabled/facilitated ME may include the ability for health systems to treat more patients with enhanced overall safety and quality, significantly greater throughput at a lower overall cost, and improved provider job satisfaction.[17,20,21] In addition, AI-based efficiency enhancements would be inherently synergistic with the emerging field of precision medicine and personalized health care.[22,23] One can speculate that in the setting of LMIRs, these developments could create a literal “technological leap” from “severely restricted health-care resources” to “precision and personalized medicine” within a relatively short period of time (assuming that resources and other implementation considerations are favorable). In terms of participants who may gain from the implementation of AI-based/enhanced ME, the list is long and encompasses multiple stakeholder types across the entire society.[24,25] Obvious primary beneficiaries of a system where ME can take place locally will be local communities.[26,27] Here, the ability of harness AI-based/enhanced ME frameworks, augmented by Internet-based video conferencing and virtual medical visits, should theoretically translate into very real and relatively rapid increases in local provider (and expertise) availability. Medical students and subsequently graduate medical education (GME) trainees would be both empowered and enabled to act locally, participate more intimately in building and strengthening local capacity within communities, and would be less reliant on external resources to deliver equivalent or better care.[28-31] Equally important, medical trainees and their families are also positioned to benefit from AI-based/enhanced education. The assurance of high-quality ME, delivered and applied locally, also promises to improve both access and affordability of medical training, traditionally restricted to those who could overcome various nonacademic constraints (e.g., travel, tuition costs, and personal and professional networking limitations). There are likely more downstream benefits to the above paradigm; however, due to the space restrictions of this editorial, only superficial exploration is possible. Given many positives of the proposed paradigm, one naturally wonders, “who stands to lose from all this?” Certainly, the established structures, especially if insufficiently flexible, may be put under severe stress. Without adequate support and/or effective regulatory frameworks, it will be easy for the existing talent base to lose relevance and erode over time – something that should be avoided. Given the tremendous amount of investment in such traditional health-care education infrastructure, it will be critical to ensure a smooth transition, possibly leading to further growth and sustainable strengthening of existing local and regional ME institutions.[32-35] A corresponding transition period could ensure that existing educational resources are optimally utilized while new, AI-centric resources are being developed and implemented responsibly. At the societal level, the emergence of AI-based/enhanced ME may pose a difficult challenge to structures responsible for allocating scarce resources. Such centralized/traditional gatekeeper organizations could be adversely affected by the more decentralized approach where the majority of administrative and oversight operations would be conducted locally.[36-39] Of great importance, AI-assisted/facilitated ME models must also operate within a relatively narrowly defined and strictly controlled framework of knowledge verification, propagation, and re-evaluation, especially in the context of the so-called black-box large language models (LLMs) which lack source attribution transparency.[40,41] Consequently, all knowledge inputs should be pre-vetted carefully, and more importantly, outdated information must be periodically purged from existing models. Finally, systematic biases in LLM sources/data inputs may translate into suboptimal, inefficient, and/or potentially harmful model outputs without appropriate and ongoing oversight (e.g., LLM model providing incorrect answers and/or medical advice, out of context, or fictitious answers).[42-44] Briefly discussed above, the very important aspect of the effect of decentralization of ME resources within a traditionally centralized system cannot be overstated. Due to the overall high complexity of our current ME frameworks, regardless of level of income, geographic location, and/or other resource-based considerations, any new implementations and/or reforms based on the current status quo will need to be well researched, intricately planned, and masterfully executed.[45-47] Although successful implementations of “overnight changes” – even very drastic ones – are possible, as shown during the coronavirus 2019 (COVID-19) pandemic, the associated experiences do not necessarily translate into or apply within the context of our current discussion.[48,49] Consequently, it will be important for local collaboratives, consisting of community thought leaders, acting at a micro/local level, but also coordinating with other regional stakeholders at a more macro/global level, to comprehensively consider all available options and to select actions that best address local needs, with focus on culturally acceptable solutions. Of importance, based on historical experiences, any new technology platforms intended for LMIRs should not represent hastily adapted solutions that are primarily designed for HIRs and HIR learning environments.[50-53] In general, HIR participants in educational technology implementations for LMIRs should have a primarily consultative presence, especially at the local/community level. Intentional empowerment of local stakeholders, including governments, communities, students/trainees, and patient-facing health-care personnel in general, is the best way to ensure successful technology roll-outs/implementations, long-term sustainability, effective system strengthening, and ultimately safe/unbiased delivery of optimized medical care. Within this context, AI-enabled/facilitated ME may become one of the cornerstones of such intentional and sustainable empowerment. Another important consideration here is the parallel development of highly competent medical educators using similar AI-driven approaches. Although much work remains to be done before any future durable and high-quality implementations of AI-based ME are feasible within LMIRs, all of the fundamental building blocks are already in place. It is up to us – the collective academic international medical community – to ensure that appropriate local solutions, with predominantly local drivers and inputs, are thoughtfully developed and put into locally relevant practice. The ultimate prize is very tempting – the potential for highly accelerated attainment of global parity in ME and subsequently medical care – realized through technologically driven “quantum leap” similar to that observed after the mass introduction of cellular telephony across LMIRs. Although not without risks, the benefits of such a promising prospect are likely to significantly outweigh any downsides. The future is full of promise – it is up to us to keep it that way! Ethical conduct of research The authors verify that preparatory activities related to this Editorial meet the institutional standards of ethical research. This article does not contain any data or studies involving human participants. The authors declare that this Editorial did not require Institutional Review Board/Ethics Board approval.

Ähnliche Arbeiten