JMIR Medical Informatics最新文献

{"title":"Performance of Large Language Models in the Non-English Context: Qualitative Study of Models Trained on Different Languages in Chinese Medical Examinations.","authors":"Zhong Yao, Liantan Duan, Shuo Xu, Lingyi Chi, Dongfang Sheng","doi":"10.2196/69485","DOIUrl":"10.2196/69485","url":null,"abstract":"<p><strong>Background: </strong>Research on large language models (LLMs) in the medical field has predominantly focused on models trained with English-language corpora, evaluating their performance within English-speaking contexts. The performances of models trained with non-English language corpora and their performance in non-English contexts remain underexplored.</p><p><strong>Objective: </strong>This study aimed to evaluate the performances of LLMs trained on different languages corpora by using the Chinese National Medical Licensing Examination (CNMLE) as a benchmark and constructed analogous questions.</p><p><strong>Methods: </strong>Under different prompt settings, we sequentially posed questions to 7 LLMs: 2 primarily trained on English-language corpora and 5 primarily on Chinese-language corpora. The models' responses were compared against standard answers to calculate the accuracy rate of each model. Further subgroup analyses were conducted by categorizing the questions based on various criteria. We also collected error sets to explore patterns of mistakes across different models.</p><p><strong>Results: </strong>Under the zero-shot setting, 6 out of 7 models exceeded the passing level, with the highest accuracy rate achieved by the Chinese LLM Baichuan (86.67%), followed by ChatGPT (83.83%). In the constructed questions, all 7 models exceeded the passing threshold, with Baichuan maintaining the highest accuracy rate (87.00%). In few-shot learning, all models exceeded the passing threshold. Baichuan, ChatGLM, and ChatGPT retained the highest accuracy. While Llama showed marked improvement over previous tests, the relative performance rankings of other models stayed similar to previous results. In subgroup analyses, English models demonstrated comparable or superior performance to Chinese models on questions related to ethics and policy. All models except Llama generally had higher accuracy rates for simple questions than for complex ones. The error set of ChatGPT was similar to those of other Chinese models. Multimodel cross-verification outperformed single model, particularly improving accuracy rate on simple questions. The implementation of dual-model and tri-model verification achieved accuracy rates of 94.17% and 96.33% respectively.</p><p><strong>Conclusions: </strong>At the current level, LLMs trained primarily on English corpora and those trained mainly on Chinese corpora perform similarly well in CNMLE, with Chinese models still outperforming. The performance difference between ChatGPT and other Chinese LLMs are not solely due to communication barriers but are more likely influenced by disparities in the training data. By using a method of cross-verification with multiple LLMs, excellent performance can be achieved in medical examinations.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e69485"},"PeriodicalIF":3.1,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12227152/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144512869","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Health Information Systems' Support for Management and Changing Work: Survey Study Among Physicians.","authors":"Tarja Heponiemi, Lotta Virtanen, Emma Kainiemi, Petra Saukkonen, Jarmo Reponen, Tinja Lääveri","doi":"10.2196/65913","DOIUrl":"10.2196/65913","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;The digitalization of health care has advanced significantly in recent years. Consequently, physicians have needed to increasingly adopt new digital health technologies such as electronic health record systems and other health information systems. Digitalization has changed physicians' clinical work, work environment, management work, and use of tools for leadership. Many physician leaders have been critical of the capabilities of health information systems (HISs) to support leadership, management, and knowledge management.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;We aimed to examine the association between leadership position and perceived changes in clinical work due to digitalization among a nationally representative sample of Finnish physicians and physician leaders. In addition, we examined physician leaders' perceptions of HISs as a support for management and whether their opinions differed based on their perceptions on changes in clinical work due to digitalization.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Altogether 4630 Finnish physicians (2960/4586, 64% women) responded to a cross-sectional nation-wide web-based survey conducted in spring 2021. Perceptions of improved preventive work, facilitated access to patient information, progressed interprofessional collaboration, and accelerated clinical encounters were used as measures of changes due to digitalization. First, we examined with multivariable logistic regression analyses whether being in a leadership position was associated with perceived changes in work due to digitalization (improved preventive work, facilitated access to patient information, progressed interprofessional collaboration, and accelerated clinical encounters in separate analyses) in the total sample. Second, we examined with analyses of covariance whether the variables related to perceived changes in work due to digitalization were associated with perceived management support from HISs among those who had administrative or management responsibilities (n=817). All analyses were adjusted for gender, age, and sector.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Physician leaders had greater odds of agreeing that digitalization had improved preventive work (odds ratio [OR] 1.62, 95% CI 1.33-1.98), facilitated access to patient information (OR 1.28, 95% CI 1.09-1.51), progressed interprofessional collaboration (OR 1.81, 95% CI 1.53-2.14), and accelerated clinical encounters (OR 1.31, 95% CI 1.01-1.70) than those in nonleadership positions. Furthermore, leaders who perceived these changes in work due to digitalization positively also considered that health information systems supported their management work.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Physician leaders appeared to view the changes in work due to digitalization more positively than other physicians. In addition, those leaders who perceived these changes positively also perceived that HISs supported their management work. Thus, leaders should thoroughly evaluate ","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e65913"},"PeriodicalIF":3.1,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12226959/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144509752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Responsible Framework for Assessing, Selecting, and Explaining Machine Learning Models in Cardiovascular Disease Outcomes Among People With Type 2 Diabetes: Methodology and Validation Study.","authors":"Yang Yang, Che-Yi Liao, Esmaeil Keyvanshokooh, Hui Shao, Mary Beth Weber, Francisco J Pasquel, Gian-Gabriel P Garcia","doi":"10.2196/66200","DOIUrl":"10.2196/66200","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Building machine learning models that are interpretable, explainable, and fair is critical for their trustworthiness in clinical practice. Interpretability, which refers to how easily a human can comprehend the mechanism by which a model makes predictions, is often seen as a primary consideration when adopting a machine learning model in health care. However, interpretability alone does not necessarily guarantee explainability, which offers stakeholders insights into a model's predicted outputs. Moreover, many existing frameworks for model evaluation focus primarily on maximizing predictive accuracy, overlooking the broader need for interpretability, fairness, and explainability.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study proposes a 3-stage machine learning framework for responsible model development through model assessment, selection, and explanation. We demonstrate the application of this framework for predicting cardiovascular disease (CVD) outcomes, specifically myocardial infarction (MI) and stroke, among people with type 2 diabetes (T2D).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We extracted participant data comprised of people with T2D from the ACCORD (Action to Control Cardiovascular Risk in Diabetes) dataset (N=9635), including demographic, clinical, and biomarker records. Then, we applied hold-out cross-validation to develop several interpretable machine learning models (linear, tree-based, and ensemble) to predict the risks of MI and stroke among patients with diabetes. Our 3-stage framework first assesses these models via predictive accuracy and fairness metrics. Then, in the model selection stage, we quantify the trade-off between accuracy and fairness using area under the curve (AUC) and Relative Parity of Performance Scores (RPPS), wherein RPPS measures the greatest deviation of all subpopulations compared with the population-wide AUC. Finally, we quantify the explainability of the chosen models using methods such as SHAP (Shapley Additive Explanations) and partial dependence plots to investigate the relationship between features and model outputs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Our proposed framework demonstrates that the GLMnet model offers the best balance between predictive performance and fairness for both MI and stroke. For MI, GLMnet achieves the highest RPPS (0.979 for gender and 0.967 for race), indicating minimal performance disparities, while maintaining a high AUC of 0.705. For stroke, GLMnet has a relatively high AUC of 0.705 and the second-highest RPPS (0.961 for gender and 0.979 for race), suggesting it is effective across both subgroups. Our model explanation method further highlights that the history of CVD and age are the key predictors of MI, while HbA1c and systolic blood pressure significantly influence stroke classification.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This study establishes a responsible framework for assessing, selecting, and explaining machine learning models, emphasizing ","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e66200"},"PeriodicalIF":3.1,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12256707/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144512868","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring the Potential of Electroencephalography Signal-Based Image Generation Using Diffusion Models: Integrative Framework Combining Mixed Methods and Multimodal Analysis.","authors":"Chi-Sheng Chen, Shao-Hsuan Chang, Che-Wei Liu, Tung-Ming Pan","doi":"10.2196/72027","DOIUrl":"10.2196/72027","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Electroencephalography (EEG) has been widely used to measure brain activity, but its potential to generate accurate images from neural signals remains a challenge. Most EEG-decoding research has focused on tasks such as motor imagery, emotion recognition, and brain wave classification, which involve EEG signal analysis and classification. Some studies have explored the correlation between EEG and images, primarily focusing on EEG-image pair classification or transformation. However, EEG-based image generation remains underexplored.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;The primary goal of this study was to extend EEG-based classification to image generation, addressing the limitations of previous methods and unlocking the full potential of EEG for image synthesis. To achieve more meaningful EEG-to-image generation, we developed a novel framework, Neural-Cognitive Multimodal EEG-Informed Image (NECOMIMI), which was specifically designed to generate images directly from EEG signals.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We developed a 2-stage NECOMIMI method, which integrated the novel Neural Encoding Representation Vectorizer (NERV) EEG encoder that we designed with a diffusion-based generative model. The Category-Based Assessment Table (CAT) score was introduced to evaluate the semantic quality of EEG-generated images. In addition, the ThingsEEG dataset was used to validate and benchmark the CAT score, providing a standardized measure for assessing EEG-to-image generation performance.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The NERV EEG encoder achieved state-of-the-art performance in several zero-shot classification tasks, with an average accuracy of 94.8% (SD 1.7%) in the 2-way task and 86.8% (SD 3.4%) in the 4-way task, outperforming models such as Natural Image Contrast EEG, Multimodal Similarity-Keeping Contrastive Learning, and Adaptive Thinking Mapper ShallowNet. This highlighted its superiority as a feature extraction tool for EEG signals. In a 1-stage image generation framework, EEG embeddings often resulted in abstract or generalized images such as landscapes instead of specific objects. Our proposed 2-stage NECOMIMI architecture effectively extracted semantic information from noisy EEG signals, showing its ability to capture and represent underlying concepts derived from brain wave activity. We further conducted a perturbation study to test whether the model overly depended on visual cortex EEG signals for scene-based image generation. The perturbation of visual cortex EEG channels led to a notable increase in Fréchet inception distance scores, suggesting that our model relied heavily on posterior brain signals to generate semantically coherent images.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;NECOMIMI demonstrated the potential of EEG-to-image generation, revealing the challenges of translating noisy EEG data into accurate visual representations. The novel NERV EEG encoder for multimodal contrastive learning reached state-of-t","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e72027"},"PeriodicalIF":3.1,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12242056/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144499582","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Relationship Between Stroke Knowledge, Health Information Literacy, and Health Self- Management Among Patients with Stroke: Multicenter Cross-Sectional Study.","authors":"Mengxue Zeng, Yanhua Liu, Ying He, Wenxia Huang","doi":"10.2196/63956","DOIUrl":"10.2196/63956","url":null,"abstract":"<p><strong>Background: </strong>The World Health Organization highlights the essential role of effective self-management in the prevention and control of chronic diseases, noting that proper self-management can significantly slow disease progression. Stroke, which ranks as the third leading cause of death worldwide, is often accompanied by inadequate self-management among patients. While health information literacy (HIL) has been shown to influence self-management in conditions such as diabetes and hypertension, its role as a mediating factor linking disease perception and health behavior in patients with stroke remains insufficiently explored. Addressing this research gap is vital for developing targeted interventions.</p><p><strong>Objective: </strong>The aim of this study was to investigate the current status of HIL, stroke knowledge, and self-management abilities among patients with stroke in Southwest China. Additionally, the study analyzed the relationships among these three factors and their mechanisms of action, providing evidence to inform the development of effective health education strategies for enhancing self-management in patients with stroke.</p><p><strong>Methods: </strong>A multicenter cross-sectional design was employed, enrolling 514 patients with stroke from the neurology departments of three tertiary general hospitals in Chengdu between September 2022 and March 2023. Data collection used a standardized set of scales: the health information literacy questionnaire for stroke assessed HIL, the stroke prevention questionnaire evaluated knowledge levels, and the stroke self-management assessment scale measured self-management abilities. Regression analysis and bootstrap mediation effect testing were conducted using SPSS software (version 26.0).</p><p><strong>Results: </strong>Patients with stroke had a mean (SD) score of 17.61 (6.46) for stroke knowledge, 61.17 (13.58) for HIL, and 158.70 (19.07) for self-management skills. Correlation analysis indicated a positive correlation of stroke knowledge with both self-management (r=0.668; P<.001) and HIL (r=0.138; P<.001). The mediation test showed a significant mediating effect of HIL between stroke knowledge and self-management (β=0.543; 95% CI: 0.431-0.663), with an effect share of 82.77%.</p><p><strong>Conclusions: </strong>There is a correlation between HIL and self-management; the higher the HIL, the better is the self-management behavior. Furthermore, HIL partially mediates the effect of stroke knowledge on self-management.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e63956"},"PeriodicalIF":3.1,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12208508/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144477994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluating and Improving Syndrome Differentiation Thinking Ability in Large Language Models: Method Development Study.","authors":"Chunliang Chen, Xinyu Wang, Ming Guan, Wenjing Yue, Yuanbin Wu, Ya Zhou, Xiaoling Wang","doi":"10.2196/75103","DOIUrl":"10.2196/75103","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;A large language model (LLM) provides new opportunities to advance the intelligent development of traditional Chinese medicine (TCM). Syndrome differentiation thinking is an essential part of TCM and equipping LLMs with this capability represents a crucial step toward more effective clinical applications of TCM. However, given the complexity of TCM syndrome differentiation thinking, acquiring this ability is a considerable challenge for the model.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aims to evaluate the ability of LLMs for syndrome differentiation thinking and design a method to effectively enhance their performance in this area.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We decomposed the process of syndrome differentiation thinking in TCM into three core tasks: pathogenesis inference, syndrome inference, and diagnostic suggestion. To evaluate the performance of LLMs in these tasks, we constructed a high-quality evaluation dataset, forming a reliable foundation for quantitative assessment of their capabilities. Furthermore, we developed a methodology for generating instruction data based on the idea of an \"open-book exam,\" customized three data templates, and dynamically retrieved task-relevant professional knowledge that was inserted into predefined positions within the templates. This approach effectively generates high-quality instruction data that aligns with the unique characteristics of TCM syndrome differentiation thinking. Leveraging this instruction data, we fine-tuned the base model, enhancing the syndrome differentiation thinking ability of the LLMs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;We collected 200 medical cases for the evaluation dataset and standardized them into three types of task questions. We tested general and TCM-specific LLMs, comparing their performance with our proposed solution. The findings demonstrated that our method significantly enhanced LLMs' syndrome differentiation thinking. Our model achieved 85.7% in Task 1 and 81.2% accuracy in Task 2, surpassing the best-performing TCM and general LLMs by 26.3% and 15.8%, respectively. In Task 3, our model achieved a similarity score of 84.3, indicating that the model was remarkably similar to advice given by experts.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Existing general LLMs and TCM-specific LLMs continue to have significant limitations in the core task of syndrome differentiation thinking. Our research shows that fine-tuning LLMs by designing professional instruction templates and generating high-quality instruction data can significantly improve their performance on core tasks. The optimized LLMs show a high degree of similarity in reasoning results, consistent with the opinions of domain experts, indicating that they can simulate syndrome differentiation thinking to a certain extent. These findings have important theoretical and practical significance for in-depth interpretation of the complexity of the clinical diagnosis and treatment process o","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e75103"},"PeriodicalIF":3.1,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12204376/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144337292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development and Validation of a Predictive Model for Activities of Daily Living Dysfunction in Older Adults: Retrospective Analysis of Data From the China Health and Retirement Longitudinal Study.","authors":"Fangbo Lin, Chao Liu, Hua Liu","doi":"10.2196/73030","DOIUrl":"10.2196/73030","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;The global aging crisis has precipitated significant public health challenges, including rising chronic diseases, economic burdens, and labor shortages, particularly in China. Activities of daily living (ADL) dysfunction, affecting over 40 million Chinese older adults (16% of the aging population), severely compromises independence and quality of life while increasing health care costs and mortality. ADL dysfunction encompasses both basic ADL (BADL) and instrumental ADL (IADL), which assess fundamental self-care and complex environmental interactions, respectively. With projections indicating 65 million cases by 2030, there is an urgent need for tools to predict ADL impairment and enable early interventions.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to develop and validate a predictive nomogram model for ADL dysfunction in older adults using nationally representative data from the China Health and Retirement Longitudinal Study (CHARLS). The model seeks to integrate key risk factors into an accessible clinical tool to facilitate early identification of high-risk populations, guiding targeted health care strategies and resource allocation.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A retrospective analysis was conducted on 5081 CHARLS wave 3 participants (2015-2016) aged 60-80 years. Participants were categorized into ADL dysfunction (n=1743) or normal (n=3338) groups based on BADL and IADL assessments. Forty-six variables spanning demographics, health status, biomeasures, and lifestyle were analyzed. After addressing missing data via multiple imputation, Least Absolute Shrinkage and Selection Operator (LASSO) regression and multivariable logistic regression identified 6 predictors. Model performance was evaluated using receiver operating characteristic curves, calibration plots, decision curve analysis, and Shapley additive explanations (SHAP) for interpretability.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The final model incorporated 6 predictors: the 10-item Center for Epidemiologic Studies Depression Scale depression score, number of painful areas, left-hand grip strength, 2.5-m walking time, weight, and cystatin C level. The nomogram demonstrated robust discriminative power, with area under the curve values of 0.77 (95% CI 0.76-0.79) in both the training and testing sets. Calibration curves confirmed strong agreement between predicted and observed outcomes, while decision curve analysis highlighted superior clinical use over \"treat-all\" or \"treat-none\" approaches. SHAP analysis revealed depressive symptoms and physical frailty markers (eg, slow walking speed and low grip strength) as dominant predictors, aligning with existing evidence on ADL decline mechanisms.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This study presents a validated nomogram for predicting ADL dysfunction in older adult populations, combining psychological, physical, and biochemical markers. The tool enables risk stratification, supports personalized interventi","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e73030"},"PeriodicalIF":3.1,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12202242/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144334500","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The Use of Machine Learning for Analyzing Real-World Data in Disease Prediction and Management: Systematic Review.","authors":"Norah Hamad Alhumaidi, Doni Dermawan, Hanin Farhana Kamaruzaman, Nasser Alotaiq","doi":"10.2196/68898","DOIUrl":"10.2196/68898","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Machine learning (ML) and big data analytics are rapidly transforming health care, particularly disease prediction, management, and personalized care. With the increasing availability of real-world data (RWD) from diverse sources, such as electronic health records (EHRs), patient registries, and wearable devices, ML techniques present substantial potential to enhance clinical outcomes. Despite this promise, challenges such as data quality, model transparency, generalizability, and integration into clinical practice persist.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This systematic review aims to examine the use of ML for analyzing RWD in disease prediction and management, identifying the most commonly used ML methods, prevalent disease types, study designs, and the sources of real-world evidence (RWE). It also explores the strengths and limitations of current practices, offering insights for future improvements.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A comprehensive search was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to identify studies using ML techniques for analyzing RWD in disease prediction and management. The search focused on extracting data regarding the ML algorithms applied; disease categories studied; types of study designs (eg, clinical trials and cohort studies); and the sources of RWE, including EHRs, patient registries, and wearable devices. Studies published between 2014 and 2024 were included to ensure the analysis of the most recent advances in the field.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;This review identified 57 studies that met the inclusion criteria, with a total sample size of &gt;150,000 patients. The most frequently applied ML methods were random forest (n=24, 42%), logistic regression (n=21, 37%), and support vector machines (n=18, 32%). These methods were predominantly used for predictive modeling across disease areas, including cardiovascular diseases (n=19, 33%), cancer (n=9, 16%), and neurological disorders (n=6, 11%). RWE was primarily sourced from EHRs, patient registries, and wearable devices. A substantial portion of studies (n=38, 67%) focused on improving clinical decision-making, patient stratification, and treatment optimization. Among these studies, 14 (25%) focused on decision-making; 12 (21%) on health care outcomes, such as quality of life, recovery rates, and adverse events; and 11 (19%) on survival prediction, particularly in oncology and chronic diseases. For example, random forest models for cardiovascular disease prediction demonstrated an area under the curve of 0.85 (95% CI 0.81-0.89), while support vector machine models for cancer prognosis achieved an accuracy of 83% (P=.04). Despite the promising outcomes, many (n=34, 60%) studies faced challenges related to data quality, model interpretability, and ensuring generalizability across diverse patient populations.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This systematic rev","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e68898"},"PeriodicalIF":3.1,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12226786/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144334501","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Patient Perspectives on Digital Technology and Experiences of Computerized History-Taking for Chest Pain Management in the Emergency Department: CLEOS-CPDS Prospective Cohort Study.","authors":"Kaisa Fritzell, Helge Brandberg, Jonas Spaak, Sabine Koch, Carl Johan Sundberg, David Zakim, Thomas Kahan, Kay Sundberg","doi":"10.2196/65568","DOIUrl":"10.2196/65568","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Automated, self-reported medical history-taking has the potential to provide comprehensive patient-reported data across a wide range of clinical issues. In the Clinical Expert Operating System-Chest Pain Danderyd Study (CLEOS-CPDS), medical history data were entered by patients using tablets in an emergency department (ED). Since successful implementation of this technology depends on understanding patients' views and willingness to use it, we have studied these factors following patients' use of the CLEOS program.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to develop and use a questionnaire to investigate patients' attitudes, perceptions and skills related to using digital technology in health care in general, and specifically their experiences with the CLEOS program during their visit to an ED with a chief complaint of chest pain.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;The study included the development of a questionnaire, followed by a cross-sectional study. Questionnaire design and the technology acceptance model underpinned the development of the questionnaire. The think-aloud method was used to test the questionnaire. Adults who participated in the CLEOS-CPDS were invited consecutively to respond to the questionnaire. Descriptive and correlational analyses were performed.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The refinement of the questionnaire included language revision, removal of similar items, and replacement of some response formats. The final questionnaire consisted of 16 items and one free text comment that assessed attitudes, perceptions, and skills related to the use of digital technology in health care in general and the specific experience of using self-reported history-taking by CLEOS. The majority of the 129 patients (mean age 56, SD=17.3 y) who answered the questionnaire found it easy to use digital technology in general (118/129, 91%), that digital technology has a role when seeking health care (115/129, 91%), and that patient-reported symptoms are helpful in making a diagnosis (83/129, 65%). There were some concerns that the patient-physician interaction would be disrupted when using digital technology (48/129, 38%). The overall experience of using CLEOS was positive and most felt confident in answering the questions on a tablet (118/129, 91%). Older age was associated with less ease (P&lt;.001), confidence (P&lt;.001), and trust (P=.002) when using digital technology, as well as less confidence in answering the questions in CLEOS (P=.019). Moreover, older age was associated with more worry about the potential disruption of the patient-physician personal contact when using digital technology (P&lt;.001).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This study suggests strong approval of usefulness and trust in digital technology among patients with chest pain visiting a cardiology ED, but the concern for lack of personal contact should be acknowledged. End users found the CLEOS program to perform well but recommend ","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e65568"},"PeriodicalIF":3.1,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12187027/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144318846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Alert Reduction and Telemonitoring Process Optimization for Improving Efficiency in Remote Patient Monitoring Programs: Framework Development Study.","authors":"Job van Steenkiste, Niki Lupgens, Martijn Kool, Daan Dohmen, Iris Verberk-Jonkers","doi":"10.2196/66066","DOIUrl":"10.2196/66066","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Telemonitoring can enhance the efficiency of health care delivery by enabling risk stratification, thereby allowing health care professionals to focus on high-risk patients. Additionally, it reduces the need for physical care. In contrast, telemonitoring programs require a significant time investment for implementation and alert processing. A structured method for telemonitoring process optimization is lacking.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;We propose a framework for optimizing efficient care delivery in telemonitoring programs based on alert data analysis and scenario analysis of a telemonitoring program for hypertension combined with a narrative literature review on methods to improve efficient telemonitoring care delivery.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We extracted 1-year alert processing data from the telemonitoring platform and electronic health records (June 2022-May 2023) from all users participating in the hypertension telemonitoring program in the outpatient clinic of the Department of Internal Medicine of the Maasstad Hospital. We analyzed the alert burden and alert processing data. Additionally, a scenario analysis with different threshold values was conducted for existing blood pressure alerts to assess the impact of threshold adjustments on the overall alert burden and processing. We searched for English language academic research papers and conference abstracts reporting clinical alert or workflow optimization in telemonitoring programs on May 24, 2024 in Embase, Medline, Cochrane, Web of Science, and Google Scholar.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;In total, 174 users were included and analyzed. On average, each user was active in the telemonitoring program for 207 days and a total of 30,184 measurements were performed. These triggered a total of 17,293 simple, complex, and inactive or overdue alerts: 13,647 were processed automatically by the telemonitoring platform, and 3646 were processed manually by e-nurses from the telemonitoring center, equivalent to 21 manually processed alerts per user. Additional analysis of the manually processed alerts revealed that 25 (15%) users triggered more than 50% of these specific alerts. Furthermore, scenario analysis of the alert thresholds revealed that a single increase of 5 and 10 mmHg for the diastolic and systolic blood pressure alerts would reduce the number of alerts by about 50%, resulting in a total reduced time investment for the e-nurse of 5973 minutes over 1 year. Literature search yielded 251 articles, of which 7 studies reported methods to improve efficiency in telemonitoring programs, including the introduction of complex alerts and clinical algorithms to triage alerts, scenario analysis with alert threshold adjustments, and a qualitative analysis to create an alert triage algorithm.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Based on the data analysis and literature review, a 4-step framework was developed to optimize the efficiency of telemonitori","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e66066"},"PeriodicalIF":3.1,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12206671/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144289766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}