Potential role of large language models and personalized medicine to innovate cardiac rehabilitation

Abstract

Cardiac rehabilitation is a crucial multidisciplinary approach to improve patient outcomes. There is a growing body of evidence that suggests that these programs contribute towards reducing cardiovascular mortality and recurrence. Despite this, cardiac rehabilitation is underutilized and adherence to these programs has been a demonstrated barrier in achieving these outcomes. As a result, there is a growing focus on innovating these programs, especially from the standpoint of digital health and personalized medicine. This editorial discusses the possible roles of large language models, such as their role in ChatGPT, in further personalizing cardiac rehabilitation programs through simplifying medical jargon and employing motivational interviewing techniques, thus boosting patient engagement and adherence. However, these possibilities must be further investigated in the clinical literature. Likewise, the integration of large language models in cardiac rehabilitation will be challenging in its nascent stages to ensure accurate and ethical information delivery.

Key Words: Cardiac rehabilitation; Large language models; Patient education; Motivational interviewing; Artificial intelligence

Core Tip: Large language models may help innovate cardiac rehabilitation programs on a larger scale, but there is a large paucity in evidence to support its utility and evaluating the validity of these innovative proposals. Likewise, this new innovation may be able to assist in developing more personalized medicine for patients and clinical research.

INTRODUCTION

Cardiac rehabilitation is a multidisciplinary approach aimed to improve cardiovascular outcomes via emphasizing a patient’s functional capacity and quality of life. Major components of cardiac rehabilitation include prescriptive exercise-based therapy, behavioral modifications, psychosocial counseling, and medical risk factor stratification[1-3]. While pharmacologic therapy is the initial choice of treatment for current cardiovascular morbidities, cardiac rehabilitation is somewhat of an underutilized tool that can serve as an adjunct to enhance patient outcomes, with current literature showing participation in exercise-based cardiac rehabilitation reducing cardiovascular mortality, recurrent cardiac events, and improving overall quality of life[4]. As cardiac rehabilitation is a rapidly growing field of interest that has shown evidence-based benefits, the current literature has been expanding on the potential implications given that only about 25% of individuals with eligible cardiovascular events will participate in cardiac rehabilitation[2,5] with some literature showing attendance rates at less than 20% over the past two decades[6,7]. Although cardiac rehabilitation can be valuable, adherence can be difficult for even the most motivated patients due to extraneous factors like health literacy, socioeconomic status, and personal obligations. This editorial aims to expand upon a recent publication by Kourek et al[6], which describes the latest updates on cardiac rehabilitation, including its indications, program structure, clinical outcomes, and potential limitations, while initiating a discussion regarding the likely innovations in the field of cardiac rehabilitation with patient adherence at its forefront. In particular, this editorial aims to build upon advancing cardiac rehabilitation by highlighting the possible incorporation of a large language model (LLM) as an educational tool to personalize further a patient’s experience undergoing cardiac rehabilitation[6,7].

LLMS

LLMs are an expanding area of research interest as they have the capability to leverage deep learning to train and learn responses. These models are employed via chatbots (i.e., ChatGPT). From a health literacy standpoint, LLMs may have the technological capabilities to directly impact health literacy by improving the readability of standard patient education methods, including pamphlets, action plans, and after-visit summaries[8,9]. However, there is currently a paucity of literature to support this claim. Let alone, studies have shown the potential benefits of artificial intelligence (AI) implementation on patient adherence with direct conversational influence on behavioral modification, particularly in vaccination adherence and weight-management in the overweight and obese patient population[10-13]. Although these findings are not directly in the field of cardiac rehabilitation, similar techniques can be offered to this patient population. For example, LLMs could be employed to learn and implement motivational interviewing concepts to initiate and maintain patient adherence to cardiac rehabilitation. An additional hypothetical situation where LLMs could aid the patients in cardiac rehabilitation could be in the ability to generate a personalized summary of the cardiac rehabilitation program for a patient. This personalized summary could adapt to the healthy literacy level for each patient, generating text at a grade reading level that provides the highest degree of comprehension for the patient. However, there is still a growing development of literature that demonstrates this specific situation in the field of rehabilitation[14]. Another role for LLMs could be in translating cardiac rehabilitation instructions to patients into their language of choice[15]. This can help facilitate the delivery of more equitable healthcare among those faced with language barriers.

While conversational methods in AI have been shown to improve patient adherence in preventative care in terms of improving health literacy, another potential avenue to explore is the impact of medical jargon on patient commitment to therapy. To expand, patients can find it difficult to follow medical jargon, which affects their awareness of their medical condition, its temporality, and treatment plans, creating a communication barrier between the patient and physician. A systematic review by Nickel et al[16] found that medical jargon resulted in an increased perception of disease severity, patient anxiety, and a desire for invasive management for their condition, highlighting that increased complexity of medical terms can influence the pattern of patient adherence. LLMs are a potential route that can assist how a physician communicates with a patient to advance health literacy by helping to translate medical jargon into a level that patients can understand. However, it is imperative for physicians and cardiac rehabilitation specialists to maintain a high degree of oversight with using these tools as minimizing medical jargon may blur the true meaning or severity of a condition[17,18].

CONCLUSION

While this editorial described the possible utility of LLMs in assisting in motivational conversation techniques and simplification of terminology for patients, there remain major limitations to this technology. From an evidence standpoint, the possible utilities of LLMs must continue to be investigated at a larger scale. The current direction of literature has largely utilized few LLMs, including ChatGPT, Copilot, and Gemini[19,20]. There may be an avenue for the development of LLMs which primarily learn from neural networks dedicated to using peer-reviewed, evidence-based resources, which may perform differently from these current models[21]. Secondly, there must be an ever-present need to develop further ethical and regulatory infrastructure on the use of LLMs in medicine overall. Currently, there remains a need for profound oversight by physicians and researchers to develop these models in order to effectively reduce and remove “hallucination” or the generation of inaccuracy and fabricated responses by LLMs based on the training of the model[22-24]. Additionally, it must be communicated to patients that LLMs are not medical sources of knowledge or healthcare providers but rather a communication tool to help deliver more equitable healthcare to individuals. This delivery can be in the form of personalizing the grade-reading level of the cardiac rehabilitation materials provided to patients or possibly transplanting these tools into different languages. Overall, these promising models will continue to develop given the great degree of resource allocation on advancing these promising tools[25].