Page 12 - AIH-2-2

Artificial Intelligence in Health Role of LLMs in improving patient experience

that allowed patients to inquire about their health in precise responses to queries, schedule appointments, and
general or their specific disease state and its treatment provide guidance on healthcare services can be valuable
by interacting with a chatbot-like LLM. This use most in enhancing communication between healthcare workers
closely resembled the ChatGPT-like interface that is likely and doctors as well as improving patient satisfaction
to be familiar to most patients. Evaluation of these LLMs and engagement by promptly and effectively addressing
was largely positive, with all articles employing objective their concerns. This ability to analyze vast amounts of
assessments to ascertain the accuracy of the results, which patient data, including electronic health records, medical
were rated on an ordinal scale by providers based on the histories, and treatment plans, enables them to leverage
response of ChatGPT to patient queries. This context may machine learning algorithms to identify patterns, predict
be particularly useful when there is limited availability of outcomes, and suggest tailored treatment options based on
provider access, when there is a language barrier, or when a patient’s unique characteristics and medical history. This
there are cross-cultural barriers. Notably, these studies personalized approach has the potential to enhance health
had human involvement before sharing information with outcomes as well as patient experience and satisfaction.
the patients, underscoring the fact that none of these Patient education is also essential for empowering
models have completed the rigorous testing required for individuals to make informed decisions regarding their
unsupervised deployment. health and well-being. LLMs can generate easy-to-
Another study investigated the role of LLMs during understand educational materials, such as articles, videos,
18
teleconsultation, where providers would use LLMs as a and interactive tools, to educate patients about their
prompt-like tool to facilitate interpersonal as well as aid in medical conditions, treatment options, and preventive
diagnosis establishment and documentation and sharing of care measures. By providing accurate and accessible
patient records among members of the treating team. This information, LLMs can assist patients in navigating complex
study examined the provider perception of a hypothetical healthcare information and promoting health literacy.
LLM rather than evaluating or repurposing an existing Furthermore, healthcare organizations encounter multiple
available LLM for this purpose. administrative challenges, such as appointment scheduling,
medical invoicing, and insurance claim processing. LLMs
The other aspect that has been explored was NLP, where
LLM automatically extracts data from existing clinical can automate these administrative processes by analyzing
and generating text-based documents, filling out forms,
reports to infer the context. 44-45 Here, the machine learning and facilitating communication between patients and
functions were employed to identify strings of keywords, healthcare providers. By streamlining administrative
better understand the context of the provider drafting these processes, LLMs can improve operational efficiency,
notes, and perform pattern recognition of biochemical reduce errors, and enhance the overall patient experience.
or radiological reports to enable a comprehensive risk
assessment. 46-47,51,53 This approach is not novel in medicine; However, LLMs are not without concerns, and these
risk assessment tools or nomograms have been in use for need to be resolved before the routine use of LLMs.
decades. However, LLMs facilitate the passive extraction Understanding these concerns better will enhance trust
of these data from electronic medical records. This and ensure that the limitations that they possess are
permits the simultaneous execution of screening and risk acknowledged. Using generic platforms may not be
assessments for numerous outcomes without the provider’s appropriate; it is possible that they must be modified
active involvement. Researchers have also employed NLP to and streamlined for the intended purpose. This concern
investigate social determinants of health, 42,55 demonstrating is expected to be alleviated by the increased availability
that it can be implemented in a context where objective of medical data-trained LLMs; however, it is necessary
data, such as laboratory reports, is unavailable and only to conduct robust and repeated training and validation
relatively objective data, such as clinical notes, are available. to identify potential failures and pitfalls. This can be
accomplished only by emphasizing that, similar to medical
4. Discussion devices or pharmaceutical trials, we establish explicit
LLMs play a distinct role in improving provider–patient validation and approval pathways and regulations before
experience and interaction efficiency, as demonstrated by they are available outside of a trial or research setting.
the volume and quality of data supporting the same. There One significant question remains unresolved—what
have been concerted efforts to improve their suitability for is the current status of LLMs and how close are we to
routine patient care across multiple specialties. There are integrating them into daily practice? The robust, reliable,
numerous advantages of LLM. The capacity to analyze and and consistent nature of LLMs is suggested by the substantial
interpret natural language inputs from patients, provide number of articles available from various regions of the

Volume 2 Issue 2 (2025) 6 doi: 10.36922/aih.4808