Objective To develop a data science pipeline for data extraction and collection to conduct an observational study using institutional medical informatics and AI tools.Patients and Methods: skin excision cases exposed to intra-incisional clindamycin (from February 5 to October 14, 2025) and non-exposed (from January 1 to December 31, 2022) were defined as our cohorts of interest. The outcome of our study was surgical site infection (SSI) 30-days following surgery. An LLM-supported data science pipeline was used for cohort identification, case screening, and extraction of procedural information. LLM screening and data extraction were validated in 300 random cases. A GEE model was used to analyze the effect of intra-incisional clindamycin on SSI. Results The LLM achieved high accuracy for screening cases for procedures done in the head (accuracy: 99.0%, 95%CI:[97.1-99.8%]) and skin excisions (97.0%, 95%CI:[94.4-98.6%]). Additionally, it achieved remarkable performance for extraction of clinical information, with the highest accuracy observed for extraction of anatomical location associated with the procedure (accuracy: 100.0%, 95%CI:[98.8-100.0%]). The final dataset included 2247 skin excision cases (594 exposed; 1653 non-exposed) from 1923 patients. Cases that received intra-incisional clindamycin had lower odds of SSI, however, this association was not statistically significant (aOR: 0.57, 95%CI:[0.31-1.07]). Conclusion The implementation of medical informatics tools and robust AI algorithms will enable the conduct of end-to-end research pipelines. However, it is important to always consider humans in the loop, as oversight ensures reliability and accuracy, and robustness.
Objective To test whether AI-generated, clinician-edited after-visit summaries (AVS) are more patient-friendly than clinician-generated AVS, without compromising safety, when implemented in a live inpatient setting. Patients and Methods Patients being discharged from inpatient medicine units between January 22, 2025 and May 23, 2025 were randomized to the intervention arm, which included AI-generated, clinician-edited AVS, or the control arm, which included usual care of clinician-only AVS. Outcomes include text between 6th and 8th grade reading levels, and presence of a “simplified” description of both the diagnosis and treatment/interventions. Our composite outcome was positive if all three of these outcomes were positive. We also surveyed patients, nurses, and clinicians on perceptions of safety, empathy, and understandability. Results Among 61 patients enrolled in the trial, the AI-generated, clinician-edited AVS were significantly more likely to have a mean reading level closer to 6th – 8th grade than the control group (8.5 vs 10.6, p < 0.001). The AI-generated, clinician-edited AVS were also more likely to contain simplified explanation of diagnoses (96.7% versus 12.9%, p < 0.001), and hospital interventions (86.7% versus 12.9%, p < 0.001). Our composite outcome favored AI-generated, clinician-edited summaries (13.3% vs 6.5%, p = 0.637). Conclusion In this small, randomized trial, exploratory and component measures suggest improved patient-friendliness in AI-generated, clinician-edited AVS, although no significant difference was observed in the primary outcome.
Objective To characterize methodological features, intervention characteristics, and outcome trends of medical extended reality (MXR) applications for perioperative patients. Patients and Methods This scoping review was conducted in accordance with the JBI Manual for Evidence Synthesis. Searches were conducted in January 2025 and updated on August 12, 2025, across MEDLINE, Embase, Scopus, IEEE, Google Scholar, Cochrane Library, arXiv, and ProQuest Dissertations and Theses. Primary studies evaluating immersive MXR interventions in perioperative patients were included. Results Of 5883 records identified, 190 studies published between 2012 and 2026 were included. Most studies involved adult populations (69.5%), used virtual reality (95.8%), and delivered interventions preoperatively, most often targeting anxiety or pain. Studies were predominantly randomized controlled trials evaluating efficacy (48.4%), with few early developmental studies incorporating user-centered design (2.6%). Early-stage feasibility studies (47.9%) showed substantial methodological and outcome heterogeneity, with patient satisfaction (25.3%) and feasibility (18.4%) infrequently prioritized as primary outcomes. Studies commonly evaluated multiple intervention mechanisms, such as distraction and education, within single designs (35.3%), and some delivered interventions across multiple perioperative phases (10.5%). Few studies included MXR comparator conditions (6.8%). Conclusion Perioperative MXR shows promising signals of efficacy and feasibility based on reported study outcomes; however, these favourable trends should be interpreted cautiously given possible publication bias and the absence of a formal risk-of-bias assessment. Limited use of user-centered design, inconsistent and non-validated measurement of feasibility outcomes, and mechanistic ambiguity may constrain translation. Greater alignment with staged development frameworks and more rigorous feasibility evaluation are needed to support implementation. Pre-registration: https://osf.io/nkwfy/wiki?wiki=yqanp
Objective To systematically review the development, validation, and performance of traditional and machine learning (ML) prediction models for periprosthetic joint infection (PJI) risk, diagnosis, and clinical outcomes. Methods We conducted a systematic review according to PRISMA guidelines to identify studies that developed or validated multivariable prediction models for PJI using traditional statistical or ML approaches. Searches were performed on December 10, 2025, across Ovid Embase, MEDLINE, Scopus, Web of Science, ClinicalTrials.gov, and CENTRAL. Results Among 4,565 identified records, 34 studies met inclusion criteria. Reported discrimination varied across model types and clinical endpoints. Among traditional risk prediction models, 54.5% demonstrated acceptable/good discrimination (0.7 ≤ AUROC < 0.8), and 27.3% demonstrated excellent discrimination (0.8 ≤ AUROC < 0.9). Traditional diagnostic and outcome prediction models showed good to outstanding discrimination. Among ML models, 50.0% of risk models demonstrated 0.7 ≤ AUROC < 0.8, while 60.0% of diagnostic models reported AUROC ≥ 0.9. However, calibration reporting was inconsistent, particularly among ML studies, and external validation was limited. Model performance generally decreased in external validation cohorts compared with derivation cohorts. Substantial heterogeneity among included studies and model designs precluded direct cross-model comparisons; therefore, differences in AUROC should not be interpreted as evidence of superiority between approaches. Conclusion Prediction models for PJI demonstrate variable performance across risk prediction, diagnosis, and outcome prediction settings. However, substantial clinical and methodological heterogeneity, limited calibration reporting, and inadequate external validation restrict assessment of model reliability and generalizability. Current evidence does not support conclusions regarding the superiority between approaches.
Objective To evaluate the feasibility, usability, and acceptability of a conversational artificial intelligence (CAI) agent designed to conduct pre-visit patient intake in a cardiovascular outpatient clinic. Patients and Methods This prospective quality-improvement pilot study was conducted from July to September 2025 at a tertiary center. Adult patients attending an outpatient arrhythmia clinic were invited to interact with an embodied CAI agent prior to clinician review. The agent guided patients through protocol-driven dialogue, dynamically adjusting questions collect medical history, treatment adherence, and current concerns, and generated a structured summary. Participants completed a survey assessing usability, comprehension, trust, communication impact, and intention to reuse. Participating clinicians also completed a feedback survey. Responses were recorded on 5-point Likert scales and summarized descriptively. Results Of 240 invited patients, 235 (97.9%) completed the CAI session and post-interaction survey. Majority found the system easy to use (93%) and understand (91%), most (82%) expressed comfort sharing health information. Most patients (79%) reported improvement in the quality of the patient–provider discussion on decision-making and majority (82%) indicated intention to reuse. All participating clinicians reported favorable impressions with no safety concerns noted, and estimated time savings between 1-10 minutes per visit. Conclusions Implementation of a CAI agent for outpatient intake was feasible and well received by both patients and clinicians, demonstrating strong usability, trust, and communication benefits. These results suggest that CAI may meaningfully enhance patient engagement and streamline clinical workflows, supporting broader adoption of these systems.
Objective To examine primary care (PC) and mental health (MH) telehealth utilization and satisfaction across nine high-income countries and identify associated demographic and clinical factors. Patients and Methods We conducted a cross-sectional analysis of the Commonwealth Fund International Health Policy Survey, which was fielded from March through August 2023 and surveyed nationally representative samples of adults aged 18 years or older in Australia, Canada, France, Germany, the Netherlands, New Zealand, Sweden, Switzerland, the United Kingdom, and the United States. Analysis from December 2024 to November 2025 included 17,160 of 21,341 respondents after exclusions. Survey-weighted multivariable logistic regression assessed PC and MH telehealth use and satisfaction. Results Overall, an estimated 25.6% used PC telehealth and 9.0% used MH telehealth (weighted estimates). PC telehealth use ranged from 6.6% (Germany) to 39.7% (Australia) and was associated with female gender (adjusted odds ratio [aOR], 1.47; 95% CI, 1.34-1.63), higher education (aOR, 1.56; 95% CI, 1.40-1.74), above-average income (aOR, 1.28; 95% CI, 1.15-1.43), and chronic conditions (aOR, 1.75; 95% CI, 1.55-1.97). MH telehealth use decreased with age (aOR for ≥65 vs 18-24 years, 0.23; 95% CI, 0.16-0.33) but was not associated with income. Among users, 80.4% were satisfied with PC and 77.6% with MH telehealth. Older adults (≥65 years) using MH telehealth reported higher satisfaction (aOR, 2.79; 95% CI, 1.20-6.47). Conclusion Telehealth adoption varied substantially by country and service type. Income disparities in PC telehealth and low adoption but high satisfaction among older MH telehealth users suggest that equity-focused and age-friendly implementation strategies are needed.
Objective To characterize real-world body composition and blood pressure changes following glucagon-like peptide (GLP)-1 receptor agonist (RA) initiation and identify predictors of weight loss quality using connected device data. Patients and Methods In this retrospective cohort study, adult US users completing an in-application GLP-1 RA survey (June 8, 2023, to July 4, 2024) were followed up using longitudinal device data (December 2021 to September 2025). We used a distributional causal inference framework to match GLP-1 RA users to nonusers. Primary estimands were the distributional average treatment effect on the treated and median differences over months 10 to 14. Results Among 2031 users for body composition (396 exposed) and 672 for blood pressure (BP; 148 exposed), GLP-1 RA initiation was associated with a 9.8% median weight reduction (95% CI, 9.8-9.9), driven by fat mass loss (7.5%; 95% CI, 7.5-7.8) with smaller muscle mass loss (2.3%; 95% CI, 2.1-2.4). Sustained reductions were observed in systolic (2.5 mm Hg; 95% CI, 1.5-2.9) and diastolic BP (1.5 mm Hg; 95% CI, 1.2-1.9). Baseline muscle-to-fat ratio (MFR) was the strongest predictor of weight loss quality, with lower MFR individuals losing more fat and sparing more muscle. Conclusion In a real-world setting, GLP-1 RA initiation was associated with clinically significant fat-predominant weight loss and sustained BP improvements. Baseline MFR predicted weight loss quality, suggesting that pretreatment body composition may help clinicians anticipate individual responses—a step toward precision medicine in GLP-1 RA therapy.
Objective To determine if daily physical performance measures collected remotely through a wrist-worn activity monitor (AM) could serve as a reliable alternative to the clinic-based 6-minute walk test (6MWT) for assessing physical capacity after critical illness. Patients and Methods Participants were recruited from a group of critically ill adults aged ≥18 years with moderate-to-severe anemia as they enrolled in a parallel-group, single-center randomized clinical trial assessing the impact of a multifaceted anemia prevention and treatment strategy vs standard care at intensive care units at a large US medical center. The nested prospective observational study was conducted from September 2022 to October 2025. Participants were instructed to wear a wrist-worn AM for 4 days in their free-living environments at 1 month and 3 months post-hospitalization to coincide with their clinic-based 6MWTs. Mean daily step counts and maximum daily cadences were calculated from the AM. Results Fifty-two participants (median [interquartile range] age, 66 [60-71]; 29 men) were enrolled. Over 80% of participants completed monitoring and provided valid data. Mean daily step count (rs=0.557, P = 0.001) and mean maximum daily cadence (rs=0.830, P < 0.001) were positively correlated with 6MWT distance. There were no marked differences between the trial’s intervention groups. Conclusion Home-based wrist-worn AM use is feasible after critical illness with high retention and adherence rates. The strong correlation between daily maximum cadence and 6MWT distance suggests this device can effectively evaluate physical performance and recovery in intensive care unit survivors remotely. These findings provide valuable preliminary data to guide the design of large clinical trials. Trial Registration ClinicalTrials.gov Identifier: NCT05167734 (2021-12-09)
Holographic transmission represents a transformative convergence of advanced imaging, augmented reality, and extended reality technologies for real-time education and medical care. We reviewed the current state of holographic transmission, emphasizing its applications in intraoperative guidance, telementoring, teleteaching, and telesupervision. By enabling real-time, life-sized 3-dimensional interaction across distances, holographic systems can democratize access to expertise, especially in underserved or remote areas. We analyzed technical enablers, including high-speed digital cameras, CoaXPress data transmission, graphics processing unit-based reconstruction, and 5G low-latency networks for their role in achieving near real-time fidelity. Two proof of concept demonstrations validate the technology's feasibility: a transcontinental surgical telementoring session between Boston, MA, and São Paulo, Brazil, and a large-scale educational holographic transmission connecting Mayo Clinic (United States) and the G7 Summit (Brazil). Both achieved seamless, bidirectional volumetric communication without perceptible delay, showcasing immersive copresence for clinical and educational collaboration. Despite these advances, barriers persist, most notably high implementation costs, latency sensitivity, limited field of view, and the digital divide that restricts broadband and electrical infrastructure in low-resource regions. Ethical issues surrounding privacy, data ownership, and equitable access also require structured governance. Our discussion highlights holographic telemedicine's potential to establish a borderless educational landscape where knowledge flows freely and health care equity is strengthened globally. Continued innovation, investment in infrastructure, and development of ethical and regulatory frameworks are necessary to create truly global, real-time, and immersive health care without borders.
Access to high-quality primary and specialty care remains a persistent challenge in the United States, with millions of individuals facing barriers due to geography, provider shortages, and health inequities. As health care systems seek innovative solutions, virtual care providers have emerged as critical partners in expanding access, enhancing continuity, and supporting transitions across the health care continuum. This manuscript explores the integration of virtual care providers within an academic medical center, highlighting the dual roles of collaborative and referral models in bridging gaps between traditional outpatient care and virtual health services. Drawing on practical experience, we identify core design principles (acute access, continuity, convenience, cost, reach, and transitions) that underpin successful partnerships and sustainable adoption. These insights underscore the importance of aligning virtual care provider relationships with existing organizational culture, secure data sharing, and supportive reimbursement structures. As academic medical centers and health systems nationwide navigate the evolving landscape of virtual health, thoughtful integration of virtual care providers offers a pathway to address disparities, improve patient outcomes, and advance the quality and reach of care delivery.
Objective:To use a motion capture technique in a longitudinal study design to study movement performance in older women and it's association with incipient dementia. Patients and Methods:This study is part of the Gothenburg H70 birth cohort studies and the prospective population study of women in Gothenburg. Altogether 692 nondemented women aged 62-84 years were followed for 20 years. Movement performance was measured by the Posturo-Locomotion-Manual test using optoelectronic technology. Dementia diagnoses were based on information from neuropsychiatric examinations, informant interviews, hospital records, and registry data. Results:Total 170 women developed dementia during the follow-up. Slower gait at baseline (1992/1993) was associated with higher risks of dementia (1992-2012). Individuals in the second and third tertiles of the locomotor phase were associated with dementia. Conclusion:Our findings suggest that motor impairment is an important symptom of preclinical dementia in women.
Objective: To validate the United Kingdom Conformity Assessed-marked PinPoint blood tests, which use machine learning models and routinely available blood analytes to estimate cancer risk in adults referred on urgent suspected cancer pathways in National Health Service (NHS) England. Patients and Methods: This work comprises a large-scale, prospective, observational, real-world NHS service evaluation of 9 blood tests, carried out from December 21, 2020 to July 31, 2025. Total of 16,481 patients with urgent suspected cancer referrals were enrolled across 5 secondary care Trusts and 170 General Practitioner surgeries. Real-world performance of the tests was evaluated using a range of diagnostic accuracy statistics. Results: Five tests have performance indicating potential clinical utility. Receiver operating characteristic area-under-curve scores (95% CI) for these were: Upper gastrointestinal=0.86 (0.81-0.90), Gynecological=0.81 (0.77-0.85), Lung=0.79 (0.74-0.84), Head & Neck=0.73 (0.68-0.78), and Lower gastrointestinal=0.72 (0.67-0.78), Prioritization of the 10% of highest-risk patients would reduce the number needed to investigate to detect one cancer by a factor of 2.6-6.1. Conclusion: This work shows the potential of these tests to improve urgent suspected cancer referral pathways. High-risk patients could be diagnosed more rapidly, leading to potential earlier-stage diagnosis and a better diagnostic experience. Low-risk patients could avoid unnecessary invasive medical testing for cancer. The software can be deployed rapidly across the NHS, without the need for additional hardware.
