Unlabelled: In our study, a commercially available continuous ketone monitoring device captured β-Hydroxybutyrate (BHB) dynamics during exogenous ketosis but revealed a gradual decline day-to-day BHB concentrations over 14 days in both ketone ester and placebo groups, likely reflecting sensor drift.
Background: Diabetic kidney disease (DKD) is a major complication of diabetes and the leading cause of end-stage renal disease globally. Artificial intelligence (AI) technologies have shown increasing potential in DKD research for early detection, risk prediction, and disease management. However, the landscape of AI applications in this field remains incompletely mapped, especially in terms of collaboration networks, thematic evolution, and clinical translation.
Objective: This study aims to perform a comprehensive bibliometric and translational analysis of AI-related DKD research published between 2006 and 2024, identifying publication trends, research hotspots, key contributors, collaboration patterns, and the extent of clinical validation and explainability.
Methods: A systematic search of the Web of Science Core Collection was conducted to identify English-language original articles applying AI technologies to DKD. Articles were screened following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines. Bibliometric visualization was performed using CiteSpace and VOSviewer to assess coauthorship, institutional and country collaboration, keyword evolution, and citation bursts. A qualitative review was conducted to evaluate clinical validation, model explainability, and real-world implementation.
Results: Out of 1158 retrieved records, 384 studies met the inclusion criteria. Global publications on AI in DKD increased rapidly after 2019. China led in publication volume, followed by the United States, India, and Iran. Keyword analysis showed a thematic transition from early biomarker and proteomic research to deep learning, clinical prediction models, and management tools. Despite methodological advances, few studies included external validation or explainability frameworks. Notable translational efforts included DeepMind's acute kidney injury predictor and a chronic kidney disease prediction model developed by Sumit, yet widespread real-world integration remains limited.
Conclusions: AI research in DKD has grown substantially over the past 2 decades, with expanding international collaboration and diversification of research themes. However, challenges persist in clinical applicability, model transparency, and global inclusivity. Future research should prioritize explainable AI, multicenter validation, and integration into clinical workflows to support effective translation of AI innovations into DKD care.
Background: Gestational diabetes mellitus (GDM) is a prevalent chronic condition that affects maternal and fetal health outcomes worldwide, increasingly in underserved populations. While generative artificial intelligence (AI) and large language models (LLMs) have shown promise in health care, their application in GDM management remains underexplored.
Objective: This study aimed to investigate whether retrieval-augmented generation techniques, when combined with knowledge graphs (KGs), could improve the contextual relevance and accuracy of AI-driven clinical decision support. For this, we developed and validated a graph-based retrieval-augmented generation (GraphRAG)-enabled local LLM as a clinical support tool for GDM management, assessing its performance against open-source LLM tools.
Methods: A prototype clinical AI assistant was developed using a GraphRAG constructed from 1212 peer-reviewed research articles on GDM interventions, retrieved from the Semantic Scholar API (2000-2024). The GraphRAG prototype integrated entity extraction, KG construction using Neo4j, and retrieval-augmented response generation. The performance was evaluated in a simulated environment using clinical and layperson prompts, comparing the outputs of the systems against ChatGPT (OpenAI), Claude (Anthropic), and BioMistral models across 5 common natural language generation metrics.
Results: The GraphRAG-enabled local LLM showed higher accuracy in generating clinically relevant responses. It achieved a bilingual evaluation understudy score of 0.99, Jaccard similarity of 0.98, and BERTScore of 0.98, outperforming the benchmark LLMs. The prototype also produced accurate, evidence-based recommendations for clinicians and patients, demonstrating its feasibility as a clinical support tool.
Conclusions: GraphRAG-enabled local LLMs show much potential for improving personalized GDM care by integrating domain-specific evidence and contextual retrieval. Our prototype proof-of-concept serves two purposes: (1) the local LLM architecture gives practitioners from underserved locations access to state-of-the-art medical research in the treatment of chronic conditions and (2) the KG schema may be feasibly built on peer-reviewed, indexed publications, devoid of hallucinations and contextualized with patient data. We conclude that advanced AI techniques such as KGs, retrieval-augmented generation, and local LLMs improve GDM management decisions and other similar conditions and advance equitable health care delivery in resource-constrained health care environments.
Background: Continuous glucose monitors (CGM) reduce the burden of glycemic monitoring and improve glycemic control, quality of life, and decreased health care use. Despite expanded insurance coverage and adoption, barriers remain, especially in primary care. Existing research largely evaluates specific populations or interventions, leaving limited insight into the broader primary care experience.
Objective: This study aims to examine the experiences of adults with type 2 diabetes mellitus (T2DM) using CGM in primary care, guided by the Health Belief Model and Technology Acceptance Model.
Methods: This qualitative study included in-person semistructured sessions (interviews or a focus group), surveys, and electronic health record data. Participants were recruited from 3 urban primary care (internal medicine and internal medicine-pediatrics) clinics affiliated with a large academic health system in Southwest Ohio, United States, with high rates of public insurance (Medicare or Medicaid). Eligible participants were adults (≥18 y) with T2DM and a CGM prescription. Data were analyzed using theme generation guided by directed content analysis in MAXQDA (VERBI Software GmbH) with codes derived from Health Belief Model and Technology Acceptance Model constructs. Survey data were used to triangulate to enhance validity.
Results: Overall, 16 participants (interviews: n=12; 1 focus group: n=4) were recruited for the study with a mean age of 56.9 (SD 10.5) years. In total, 69% (11/16) identified as Black, 100% (16/16) as Non-Hispanic, and 69% (11/16) as female, and 94% (15/16) used public insurance. Six themes emerged: disease susceptibility, disease severity, influential drivers, perceived ease of use, perceived usefulness, and attitude toward using CGM. All participants found CGM helpful and would recommend it to others. While affirming numerous barriers well-described in other populations, this study uniquely describes the burden of comorbidities, the trust in CGM data compared to glucometer-based monitoring, and the reliance on receivers to use CGM technology in this patient population.
Conclusions: CGM is valued by adults with T2DM in primary care, yet barriers remain. Tailored support for initiation, troubleshooting, and education (especially alarm management and data interpretation) is needed. These insights can inform scalable strategies to enhance CGM use and experience in primary care.
Background: There are many mobile apps for diabetes self-management; however, most target Western populations and lack dietary content relevant to Asian contexts. Our mobile app addresses this gap by providing self-care tools and a database of regionally relevant foods.
Objective: This study aimed to evaluate the effectiveness of the app in improving glycemic control and self-care behaviors among outpatients with uncontrolled type 2 diabetes at our hospital.
Methods: We conducted a randomized controlled trial with adults with type 2 diabetes, hemoglobin A1c (HbA1c) of >7%, and access to a smartphone. Participants were randomized to an intervention group (daily use of the Rama Diabetes Care app) or a control group (standard care), with all receiving diabetes self-management education and support. The app includes 6 features, notably a nutritional logging system with a verified database of Thai and commonly consumed foods, including Asian and Western dishes, as well as blood glucose monitoring, exercise and medication tracking, symptom screening, and weight logging. The primary outcome was HbA1c level, and secondary outcomes included fasting plasma glucose (FPG), low-density lipoprotein cholesterol, estimated glomerular filtration rate, BMI, self-care behaviors, and user satisfaction with the app. The study was conducted between November 29, 2023, and October 30, 2024.
Results: A total of 129 participants were randomized (intervention: n=64, 49.6%; control: n=65, 50.4%). Participants in the intervention group were younger (mean age 54.6, SD 14.3 years vs 61.9, SD 12.0 years; P=.002), whereas baseline HbA1c (mean 9.3%, SD 1.96%) and FPG (mean 179.5, SD 5.9 mg/dL) levels were similar between the groups. Over 6 months, the intervention group showed a greater HbA1c reduction than the control group (mean difference -0.24%), but the difference was not statistically significant (P=.13). Among participants aged <65 years, FPG at 6 months was significantly lower in the intervention group (mean difference -29.3 mg/dL; P=.03). App satisfaction was rated as moderate.
Conclusions: The mobile app achieved glycemic control comparable to that achieved through standard care, with significant improvement in FPG among participants younger than 65 years. Tailor-made apps integrating regionally relevant dietary content may support effective self-management in type 2 diabetes and warrant further evaluation in larger, long-term studies.
Background: Insulin therapy is crucial for managing type 2 diabetes mellitus, with its use steadily increasing in Indonesia and its effectiveness well established. However, prescribing insulin poses various challenges that can impact the effectiveness of insulin. Patient education is crucial for the successful implementation of insulin therapy. Proper insulin use remains insufficient in Indonesia.
Objective: This study aims to investigate physicians' knowledge and practice in providing education on insulin use to patients with type 2 diabetes mellitus in Indonesia.
Methods: This study recruited potential participants (all physicians in Indonesia) through the internet using a convenience sampling method. The participants were asked to fill out a questionnaire. The questionnaire had 32 questions divided into 4 sections: demographics and clinical practice, practice of insulin education, the Indonesian insulin injection technique guideline, and knowledge of insulin injection techniques. The instrument used in this study was developed based on the Pedoman Teknik Menyuntik Insulin Indonesia, which was adapted from the international consensus by the Forum for Injection Technique and Therapy Expert Recommendations. The survey lasted from February 2021 to March 2021. Data were analyzed using the Kruskal-Wallis tests.
Results: A total of 823 participants were included in the analysis. Out of 823 participants, 680 (82.6%) had given insulin education to patients at least once during the last 30 days. However, out of 823 participants, only 479 (58.2%) used specific guidelines in their practice, with only 280 (34.0%) aware of the Indonesian guidelines. Out of 823 participants, 815 (99.1%) agreed that insulin injection techniques would affect clinical results. The median score of knowledge about insulin injection techniques was 7 (IQR 2) among the study participants, indicating good knowledge. Profession was the only variable significantly associated with knowledge scores, with consultants in endocrinology, metabolism, and diabetes achieving the highest median scores, and other physicians the lowest (P<.001).
Conclusions: Most physicians in this study reported providing education to their patients. However, there was still a gap between the guidelines and the practice of insulin education, as indicated by the lack of awareness and a fair level of knowledge about the Indonesian guidelines.
Unlabelled: We developed an innovative bilingual toolkit comprising a personalized action plan and educational videos to encourage insulin dose self-titration by adults living with type 2 diabetes.
Background: Basal rate (BR) adjustment is crucial for managing type 1 diabetes mellitus, accounting for 30% to 50% of total daily insulin needs. All current closed-loop systems revert to the user's usual pump BR (known as manual mode) in the event of closed loop failure. Furthermore, access to closed-loop systems remains relatively low in low- and middle-income countries and among those without suitable health insurance. Accurately adjusting the BR remains challenging, leading to hypo- or hyperglycemia, and research on optimizing the BR is limited.
Objective: This study proposed an adaptive algorithm that uses continuous glucose monitoring data to identify BR inaccuracies without requiring meal intake information.
Methods: The OhioT1DM dataset formed the basis for implementing this methodology. Each composite day was generated by excluding bolus insulin profiles lacking meal intake information and by calculating hourly blood glucose (BG) relative levels along with their corresponding reliability measures, enabling assessment of deviations from the recommended BR (ie, a BG relative change of 0 mg/dL). Both a noninferiority analysis and a classification precision metric were used to assess the practicality of this approach compared to using meal data.
Results: Data from 12 participants showed noninferiority of the no-meal method: using a 20% noninferiority margin on absolute BG relative change, 9 of 12 participants met the criterion (1-sided P<.05). Classification precision was 73.9% (139/188) of meals correctly classified on average per participant (SD 11.8%; 95% CI 67.2%-79.7%). The daily cumulative BG average was 200.6 mg/dL (SD 61.7 mg/dL; 11.1 mmol/L, SD 3.4 mmol/L; 95% CI 161.4-239.8 mg/dL), with peak values reaching 270.15 mg/dL (14.99 mmol/L). Furthermore, 99.3% (286/288) of the BG relative values (SD 0.5%; 95% CI 97.5%-99.8%) that were unaffected by external factors were associated with incorrect BR settings, with deviations ranging from -25.5 to 46 mg/dL (-1.58 to 2.59 mmol/L).
Conclusions: Current strategies to optimize BR settings are inadequate, and our approach of a personalized basal tuner (PBT) helps better analyze BR without relying on meal intake information. Indeed, without an optimally set BR, in the event of the closed loop reverting to manual mode, patients may be exposed to persistent hypo- or hyperglycemia, leading to safety and efficacy issues. Future work will focus on generating BR recommendations through the application of this algorithm in clinical practice to assist clinicians in setting BR in low- and middle-income countries, where closed-loop systems are not prevalent, to help increase time in range.
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：