Macey Leigh Levan

Background: Qualitative research is critical in transplantation science for understanding stakeholder perspectives and informing implementation, ethics, and operations. However, qualitative analysis is time-intensive and requires multiple trained team members. Large language models (LLMs) such as ChatGPT have emerged as potential tools for qualitative analysis, but their performance relative to trained human coders remains unclear. The study compared LLM-generated and human thematic analyses across three transplant studies. Methods: The Naeem et al. six-step thematic analysis process was applied to three existing Datasets: (1) semi-structured interviews with United States (US) Organ Procurement Organization executives about normothermic regional perfusion (NRP) practices and ethical concerns; (2) open-ended survey responses from parents of pediatric heart transplant recipients addressing daily concerns, quality of life, and transition to adult care; and (3) semi-structured interviews with infectious disease providers exploring attitudes toward living donation from donors with human immunodeficiency virus (HIV) to recipients with HIV. These datasets had previously been analyzed by humans trained in traditional qualitative analysis. A Health Insurance Portability and Accountability Act (HIPAA)-compliant instance of OpenAI’s Generative Pre-trained Transformer 5.4.(GPT-5.4) generated thematic outputs for comparison. Results: Across all three studies, LLM-Generated themes largely overlapped with those identified by humans. However, interpretive depth and contextualization of findings still required human expertise. Conclusions: GPT-5.4 generated themes that showed conceptual overlap with previously completed human-led analyses across three transplantation datasets, supporting a hybrid model in which AI may assist with early analytic tasks while human researchers provide interpretation and rigor.

Objective: This study evaluated temporal trends and patient characteristics associated with glucagon‐like peptide‐1 (GLP‐1) receptor agonist prescribing among US adults without an apparent FDA‐approved indication. Methods: This retrospective cohort study used Cosmos electronic health records (01/2021–12/2025). Adults prescribed liraglutide, semaglutide, or tirzepatide without a documented indication (type 2 diabetes or obesity with qualifying comorbidities) were included. Patients without a recorded weight ≥ 6 months prior to first prescription were excluded. Demographic and clinical characteristics were compared using t ‐tests and χ ² tests. Results: Among 92,415,648 adults without an apparent FDA‐approved indication, 1,133,953 (1.2%) were prescribed GLP‐1 receptor agonists. Prescribing increased from 0.1% in 2021 to 1.5% in 2025 (15‐fold). Recipients had higher median BMI (25.9 vs. 24.5 kg/m ² ) and were more often female (85.5%) and White (60.0%). Hypertension and dyslipidemia were more common, while coronary artery disease and heart failure were less prevalent. Notably, 35.1% had normal BMI; eating disorders were more frequent (1.8% vs. 0.3%). GLP‐1 receptor agonist use was associated with lower social vulnerability and private insurance. Conclusions: GLP‐1 receptor agonist prescribing without an apparent FDA‐approved indication has increased rapidly and disproportionately involves socioeconomically advantaged populations, including individuals without obesity. The risk–benefit profile in these populations remains uncertain.

Background: The Increasing Organ Transplant Access (IOTA) model is a mandatory experimental payment model that aims to improve kidney transplant access using financial incentives and penalties for transplant centers randomized to model participation, based primarily on growth in transplant volume. Methods: We used national transplant registry data to conduct a retrospective cohort study of US kidney transplant centers. We identified centers randomized to IOTA participation based on the published list of center participants. Using a difference-in-differences approach, we compared IOTA participants with nonparticipants on changes in overall, deceased donor, and living donor kidney transplant volume per 30 days in the period before IOTA randomization status was announced versus the 9 months following IOTA implementation. Separately among IOTA participants and among nonparticipants, we compared the proportional change in transplant volume when stratifying centers by pre-IOTA transplant volume, organ offer acceptance rate, and transplant rate ratio. Results: Among 177 included centers, 97 (55%) were IOTA participants and 80 (45%) nonparticipants. Most IOTA participants (53%) and nonparticipants (55%) experienced a fall in overall transplant volume after model implementation. However, IOTA participants exhibited a 5% increase in transplant volume per 30 days, whereas nonparticipants exhibited a 3% decrease ( P = 0.01). The difference between groups was seen for deceased donor transplants (participants +5% versus nonparticipants −9%, P < 0.001), while living donor transplants rose in both groups (+7 versus +19%, P = 0.24). Among IOTA participants, there were no statistically significant differences in changes in transplant volume when comparing centers categorized by pre-IOTA transplant volume, organ offer acceptance rate, or transplant rate ratio. Conclusions: Centers randomized to IOTA participation outperformed model nonparticipants in changes in kidney transplant volume after model implementation, driven by an increase in deceased donor transplant volume among model participants but a fall among nonparticipants.