Journal of Medical Internet Research最新文献

{"title":"Using Synthetic Health Care Data to Leverage Large Language Models for Named Entity Recognition: Development and Validation Study.","authors":"Hendrik Šuvalov, Mihkel Lepson, Veronika Kukk, Maria Malk, Neeme Ilves, Hele-Andra Kuulmets, Raivo Kolde","doi":"10.2196/66279","DOIUrl":"https://doi.org/10.2196/66279","url":null,"abstract":"<p><strong>Background: </strong>Named entity recognition (NER) plays a vital role in extracting critical medical entities from health care records, facilitating applications such as clinical decision support and data mining. Developing robust NER models for low-resource languages, such as Estonian, remains a challenge due to the scarcity of annotated data and domain-specific pretrained models. Large language models (LLMs) have proven to be promising in understanding text from any language or domain.</p><p><strong>Objective: </strong>This study addresses the development of medical NER models for low-resource languages, specifically Estonian. We propose a novel approach by generating synthetic health care data and using LLMs to annotate them. These synthetic data are then used to train a high-performing NER model, which is applied to real-world medical texts, preserving patient data privacy.</p><p><strong>Methods: </strong>Our approach to overcoming the shortage of annotated Estonian health care texts involves a three-step pipeline: (1) synthetic health care data are generated using a locally trained GPT-2 model on Estonian medical records, (2) the synthetic data are annotated with LLMs, specifically GPT-3.5-Turbo and GPT-4, and (3) the annotated synthetic data are then used to fine-tune an NER model, which is later tested on real-world medical data. This paper compares the performance of different prompts; assesses the impact of GPT-3.5-Turbo, GPT-4, and a local LLM; and explores the relationship between the amount of annotated synthetic data and model performance.</p><p><strong>Results: </strong>The proposed methodology demonstrates significant potential in extracting named entities from real-world medical texts. Our top-performing setup achieved an F₁-score of 0.69 for drug extraction and 0.38 for procedure extraction. These results indicate a strong performance in recognizing certain entity types while highlighting the complexity of extracting procedures.</p><p><strong>Conclusions: </strong>This paper demonstrates a successful approach to leveraging LLMs for training NER models using synthetic data, effectively preserving patient privacy. By avoiding reliance on human-annotated data, our method shows promise in developing models for low-resource languages, such as Estonian. Future work will focus on refining the synthetic data generation and expanding the method's applicability to other domains and languages.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e66279"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657217","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Application of Internet Hospitals in the Disease Management of Patients With Ulcerative Colitis: Retrospective Study.","authors":"Tianzhi Yu, Wanyu Li, Yingchun Liu, Chunjie Jin, Zimin Wang, Hailong Cao","doi":"10.2196/60019","DOIUrl":"https://doi.org/10.2196/60019","url":null,"abstract":"<p><strong>Background: </strong>Ulcerative colitis (UC) is a chronic disease characterized by frequent relapses, requiring long-term management and consuming substantial medical and social resources. Effective management of UC remains challenging due to the need for sustainable remission strategies, continuity of care, and access to medical services. Intelligent diagnosis refers to the use of artificial intelligence-driven algorithms to analyze patient-reported symptoms, generate diagnostic probabilities, and provide treatment recommendations through interactive tools. This approach could potentially function as a method for UC management.</p><p><strong>Objective: </strong>This study aimed to analyze the diagnosis and treatment data of UC from both physical hospitals and internet hospitals, highlighting the potential benefits of the intelligent diagnosis and treatment service model offered by internet hospitals.</p><p><strong>Methods: </strong>We collected data on the visits of patients with UC to the Department of Gastroenterology at Tianjin Medical University General Hospital. A total of 852 patients with UC were included between July 1, 2020, and June 31, 2023. Statistical methods, including chi-square tests for categorical variables, t tests for continuous variables, and rank-sum tests for visit numbers, were used to evaluate the medical preferences and expenses of patients with UC.</p><p><strong>Results: </strong>We found that internet hospitals and physical hospitals presented different medical service models due to the different distribution of medical needs and patient groups. Patients who chose internet hospitals focused on disease consultation and prescription medication (3295/3528, 93.40%). Patients' medical preferences gradually shifted to web-based services provided by internet hospitals. Over time, 58.57% (270/461) of patients chose either web-based services or a combination of web-based and offline services for UC diagnosis and treatment. The number of visits in the combination of web-based and offline service modes was the highest (mean 13.83, SD 11.07), and younger patients were inclined to visit internet hospitals (49.66%>34.71%). In addition, compared with physical hospitals, there was no difference in testing fees and examination fees for patients with UC in internet hospitals, but medicine fees were lower.</p><p><strong>Conclusions: </strong>The intelligent diagnosis and treatment model provided by internet hospitals demonstrates the potential benefits in managing UC, including feasibility, accessibility, convenience, and economics.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e60019"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657476","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Digital Physical Activity and Sedentary Behavior Interventions for Community-Living Adults: Umbrella Review.","authors":"Eilidh Russell, Alison Kirk, Mark D Dunlop, William Hodgson, Mhairi Patience, Kieren Egan","doi":"10.2196/66294","DOIUrl":"https://doi.org/10.2196/66294","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Digital interventions hold significant potential for improving physical activity (PA) and reducing sedentary behavior (SB) in adults. Despite increasing interest, there remain surprising gaps in the current knowledge of how best to deliver these interventions, including incorporating appropriate theoretical frameworks and behavior change techniques. Following numerous systematic reviews, there is now significant potential for umbrella reviews to provide an overview of the current evidence.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This umbrella review aimed to explore digital PA and SB interventions for community-living adults across effectiveness, key components, and methodological quality.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;This review followed the Joanna Briggs Institute framework for umbrella reviews. Key search terms were developed iteratively, incorporating physical and sedentary activity alongside digital interventions. We searched 7 online databases (Web of Science Core Collection, CINAHL, APA PsycINFO, Inspec, the Cochrane Library, MEDLINE [Ovid], and PROSPERO) alongside gray literature databases. Information was extracted and tabulated from each included article on intervention effectiveness, key components, and content acknowledging both the digital and human elements. The study quality was appraised using A Measurement Tool to Assess systematic Reviews 2 (AMSTAR 2). The corrected covered area method was used to assess the overlap of primary studies included in the systematic reviews. All relevant research findings were extracted and reported.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Search terms identified 330 articles, of which 5 (1.5%) met the inclusion criteria. The most common PA outcomes identified were daily steps, moderate-to-vigorous PA, total PA, and PA change. Reviews with meta-analysis reported that digital interventions improved multiple PA outcomes (daily steps, moderate-to-vigorous PA time, and total PA time). However, findings from the remaining systematic reviews were mixed. Similarly, the findings for SB were contrasting. Regarding intervention components, monitor- and sensor-only intervention delivery methods were most frequently implemented. Eleven theoretical frameworks were identified, with social cognitive theory being the most prominent theory. In total, 28 different behavior change techniques were reported, with goal setting, self-monitoring, feedback, and social support being the most frequently used. All 5 systematic reviews were of low or critically low quality, each incorporating unique primary studies (corrected covered area=0%).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This umbrella review highlights the potential of digital interventions to increase PA and reduce SB among community-living adults. However, the disparate nature of current academic knowledge means potentially efficacious research may not realistically translate to real work impact. Our review identified a lack of consensus around out","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e66294"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prompt Framework for Extracting Scale-Related Knowledge Entities from Chinese Medical Literature: Development and Evaluation Study.","authors":"Jie Hao, Zhenli Chen, Qinglong Peng, Liang Zhao, Wanqing Zhao, Shan Cong, Junlian Li, Jiao Li, Qing Qian, Haixia Sun","doi":"10.2196/67033","DOIUrl":"https://doi.org/10.2196/67033","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Measurement-based care improves patient outcomes by using standardized scales, but its widespread adoption is hindered by the lack of accessible and structured knowledge, particularly in unstructured Chinese medical literature. Extracting scale-related knowledge entities from these texts is challenging due to limited annotated data. While large language models (LLMs) show promise in named entity recognition (NER), specialized prompting strategies are needed to accurately recognize medical scale-related entities, especially in low-resource settings.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aims to develop and evaluate MedScaleNER, a task-oriented prompt framework designed to optimize LLM performance in recognizing medical scale-related entities from Chinese medical literature.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;MedScaleNER incorporates demonstration retrieval within in-context learning, chain-of-thought prompting, and self-verification strategies to improve performance. The framework dynamically retrieves optimal examples using a k-nearest neighbors approach and decomposes the NER task into two subtasks: entity type identification and entity labeling. Self-verification ensures the reliability of the final output. A dataset of manually annotated Chinese medical journal papers was constructed, focusing on three key entity types: scale names, measurement concepts, and measurement items. Experiments were conducted by varying the number of examples and the proportion of training data to evaluate performance in low-resource settings. Additionally, MedScaleNER's performance was compared with locally fine-tuned models.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The CMedS-NER (Chinese Medical Scale Corpus for Named Entity Recognition) dataset, containing 720 papers with 27,499 manually annotated scale-related knowledge entities, was used for evaluation. Initial experiments identified GLM-4-0520 as the best-performing LLM among six tested models. When applied with GLM-4-0520, MedScaleNER significantly improved NER performance for scale-related entities, achieving a macro F&lt;sub&gt;1&lt;/sub&gt;-score of 59.64% in an exact string match with the full training dataset. The highest performance was achieved with 20-shot demonstrations. Under low-resource scenarios (eg, 1% of the training data), MedScaleNER outperformed all tested locally fine-tuned models. Ablation studies highlighted the importance of demonstration retrieval and self-verification in improving model reliability. Error analysis revealed four main types of mistakes: identification errors, type errors, boundary errors, and missing entities, indicating areas for further improvement.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;MedScaleNER advances the application of LLMs and prompts engineering for specialized NER tasks in Chinese medical literature. By addressing the challenges of unstructured texts and limited annotated data, MedScaleNER's adaptability to various biomedical contexts supports mo","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e67033"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657126","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Digital Health Education for Chronic Lung Disease: Scoping Review.","authors":"Chao Sun, Huohuo Dai, Rianne M J J van der Kleij, Rong Li, Hengchang Wu, Cynthia Hallensleben, Sofie H Willems, Niels H Chavannes","doi":"10.2196/53142","DOIUrl":"https://doi.org/10.2196/53142","url":null,"abstract":"<p><strong>Background: </strong>Chronic lung disease (CLD) is one of the most prevalent noncommunicable diseases globally, significantly burdening patients and increasing health care expenditures. Digital health education (DHE) is increasingly important in chronic disease prevention and management. However, DHE characteristics and impacts in CLD are rarely reported.</p><p><strong>Objective: </strong>This study aimed to provide an overview of the existing literature on DHE for CLD, with a focus on exploring the DHE mediums, content, mechanisms, and reported outcomes in patients with CLD.</p><p><strong>Methods: </strong>We searched PubMed, Web of Science, Embase, PsycINFO, and The Cochrane Library with the assistance of a librarian specialist. Articles were screened by the reviewer team with ASReview (Utrecht University) and EndNote X9 (Clarivate Analytics) based on predefined inclusion and exclusion criteria and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist. Quality assessment was conducted with the Critical Appraisal Skills Program tool. A descriptive analysis was used to summarize the study characteristics, DHE characteristics, and outcomes.</p><p><strong>Results: </strong>A total of 22 studies were included in this review with medium or high quality. They were published between 2000 and 2022, showing an increasing publication trend with the year, mostly in developed countries (16/22, 73%). Websites and mobile apps (10/22, 45%) were the most widely used DHE medium. Education on self-management skills of CLD was the primary topic (21/22, 95%), 4/22 (18%) of which mentioned DHE mechanisms. The majority of studies reported positive changes in CLD awareness (14/16, 88%), clinical outcomes (3/6, 50%), DHE feasibility, acceptability, and satisfaction (6/8, 75%), lifestyle outcomes (3/3, 100%), and psychosocial outcomes (7/8, 88%). Only 2 studies reported cost-effectiveness (2/22, 9%).</p><p><strong>Conclusions: </strong>Despite the heterogeneity of the study situation, some aspects can be concluded. DHE can improve disease awareness and clinical outcomes in patients with chronic lung disease, with good feasibility, acceptability, and satisfaction through different mediums and learning content. There is still relatively little research among people in low- and middle-income countries. Future research should consider the impact on cost-effectiveness, duration, frequency, and theoretical mechanisms of the DHE to maximize the potential impact. It should also be conducted in the context of health services research to better reflect the real world.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e53142"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Promoting Public Engagement in Palliative and End-of-Life Care Discussions on Chinese Social Media: Model Development and Analysis.","authors":"Yijun Wang, Han Zheng, Yuxin Zhou, Emeka Chukwusa, Jonathan Koffman, Vasa Curcin","doi":"10.2196/59944","DOIUrl":"https://doi.org/10.2196/59944","url":null,"abstract":"<p><strong>Background: </strong>In Chinese traditional culture, discussions surrounding death are often considered taboo, leading to a poor quality of death, and limited public awareness and knowledge about palliative and end-of-life care (PEoLC). However, the increasing prevalence of social media in health communication in China presents an opportunity to promote and educate the public about PEoLC through online discussions.</p><p><strong>Objective: </strong>This study aimed to examine the factors influencing public engagement in PEoLC discussions on a Chinese social media platform and develop practice recommendations to promote such engagement.</p><p><strong>Methods: </strong>We gathered 30,811 PEoLC-related posts on Weibo, the largest social media platform in China. Guided by the elaboration likelihood model, our study examined factors across 4 dimensions: content theme, mood, information richness, and source credibility. Content theme was examined using thematic analysis, while sentiment analysis was used to determine the mood of the posts. The impact of potential factors on post engagement was quantified using negative binomial regression.</p><p><strong>Results: </strong>Organizational accounts exhibited lower engagement compared to individual accounts (incidence rate ratio [IRR]<1; P<.001), suggesting an underuse of organizational accounts in advocating for PEoLC on Weibo. Posts centered on PEoLC-related entertainment (films, television shows, and books; IRR=1.37; P<.001) or controversial social news (IRR=1.64; P<.001) garnered more engagement, primarily published by individual accounts. An interaction effect was observed between content theme and post mood, with posts featuring more negative sentiment generally attracting higher public engagement, except for educational-related posts (IRR=2.68; P<.001).</p><p><strong>Conclusions: </strong>Overall, organizations faced challenges in capturing public attention and involving the public when promoting PEoLC on Chinese social media platforms. It is imperative to move beyond a traditional mode to incorporate cultural elements of social media, such as engaging influencers, leveraging entertainment content and social news, or using visual elements, which can serve as effective catalysts in attracting public attention. The strategies developed in this study are particularly pertinent to nonprofit organizations and academics aiming to use social media for PEoLC campaigns, fundraising efforts, or research dissemination.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e59944"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing Patient Outcome Prediction Through Deep Learning With Sequential Diagnosis Codes From Structured Electronic Health Record Data: Systematic Review.","authors":"Tuankasfee Hama, Mohanad M Alsaleh, Freya Allery, Jung Won Choi, Christopher Tomlinson, Honghan Wu, Alvina Lai, Nikolas Pontikos, Johan H Thygesen","doi":"10.2196/57358","DOIUrl":"https://doi.org/10.2196/57358","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;The use of structured electronic health records in health care systems has grown rapidly. These systems collect huge amounts of patient information, including diagnosis codes representing temporal medical history. Sequential diagnostic information has proven valuable for predicting patient outcomes. However, the extent to which these types of data have been incorporated into deep learning (DL) models has not been examined.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This systematic review aims to describe the use of sequential diagnostic data in DL models, specifically to understand how these data are integrated, whether sample size improves performance, and whether the identified models are generalizable.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Relevant studies published up to May 15, 2023, were identified using 4 databases: PubMed, Embase, IEEE Xplore, and Web of Science. We included all studies using DL algorithms trained on sequential diagnosis codes to predict patient outcomes. We excluded review articles and non-peer-reviewed papers. We evaluated the following aspects in the included papers: DL techniques, characteristics of the dataset, prediction tasks, performance evaluation, generalizability, and explainability. We also assessed the risk of bias and applicability of the studies using the Prediction Model Study Risk of Bias Assessment Tool (PROBAST). We used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist to report our findings.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Of the 740 identified papers, 84 (11.4%) met the eligibility criteria. Publications in this area increased yearly. Recurrent neural networks (and their derivatives; 47/84, 56%) and transformers (22/84, 26%) were the most commonly used architectures in DL-based models. Most studies (45/84, 54%) presented their input features as sequences of visit embeddings. Medications (38/84, 45%) were the most common additional feature. Of the 128 predictive outcome tasks, the most frequent was next-visit diagnosis (n=30, 23%), followed by heart failure (n=18, 14%) and mortality (n=17, 13%). Only 7 (8%) of the 84 studies evaluated their models in terms of generalizability. A positive correlation was observed between training sample size and model performance (area under the receiver operating characteristic curve; P=.02). However, 59 (70%) of the 84 studies had a high risk of bias.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;The application of DL for advanced modeling of sequential medical codes has demonstrated remarkable promise in predicting patient outcomes. The main limitation of this study was the heterogeneity of methods and outcomes. However, our analysis found that using multiple types of features, integrating time intervals, and including larger sample sizes were generally related to an improved predictive performance. This review also highlights that very few studies (7/84, 8%) reported on challenges related to generalizability and less t","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e57358"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine Learning to Assist in Managing Acute Kidney Injury in General Wards: Multicenter Retrospective Study.","authors":"Nam-Jun Cho, Inyong Jeong, Se-Jin Ahn, Hyo-Wook Gil, Yeongmin Kim, Jin-Hyun Park, Sanghee Kang, Hwamin Lee","doi":"10.2196/66568","DOIUrl":"https://doi.org/10.2196/66568","url":null,"abstract":"<p><strong>Background: </strong>Most artificial intelligence-based research on acute kidney injury (AKI) prediction has focused on intensive care unit settings, limiting their generalizability to general wards. The lack of standardized AKI definitions and reliance on intensive care units further hinder the clinical applicability of these models.</p><p><strong>Objective: </strong>This study aims to develop and validate a machine learning-based framework to assist in managing AKI and acute kidney disease (AKD) in general ward patients, using a refined operational definition of AKI to improve predictive performance and clinical relevance.</p><p><strong>Methods: </strong>This retrospective multicenter cohort study analyzed electronic health record data from 3 hospitals in South Korea. AKI and AKD were defined using a refined version of the Kidney Disease: Improving Global Outcomes criteria, which included adjustments to baseline serum creatinine estimation and a stricter minimum increase threshold to reduce misclassification due to transient fluctuations. The primary outcome was the development of machine learning models for early prediction of AKI (within 3 days before onset) and AKD (nonrecovery within 7 days after AKI).</p><p><strong>Results: </strong>The final analysis included 135,068 patients. A total of 7658 (8%) patients in the internal cohort and 2898 (7.3%) patients in the external cohort developed AKI. Among the 5429 patients in the internal cohort and 1998 patients in the external cohort for whom AKD progression could be assessed, 896 (16.5%) patients and 287 (14.4%) patients, respectively, progressed to AKD. Using the refined criteria, 2898 cases of AKI were identified, whereas applying the standard Kidney Disease: Improving Global Outcomes criteria resulted in the identification of 5407 cases. Among the 2509 patients who were not classified as having AKI under the refined criteria, 2242 had a baseline serum creatinine level below 0.6 mg/dL, while the remaining 267 experienced a decrease in serum creatinine before the onset of AKI. The final selected early prediction model for AKI achieved an area under the receiver operating characteristic curve of 0.9053 in the internal cohort and 0.8860 in the external cohort. The early prediction model for AKD achieved an area under the receiver operating characteristic curve of 0.8202 in the internal cohort and 0.7833 in the external cohort.</p><p><strong>Conclusions: </strong>The proposed machine learning framework successfully predicted AKI and AKD in general ward patients with high accuracy. The refined AKI definition significantly reduced the classification of patients with transient serum creatinine fluctuations as AKI cases compared to the previous criteria. These findings suggest that integrating this machine learning framework into hospital workflows could enable earlier interventions, optimize resource allocation, and improve patient outcomes.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e66568"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143656755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Emotion Forecasting: A Transformer-Based Approach.","authors":"Leire Paz-Arbaizar, Jorge Lopez-Castroman, Antonio Artés-Rodríguez, Pablo M Olmos, David Ramírez","doi":"10.2196/63962","DOIUrl":"https://doi.org/10.2196/63962","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Monitoring the emotional states of patients with psychiatric problems has always been challenging due to the noncontinuous nature of clinical assessments, the effect of the health care environment, and the inherent subjectivity of evaluation instruments. However, mental states in psychiatric disorders exhibit substantial variability over time, making real-time monitoring crucial for preventing risky situations and ensuring appropriate treatment.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to leverage new technologies and deep learning techniques to enable more objective, real-time monitoring of patients. This was achieved by passively monitoring variables such as step count, patient location, and sleep patterns using mobile devices. We aimed to predict patient self-reports and detect sudden variations in their emotional valence, identifying situations that may require clinical intervention.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Data for this project were collected using the Evidence-Based Behavior (eB2) app, which records both passive and self-reported variables daily. Passive data refer to behavioral information gathered via the eB2 app through sensors embedded in mobile devices and wearables. These data were obtained from studies conducted in collaboration with hospitals and clinics that used eB2. We used hidden Markov models (HMMs) to address missing data and transformer deep neural networks for time-series forecasting. Finally, classification algorithms were applied to predict several variables, including emotional state and responses to the Patient Health Questionnaire-9.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Through real-time patient monitoring, we demonstrated the ability to accurately predict patients' emotional states and anticipate changes over time. Specifically, our approach achieved high accuracy (0.93) and a receiver operating characteristic (ROC) area under the curve (AUC) of 0.98 for emotional valence classification. For predicting emotional state changes 1 day in advance, we obtained an ROC AUC of 0.87. Furthermore, we demonstrated the feasibility of forecasting responses to the Patient Health Questionnaire-9, with particularly strong performance for certain questions. For example, in question 9, related to suicidal ideation, our model achieved an accuracy of 0.9 and an ROC AUC of 0.77 for predicting the next day's response. Moreover, we illustrated the enhanced stability of multivariate time-series forecasting when HMM preprocessing was combined with a transformer model, as opposed to other time-series forecasting methods, such as recurrent neural networks or long short-term memory cells.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;The stability of multivariate time-series forecasting improved when HMM preprocessing was combined with a transformer model, as opposed to other time-series forecasting methods (eg, recurrent neural network and long short-term memory), leveraging the attention mechanisms to capture ","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e63962"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143657482","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integration of Psychiatric Advance Directives Into the Patient-Accessible Electronic Health Record: Exploring the Promise and Limitations.","authors":"Julian Schwarz, Eva Meier-Diedrich, Matthé Scholten, Lucy Stephenson, John Torous, Florian Wurster, Charlotte Blease","doi":"10.2196/68549","DOIUrl":"https://doi.org/10.2196/68549","url":null,"abstract":"<p><p>Psychiatric advance directives (PAD), also known as advance statements or advance choice documents, are legal documents that enable people with mental health conditions to specify their treatment preferences in advance for possible future crises. Subtypes of PADs include crisis cards, joint crisis plans, and self-binding directives (also known as Ulysses contracts). These instruments are intended to improve service user involvement and need orientation in the care of mental crises and to avoid traumatization through unwanted treatment. The existing evidence suggests that people who complete a PAD tend to work more cooperatively with their clinician and experience fewer involuntary hospital admissions. Nevertheless, PADs have not been successfully mainstreamed into care due to multiple barriers to the implementation of PADs, mainly around the completion of PADs and their accessibility and use in crises. The reasons for this include the lack of support in the completion process and acceptance problems, especially on the part of professionals. The research to date primarily recommends support for service users from facilitators, such as peer support workers, and training for all stakeholders. In this article, we argue that while these approaches can help to solve completion and acceptance challenges, they are not sufficient to ensure access to PADs in crises. To ensure accessibility, we propose digital PADs, which offer considerable potential for overcoming these aforementioned barriers. Embedded in national health data infrastructures, PADs could be completed and accessed by service users themselves, possibly with the support of facilitators, and retrieved by any clinic in an emergency. We highlight the strengths and limitations of digital PADs and point out that the proposed solutions must be developed collaboratively and take into account digital inequalities to be effective support for people with serious mental health conditions.</p>","PeriodicalId":16337,"journal":{"name":"Journal of Medical Internet Research","volume":"27 ","pages":"e68549"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143656752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}