Abstract
Background
Generative Artificial Intelligence (GenAI) has rapidly gained traction in medical education, yet little is known about its use among medical students in Puerto Rico. This study examines GenAI usage patterns among first- and second-year students across all four Liaison Committee on Medical Education-accredited medical schools in Puerto Rico, with comparisons based on user experience.
Methods
A cross-institutional electronic survey was conducted from January to June 2024 among first- and second-year medical students (n = 194) from Ponce Health Sciences University, University of Puerto Rico Medical Sciences Campus, San Juan Bautista School of Medicine, and Universidad Central del Caribe. Students were grouped by program year, and GenAI experience (<1 year vs ≥ 1 year). Chi-square tests were used to evaluate group differences (P < 0.05).
Results
Of the 778 first- and second-year medical students enrolled across all schools, 194 responded. Most (72.7%) reported using GenAI, primarily ChatGPT (89.9%), mainly for academic purposes (75.3%). Students with ≥1 year of GenAI experience were more likely to agree that GenAI helped them understand research papers (89.9%) compared to 51.6% of less-experienced users who disagreed (P < 0.001). From this same cohort, 72.7% also planned to use GenAI for board exam preparation, compared to 63.2% of less-experienced users who did not plan to use it. Although 72.2% believe GenAI will be integrated into Puerto Rico's healthcare system, only 52.1% felt that local medical facilities would be receptive. Challenges such as limited data access (27%) and power outages (34%) were more commonly reported by less-experienced users.
Conclusion
Generative AI adoption among Puerto Rican medical students is expanding, mainly for academic purposes. Greater experience correlates with higher perceived benefit, emphasizing the need for structured, ethical GenAI training, and institutional support within medical education.
Keywords
Introduction
Generative Artificial Intelligence (GenAI) has sparked intrigue and controversy since its introduction in the 1950s. 1 Enhancement in AI's problem-solving capabilities through advances in machine learning has driven its integration into multiple medical disciplines.2–6 GenAI operates on machine learning principles, which, through data-driven algorithms, is capable of creating new content, providing countless different responses to a singular question while maintaining its structure.7–9 In medical education, the capabilities of this new technology have been used for simulated patient cases and learning experiences, providing real-time feedback to medical students. It is because of its quick integration and practical impact that this tool has triggered considerable debate and criticism.10–14
The integration of GenAI into the medical field has sparked diverse opinions regarding its reliability and ethical implications.15–17 As GenAI solely relies on existing datasets, it may inadvertently perpetuate biases, directly affecting marginalized and underrepresented populations with limited inclusion in medical literature.18–20 This raises concerns about its accuracy and potential to reinforce disparities in healthcare outcomes.
Prior studies have assessed Hispanic medical students’ awareness and perceptions of GenAI, but they have been limited to a single institution with a small sample size, reducing the generalizability of their results. 21 Therefore, there is a need for broader, multi-institutional research that includes all accredited medical schools in Puerto Rico to better capture the diversity of student experiences and perspectives.
To our knowledge, this is the first large-scale, multi-institutional evaluation of generative AI adoption within all four Liaison Committee on Medical Education (LCME)-accredited medical schools in Puerto Rico, providing a comprehensive perspective on GenAI usage in an underrepresented Hispanic population. This study uniquely integrates demographic, geographic, and curricular variations to identify early adoption patterns, perceived benefits, and institutional challenges of GenAI in medical education. By capturing representation from the entire Puerto Rican medical education system, it offers the first national-level insight into how future Hispanic physicians are engaging with GenAI, filling a critical gap in the literature on equitable and culturally relevant AI integration in medicine.
Innovative and thorough evaluations are obtained through racially and culturally diverse teams, which simultaneously work toward addressing and managing health disparities. Despite Latino individuals making up 19% of the U.S. population, they account for only 5% of Graduate Medical Education trainees.22,23 Puerto Rico medical schools are all aligned with the LCME and, therefore, accredited by this committee. 24 Puerto Rico's LCME-accredited medical schools play a critical role in diversifying the physician workforce, contributing 14% of Latino MD graduates in the nation, which represents about 6% of all U.S. MD graduates. 25 Therefore, a high-level assessment of medical students among the four medical schools in Puerto Rico is vital to further assess the projections of GenAI usage among future generations of MD graduates.
Given that Puerto Rican medical schools contribute significantly to the U.S. Latino physician workforce, this study aims to examine GenAI usage trends among first- and second-year medical students across all four accredited medical schools in Puerto Rico. 26 Unlike previous studies, which have been limited in scope, this cross-institutional analysis provides a comprehensive perspective, considering demographic, geographic, and curricular differences in GenAI adoption; therefore, this study serves as a foundational resource to project future AI integration in medical education and medical practice.
Methods
This is a cross-institutional study encompassing all four medical schools in Puerto Rico [Ponce Health Sciences University (PHSU-SoM), University of Puerto Rico, Medical Sciences Campus (UPRRCM-SoM), San Juan Bautista School of Medicine (SJB-SoM), and Universidad Central del Caribe (UCC-SoM)]. Prior to the beginning of this study, an approval letter for the project was obtained from the administration and faculty of all institutions that were assessed. The Institutional Review Board (IRB) approved an exempt protocol from the leading site of this study, PHSU-SoM, after review of all approval letters from the four medical schools. Inclusion criteria were active first-year and second-year medical students in Puerto Rico, chosen due to their pre-rotation and pre-STEP-1 status, allowing them ample time for the use of GenAI for their coursework.
An electronic survey with sub-sections through Google Forms was developed for the purpose of this study. Prior to beginning the survey, an informative page and self-assessment question were displayed; through this, we confirmed that the respondent of the survey was a medical student from a Puerto Rican medical school. The informative page included information regarding the purpose of the study, confidentiality strategies, instructions for its completion, and emergency contact. For the development of the survey, we evaluated questions from previous studies assessing the usage and perceptions of new technology in a medical setting.27,28 These were adapted, and additional questions were added to assess specific points for medical students from Puerto Rico (Table 1). The complete questionnaire used in this study is provided as Supplementary File 1. In the first section, we address demographics among the survey participants. Secondly, we explored the use, purpose, and duration of GenAI usage among medical students. To assess the challenges medical students face while using GenAI in Puerto Rico, participants responded to 2 statements using a 5-point Likert scale (1 = strongly agree to 5 = strongly disagree) (Figure 1).
Likert scale for items 9 and 10 in survey of GenAI usage. GenAI, Generative Artificial Intelligence.
Generative Artificial Intelligence Usage-Related Questions.
To ensure a proper distribution of the survey and that a medical student was the respondent, the survey was distributed in an exclusive fashion. It was shared with medical students in both electronically [via medical students only – group chat (eg, Telegram)] and through in-person interactions prior to the beginning of classes, directed to the target group. In addition, the respondent was confirmed to be a medical student prior to the beginning of the survey. Data collection took place from January 15, 2024 to June 15, 2024. Participation was voluntary, and written informed consent was obtained electronically from all participants prior to survey initiation. Students were informed about the possible benefits, risks, or discomforts and how their information would be used.
After the closing date for questionnaire submission, results were downloaded as an Excel document. Analysis was performed comparing first- and second-year medical students. A subgroup analysis was performed comparing two categorical groups: respondents who reported using GenAI for less than a year and those who reported using GenAI for 1 year or more. For this study, we assumed that GenAI usage of 1 year or more is likely related to more expertise while navigating the platform, compared to those with less than a year of GenAI usage. However, proficiency in GenAI usage was not directly measured.
Statistical and Power Analysis
Data analysis was conducted using BlueSky Statistics software, version 10.3.4 (BlueSky Statistics LLC). Likert-scale responses were grouped into broader categories. “Strongly disagree” and “somewhat disagree” responses were grouped and classified as disagreement, and “strongly agree” and “somewhat agree” responses as agreement. (Figure 1) Crosstabulations with chi-square tests were performed to assess the association between categorical variables. The chi-square test was chosen to determine whether there were significant differences between groups. A P-value < 0.05 was considered statistically significant.
An post hoc power analysis for two independent proportions (MS1 vs MS2) was performed using Cohen's h method with α = 0.05 (two-sided). 29 Detecting a small-to-moderate effect (h = 0.30, approximately a 10% difference between groups) would require 176 participants per class to achieve 80% power. The final analyzed sample included 92 first- and 102 second-year medical students, providing approximately 50% power to detect such differences and higher power for moderate-to-large effects (h ≥ 0.40). While underpowered for very small effects, the achieved sample was sufficient to explore meaningful trends in GenAI usage across class years.
Data Exclusion
During the process of data collection, several criteria were applied to ensure the integrity and relevance of the responses included in the final analysis. To ensure that we were including responses from our desired study population, responses from participants that were not first- or second-year medical students were excluded (Figure 2). This was first verified by a pre-screening question, where the responder confirmed to be a medical student in Puerto Rico. In addition, the survey included a question about their year in medical school, validating their eligibility to participate. Furthermore, the survey was distributed only in specific class sections and Telegram chats exclusive to medical students in Puerto Rico, preventing the dissemination to non-target groups. Only complete responses were evaluated, as the platform does not share information from incomplete responses.
Participant inclusion flowchart.
Participant Recruitment and Reporting Guideline
This study aimed to include all eligible first- and second-year medical students enrolled in Puerto Rico's four LCME-accredited medical schools during the 2023-2024 academic year. Therefore, a formal sample size calculation was not performed. Recruitment efforts were maximized through electronic and in-person outreach; however, participation was ultimately limited by academic workload, timing during the semester, and response availability. Reporting guideline. The reporting of this cross-sectional study conforms to the STROBE Guidelines for observational studies [STROBE checklist provided as Supplementary File 2]. 30
Results
Demographics
The number of first- and second-year students enrolled in the four medical schools of Puerto Rico during the 2023-2024 academic year at the time of data collection was 778. During the collection period, a total of 217 medical students responded to the survey, of whom 194 were included in the analysis (Figure 2). Among the 194 participants, 92 were first-year medical students (MS1) and 102 were second-year medical students (MS2). Additionally, 126 were female and 67 were male (1 respondent identified as other). Over two-thirds (68.0%) of respondents were between 18 and 25 years old. More than half of the respondents identified as first-generation medical students (79.4% yes vs 20.6% no). Among respondents, 39.7% had a household income of less than $25,000, 20.6% between $25,000 and $50,000, and 39.7% of more than $51,000 (Table 2). Given the relatively even distribution and lack of observed group differences, no further analysis was conducted.
Demographics.
Generative Artificial Intelligence Usage
Of all the respondents, 72.7% had used GenAI either for less than a year, 1 year, 2 years, or more than 3 years, with the highest use being among first-year medical students (41.8%) compared to second-year medical students (30.9%). More specific information regarding the weekly use of GenAI among first- and second-year medical students can be found in Table 1. ChatGPT was reported by 89.9% of respondents as their primary GenAI platform, while Google Gemini (formerly Bard) was only used by 5.7% of respondents. The most reported purpose of AI usage among first-year medical students and second-year medical students was academic use (75.0% and 75.5%, respectively). A higher proportion of second-year medical students reported that AI has helped them better understand scientific literature (Table 1). Of all respondents, over half plan to use AI for board examinations (55.2% yes vs 44.8% no). Almost three-fourths of respondents saw the possibility of implementing AI within the medical context in Puerto Rico (72.2% yes vs 27.8% no). However, only approximately half of the participants said that most medical facilities would accept AI services (52.1% yes vs 47.9% no).
Challenges regarding GenAI use, such as usage limitations due to data concerns or unexpected power outages, were also explored. Among first-year medical students, 28.3% agreed, 32.6% remained neutral, and 40.2% disagreed regarding the statement that GenAI has been challenging due to data velocity limitations. A higher proportion of second-year medical students reported neutrality to this statement (41.2%). When assessing limited use of GenAI due to power outages, we found that a higher proportion of first-year medical students disagreed with the statement compared to second-year medical students (Table 1).
When it comes to subgroup analysis, 51.6% of those who had been using AI for less than a year reported disagreement regarding GenAI's utility to better understand research papers, while 89.9% of those who had been using GenAI for more than a year reported agreement (P-value <0.0001). Among those with less than a year of using AI, 63.2% reported not planning to use GenAI for their board exam preparation, while 72.7% of those with more than a year of usage reported they did plan to use it (P < 0.0001). With regard to challenges involving GenAI usage, those with less than a year of using GenAI reported higher neutrality to data velocity limitations, while 43.4% of those who had used GenAI for more than a year reported disagreement with a P-value of <0.0001. Students’ response to the statement regarding the limited use of GenAI due to power outages showed a high proportion of neutrality across both groups of experienced users. Further information regarding this finding can be found in Table 1.
Discussion
GenAI has shown an increasing trend in its evolution, practice, and integration in the medical field.2–6 This study aimed to assess GenAI usage among first- and second-year medical students in Puerto Rico, while also evaluating differences between those with less than a year of GenAI experience and those with more than a year of experience. We found that most medical students, both first and second-year medical students, share similar usage of GenAI platforms, namely for academic, personal, and research purposes (in descending order of reported use). Similarly, we found that both cohorts reported using GenAI at least once per week. These findings parallel previously reported use of GenAI in the literature. 31
When considering the use of GenAI for research purposes, both groups reported that GenAI helped them understand the research literature. Markowitz (2024) established the benefits of GenAI tools for these purposes, noting that GenAI enhances perceptions of scientists and the public's understanding of science, making the literature more accessible to the general public. 32 However, we found a significant difference in the use of GenAI as a tool for understanding research papers between students who had used GenAI for more than a year and those who had used GenAI for less than a year. The higher percentage reported among students who had used GenAI for more than a year regarding the benefit of GenAI to better understand research articles can be directly related to the students’ expertise while navigating these platforms. Those with experience using GenAI tools for more than a year have had enough exposure to become familiarized with the tool and to be able to strategize or format their request to obtain the information they desire from GenAI. This contrasts with those with less experience, in which the prompts to obtain the intended responses may mislead the platform. Based on this reasoning, advocates have reinforced the need to foster an understanding of GenAI platforms in the medical curriculum. As proponents have stated, there is a discrepancy in GenAI usage, considerations, and ethical perspectives, which should be addressed through its integration into the medical curriculum.33–35
The majority of first-year and second-year medical students reported they were planning to use GenAI as a resource to study for the USMLE Step 1 and Step 2 board exams. However, responses varied according to the duration of GenAI usage; those students who had experience using GenAI platforms for more than a year reported an overwhelming yes to the use of GenAI for board exam studying purposes, while a greater proportion of students with less experience reported they did not plan to use it for board exam preparation. This, we believe, follows the same pattern as before, where those students with less experience perhaps lack the understanding of how to properly use these GenAI platforms. Therefore, appropriate training addressing how to use GenAI may offer clarity for medical students to further determine if the tool is appropriate or beneficial for their individual studying purposes. Importantly, the present study refers only to permissible educational uses of GenAI, such as summarizing concepts, generating practice questions, or clarifying explanations, and does not endorse or imply any use that would violate examination integrity, testing security, or institutional ethical standards. 36 Ensuring students understand these boundaries will be essential for the responsible integration of GenAI into medical education.
When evaluating whether medical students view GenAI as being implemented in Puerto Rico, we found that the majority agreed with its future integration. First and second-year medical students, as well as those who had used GenAI for more than a year or less than a year, agreed that its integration will take place in the medical setting in Puerto Rico. However, when directly asked if medical facilities would welcome GenAI, opinions were more divided. First-year medical students and those who had experience of more than a year using GenAI reported that it would be welcomed by most medical facilities. However, those in the second year and those with less than a year of experience with GenAI were reluctant to see its projection as being accepted as an asset by most medical facilities. These results follow the considerations from a historical trend perspective, in which the tools of the future represent the challenges of today. 37
This study is not without limitations. Although the response rate was 25%, subgroup analyses by GenAI usage duration helped mitigate concerns regarding representativeness. Nonetheless, nonresponse bias cannot be ruled out, as students with stronger opinions about GenAI may have been more likely to participate, potentially leading to an overestimation of adoption rates and perceived benefits. Variations in curriculum across institutions and the absence of a validated or pilot-tested questionnaire may also limit the precision and comparability of findings. Participation was further constrained by academic workload, semester timing, and a temporary survey platform outage. An a priori sample size calculation was not conducted, as all eligible students were invited to participate. However, a post hoc power analysis indicated that while the final response rate may limit generalizability and statistical power to detect small effects, the achieved sample size provided sufficient power to explore meaningful trends. Despite these limitations, this study offers valuable insights into GenAI adoption among medical students and underscores the need for future longitudinal investigations to evaluate its long-term educational and clinical implications.
Our study results suggest that the use of Generative AI among medical students will continue to grow, particularly for academic and research purposes. However, its effective integration may be hindered by a lack of structured training, limited exposure, and difficulty in distinguishing trustworthy responses. To ensure that medical students maximize GenAI potential while maintaining academic and ethical integrity, it is essential to incorporate GenAI training into medical curricula. Although the duration of use does not necessarily indicate proficiency in using AI effectively, receiving targeted training and exposure to the platform can play a crucial role in developing AI-related skills. Such training, led by certified and experienced professionals, would equip students with the competencies needed to critically and objectively evaluate GenAI information output and optimize its use for medical education and future clinical practice.
Conclusion
This multi-institutional study provides the first comprehensive evaluation of GenAI usage among medical students in Puerto Rico. Most first- and second-year students reported using GenAI, primarily for academic and research purposes, with greater perceived benefits among those with over 1 year of experience. These findings indicate that familiarity with GenAI enhances confidence and educational value. However, differences in perceived institutional acceptance and limited formal training highlight the need for structured curricular integration. Incorporating standardized GenAI education within medical programs could foster responsible, ethical, and effective use of AI tools, improving students’ research literacy and critical appraisal skills. By capturing perspectives from all LCME-accredited medical schools in Puerto Rico, this study offers essential baseline data on GenAI adoption among Hispanic medical students and expands on the importance of equitable, culturally relevant AI implementation in medical education and future clinical practice.
Supplemental Material
sj-pdf-1-mde-10.1177_23821205251398923 - Supplemental material for Understanding Generative Artificial Intelligence Adoption in Puerto Rican Medical Schools: A Cross-Institutional Survey of First- and Second-Year Students
Supplemental material, sj-pdf-1-mde-10.1177_23821205251398923 for Understanding Generative Artificial Intelligence Adoption in Puerto Rican Medical Schools: A Cross-Institutional Survey of First- and Second-Year Students by Adrián E. González-Bravo, Carolina Batlle, Valeria S. Fadhel-Hernández, Pedro E. Maldonado-Núñez, Gustavo Christian, Corally López-Vega, Wilfredo De Jesús-Rojas, Norman Ramírez and Marcos J. Ramos-Benitez in Journal of Medical Education and Curricular Development
Supplemental Material
sj-pdf-2-mde-10.1177_23821205251398923 - Supplemental material for Understanding Generative Artificial Intelligence Adoption in Puerto Rican Medical Schools: A Cross-Institutional Survey of First- and Second-Year Students
Supplemental material, sj-pdf-2-mde-10.1177_23821205251398923 for Understanding Generative Artificial Intelligence Adoption in Puerto Rican Medical Schools: A Cross-Institutional Survey of First- and Second-Year Students by Adrián E. González-Bravo, Carolina Batlle, Valeria S. Fadhel-Hernández, Pedro E. Maldonado-Núñez, Gustavo Christian, Corally López-Vega, Wilfredo De Jesús-Rojas, Norman Ramírez and Marcos J. Ramos-Benitez in Journal of Medical Education and Curricular Development
Footnotes
Acknowledgements
The authors gratefully acknowledge the collaboration of the four participating institutions, as well as the students who generously contributed their time to complete the survey.
Ethical Considerations
The Institutional Review Board (IRB) at Ponce Health Sciences University, School of Medicine, approved our questionnaire (approval: 2403188141) on December 20, 2023. Respondents gave written consent for review and signature before starting interviews.
Consent for Publication
Not applicable. Respondents provided informed consent at the beginning of the questionnaire, authorizing the use of their responses for research purposes. No identifiable personal data were collected, and no additional documentation was required.
Authors' Contributions
AEG contributed to conception and design of the study, data collection, statistical analysis, interpretation of data, drafting of the manuscript, and critical revision for important intellectual content. CB contributed to data acquisition, statistical analysis support, interpretation of data, drafting sections of the manuscript, and critical manuscript revisions. VSF contributed to assistance with statistical analysis, interpretation of findings, manuscript editing, and preparation of figures/tables. PEM contributed to data acquisition, quality assurance of study variables, and manuscript drafting. GC contributed to data acquisition, quality assurance of study variables, and drafted the manuscript. CLV contributed to assistance with study design, data verification, interpretation of results, and drafted the manuscript. WdJ contributed to oversight of study design, critical revision of the manuscript, and educational mentorship throughout the project. NR contributed to substantial contributions to conception and design, interpretation of data, critical revisions of the manuscript, and guidance in scholarly development. MJR contributed to study supervision, project oversight, substantial contributions to study conception and design, interpretation of data, and critical revision of the manuscript for important intellectual content. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Ponce Research Institute (PRI), Ponce Health Sciences University, and by the School of Medicine Student Academic Committee (STAC), Ponce Health Sciences. University.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Data supporting the findings of the study are available from the corresponding author upon reasonable request.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
