Page 9 - AIH-2-2

Artificial Intelligence in Health Role of LLMs in improving patient experience

Records identified through database
screening (N = 3568)
Records excluded as duplicates
(N = 548)

Records after duplicates removed
(N = 3020) Records excluded as titles and abstracts did not match.
(N = 2973)
• Studies focusing on non-LLMs AI models or other machine
learning techniques unrelated to patient experience.
• Studies using LLMs solely for generating manuscripts or
text without commentary on patient outcomes
Full text articles assessed for • Articles published before January 2015
eligibility (N = 47) • Articles published in languages other than English
• Studies focused on administrative tasks without
discussing the patient-facing impact or satisfaction.
• Studies lacking specific outcome measures related
Studies included (N = 47) to patient experience or satisfaction
Figure 1. Preferred reporting items for systematic reviews and meta-analyses for scoping reviews flow diagram of the search and study selection process

“How can LLMs enhance doctor-patient interaction and • Type of LLM used
satisfy patients?” LLMs are a subset of AI that are rapidly • Purpose and application of the LLM
developing and attracting the attention of all sections • Key outcomes (e.g., patient satisfaction and
of the population. To identify relevant articles, Ovid improvements in communication)
MEDLINE, PUBMED, and Google Scholar databases • Challenges and limitations reported
were searched for potential citations of interest. We have A single researcher (ABV) extracted data from all 47
included all articles that have been published since January articles, which were subsequently reviewed by all authors.
2015 and discussed the role of LLMs in enhancing patient
satisfaction and experience. Furthermore, we excluded 2.5. Collating, summarizing, and reporting
cases where an LLM was used to prepare the manuscript the results
without any additional commentary. Qualitative and quantitative analyses were used to

2.3. Study selection synthesize and collate the data. The descriptive summaries
of each article were tabulated using Microsoft Excel, and
Citations were managed using Zotero, and the relevant the results of the included studies were reported.
articles were included if they addressed the use of LLM in
patient satisfaction, including applications in pre-hospital 3. Results
and preadmission settings as well as maintenance of
electronic health records. Furthermore, we incorporated 3.1. Article characteristics
articles that discussed teleconsultation, patient education, Our search yielded 47 unique articles that were published
and patient care and health outcomes. We excluded studies across 13 countries (Figure 2). 12-17 The public and
that employed alternative forms of machine learning or healthcare providers’ access to LLMs has considerably
NLP that were not related to LLMs or patient experience increased, as evidenced by the fact that the majority of
and/or satisfaction. articles were published between 2022 and 2024, with the
earliest publication dating back to 2020. The journals in
2.4. Data extraction which they are published cover a wide range of topics, from
A data extraction tool was developed using Microsoft primary care to health technology. The articles were also
Excel to systematically extract critical information from of varying structures, such as prospective studies, cross-
each included study. The data extraction categories were sectional studies, retrospective reviews, and systematic
as follows: reviews (Table 1).
• Publication year
• First author(s) 3.2. Area of health research
• Study location Broadly, the thematic areas that were addressed by
• Study design the articles included enhancing communication
• Participant demographics (if applicable) between patients and healthcare providers, improving

Volume 2 Issue 2 (2025) 3 doi: 10.36922/aih.4808