JMIR Medical Informatics

Prediction of Early Hospital Admission (≤24 Hours) After Stroke Using Machine Learning and Deep Learning: Multicenter Study From China

2026-06-24T17:15:03-04:00

Background: Timely hospital admission is a prerequisite for effective acute stroke management, yet a substantial proportion of patients fail to reach medical facilities within the optimal therapeutic window. Existing prediction models often lack temporal robustness and clinical interpretability, limiting their utility in real-world, evolving health care systems. Objective: This study aimed to develop and temporally validate machine learning and deep learning models using multicenter clinical data to predict early hospital admission (≤24 h) after acute stroke. Methods: In this multicenter retrospective study, we analyzed routinely collected electronic medical record data from 1327 patients across 6 hospitals in China. We developed and compared 6 predictive models: logistic regression, support vector machine, random forest, multilayer perceptron (MLP), convolutional neural network, and long short-term memory, for early admission (≤24 h from symptom onset). Model training was performed on a train set (2019-2022), followed by independent temporal validation on a testing set (2023-2025). Model prediction performance was evaluated using discrimination metrics, sensitivity, and robustness under temporal distribution shift. Model interpretability was assessed using Shapley additive explanations. Results: A total of 1327 patients were included, of whom 821 were assigned to the train set and 506 to the independent temporal testing set. Among the 6 candidate models, the MLP showed the best overall performance in the independent temporal testing set, achieving an area under the receiver operating characteristic curve of 0.9020 (95% CI 0.8718-0.9283), sensitivity of 91.5%, specificity of 75.6%, and F₁-score of 0.9033. Formal statistical comparisons showed that the MLP achieved significantly higher area under the receiver operating characteristic curve values than logistic regression, support vector machine, random forest, and one-dimensional convolutional neural network after false discovery rate correction, with a smaller but still statistically significant improvement over the long short-term memory. Calibration analysis further showed that the MLP had the most favorable overall calibration profile among the candidate models. Conclusions: In this multicenter Chinese cohort, the MLP showed favorable temporal performance for predicting early hospital admission after stroke. The model may support future risk stratification and targeted public health interventions, although further external validation and calibration refinement are needed before deployment-oriented use. Trial Registration:

Assessing the Feasibility and Acceptability of Implementing a Preclinic Vital Signs Assessment in Primary Care: Cross-Sectional Pilot Study

2026-06-24T16:15:13-04:00

Background: Vital signs are objective measurements of the body’s most basic, essential functions, indicating overall health status. However, such assessments are time-consuming and so are not always prioritized. Measuring vital signs before doctor visits may, therefore, be an effective and efficient strategy. Objective: We piloted a preclinic vital signs assessment (PCVSA) procedure within a primary care center to determine its feasibility and acceptability. Methods: A mixed methods cross-sectional design was used for piloting the PCVSA procedure. Study participants included adult patients and practice staff. Patients had vital signs assessed by a primary care assistant before general practitioner (GP) visits. Collected data concerned participants’ study engagement, the timings of PCVSA/GP visits, and surveys/interviews investigating participants’ experiences. Implementation (Results): A total of 16 patients and 4 staff participated. The mean duration for PCVSAs was 2 minutes and 23 seconds (SD 38.8 s), and the mean duration for GP visits was 9 minutes and 21 seconds (SD 252.4 s). Patients said the PCVSA was a “Positive experience” (n=14, 88%), “Helpful” (n=13, 81%), “Valuable” (n=7, 44%), and “Interesting” (n=6, 38%). The GP said the PCVSAs were either “Helpful” (8/15, 53%) or “Extremely Helpful” (7/15, 47%) in each of their consultations and that the PCVSAs improved engagement with patients (12/15, 80%), allowed them to spend more time gaining an understanding of the conditions of patients (14/15, 93%), and enhanced productivity during consultations (11/15, 73%). The GP strongly agreed that collecting PCVSA data before appointments would benefit patients over time. Qualitative interviews with practice staff yielded three themes: (1) improved patient engagement and efficient consultation, (2) time-saving potential, and (3) practicing in general practice and associated challenges. Conclusions: The PCVSA pilot showed good feasibility and acceptability as indicated by high participant engagement, short PCVSA and GP visit times (albeit GP visit times did not measure non–patient-facing clinical activity), and positive feedback from patients and staff. Introducing PCVSAs in health care settings may have potential in terms of improving the standard and efficiency of care.

Candidate Passive Sensor Suite Technologies for Tactical Combat Casualty Care Environments: Comparative Assessment Study

2026-06-24T14:00:18-04:00

Background: The United States Army Institute for Surgical Research conducted an analysis of 3 prototype sensor suites; all candidates were designed to passively document care delivery in tactical combat casualty care environments. Objective: This study aims to ensure sensor suites remain resilient and adaptive in complex battlefield environments. This research effort conducts a systematic comparative assessment of prototype solutions. Methods: The assessment methodology prioritized functionality, usability, and performance. The assessment consisted of three phases: (1) tabletop evaluations, (2) simulated use testing, and (3) a sensor suite rodeo simulation event. The second and third phases included human participants leveraging the technology prototypes in hyperrealistic tactical combat casualty care simulation environments. Additionally, the third phase allowed the researchers to assess the performance of each prototype in a range of operational environmental lighting conditions. Results: During the tabletop evaluation phase, all 3 prototype sensor suite solutions demonstrated acceptable results (≥1) in the technical specification assessment. The 2-part heuristic analysis revealed variability, where the least complex configurations received the highest assessment scores. To capture and record raw data, scores ranged from 44.6 to 87 on a 100-point scale. To offload and export the raw data, scores ranged from 22.9 to 87.5 on a 100-point scale. During simulated user testing, all 3 sensor suites achieved passing quantitative scores (≥60); the system usability scores (SUS) ranged from 60 to 85 on a 100-point scale. More complex technology configurations received higher usability scores. From a qualitative perspective, vital sign monitor latency display issues led to reliability concerns. All 3 prototypes successfully generated raw data; the individual outputs ranged from 0.06 to 0.13 GB/minute. During the sensor suite rodeo simulation event, all 3 sensor suites achieved passing quantitative scores (≥60); the SUS ranged from 66.7 to 86.7 on a 100-point scale; the most complex technology prototype configuration scored higher. Qualitative findings identified data transfer issues with large file sizes and pairing issues with vital sign monitors. All 3 prototypes successfully generated raw data; the individual outputs varied (ranging from 0.012 to 0.24 GB/min) based on the environmental lighting conditions (full sun, indoor lighting, and low light). However, from a data quality perspective, only 1 camera component produced viable video data in all 3 environments. Conclusions: The comparative assessment revealed opportunities to combine the strengths of both approaches in a next-generation implementation. This preliminary assessment was constrained by several factors: (1) effective tracking of consumable medical supplies, (2) advancement of artificial intelligence algorithms to process the raw data, and (3) ability to manage multiple casualties or patients. Follow-on evaluations are needed to address these limitations. This systematic, 3-part methodology evaluates early-stage sensor suite prototypes and provides a reproducible framework for advancing battlefield medical technologies. International Registered Report Identifier (IRRID): RR2-10.2196/67673

Machine Learning–Based Prognostic Models for Functional Outcomes in Spinal Cord Injury: Systematic Review

2026-06-23T17:00:03-04:00

Background: Machine learning is increasingly used to develop prognostic prediction models for spinal cord injury. Nevertheless, current studies exhibit heterogeneity in outcome measures, predictors, modeling strategies, and validation methods. Moreover, the reporting quality, risk of bias, and clinical applicability of these models have not been systematically evaluated using assessment tools specific to prediction models. Objective: This review aimed to assess the reporting quality and risk of bias of machine learning–based prognostic models for spinal cord injury, and evaluate their clinical applicability, model features, validation, and implementation barriers. Methods: We searched the China National Knowledge Infrastructure (CNKI), Wanfang Data, VIP Database, Sinomed, PubMed, Web of Science, Embase, and Scopus databases from their inception up to May 14, 2025. Two investigators independently screened studies, extracted data, and assessed risk of bias. Reporting quality and risk of bias were evaluated using the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement and the Prediction Model Risk of Bias Assessment Tool (PROBAST). Descriptive statistics and narrative synthesis were used to summarize the findings. Results: In total, 19 cohort studies were included. TRIPOD adherence ranged from 54.8% (17/31) to 81.1% (30/37), with a median of 74.2% (IQR 64.5%-77.4%). Overall, all 19 studies were judged to have a high risk of bias, mainly because of limitations in the analysis domain. Only 1 (5.3%) study included external validation, while 16 (84.2%) studies used internal validation and 2 (10.5%) studies reported model development only. No study justified the sample size; 6 (31.6%) studies reported imputation or other methods for handling missing data, and calibration was rarely reported. Conclusions: Machine learning shows potential for spinal cord injury prognostic modeling, especially when complex clinical or imaging data are available. However, existing evidence is limited by incomplete reporting, high risk of bias, substantial heterogeneity, and limited external validation. Larger, methodologically robust studies with standardized outcomes, external validation, and evaluation of clinical usefulness are necessary before these models can be implemented in routine clinical practice. Clinical Trial: PROSPERO CRD420251071502; https://www.crd.york.ac.uk/PROSPERO/view/1071502

Implementation and Evaluation of a Social Networking Service–Based Mobile Patient-Generated Health Data System With Direct Electronic Medical Record Integration: Prospective Observational Study

2026-06-23T16:45:15-04:00

Background: Patient-generated health data (PGHD) can enhance patient-centered care by improving disease awareness and preparedness for clinical encounters. However, automated incorporation of PGHD into electronic medical records (EMRs), which is a prerequisite for broader clinical implementation, remains technically and administratively challenging. Objective: This study describes the development of , a PGHD collection platform that delivers mobile social networking service–based previsit questionnaires with automated transfer of structured patient responses into the EMR, and evaluates patient participation, EMR documentation quality, and user satisfaction in a cardiology outpatient clinic. Methods: This single-center observational study was conducted between August and November 2024 and included 751 consecutive cardiology outpatients, comprising 282 first-visit patients and 469 patients attending follow-up visits for heart failure. All eligible patients received a previsit electronic questionnaire link via KakaoTalk or multimedia messaging service prior to their scheduled visit. The primary outcomes were the overall survey response rate among all enrolled patients and EMR documentation completeness among follow-up patients with heart failure. Documentation quality was evaluated based on 3 prespecified parameters relevant to routine heart failure care—dyspnea, peripheral edema, and medication adherence status—and was quantified using an EMR completeness score ranging from 0 to 3. Secondary outcomes included patient and provider satisfaction assessed using postvisit 5-point Likert-scale surveys. Firth penalized logistic regression was used to evaluate the association between survey response status and EMR completeness, with adjustment for age and sex. Results: The response rate was 38.5% (289/751), including 48.9% (138/282) of new patients and 32.2% (151/469) of follow-up patients with heart failure. Responders were younger than nonresponders (mean 62.0, SD 15.7 years vs mean 69.8, SD 12.5 years; <.001). Among the follow-up patients with heart failure, EMR completeness was higher among responders (median score 3, IQR 3‐3) than among nonresponders (median score 0, IQR 0‐1; <.001). Patient satisfaction was high: 82.9% (63/76) to 92.1% (70/76) agreed that the system was appropriate, easy to use, and helpful, and 78.9% (60/76) completed the survey in <10 minutes. Both cardiologists and 7 of the 8 participating nurses supported continued use of the system, citing workflow efficiency gains. Conclusions: Miri-Alimi enabled patient-friendly PGHD collection without requiring log-ins or a dedicated app and demonstrated direct transfer of patient responses into the EMR. Its use was associated with effective transfer and structured integration of PGHD into the EMR, as well as high satisfaction among survey respondents and participating staff. Further studies should evaluate sustainability and associations with long-term clinical outcomes across diverse care settings. Trial Registration: Clinical Research Information Service of the Korea Disease Control and Prevention Agency KCT0011327;

Computational Insights Into Smart Bioelectronics in Digital Health Care (2020-2024): Topic Modeling Study

2026-06-23T16:30:19-04:00

Background: Smart bioelectronics are electronic medical devices that combine hardware and artificial intelligence (AI)–based software. These convergent medical devices analyze bio-signals measured through hardware using AI algorithms and deliver physical stimulation to enhance therapeutic effects. Objective: This study aimed to systematically analyze recent research trends in smart bioelectronics to understand their evolving role in digital health care and to provide evidence-based insights for shaping future research and development strategies. Methods: A total of 92 publications indexed in PubMed between 2020 and 2024 were analyzed. Latent Dirichlet allocation–based topic modeling, optimized using coherence scores, was applied to identify latent research themes. Results: The results indicate a steady increase in related research over the past 5 years, along with a clear shift in research focus from bio-signal sensing and bioelectronic device materials toward AI-driven analysis and disease-oriented applications, ultimately evolving into intelligent and adaptive bioelectronic therapeutic systems. Three major research topics were identified: bio-signal–based neuromodulation (n=23, 25%), AI-driven neurological disease analysis (n=32, 34.7%), and implantable bioelectronics and biomaterials (n=37, 40.2%). Conclusions: By mapping the evolving landscape of smart bioelectronics, this study provides valuable insights into their multidisciplinary development and highlights their potential applications in clinical decision support, personalized rehabilitation, and next-generation medical device innovation.

US Virgin Islands Launches Modernized NBS 7 Disease Surveillance System to Transform Public Health: Implementation Report

2026-06-23T15:15:13-04:00

Background: During January 2024, the US Virgin Islands (USVI) Department of Health (VIDOH) identified a critical need to maintain the cloud-hosted National Electronic Disease Surveillance System Base System (NBS) instance and support the local data modernization initiative. After consulting with federal partners and subject matter experts, VIDOH’s leadership chose to migrate the integrated disease surveillance system to a new platform hosted on Amazon Web Services (AWS) and update the NBS instance to the most advanced version, NBS 7. Objective: The primary aim was to support a USVI disease surveillance system that is modern, functional, and cost-efficient by migrating the VIDOH NBS instance from a vendor-managed environment to a jurisdiction-managed AWS cloud-based infrastructure while upgrading to NBS 7. Methods: The VIDOH implemented a phased migration strategy that included planning and cost-benefit assessment, deployment of NBS 7 within AWS, database migration, validation and optimization, and staged reonboarding of electronic reporting facilities. Implementation (Results): The USVI NBS 7 instance went live on May 6, 2025, with USVI becoming the first US jurisdiction using AWS for implementation of NBS 7 and the second using NBS 7 in production, overall. Benefits of this change included nearly 90% cost savings (preliminarily estimated at 80%), additional bandwidth, real-time data ingestion and updates, an opportunity to build local informatics capacity, and the ability to have greater autonomy over the data and its end points. To date, the VIDOH successfully reonboarded 106 of 109 (97%) previously connected electronic reporting facilities and onboarded 1 new reporting laboratory previously unable to connect due to interoperability barriers. Conclusions: Updating the USVI database to NBS 7 in a locally owned, cloud-hosted, AWS environment has improved disease surveillance specifically by providing the most up-to-date Centers for Disease Control and Prevention–supported data system, improving timeliness of reporting by offering local providers more flexibility in electronic reporting options, and giving USVI direct control over workflow decision functionality. Furthermore, improved interoperability and maintaining a cloud-based platform were additional benefits of the database migration. This important investment in public health infrastructure will allow USVI public health professionals, clinicians, policymakers, and other stakeholders to be able to monitor and respond to disease threats quickly and inform appropriate public health action.

Ambient AI Scribe Implementation in an Ambulatory Setting in a Single Medical Group: Prospective Study

2026-06-23T15:00:03-04:00

Background: Health care providers spend an excessive amount of time within electronic medical record (EMR) systems documenting patient encounters, often amounting to hours of work outside of regular office hours. This affects physician productivity and directly contributes to burnout. Artificial intelligence (AI) is becoming more integrated into medical care, including the development of speech recognition and note generation algorithms. Limited studies exist on how these AI tools affect provider satisfaction, work-life balance, and patient satisfaction. Objective: The aim of this study was to assess the use of ambient AI in medical documentation and its effects on time spent in the EMR and on provider burnout, with a secondary focus on note quality and patient satisfaction. Methods: This prospective study was conducted at the Hawaii Pacific Health Medical Group to pilot an AI note writer. Abridge was chosen as the AI platform and integrated with the Epic EMR. A goal of 75 providers for a 3-month pilot period was established from December 2024 through February 2025. Surveys were distributed to providers before and during the trial period. Epic Signal and Abridge data were used to correlate provider-perceived outcomes with EMR-recorded outcomes. Users were then divided into groups based on frequency of AI use, with high use defined as ≥60% AI scribe use in patient encounters. The primary outcome was time spent on documentation per appointment. Results: A total of 80 providers were recruited, with 79 completing the pilot. More than 25,000 notes were generated across 23 specialties. Signal metrics found a 21% decrease in time spent on notes per day (−13.6 minutes) and a 13% decrease in pajama time (−3.6 minutes) among high users. Among 79 providers using ambient AI, 6 (7.6%) reported spending ≥8 hours per week on notes outside of clinic hours, a 76% decrease from 25 (31.6%) providers before the pilot. With ambient AI, 39 (49.4%) physicians reported no burnout symptoms, representing a 22% increase. Provider-perceived workload decreased, and self-reported note quality remained favorable. Providers reported that 84% of notes required edits to less than one-quarter of the note content. Patient experience, as measured by “provider listened to me” scores on patient satisfaction surveys, was not significantly affected by ambient AI use (=.39). Conclusions: This ambient AI scribe decreased the time providers spent writing notes in the clinic and decreased time spent in the EMR outside of work hours. There was no significant difference in symptoms of burnout.

Leveraging Large Language Models to Integrate Clinical Knowledge and Machine Learning Predictions for Lymph Node Metastasis Prediction: Development of a Knowledge-Augmented Framework

2026-06-22T17:00:27-04:00

Background: Lymph node metastasis (LNM) is a critical clinical indicator for determining the initial treatment strategy for patients with lung cancer. However, accurately diagnosing LNM preoperatively remains a significant challenge. Data-driven predictive modeling has become a mainstream approach to address this issue, yet it often overlooks existing clinical knowledge. Large language models (LLMs) have demonstrated the potential to predict clinical risks in a zero-shot manner based on the extensive clinical knowledge learned from large-scale corpora. Objective: LLMs have demonstrated the potential to predict clinical risks in a zero-shot manner based on the extensive clinical knowledge learned from large-scale corpora. This study aims to investigate the integration of LLM-derived knowledge with data-driven patterns to enhance the accuracy of LNM prediction. Methods: We propose a novel ensemble framework that combines the strengths of LLMs and machine learning (ML) models for LNM prediction in lung cancer. Specifically, 3 ML models were trained using clinical data, and their predicted probabilities, along with the original clinical features, were incorporated into prompts for LLMs. Three LLMs—GPT-5.4, GPT-5.4-nano, and DeepSeek-V3.2—were used to independently predict LNM risk 5 times, and 4 ensemble strategies were applied to aggregate their predictions into a final outcome. Results: The proposed approach was evaluated on clinical data from 767 patients with lung cancer at Peking University Cancer Hospital. Experimental results show that our proposed framework significantly outperforms base ML models, achieving an area under the curve of 0.781 and an average precision of 0.420. Compared with the no reasoning English setting, both the reasoning English setting and nonreasoning Chinese setting showed a lower area under the curve but higher average precision. Conclusions: This study presents a novel knowledge-augmented strategy for integrating the clinical knowledge embedded in LLMs with the statistical patterns captured by ML models to improve the LNM prediction of lung cancer, offering a new paradigm for integrating medical knowledge and patient data in clinical predictions.

Blockchain-Based Dynamic and Revocable Consent for Secondary Health Data Use: Systematic Review

2026-06-22T16:30:19-04:00

Background: The secondary use of health data holds substantial potential for advancing biomedical research, strengthening population health analytics, and enabling artificial intelligence–driven decision-making support. Yet, ensuring that such reuse respects patient autonomy, privacy, and regulatory obligations remains a major challenge. Conventional consent mechanisms are typically static, difficult to revoke, and offer limited transparency or accountability after data disclosure. Objective: This review aimed to systematically examine blockchain-based frameworks that enable dynamic, auditable, and revocable consent for the secondary use of health data. Methods: A structured literature search was conducted in PubMed, Scopus, and Web of Science covering the period 2020 to 2025. Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, 55 peer-reviewed studies meeting predefined inclusion criteria were analyzed. Data extraction focused on four dimensions: (1) consent life cycle management, (2) auditability and traceability, (3) usability and patient empowerment, and (4) legal and ethical alignment. Results: Findings indicate that blockchain technologies provide a robust foundation for automating consent life cycles, ensuring immutable auditability, and enabling decentralized patient control. Most frameworks used smart contracts, decentralized identifiers, and verifiable credentials to implement programmable and verifiable consent processes. Nevertheless, key challenges persist, including limited usability testing, complexities in real-time revocation propagation, interoperability gaps with clinical systems, and tensions with regulatory requirements such as the General Data Protection Regulation right to erasure. Only a small subset of studies reported real-world deployments or user-centered evaluations. Conclusions: Blockchain offers substantial promise for improving the trustworthiness, transparency, and accountability of consent management for secondary health data use. However, wider adoption requires human-centered design approaches, stronger interoperability through standards such as Fast Healthcare Interoperability Resources, verifiable credentials, and consent receipts, and clearer legal guidance for compliance. Future research should prioritize integrating blockchain-enabled consent infrastructures into national and cross-border digital health ecosystems such as the European Health Data Space to support secure, patient-controlled, and ethically governed secondary data use.