Page 13 - AIH-2-2

Artificial Intelligence in Health Role of LLMs in improving patient experience

globe and the credible, objective results. However, these currently evident. High-quality deidentified health care data
articles have investigated different LLMs with minimal or are needed from all populations to ensure adequate patient
no customization, except for a single study that developed representation in the resultant technology. Providers need
an LLM from scratch. The increasing availability of medical to collaborate with researchers to ensure this, not only from
data-specific LLMs is expected to significantly address a commercial perspective but also because high-quality,
this heterogeneity. Much larger validation is required on a relevant data are essential for the development of open-
geographically, socially, and culturally diverse population access or free tools. Researchers, providers, professional
to ensure that these can be accepted for routine clinical bodies and the industry need to work closely to comprehend
use. 56-57 Given that any of these LLMs are regarded as the scope of LLM-based tools, both at the population and
decision-making aids or adjuncts, it is unclear which level patient level, to collect the data and conduct the research
of stringency needs to be adopted for their validation required to develop high-quality tools focused on patient-
process. Consensus among large professional bodies and specific outcomes that help enhance the quality of care
organizations to collaborate on these solutions to ensure delivery or patient experience. Technology can serve as a
that the datasets on which they are trained and validated are key force multiplier, particularly in regions where healthcare
large and homogenous may be beneficial; this may also help delivery and providers are scarce. Although recent regional
ensure equitable and universal access to these technologies. guidelines for the implementation of AI in healthcare have
60
A clear distinction exists between LLMs that are been put in place, there is still a significant absence of
designed with a large population in mind for the purpose consensus regarding the appropriate use and application of
of screening or engagement and those designed to achieve dozens of emerging tools and technologies in the healthcare
a clear, objectively actionable outcome, such as surgical sector. Regional and international consensus is crucial to
intervention, based on a comprehensive risk assessment. ensure appropriate use. Finally, researcher collaboration
The latter will require multiple specific, objective, and is the need of the hour; rather than performing numerous
accurate parameters as inputs to facilitate inference, independent small-scale validation of these tools, large trials
while the former may permit a lower quantity and quality need to be run on thousands of patients in a multicentric
of data, as well as a lower specificity and sensitivity of manner to assess them. With the progress in technology
outcomes. When evaluating the appropriateness of these and access, such tools are now feasible at a reasonable cost.
tools for their ecosystems, practitioners must possess a The buy-in of governments and payor systems is essential to
comprehensive understanding of the training data and the leverage this technology for the public interest.
patient population under investigation. This is particularly 5. Conclusions
true if it is a population-focused tool, such as a screening
test, or a patient-centric tool, like a diagnostic test. LLMs are becoming an increasingly common part of
Another significant consideration is the cost-benefit daily life. Health systems and professionals worldwide
ratio of these technologies and solutions. With spiraling can capitalize on their influence by gaining a more
healthcare costs around the world, it is now more crucial comprehensive understanding of LLMs and utilizing
than ever to be prudent with healthcare expenditure. The their potential to develop a safe and suitable tool that can
economic implications of incorporating LLMs into routine enhance patient care and experience. Significant efforts
practice depend considerably on how they are planned to are needed to conduct large multicentric validations of
be implemented. For an institutional LLM that is a white- these tools before they are routinely applied in clinical
labeled, optimized version of a commercially available practice. It is imperative that providers should gain a
product, the cost addition to the patient may be negligible, better understanding of these technologies and tools in a
whereas a subscription to a technology product linked to much more comprehensive manner to ensure that patients
the electronic medical record system may be too expensive benefit from them in a demonstrable and meaningful way.
or inaccessible for most. Additional research is required to Acknowledgments
better comprehend the needs of the local patient, provider,
and healthcare system, for whom these factors can vary None.
widely. Notably, none of these articles utilized tools that
are fully validated or approved for immediate clinical use, Funding
reinforcing the fact that this technology, despite its high None.
potential in healthcare, remains a work in progress.
The next steps to ensure that these tools are utilized in Conflict of interest
an appropriate, acceptable, and patient-centric manner are The authors declare they have no competing interests.

Volume 2 Issue 2 (2025) 7 doi: 10.36922/aih.4808