Citations

Abstract Background Personalized meal planning by registered dietitian nutritionists (RDNs) is time‐intensive. Large language models (LLMs) may automate drafting meal plans, but their nutritional accuracy in clinical practice is uncertain. Methods In this proof‐of‐concept study, five outpatient RDNs and four LLMs (Gemini, CoPilot, ChatGPT 4.0, and customized ChatGPT 4.0) each generated 3‐day meal plans for five validated clinical scenarios. Effectiveness was defined as accuracy in meeting pre‐specified energy, protein, carbohydrate, fat, and sodium targets. Time to create plans and RDN comfort (self‐rated confidence in nutritional accuracy and clinical appropriateness on 1–5 Likert scale) were recorded. Three independent RDNs, blinded to source, analyzed nutrient content using Nutritionist Pro. Group differences were assessed with t ‐test and ANOVA. Results All LLMs and RDNs produced feasible meal plans. LLMs generated meal plans in under 1 min, whereas RDNs required a mean of 44 min per scenario. RDNs reported comfort levels ranging from 3.8 to 4.8. Across most scenarios, LLM plans delivered a smaller proportion of requested energy than RDN plans, which more consistently approached energy targets. Both groups performed similarly for the Mediterranean diet scenario. Overall, protein accuracy did not differ. However, in chronic kidney disease, LLMs undershot the guideline‐based protein target, while RDNs tended to modestly exceed it. Accuracy for low‐carbohydrate, fat, and sodium diets was comparable. Conclusion LLMs can rapidly generate clinically plausible meal plans but are less reliable than RDNs in achieving prescribed energy and selected macronutrient goals. Prompt precision is essential for nutrient‐specific targets. A hybrid model in which RDNs refine LLM‐generated drafts may leverage efficiency without sacrificing clinical accuracy.

Introduction/Objective Opportunistic salpingectomy (OS) has been shown to reduce high-grade serous ovarian cancer (HGSOC) by 42-80%, yet awareness of this strategy outside of obstetrics and gynecology remains limited. The objective of this study was to assess awareness and counseling practices regarding OS for HGSOC prevention among primary care providers (PCPs). Methods We conducted a cross-sectional survey study using a 6-item electronic survey evaluating PCP knowledge and counseling practices related to HGSOC prevention through OS. The study was conducted at a single tertiary care center with gynecologic oncology services. Survey items assessed awareness of HGSOC origin, awareness of OS as a preventative strategy, counseling frequency in the setting of abdominal or pelvic surgery, counseling comfort, and resources needed to increase confidence in recommending OS. Response formats included Likert scales and Yes/No/Unsure options. The survey was distributed by email to attending physicians and advanced practice providers (APPs) in the Department of Family Medicine (FM) and Internal Medicine (IM). Results Of 398 providers surveyed, 136 responded (34% overall response rate; FM 44%, IM 28%), including 96 attending physicians and 40 APPs. Most PCPs were unaware that HGSOC originates in the fallopian tubes (60%) and were unaware of OS as a preventive strategy (71%). Counseling was uncommon: 11% had ever counseled on OS, while 74% had not and 15% reported it was outside their scope. Most respondents (63%) reported being not at all comfortable counseling patients about OS. Providers identified clinical guidelines (60%), CME or training modules (50%), eligibility criteria (48%), gynecology collaboration pathways (48%), and patient-friendly materials (44%) as key supports to improve counseling confidence. Conclusion PCP awareness and counseling regarding OS are limited. Given that guideline support for OS has solidified only in recent years, these gaps are not surprising—and they represent a clear, actionable opportunity. PCPs are well-positioned to lead the expansion of ovarian cancer prevention beyond specialty care with the right tools and support.

Home enteral nutrition (HEN) is essential for patients with a functioning gastrointestinal tract who cannot meet nutritional needs orally, yet outpatient management remains complex, resource-intensive, and supported by a limited evidence base. Artificial intelligence (AI) may augment HEN care by extending monitoring, education, risk assessment, documentation support, and operational coordination into the home environment. Consistent with the narrative review format, this article uses a pragmatic, transparent synthesis of influential HEN-specific literature, relevant clinical nutrition evidence, and background knowledge from adjacent fields, including home healthcare, chronic disease management, oncology nutrition, telehealth, and software regulation. Direct evidence in established HEN populations remains scarce; therefore, most AI applications should be considered hypotheses or early implementation opportunities rather than proven standards of care. The strongest near-term opportunities are clinician-supervised patient education, symptom triage, adherence support, remote monitoring, and workflow automation. Predictive analytics, smart pumps, and precision enteral prescription tools are promising but require prospective HEN-specific validation, interoperability with electronic health records and home-infusion systems, reimbursement pathways, and governance safeguards. Key barriers include dataset bias, limited external validation, alert fatigue, privacy and regulatory concerns, unclear accountability, digital equity, cost uncertainty, and the risk of dehumanizing care. AI should be viewed as a complement to multidisciplinary HEN expertise. Priorities for the near future include HEN registries, standardized outcomes, prospective validation, pragmatic implementation trials, health-economic evaluation, and transparent oversight that preserves clinician accountability and patient-centered care.

Objectives Long-term symptoms following infection with acute respiratory syndrome coronavirus-2 (SARS-CoV-2), termed post-acute sequelae of SARS-CoV-2 infection (PASC) or Long COVID (LC), persist beyond 90 days and may impact at least 10% of patients with acute infection. The exact cause of LC remains unknown, but several theories include persistent viral particles, immune dysfunction, and autonomic nervous system (ANS) imbalance, particularly involving the vagus nerve. Given the role of the vagus nerve in inflammation regulation, vagal nerve stimulation (VNS) is being explored as a treatment option. The present study aimed to evaluate the feasibility, safety, and exploratory effects of a handheld external transcutaneous cervical non-invasive VNS (tcVNS) on the primary endpoint of Post-COVID Functional Status (PCFS) grade, alongside secondary outcomes such as LC-related fatigue and other symptoms. Methods Participants from our Post COVID Care Clinic were randomized into treatment (tcVNS treatments two minutes, three times a day) and control (no VNS) groups for a study period of 12 weeks. Participants completed validated questionnaires, PET-CT scans, and inflammatory cytokines at baseline and 12 weeks. Results All randomized patients (8 tcVNS, 10 control) completed the 12-week study. Participants randomized to tcVNS were observed to have improvements in PHQ-9, PROMIS-F, and COMPASS-31 when compared to controls, but these differences were not statistically significant. Cortisol and DHEA were also observed to increase less in the tcVNS group when compared to the control group, but these changes were also not statistically significant. Conclusions tcVNS was well tolerated and may be a promising therapeutic approach for LC, a chronic disease with no approved treatments and very low rates of complete recovery.

Conventional cigarette smoking remains a leading preventable cause of morbidity and mortality despite declining prevalence, yet evidence-based treatments remain underused. This narrative review examines guideline-supported therapies and emerging approaches, including digital tools, neuromodulation, and harm-reduction strategies. Standard care typically combines FDA-approved pharmacotherapy with behavioral counseling. Digital interventions can extend access and support adherence, particularly when integrated with pharmacotherapy and follow-up. Repetitive transcranial magnetic stimulation (rTMS) can be considered as an adjunctive option for patients who do not respond to pharmacotherapy and is FDA-approved. For patients unable to quit, structured harm-reduction strategies may help reduce risk while supporting progression toward eventual abstinence. These approaches should be used alongside, not in place of, guideline-recommended pharmacotherapy and behavioral counseling, which remain the gold standard for care. This review outlines a pragmatic, primary care–oriented model that integrates stepped-care with harm-reduction strategies.

Primary care is a language-intensive environment where clinicians frequently synthesize, document, and communicate large amounts of information. Given the amount of information in primary care, large language models (LLMs) have emerged as tools to support primary care clinician workflows, aiming to reduce administrative burden and augment patient care. LLMs are artificial intelligence tools that can generate text and assist with chart summarization, message drafting, and clinical decision support. However, because LLMs are probabilistic, identical inputs may yield inconsistent responses. The LLMs output can be significantly improved through prompt engineering. Prompt engineering is the intentional design of text entered into the LLM to help produce better LLM results. Prompting is best approached as an iterative process where a clinician defines the task, the LLM generates output, and the clinician critically reviews and refines the results. Low-quality prompts are often vague and produce generic or misleading responses. High-quality prompts are specific, context-rich, and well-structured. Using structured prompting frameworks can improve the LLMs response. Including core elements, such as the LLMs role, the clinical context, the requested task, and the desired output formatting, can help improve the LLMs response by providing situational background and framing response formatting. Additional frameworks, such as Ask-Context-Expectation and PICO+O, can be utilized for specific patient circumstances and cases. However, despite their promise, LLMs have important limitations. They may generate inaccurate or hallucinated information and produce inconsistent results. Given these limitations, LLMs should not be viewed as knowledge authorities but rather as cognitive assistants. Thoughtful prompt design can help reduce these limitations and improve response accuracy, always keeping in mind that LLMs should not replace human judgment or clinical reasoning. As LLMs continue to be integrated into primary care, prompt engineering will become an increasingly important clinical skill. Careful prompt design and critical appraisal of LLM responses will be key to maintaining excellent patient care in the AI era.

Objective To evaluate patient satisfaction with the experience of using an autonomous wheelchair to transport patients in a large outpatient clinical environment. Methods The autonomous wheelchair pilot was approved as a feasibility pilot by the institutional committees and deemed a quality improvement project by the Institutional Review Board (IRB). A total of 409 adult patients using an autonomous wheelchair at a large academic medical center who volunteered to complete a paper survey were included. The survey was administered immediately after autonomous wheelchair use, using a cross-sectional, anonymous survey, between 15 Oct 2025 and 14 Jan 2026. Of 409 completed surveys, six were excluded because participants did not identify their endpoint for stratification purposes. Descriptive analysis included frequencies and percentages of responses. Results No collisions or adverse events were observed during the pilot, and the system operated reliably within the predefined routes. Most survey respondents were first-time users (335/402 [83.3%]). A majority reported they would use the autonomous wheelchair again (341/395 [86.3%]) and would recommend it to others (364/397 [91.7%]). Overall, the experience was rated better than expected by 293 of 393 participants (74.6%). When given a choice, 271 of 379 respondents (71.5%) preferred the autonomous wheelchair over a staff-operated wheelchair. Conclusion These findings suggest that autonomous wheelchairs are feasible and acceptable to patients in a controlled outpatient setting and support continued piloting and prospective evaluation.

Objective Long COVID (LC) refers to the long-term symptoms occurring after SARS-CoV-2 infection and resolution of the initial disease prodrome. While LC is increasingly recognized, knowledge about its biomarkers is still limited. This study examines imaging findings in 18 F-FDG PET-CT scans in 40 patients with LC from an academic LC Clinic, for potential imaging biomarkers. Methods This study included patients aged 18 and older with confirmed positive SARS-CoV-2 PCR test and at least 28 days post-symptom onset. 18 F-FDG PET-CT scans were performed, and brain metabolic activity was compared to a control database to generate Z-scores. Participants were grouped based on their clinical phenotype and Z-scores of different brain regions were analyzed using t-tests. Results The study group was mainly female (70%), with a median age of 53 years, and predominantly non-Hispanic White (90%). Most had pre-existing conditions such as gastrointestinal, cardiovascular, endocrine, and psychiatric disorders. The findings showed significant cerebral hypometabolism in 29 patients with fatigue and post-exertional malaise (PEM), particularly in the left sensorimotor cortex (p=0.0253) and bilateral primary visual cortex (right: p=0.0096, left: p=0.0016), which persisted up to two years after infection. Conclusions In summary, this study identified persistent cerebral hypometabolism in LC patients, especially those with fatigue with PEM, up to two years post-infection. These results suggest that 18 F-FDG PET-CT could be a valuable tool for diagnosing and managing LC. Further research is essential to confirm these findings and improve treatment strategies for patients with LC.

10578 Background: Multicancer detection (MCD) testing using cell-free DNA methylation patterns is a novel blood-based approach to cancer screening. In 2022, Mayo Clinic implemented offering MCD testing into routine clinical practice. Prior real-world reports have been limited to single-site experiences. We evaluated enterprise-wide implementation of MCD testing across all three Mayo Clinic sites using the Mayo Clinic Platform. Methods: We performed a retrospective analysis of de-identified clinical data from Mayo Clinic Platform Discover, including GRAIL Galleri tests performed at Mayo Clinic sites in Rochester, MN; Scottsdale, AZ; and Jacksonville, FL. This activity did not require IRB review per 45 CFR 46.102. Demographics, test results, cancer diagnoses, and prior cancer history were extracted using a privacy-preserving protocol, with data de-identified by expert determination. ICD-10 diagnostic codes were used to identify subsequent cancer diagnoses. Results: A total of 8,201 Galleri tests were performed in 7,300 patients; some patients underwent repeat testing. Fifty-seven tests (0.7%) were positive. Among MCD-positive patients, cancer diagnostic codes were identified in 32 patients, yielding a positive predictive value (PPV) of 58.7%. The most common cancers identified were oropharyngeal (34.4%), lymphoid and hematopoietic (34.4%), and gastrointestinal malignancies (18.8%).Among 8,144 MCD-negative tests (99.3% of all testing performed), 423 patients (5.2%) subsequently received a cancer diagnostic code, yielding a negative predictive value of 92.5%. The most commonly identified cancers in this group were male genital organs (36.6%), urinary tract (10.6%), and thyroid or other endocrine glands (7.3%). Overall, 10.1% of patients had a prior cancer history, which was more common in MCD-positive than MCD-negative patients (22.6% vs 10.0%). Patients with positive tests were older (mean age 68.4 vs 61.5 years) and more often male (67% vs 62%). Race distribution was similar. Conclusions: Enterprise-wide implementation of MCD testing across diverse clinical settings is feasible within routine workflows. Over half of positive tests were associated with cancer diagnostic codes, supporting the clinical promise of MCD testing. Limitations include incomplete capture of externally diagnosed cancers and lack of cancer signal-of-origin data. The Mayo Clinic Platform enables scalable, secure analytics to support future cancer interception strategies.