Andrew Gregg, MD, PhD

Investigating the effects of colonoscopy in patients with inflammatory bowel disease: Are we prepping for complications?
Fellow: Andrew Gregg, MD, PhD
Mentors: Berkeley Limketkai, MD, PhD; Jonathan P. Jacobs, MD, PhD

We hypothesize that patients with inflammatory bowel disease (IBD) undergoing bowel preparation for colonoscopy are at increased risk of worsening IBD related symptoms due to alterations in gut microbiome composition. In this study we will enroll patients with active IBD (UC or CD), who are scheduled to undergo colonoscopy. We plan to collect stool samples before bowel prep, and 2 weeks after completion of colonoscopy, and assess for changes in microbiota composition. These findings will be correlated with symptoms before and after colonoscopy using patient reported simple clinical colitis activity index for patients with UC, and self-reported Harvey-Bradshaw Index for patients with CD.


Soonwook Hong, MD

Analyzing performance and accuracy of foundational large language models in the clinical documentation of patients with inflammatory bowel disease
Fellow: Soonwook Hong, MD
Mentor: Berkeley Limketkai, MD, PhD

Foundational large language models (LLMs) are a novel class of artificial intelligence (AI) models designed for natural language processing (NLP) tasks. There is significant interest in applying LLMs to a broad range of clinically relevant tasks such as medical documentation, responding to patient inquiries, differential diagnosis of medical conditions, medical education, and research. However, there are broad concerns about LLM safety and reliability, such as “hallucination,” a phenomenon in which they generate seemingly coherent yet inaccurate content, making their use in sensitive applications such as healthcare uncertain. Furthermore, there are also concerns with data safety and privacy with commercially available LLMs such as OpenAI’s ChatGPT suite. Recently, open source LLMs such as Llama 2 have been released, which are available for public use and research with no collection of data by outside agencies. In fact, these LLMs can even be implemented on local air-gapped computers disconnected from the Internet. There is a significant lack of research on the behavior and accuracy of LLMs in medically relevant applications. Hence, we propose to study the behavior of LLMs and evaluate their accuracy and reliability when applied to medical data from patients with gastrointestinal disease.


Philip Kozan, MD, MS

Impact and outcomes of endoscopic bariatric therapy in the Veterans Health Administration
Fellow: Philip Kozan, MD
Mentor: Jennifer Kolb, MD, MS

Obesity is a growing public health concern and is associated with important medical comorbidities including cardiovascular disease and diabetes. Treatment of obesity can prevent the associated morbidities and mortality, however effectiveness is limited in current practice largely due to lack of success with diet and lifestyle changes and a relatively small proportion of eligible patients undergoing bariatric surgery. Endoscopic bariatric therapies (EBT) such as gastric remodeling procedures provide an additional option for weight loss among patients who either do not want or are not candidates for bariatric surgery. EBTs were recently introduced at the West LA VA and therefore the goal of this project is to contribute knowledge and insight from our experience in a VA setting. We will evaluate outcomes locally and compare these to national benchmarks for success in EBT.


Selena Zhou, MD

Determining why Asian Americans do not participate in mailed FIT outreach for colorectal cancer screening
Fellow: Selena Zhou, MD
Mentor: Folasade P. May, MD, PhD, MPhil

Data show that Asian Americans have one of the lowest percentages of up-to-date colorectal cancer (CRC) screening in the United States. There have been many studies evaluating different interventions to improve the CRC screening rates in the Asian American population, including mailed fecal immunochemical test (FIT) kits. The University of California, Los Angeles’ (UCLA’s) mailed FIT outreach program mails FIT kits to all patients seen in primary care clinics who are overdue for CRC screening. Despite this intervention, 48.3% of Asians (8,786 individual patients) seen in 2022 were still overdue for their colorectal cancer screening. We plan to administer a survey to Asian American patients who were mailed a FIT kit for CRC screening but did not complete it to identify the reasons why they did not return it with the goal of tailoring and improving future interventions.