Researchers from the National Institutes of Health are making significant strides in the field of healthcare by leveraging large language models to develop an artificial intelligence framework. This innovative approach aims to streamline the clinical trial matching process and facilitate the quicker linking of potential volunteers to relevant trials listed on ClinicalTrials.gov. According to an NIH announcement this month, when benchmarked against three human clinicians, the tool, TrialGPT, achieved nearly the same level of accuracy. This discovery holds great promise for improving the efficiency and effectiveness of clinical trials.

Revolutionizing Clinical Trial Matching with AI

Streamlining the Clinical Trial Matching Process

Researchers at the National Library of Medicine and National Cancer Institute have developed the TrialGPT framework to address the time and resource-intensive nature of finding the right clinical trial for a patient. The new clinical trial matching algorithm analyzes patient summaries for relevant medical and demographic information. It then identifies clinical trials for which a patient is eligible and excludes those for which they are not. This annotated list of clinical trials, ranked by relevance and eligibility, allows clinicians to discuss trial opportunities with their patients more effectively. The AI tool also explains how a person meets the study enrollment criteria, which is crucial for its efficacy.

To assess the tool's performance, researchers compared TrialGPT's results to those of three human clinicians who evaluated over 1,000 patient-criterion pairs. This comparison showed that TrialGPT could help clinicians connect their patients to clinical trial opportunities more efficiently, saving precious time that can be better spent on harder tasks that require human expertise.

Advancing Health Equity through AI in Clinical Trials

The use of AI to improve patient recruitment, retention, and outcomes of clinical trials began before the launch of OpenAI's ChatGPT generative AI model. During the COVID-19 pandemic, oncology organizations sought ways to find patients across the country who would qualify for trials, even if they weren't physically present. Through healthcare data, increased AI adoption helped drive decentralized clinical trials and advance health equity and trial diversity.

Jeff Elton, CEO of ConcertAI, a vendor of data and AI SaaS platforms for clinical trial optimization, emphasized the importance of integrated digital trials. "With integrated digital trials, clinical studies are integral to the care process itself, versus being imposed on it," he said. "Trials don't need to place a higher burden on providers and patients than the standard of care."

Reducing Friction in the Clinical Trial Lifecycle

Reducing friction throughout the clinical trial lifecycle is critical to helping patients access trial therapies. Epic, an electronic health record vendor, implemented data-driven clinical trial matchmaking two years ago. Using its de-identified Cosmos data set, Epic allows providers to match clinical trial opportunities from sponsors and count their organization's eligible patients.

Many health systems have also tested using analytical applications that can surface clinical trial opportunities for patients using their organizations' EHR data. In October, Microsoft announced new AI tools that will enable health systems to build their custom AI tools for various administrative needs, including clinical trial matching. However, bias in AI remains a concern for clinical outcomes. Yale School of Medicine researchers have highlighted how bias can surface in any algorithm development pipeline and worsen healthcare disparities.

Conclusion

The research conducted by the National Institutes of Health and their collaborators has shown the potential of AI in clinical trial matching. TrialGPT has demonstrated its ability to save time and improve the efficiency of connecting patients to relevant trials. As AI continues to advance, it holds the promise of further improving healthcare outcomes and addressing the challenges faced in clinical trials.

The Introduction to Medical Spanish Elective stands as a remarkable 10-lecture online course, gracefully offered multiple times annually through the seamless platform of Zoom. Each session lasts a generous 1 hour, providing students with a consistent and immersive learning experience. From 2022 to 2023, an impressive three cohorts of students have successfully completed this course, each embarking on a journey of language acquisition and clinical skill enhancement.

Course Curriculum: A Treasure Trove of Knowledge

This course curriculum is a comprehensive blend of grammar, anatomy, cultural competency, and standardized patient case videos accompanied by Spanish/English scripts. [Refer to Table 1 for a detailed breakdown.] The lecture videos serve as the bedrock of teaching, covering a wide range of common chief complaints such as vaginal discharge, back pain, abdominal pain, headache, chest pain, shortness of breath, and a general pediatric encounter. Prior to each class, students are provided with the lecture video link, allowing them ample time to familiarize themselves with the content. During the lecture, a dynamic interaction takes place as students play both the patient and the doctor roles, watching the video in perfect synchrony. They receive real-time feedback on grammar, pronunciation, and alternative phrases, enabling them to refine their communication skills. These videos adhere strictly to the standard United States Medical Licensure Examination Clinical Skills (USMLE Step 2 CS) vignette format, ensuring their relevance and practicality. Moreover, they are readily available on YouTube, making them easily accessible to all.

Session Timeline and Class Segments

The session timeline of this 1-hour online class is meticulously designed to maximize learning. Each segment plays a crucial role in shaping the students' understanding and proficiency. From grammar drills to cultural insights, every minute is utilized effectively. The variety of topics covered ensures that students are well-prepared to handle a wide array of clinical scenarios.The anatomy section, for instance, provides a detailed exploration of the human body, helping students connect medical terms with their corresponding anatomical structures. Cultural competency lessons delve into the nuances of different Spanish-speaking cultures, enabling students to interact with patients more empathetically. And the standardized patient case videos bring the learning to life, allowing students to practice their skills in a realistic setting.

Student Selection: Opening Doors to Learning

This course is open to all interested medical students, whether they are in the pre-clinical or clinical years, as well as residents. It is promoted through the vibrant St. George’s University School of Medicine (SGUSOM) “Clinical Students” Facebook Page. The initial post about the course provides a clear description and a link to a Google form for prospective students. [Refer to Supplemental Survey 1 for more details.] A remarkable 126 medical students responded to the initial invitation through this Google survey, sharing their availability in the schedule. This valuable information is collected in an excel sheet, which is then used to propose tentative times for the sessions when both instructors and participants are available. [See Supplemental Copy of IMS Reponses_Cohort1_2022 for reference.]The course description explicitly states that certification is not available at this time, emphasizing that participation is entirely voluntary. Students are asked to provide their contact information along with demographic details such as their level of medical education (pre-clinical vs. clinical years). Additionally, they are given the option to select their meeting preference – a fixed schedule (e.g., ‘every Saturday’) or a flexible schedule (e.g., a non-fixed day that suits the majority of students). They can also indicate their preferred days and times of the day for class. The student selection criteria include availability to participate in at least 80% of lectures, a commitment to take the final exam, and a preference for students in their clinical years (as it is assumed that this group aims to enhance their interview skills and patient interaction). Selected students receive an email confirming their acceptance to the course, along with a link to the YouTube channel containing the lecture videos and the tentative meeting days/time(s). The cohort size is limited to 15 students (or fewer) to ensure adequate participation and success.In addition to the Facebook announcement, students doing their rotation at Wyckoff Heights Medical Center in Brooklyn are also informed about the course and given the opportunity to apply via a Google form. Given the prevalent Spanish-speaking population at this medical institution and other SGUSOM clinical affiliate locations, it offers an ideal environment for students to practice medical Spanish.

Comfort Measures: Nurturing Confidence

During the pre-course and post-course surveys, students are asked about their comfort level when taking a patient's history and chief concern in Spanish using appropriate medical terminology. [Refer to Supplemental Survey 2 and 3 for details.] The comfort score level is measured on a scale from 0 to 5, where 0 indicates an inability to take the patient's history and 5 represents a high level of comfort. This measure helps students track their progress and identify areas where they need further improvement.

Final Exam Evaluation: Measuring Success

Only students who achieve an 80% participation attendance are eligible to take the exam. Those who meet the eligibility criteria are notified and encouraged to participate voluntarily. The final exam consists of a 20-minute patient encounter conducted via Zoom, with the course instructor playing the role of the patient and the student assuming the role of the physician. A standardized checklist [Refer to Supplemental Table 2] is used to evaluate the participant's performance. The questions are divided into subcategories, and each question is awarded 3 points, resulting in a total of 63 potential points. This total is then divided by 21 (the total number of questions), and a passing score ratio of 2 (equivalent to a 66.7% passing score) is set. This rigorous evaluation system ensures that students are truly prepared to handle real-life clinical situations.