Mayo Clinic Proceedings. Digital health最新文献

{"title":"Implementation and Updating of Clinical Prediction Models: A Systematic Review","authors":"Alexander Saelmans MD ,&nbsp;Tom Seinen PhD ,&nbsp;Victor Pera PharmD ,&nbsp;Aniek F. Markus PhD ,&nbsp;Egill Fridgeirsson PhD ,&nbsp;Luis H. John MSc ,&nbsp;Lieke Schiphof-Godart PhD ,&nbsp;Peter Rijnbeek PhD ,&nbsp;Jenna Reps PhD ,&nbsp;Ross Williams PhD","doi":"10.1016/j.mcpdig.2025.100228","DOIUrl":"10.1016/j.mcpdig.2025.100228","url":null,"abstract":"<div><h3>Objective</h3><div>To summarize the implementation approaches and updating methods of clinically implemented models and consecutively advise researchers on the implementation and updating.</div></div><div><h3>Patients and Methods</h3><div>We included studies describing the implementation of prognostic binary prediction models in a clinical setting. We retrieved articles from Embase, Medline, and Web of Science from January 1, 2010, to January 1, 2024. We performed data extraction, based on Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis and Prediction Model Risk of Bias Assessment guidelines, and summarized.</div></div><div><h3>Results</h3><div>The search yielded 1872 articles. Following screening, 37 articles, describing 56 prediction models, were eligible for inclusion. The overall risk of bias was high in 86% of publications. In model development and internal validation, 32% of the models was assessed for calibration. External validation was performed for 27% of the models. Most models were implemented into the hospital information system (63%), followed by a web application (32%) and a patient decision aid tool (5%). Moreover, 13% of models have been updated following implementation.</div></div><div><h3>Conclusion</h3><div>Impact assessments generally showed successful model implementation and the ability to improve patient care, despite not fully adhering to prediction modeling best practice. Both impact assessment and updating could play a key role in identifying and lowering bias in models.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 3","pages":"Article 100228"},"PeriodicalIF":0.0,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144297539","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Battle of the Bots: Solving Clinical Cases in Osteoarticular Infections With Large Language Models","authors":"Fabio Borgonovo MD ,&nbsp;Takahiro Matsuo MD ,&nbsp;Francesco Petri MD ,&nbsp;Seyed Mohammad Amin Alavi MD ,&nbsp;Laura Chelsea Mazudie Ndjonko ,&nbsp;Andrea Gori MD ,&nbsp;Elie F. Berbari MD, MBA","doi":"10.1016/j.mcpdig.2025.100230","DOIUrl":"10.1016/j.mcpdig.2025.100230","url":null,"abstract":"<div><h3>Objective</h3><div>To evaluate the ability of 15 different large language models (LLMs) to solve clinical cases with osteoarticular infections following published guidelines.</div></div><div><h3>Materials and Methods</h3><div>The study evaluated 15 LLMs across 5 categories of osteoarticular infections: periprosthetic joint infection, diabetic foot infection, native vertebral osteomyelitis, fracture-related infections, and septic arthritis. Models were selected systematically, including general-purpose and medical-specific systems, ensuring robust English support. In total, 126 text-based questions, developed by the authors from published guidelines and validated by experts, assessed diagnostic, management, and treatment strategies. Each model answered individually, with responses classified as correct or incorrect based on guidelines. All tests were conducted between April 17, 2025, and April 28, 2025. Results, presented as percentages of correct answers and aggregated scores, highlight performance trends. Mixed-effects logistic regression with a random question effect was used to quantify how each LLM compared in answering the study questions.</div></div><div><h3>Results</h3><div>The performance of 15 LLMs was evaluated, with the percentage of correct answers reported. OpenEvidence and Microsoft Copilot achieved the highest score (119/126 [94.4%]), excelling in multiple categories. ChatGPT-4o and Gemini 2.5 Pro scored 117 of the 126 (92.8%). When used as references, OpenEvidence was not inferior to any comparator and was superior to 5 LLMs. Performance varied across categories, highlighting the strengths and limitations of individual models.</div></div><div><h3>Conclusion</h3><div>OpenEvidence and Miccrosoft Copilot achieved the highest accuracy among evaluated LLMs, highlighting their potential for precisely addressing complex clinical cases. This study emphasizes the need for specialized, validated artificial intelligence tools in medical practice. Although promising, current models face limitations in real-world applications, requiring further refinement to support clinical decision making reliably.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 3","pages":"Article 100230"},"PeriodicalIF":0.0,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144280102","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Using Social Media to Combat Influenza Vaccine Misinformation and Improve Uptake: A Social Media Campaign and Repeated Cross-sectional Survey Analysis","authors":"Jessica B. Steier DrPH","doi":"10.1016/j.mcpdig.2025.100229","DOIUrl":"10.1016/j.mcpdig.2025.100229","url":null,"abstract":"<div><h3>Objective</h3><div>To combat influenza (flu)-vaccine misinformation and improve vaccine uptake using social media.</div></div><div><h3>Patients and Methods</h3><div>Unbiased Science used an online survey to identify flu vaccine-hesitant demographic groups and their specific objections to vaccination. Targeted educational content was then created and deployed through a variety of media formats, including podcasts, newsletters, reels, and infographics. A postcampaign survey determined the proportion of individuals who changed their minds about vaccination as a result of the educational content. The study was conducted between October 28, 2022 and February 7, 2023.</div></div><div><h3>Results</h3><div>In 3626 precampaign surveys, 187 individuals (5.1%) reported being unvaccinated and not planning to get the flu vaccine (the unvaccinated group). Multivariable analysis showed that geographic region (Northeast and Southeast), gender identity (male and other), race–ethnicity (non-Hispanic Black, and non-Hispanic other), and education level (high-school or less and some college) were independently associated with being unvaccinated. The main reasons were needlephobia, dismissal of flu severity, and concerns about vaccine components, multiple vaccines, and side effects. In 838 postcampaign surveys, 39 individuals (4.7%) indicated changing their mind about vaccination: of these, 27 (69.2%) said they were more likely to get vaccinated and 22 (56.4%) had gotten vaccinated. Twenty individuals (51.3%) said they changed their mind at least in part because of the targeted educational content.</div></div><div><h3>Conclusion</h3><div>Social media has the potential to change attitudes and behaviors around vaccination. When science messaging is deployed across several platforms and targeted to key demographic characteristics, it has the ability to combat misinformation and influence vaccine uptake.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 3","pages":"Article 100229"},"PeriodicalIF":0.0,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144297540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Model for Rapid Innovation for Engagement, Enrollment, and Data and Sample Collection in a Diverse Cohort Study: Insights from All of Us Participant Labs","authors":"Janna Ter Meer PhD ,&nbsp;Jessica Chen BS ,&nbsp;Romina Foster-Bonds MHS ,&nbsp;Andrea Goosen RN ,&nbsp;Gayle Valensky MPH ,&nbsp;Ethan Dinh-Luong BS ,&nbsp;Rachele Peterson MBA ,&nbsp;Geoffrey Ginsburg MD, PhD ,&nbsp;Chris Lunt BS ,&nbsp;Vik Kheterpal MD ,&nbsp;Eric Topol MD ,&nbsp;Allison Mandich MS ,&nbsp;Yentram Huyen PhD ,&nbsp;Julia Moore Vogel PhD","doi":"10.1016/j.mcpdig.2025.100227","DOIUrl":"10.1016/j.mcpdig.2025.100227","url":null,"abstract":"<div><h3>Objective</h3><div>To improve engagement and retention of a cohort that reflects the US population within the <em>All of Us</em> Research Program, we created and implemented an innovation infrastructure and initiatives.</div></div><div><h3>Participants and Methods</h3><div><em>All of Us</em> participant laboratories (APLs) established innovation-specific processes to rapidly ideate, select, implement, and evaluate cost-effective innovative initiatives, while mitigating risks. This was done within 4 priority areas: accelerating enrollment, enhancing engagement and retention, improving biospecimen collection, and broadening data types. Participants within the <em>All of Us</em> Research Program were engaged in this research between April 6, 2022 and May 6, 2024.</div></div><div><h3>Results</h3><div>We present a summary of APL processes and portfolio along with 5 specific initiatives that rapidly tested innovative ways to increase task completion and broaden biospecimen submission accessibility. Each initiative’s cost–benefit profile was evaluated by a committee of program leadership. Findings include the following: (1) offering compensation increased task completion, the degree of which was dependent on the context and amount of compensation; and (2) adding evening and weekend blood donation appointment times and distributing saliva collection kits through community partners increased donations from participants who have been historically underrepresented in biomedical research. On average, program staff predicted initiative effect sizes would be more than double their actual effect.</div></div><div><h3>Conclusion</h3><div>We found that large research studies can rapidly innovate to meet program goals, including a focus on diversity. We identified specific strategies and tactics to improve health research engagement and retention, with a focus on historically underrepresented in biomedical research communities, which can be used by numerous health research studies.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 3","pages":"Article 100227"},"PeriodicalIF":0.0,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144221505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Much More Than the Malady: The Promise of a Web-Based Digital Platform Incorporating Self-Report for Research and Clinical Care in Mild Cognitive Impairment","authors":"Andrew McGarry MD ,&nbsp;Oliver Roesler PhD ,&nbsp;Jackson Liscombe PhD ,&nbsp;Michael Neumann PhD ,&nbsp;Hardik Kothare PhD ,&nbsp;Abhishek Hosamath MBM ,&nbsp;Lakshmi Arbatti MS ,&nbsp;Anusha Badathala MD ,&nbsp;Stephen Ruhmel BS, MPH ,&nbsp;Bryan J. Hansen PhD ,&nbsp;Madeline Quall BA ,&nbsp;Sandrine Istas MBio ,&nbsp;Arthur Wallace MD, PhD ,&nbsp;David Suendermann-Oeft PhD ,&nbsp;Vikram Ramanarayanan PhD ,&nbsp;Ira Shoulson MD","doi":"10.1016/j.mcpdig.2025.100224","DOIUrl":"10.1016/j.mcpdig.2025.100224","url":null,"abstract":"<div><div>Traditional clinical trials in neurodegenerative disorders have utilized combinations of examination-based outcomes, global assessments by investigators and participants, and scales aimed at function, some of which are patient-reported outcomes. It is debatable whether these tools optimally convey therapeutic efficacy. A complementary approach using digital biomarkers to surpass exam-based limitations for detecting physical change coupled with a direct report from participants on what their sources of suffering are could be a useful advance in reporting beneficial effects of interventions, particularly if changes track together. We sought to determine the feasibility of remotely assessing speech, facial features, and cognition in an mild cognitive impairment (MCI) population, whether those extracted features could distinguish MCI from controls, and to explore what self-reported problems could reveal about the MCI experience. Our web-based platform was easy to use and revealed facial features in particular as capable of discriminating MCI from controls. Using the features that showed a statistically significant difference between cohorts (<em>P</em>&lt;.01) produced an area under the receiver operating curve of 0.75. Self-reported problems with cognition, gait, sleep, and behavior were more common in the MCI group. The MCI was associated with 6 times more difficulty with falls (n=6 vs 1). These data support the feasibility and discriminative utility of using remote monitoring technology in combination with participant self-report in an MCI population. Future work will investigate the extent to which multimodal biomarkers combined with self-report can characterize MCI longitudinally and for potential research applications as a measure of therapeutic effect.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100224"},"PeriodicalIF":0.0,"publicationDate":"2025-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134724","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring Evaluation of eHealth Lifestyle Interventions for Preschool Children: A Scoping Review","authors":"Marissa C.J. Kooij MD ,&nbsp;Ashley J.P. Smit MSc ,&nbsp;Linda D. Breeman PhD ,&nbsp;Lieke Schiphof-Godart PhD ,&nbsp;Isra Al-Dhahir MSc ,&nbsp;Andrea W.M. Evers PhD ,&nbsp;Koen F.M. Joosten MD, PhD","doi":"10.1016/j.mcpdig.2025.100223","DOIUrl":"10.1016/j.mcpdig.2025.100223","url":null,"abstract":"<div><div>EHealth lifestyle interventions can promote positive lifestyle changes in preschool children, but they need to be evaluated to assess their effectiveness and identify areas for improvement. This scoping review aimed to examine evaluation methods, outcome measures, and methodologic strengths and weaknesses, to provide recommendations for the evaluation of eHealth lifestyle interventions for preschool children. A comprehensive literature search was conducted across 6 databases for articles published up to September 29, 2023. We identified 48 articles describing 31 interventions that met our predefined eligibility criteria. These interventions predominantly targeted children’s diet. The most frequently evaluated outcomes were effectiveness, acceptability, and usage. Effectiveness outcomes included, among others, dietary intake, anthropometrics, and child and parental behaviors. Acceptability was evaluated primarily as user satisfaction. Evaluation methods for effectiveness and acceptability included questionnaires, interviews, focus groups, and portable devices. Intervention usage was evaluated via logged use and self-reported data. On the basis of our findings, we present recommendations for future evaluation of eHealth interventions for preschool children. These recommendations focus on selecting relevant outcome measures and appropriate evaluation methods and on integrating and applying evaluation results.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100223"},"PeriodicalIF":0.0,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143927368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Validation Study on Iatrogenic Nerve Damage Reduction Using Augmented Reality on Elbow Phantom","authors":"Giacomo Riberi MD ,&nbsp;Antonio Cangelosi MSc ,&nbsp;Paolo Titolo MD ,&nbsp;Elisa Dutto MD ,&nbsp;Massimo Salvi PhD ,&nbsp;Filippo Molinari PhD ,&nbsp;Luca Ulrich PhD ,&nbsp;Marco Agus PhD ,&nbsp;Corrado Calì PhD","doi":"10.1016/j.mcpdig.2025.100221","DOIUrl":"10.1016/j.mcpdig.2025.100221","url":null,"abstract":"<div><h3>Objective</h3><div>To compare augmented reality (AR) and classical intraoperative C-arm surgical navigation and evaluate whether head-mounted display improves surgical accuracy in the placement of a rod-like object, such as K-wire, using an anatomically accurate elbow phantom.</div></div><div><h3>Participants and Methods</h3><div>Data were collected between January 10, 2024, and March 15, 2024. We developed an AR system, X-ray simulation system and surgical phantom to test K-wire placement in 3 locations of the distal humerus and proximal ulnar bones. An initial phase with only X-ray as guidance was performed as case control; in later phases, the candidates were allowed to also use the head-mounted display. The evaluation parameters were time, placement angle, number of X-ray images taken, number of attempts, and distance from anatomical structures.</div></div><div><h3>Results</h3><div>In total, 19 physicians participated in the study. We analyzed 193 K-wire placements attempts that resulted in 150 estimated correct positions. This reflects a real-world scenario where multiple placements might be attempted to correctly place a K-wire. Compared with standard procedure, the use of AR resulted in −53.8 seconds in K-wire placement time, −47% of angular error from the K-wire target, −80% X-ray images taken to reach the estimate correct position, and decrease in distance variability of −81%, of the K-wire from anatomical structures of interest.</div></div><div><h3>Conclusions</h3><div>Compared with C-arm, AR navigation improved time, and angle of placement, requiring less X-ray images.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100221"},"PeriodicalIF":0.0,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143942075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimizing Digital Management of Research and Collaboration With Academic Information Manager","authors":"Peyman Nejat MD ,&nbsp;Vitali Fedosov MD, PhD ,&nbsp;Chady Meroueh MD ,&nbsp;Hugo Botha MB, ChB ,&nbsp;Svetlana Herasevich MD, MS ,&nbsp;Ing Tiong MS, MA ,&nbsp;David Martin MD, PhD ,&nbsp;Brian W. Pickering MD, MS ,&nbsp;Vitaly Herasevich MD, PhD","doi":"10.1016/j.mcpdig.2025.100222","DOIUrl":"10.1016/j.mcpdig.2025.100222","url":null,"abstract":"<div><h3>Objective</h3><div>To evaluate the efficacy, efficiency, and usability of the current iteration of the fully automatic Academic Information Manager (AIM) within the Department of Anesthesiology and Perioperative Medicine.</div></div><div><h3>Participants and Methods</h3><div>AIM was designed, developed, and deployed to address the growing need for digital information management in academic research. In a randomized, unblinded crossover study from April 1, 2020 to August 1, 2020, 15 participants completed 8 tasks using both AIM and conventional information retrieval methods. We assessed task completion time (efficiency), task completion status and accuracy (efficacy), subjective mental workload using the National Aeronautics and Space Administration Task Load Index (NASA-TLX), and system usability using System Usability Scale questionnaire, with and without AIM.</div></div><div><h3>Results</h3><div>Using AIM resulted in a significant time saving, with significantly higher task completion (99% vs 57%) and accuracy (99% vs 59%) compared with conventional methods. The NASA-TLX scores with AIM showed a statistically significant decrease in mental demand, temporal demand, effort, and frustration, along with an increase in performance, compared with those without AIM. The System Usability Scale score for AIM was above the 90th percentile.</div></div><div><h3>Conclusion</h3><div>Using AIM, we observed a significant increase in efficacy and efficiency, along with a decreased mental workload, as measured by NASA-TLX, and improved usability scores. Implementing AIM will help new investigators quickly and intuitively identify ongoing research at our institution. It will also enable them to broadcast their research interests to find potential collaborators.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100222"},"PeriodicalIF":0.0,"publicationDate":"2025-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143912144","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"InfoKids+: A Validation Study of a Pediatric Acuity Risk Stratification Algorithm","authors":"Carl A. Starvaggi MD ,&nbsp;Sophie Affentranger MMed ,&nbsp;Noelie Lengeler MMed ,&nbsp;Johan N. Siebert MD ,&nbsp;Annick Galetto-Lacour MD ,&nbsp;Rainer Tan PhD ,&nbsp;Manon Jaboyedoff MD ,&nbsp;Claudia E. Kuehni MD ,&nbsp;Mary-Anne Hartley PhD ,&nbsp;Kristina Keitel PhD","doi":"10.1016/j.mcpdig.2025.100220","DOIUrl":"10.1016/j.mcpdig.2025.100220","url":null,"abstract":"<div><h3>Objective</h3><div>To prospectively validate InfoKids+, a pediatric acuity electronic risk stratification algorithm (eRSA), against a nurse-based triage standard (nbTS).</div></div><div><h3>Participants and Methods</h3><div>We conducted a prospective validation study in a Swiss university hospital pediatric emergency department to assess the performance of a pediatric acuity eRSA, InfoKids+, on the basis of a well-established parental guidance application, InfoKids. Participants completed the eRSA once seated in a consultation booth. We compared the acuity levels from InfoKids+ (urgent, &lt;4 hours; nonurgent, &lt;24 hours; and no emergency, ≥24 hours) against an nbTS. The primary outcome was the level of agreement and rate of alignment between InfoKids+ and the reference standard.</div></div><div><h3>Results</h3><div>We included 1990 participants from June 3, 2020, through January 31, 2022. InfoKids+ showed a slight level of agreement with the nbTS (κ_lw=0.08; 95% CI, 0.06-0.10). InfoKids+ triaged 1762 (89%) cases as urgent (&lt;4 hours), 106 (5%) as nonurgent (≤24 hours), and 122 (6%) as no emergency (≥24 hours), compared with 810 (41%), 843 (42%), and 337 (17%) triages by the nbTS, respectively (<em>P</em>&lt;.001). InfoKids+ acuity level aligned with the reference standard in 888 (45%) cases, whereas it overreferred and underreferred in 999 (50%) and 103 (5%) cases, respectively (<em>P</em>&lt;.001).</div></div><div><h3>Conclusion</h3><div>In summary, our study uncovered notable discrepancies between the InfoKids+ algorithmic triage and conventional nurse-based triage. Our results highlight the critical need for rigorous validation of such tools for accuracy and safety before public release to ensure these tools are beneficial and do not inadvertently cause harm or misallocation of resources.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100220"},"PeriodicalIF":0.0,"publicationDate":"2025-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143927369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gender Disparities in Artificial Intelligence–Generated Images of Hospital Leadership in the United States","authors":"Mia Gisselbaek MD ,&nbsp;Joana Berger-Estilita MD, PhD ,&nbsp;Laurens Minsart MD ,&nbsp;Ekin Köselerli MD ,&nbsp;Arnout Devos PhD ,&nbsp;Francisco Maio Matos PhD ,&nbsp;Odmara L. Barreto Chang MD, PhD ,&nbsp;Peter Dieckmann PhD ,&nbsp;Melanie Suppan MD ,&nbsp;Sarah Saxena MD, PhD","doi":"10.1016/j.mcpdig.2025.100218","DOIUrl":"10.1016/j.mcpdig.2025.100218","url":null,"abstract":"<div><h3>Objective</h3><div>To evaluate demographic representation in artificial intelligence (AI)–generated images of hospital leadership roles and compare them with real-world data from US hospitals.</div></div><div><h3>Patients and Methods</h3><div>This cross-sectional study, conducted from October 1, 2024 to October 31, 2024, analyzed images generated by 3 AI text-to-image models: Midjourney 6.0, OpenAI ChatGPT DALL-E 3, and Google Gemini Imagen 3. Standardized prompts were used to create 1200 images representing 4 key leadership roles: chief executive officers, chief medical officers, chief nursing officers, and chief financial officers. Real-world demographic data from 4397 US hospitals showed that chief executive officers were 73.2% men; chief financial officers, 65.2% men; chief medical officers, 85.7% men; and chief nursing officers, 9.4% men (overall: 60.1% men). The primary outcome was gender representation, with secondary outcomes including race/ethnicity and age. Two independent reviewers assessed images, with interrater reliability evaluated using Cohen κ.</div></div><div><h3>Results</h3><div>Interrater agreement was high for gender (κ=0.998) and moderate for race/ethnicity (κ=0.670) and age (κ=0.605). DALL-E overrepresented men (86.5%) and White individuals (94.5%). Midjourney showed improved gender balance (69.5% men) but overrepresented White individuals (75.0%). Imagen achieved near gender parity (50.3% men) but remained predominantly White (51.5%). Statistically significant differences were observed across models and between models and real-world demographics.</div></div><div><h3>Conclusion</h3><div>Artificial intelligence text-to-image models reflect and amplify systemic biases, overrepresenting men and White leaders, while underrepresenting diversity. Ethical AI practices, including diverse training data sets and fairness-aware algorithms, are essential to ensure equitable representation in health care leadership.</div></div>","PeriodicalId":74127,"journal":{"name":"Mayo Clinic Proceedings. Digital health","volume":"3 2","pages":"Article 100218"},"PeriodicalIF":0.0,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143936608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}