Sage Journals: Discover world-class research

Abstract

The escalating prevalence of diabetes, along with its complications and mortality risks, imposes a substantial disease burden worldwide. The current suboptimal medical conditions and poor self-management among diabetic patients have exacerbated the deterioration of diabetes globally, particularly in economically underdeveloped countries. However, this situation may now be approaching a turning point. With the constantly advancement of intelligent technologies, the widespread adoption of information management systems and the rise of artificial intelligence have made it possible to enhance the efficiency of diabetes treatment and reduce management costs. Therefore, we have reviewed the relevant literature and conducted a narrative review following the guidance of the Scale for the Assessment of Narrative Review Articles (SANRA). The present paper provides a narrative review of research advances from information management to artificial intelligence in the field of diabetes treatment and management, while also discussing the opportunities and challenges in clinical translation and application. The present review offers a conceptual framework to inform future research and development in intelligent diabetes care.

Keywords

Diabetes artificial intelligence diagnosis informatization management precision prevention

Introduction

Nowadays, unhealthy lifestyles including prolonged sedentary behavior and high-fat and high sugar diets have led to the prevalence of diabetes increases in a staggering manner.¹ According to the latest World Health Organization (WHO) survey, approximately 830 million people worldwide suffer from diabetes in 2025.² Moreover, nearly half of all global diabetes cases remain undiagnosed, with 60% in low-income countries.² Additionally, about 70% of individuals with prediabetes will progress to diabetes, yet only a minority receive lifestyle interventions.³ Reports indicate that diabetes directly causes 6.7 million deaths annually.⁴ Furthermore, diabetes ranks as the ninth leading cause of disability globally, severely impacting quality of life.⁵

Notably, one-third of type 1 diabetes patients worldwide still lack stable access to insulin and other effective treatment options (particularly in Africa and South Asia).⁶ Innovative drugs such as glucagon-like peptide-1 (GLP-1) receptor agonists and sodium-glucose co-transporter-2 (SGLT-2) inhibitors are difficult to be popularized in low-income countries.^7–9 While continuous glucose monitoring has an adoption rate exceeding 30% in Europe and America, it remains below 5% in developing countries.^10–12 Additionally, traditional fingerstick blood sampling remains predominant, yet pain and operational complexity compromise adherence.^13,14 Diabetes patients face 2–3 times higher risks of depression and anxiety compared to the general population.^15,16 Only 30%–40% of patients can sustain long-term healthy lifestyles (diet, exercise, medication).^17,18 Despite continuous medical advancements and some progress in global diabetes management, challenges persist, including low diagnosis rates, treatment inequities, and insufficient complication prevention. Optimizing diabetes management has become an urgent public health priority.

Over the past period, information-based solutions have transformed diabetes management to a certain extent in five key dimensions: closed-loop of whole-cycle health management, precision medical decision support, intelligent patient self-management, efficient healthcare resource allocation, and structured data value mining.^19,20 However, current information-based solutions in diabetes management still have limitations involving only enabling static data recording, simple rule-based alerts, lacking capabilities for dynamic prediction and personalized intervention. Surprisingly, artificial intelligence is becoming increasingly prominent with core advantages in intelligent analysis and proactive intervention. For example, machine learning enables real-time analysis of multidimensional data such as blood glucose, diet, and exercise, which could dynamically predict risks of hypo- or hyperglycemia and provide early warnings.^21–24 Combined with personalized algorithms, it automatically optimizes insulin dosages or generates tailored recommendations, significantly improving glycemic control precision.^25–28 Additionally, artificial intelligence can deeply uncover hidden correlations, such as the impact of sleep on blood glucose, to assist in early screening for complications and drive behavioral changes in patients through intelligent interactions, such as voice assistants, achieving a transition from “passive recording” to “active health management”^29–31 (Figure 1).

Figure 1.

Evolution of technology-driven diabetes management paradigms. Comparative framework illustrating three progressive stages in digital diabetes care, highlighting transitions in core objectives, technological enablers, data characteristics, interaction modes, and fundamental paradigm shifts from data digitization to artificial intelligence-driven precision management.

This paper conducts a narrative review following the guidance of the Scale for the Assessment of Narrative Review Articles (SANRA) and systematically reviews the research advances in core areas of diabetes management from informatization to artificial intelligence, elucidates the key advantages of artificial intelligence in predictive intervention and personalized decision-making, explores cutting-edge directions such as federated learning and digital twins, and identifies opportunities and challenges in clinical translation and application. The literature for this narrative review was identified through searches in PubMed, Web of Science, and IEEE Xplore. The search strategy combined key terms: (a) diabetes; (b) digital health, artificial intelligence, machine learning; and (c) prevention, management. We included peer-reviewed, English-language articles (2000–present) that focus on the technological applications in diabetes care. We excluded articles that: (a) solely described general IT infrastructure without clinical or behavioral focus; (b) involved nonhuman studies; or (c) focused solely on type 1 diabetes without generalizable insights to prevention/management. The final selection aims to map the evolution from informatization to intelligent systems. The present review aims to provide technical selection references for clinical practice and promote the implementation of precision diabetes management by integrating multidisciplinary evidence.

Information technology architecture for diabetes management

Data acquisition and standardization

The data collection and standardization of diabetes information are the critical foundations for constructing an intelligent management system. Currently, the primary integration involves three categories of data sources. The electronic medical record system integrates multiple data modalities with distinct structural characteristics. Structured laboratory test results represent one key component, with common measures including hemoglobin A1c and fasting blood glucose levels and other index, all uniformly coded using the logical observation identifiers names and codes standard (LOINC).^32,33 Medication records follow rigorous classification through the World Health Organization anatomical therapeutic chemical (WHO-ATC) system.^34,35 Additionally, the system contains substantial unstructured clinical documentation, particularly physician progress notes, which require specialized natural language processing techniques for effective data extraction and subsequent analysis.^36–38 Moreover, the system incorporates diverse monitoring device data with specific preprocessing requirements. Glucometer-collected fingertip blood glucose measurements necessitate standardized unit conversion between millimoles per liter and milligrams per deciliter.^39–41 Additionally, wearable devices capture multiple physiological parameters including physical activity levels, heart rate variability, and sleep patterns.^42–44 These measurements require careful calibration to harmonize sampling frequencies across different device manufacturers and models. Finally, the patient self-reported lifestyle behavior data, which involves diet logs, symptom records, and so on.⁴⁵ To enable seamless interoperability across multisource data systems, the health level seven fast healthcare interoperability resources (HL7 FHIR) standard has emerged as the prevailing framework for healthcare data exchange.^46,47 The implementation entails formal specification of standardized resource types (such as patient and observation) and facilitates bidirectional data synchronization between institutional hospital systems and mobile health (mHealth) devices via RESTful application programming interfaces (APIs).^47,48 The data quality control requires implementation of comprehensive validation protocols. First, missing device data should be systematically addressed through appropriate imputation methods; for continuous glucose monitoring measurements, linear interpolation may be applied when data gaps are limited to 15% or less. Second, all recorded values must be screened against predefined clinical plausibility ranges, with outliers automatically flagged and removed. Finally, strict timeliness standards must be enforced, particularly requiring all outpatient data to be entered into standardized databases within 24 h of collection.^49,50 These standardization efforts lay a solid foundation for subsequent data analysis and clinical applications, yet it still faces implementation challenges such as device compatibility discrepancies and privacy protection requirements.

Data storage and management

Informatization of diabetes management requires addressing the storage, access, and security issues of massive medical data in data storage and management. Current implementations primarily utilize a hybrid database architecture approach. For structured clinical data management, relational database systems including MySQL and SQL Server are commonly implemented.^51–53 These systems provide essential ACID (Atomicity, Consistency, Isolation, Durability) transaction properties while supporting comprehensive query analysis functionalities.^51,52 This architecture is particularly suited for handling well-defined clinical data elements such as laboratory test results and medication records. Additionally, the system employs specialized database architectures to accommodate distinct data types with unique storage requirements. Time-series data from medical devices, particularly continuous glucose monitoring records, demand efficient handling of temporal sequences. Similarly, unstructured data formats including medical images require flexible storage solutions. To address these needs, distributed NoSQL database systems such as MongoDB and Cassandra are implemented, providing the necessary high-throughput capabilities and elastic scalability for large-scale clinical data management. In terms of deployment models, large medical institutions mostly adopt local private cloud storage for core medical data, while combining public cloud services to process nonsensitive data, forming a hybrid cloud architecture.^54–56 Data security and privacy protection are very important in healthcare informatization construction. Technically, implement security measures such as end-to-end encryption (AES-256), role-based access control (RBAC), and multifactor authentication.^57,58 The management framework rigorously adheres to established regulatory standards, notably the Health Insurance Portability and Accountability Act in the United States and the General Data Protection Regulation in the European Union.^59,60 This compliance encompasses several critical components: implementation of data minimization protocols to collect only essential information, establishment of robust patient informed consent procedures, and maintenance of comprehensive audit trail systems to ensure full traceability of data access and usage.⁶⁰ Cross-border data transfers require specialized security measures to ensure compliance and protect patient privacy. Two critical components must be implemented: robust data desensitization techniques including k-anonymization methods, and secure dedicated transmission channels specifically designed for international healthcare data exchange.^61–63 These safeguards are essential for maintaining data integrity while meeting international regulatory requirements. Additionally, the system architecture must incorporate comprehensive disaster recovery mechanisms to guarantee uninterrupted service continuity. This requires implementation of geographically distributed remote backup systems coupled with scheduled recovery simulation exercises. These provisions ensure operational resilience against potential system failures or catastrophic events.^64–67 These measures collectively form the data governance framework for diabetes information management, providing secure and reliable data support for clinical decision-making and scientific research analysis.

Application scenarios of informatization

The diabetes information management system has demonstrated significant clinical utility across multiple well-established applications. A key implementation involves clinical decision support functionality, where an embedded clinical decision support system provides intelligent clinical alerts.^68,69 The system generates real-time drug interaction warnings, including specific alerts regarding metformin contraindications with radiographic contrast media.^68–71 Furthermore, it delivers dynamic medication recommendations stratified according to individual patients’ renal function parameters.⁷² These clinical decision support functionalities are systematically implemented through two core technical components: (a) an integrated medical knowledge base incorporating the latest American Diabetes Association (ADA) treatment guidelines, and (b) a sophisticated rule-based inference engine. This dual-component architecture has been demonstrated to effectively reduce medication errors and improve clinical decision-making accuracy in diabetes management.^73–75 The remote monitoring platform demonstrates robust capabilities in aggregating and analyzing multisource physiological data. Specifically, it seamlessly integrates continuous data streams from both conventional blood glucose meters and modern wearable devices. A key feature involves the implementation of personalized glycemic threshold parameters, where fasting blood glucose levels exceeding 7.0 millimoles per liter automatically activate a multitiered alert system. This intelligent monitoring architecture facilitates timely clinical intervention by enabling healthcare providers to rapidly detect and respond to pathological deviations.⁵⁰ The chronic disease management module represents a core functional component of the system. Through the implementation of structured follow-up protocols incorporating glycemic control objectives and complication surveillance intervals, the platform standardizes clinical workflows. Additionally, it autonomously produces tailored patient education content based on individual clinical profiles. For research applications, the system offers two key capabilities: standardized electronic case report form implementation and a collaborative multi-institutional data sharing framework. These features collectively improve the operational efficiency and methodological rigor of clinical investigations in diabetes care. It is worth noting that the current system still has certain limitations, such as barriers to data interoperability between institutions and incomplete self-management data collection from patients.⁵⁰ These issues provide opportunities for improvement when introducing artificial intelligence technology in the future. These information applications not only optimize the diagnostic and treatment process but also lay the groundwork for comprehensive diabetes management throughout the patient's life cycle.

Artificial intelligence-driven intelligent diabetes management

Breakthrough in core technologies

The core breakthroughs of artificial intelligence technology in diabetes management are mainly reflected in three technical fields. First, machine learning technology has made significant progress in dynamic prediction of blood glucose. The prediction system architecturally integrates two distinct machine learning paradigms: a recurrent neural network with long short-term memory (LSTM) units for temporal pattern extraction, and an extreme gradient boosting (XGBoost) ensemble method for feature-based regression. This dual-model framework performs synchronous analysis on three clinically validated data modalities: (a) high-frequency continuous glucose monitoring time-series, (b) quantified nutritional intake logs with macronutrient breakdown, and (c) accelerometer-derived physical activity energy expenditure measurements. The hybrid approach demonstrates superior predictive performance in clinical validation studies, achieving a mean absolute error of 1.18 mmol/L (95% CI: 1.12–1.24) for 2– h glucose forecasting, with <15% prediction error in 89.3% of test cases.^76–81 The system implements a reinforcement learning-based decision framework for personalized therapeutic strategy optimization. By analyzing large-scale clinical datasets encompassing diverse patient populations, the algorithm dynamically adapts treatment recommendations according to individual physiological profiles. Key patient-specific parameters incorporated in this process include pancreatic β-cell functional capacity and pharmacodynamic response characteristics, enabling precision medicine approaches in diabetes care. Clinical trials show that the rate of achieving blood glucose target can be increased by 18%–22%.^76–82

Secondly, deep learning shows outstanding value in the field of medical image analysis. The retinopathy screening system based on convolutional neural network (CNN) can automatically identify the pathological features such as microaneurysm and bleeding by analyzing fundus photos. Its sensitivity and specificity have reached more than 93%, which is close to the level of specialists. For complication risk assessment, deep neural networks (DNNs) demonstrate strong predictive performance, with the area under the receiver operating characteristic curve (AUC) > 0.85, in estimating 3-year risks of diabetic nephropathy and cardiovascular disease by integrating multimodal clinical data, including electronic health records (EHRs), laboratory tests, and imaging findings, thereby facilitating early intervention strategies.^83–88 In addition, natural language processing technology has driven the innovation of intelligent interaction systems. The transformer-based artificial intelligence health assistant can understand a patient's natural language advice (such as “What should I do if my blood sugar is high after dinner”) and provide customized advice based on the individual's health profile. More advanced multimodal systems can also parse the diet pictures uploaded by patients, automatically identify food types and estimate carbohydrate content, which significantly improves the convenience of self-management.^76,89–93 The integrated application of these technologies marks an important shift in diabetes management from traditional information to intelligent and personalized direction.

Typical application scenarios

The typical application of artificial intelligence technology in diabetes management has moved from theoretical research to clinical practice, forming a number of solutions with significant clinical value. The closed-loop artificial pancreas system represents the highest level of current diabetes treatment technology. The system uses real-time continuous glucose monitoring data to drive artificial intelligence algorithms and dynamically adjust the infusion rate of the insulin pump to achieve precise control similar to physiological insulin secretion.^94,95 The latest generation of hybrid closed-loop systems, such as the MiniMed 780G, has been approved by the Food and Drug Administration (FDA) and clinical trials have shown that it can increase the time to target blood glucose (TIR) to more than 80% while significantly reducing the incidence of hypoglycemic events.^96–100 In terms of lifestyle management, the intelligent diet and exercise management system uses computer vision technology to analyze the meal images taken by patients, accurately identifies food types and estimates nutritional components based on deep learning algorithms, and then generates dietary recommendations based on personalized recommendation algorithms.^101,102 Such a system integrates the motion data collected by wearable devices to adjust insulin dosage recommendations in real time, reducing postprandial blood glucose fluctuations by 30%–40%.^101,102 Early complication screening is another important application scenario of artificial intelligence. The retinopathy screening system based on deep learning has been certified as a medical device in the United States and other countries, which can realize automatic grading diagnosis of diabetic retinopathy in primary medical institutions, and its diagnostic accuracy is comparable to that of ophthalmic experts.^103,104 In the early diagnosis of diabetic nephropathy, artificial intelligence model can predict the risk of renal disease progression 12–18 months in advance by analyzing the changing trend of microalbuminuria and glomerular filtration rate in urine, thus gaining a precious time window for clinical intervention.^105,106 The implementation of these application scenarios not only improves the efficiency and quality of diabetes management, but also reshapes the traditional diagnosis and treatment mode, promoting the development of diabetes management to the direction of precision and personalization.

Core advantages of artificial intelligence compared with informatization

Artificial intelligence technology shows significant advantages in diabetes management, which are different from traditional information solutions, mainly reflected in the paradigm shift of intervention mode and decision-making mode. In terms of intervention mode, artificial intelligence technology has achieved a leapfrog development from “passive recording” to “active intervention.” Traditional information systems are mainly limited to the collection, storage and simple alarm functions of medical data, such as recording blood glucose measurement values and issuing prompts when exceeding the threshold; while artificial intelligence systems can analyze continuous dynamic data through machine learning algorithms, predict future blood glucose changes and make autonomous intervention decisions. Traditional information systems are mainly limited to the collection, storage and simple alarm functions of medical data, such as recording blood glucose measurement values and issuing prompts when exceeding the threshold; while artificial intelligence systems can analyze continuous dynamic data through machine learning algorithms, predict future blood glucose changes and make autonomous intervention decisions.^101,102 In terms of decision-making, artificial intelligence has driven the shift from “standardized processes” to “personalized decisions.” Traditional clinical decision support systems typically offer the same advice to all patients based on fixed rules and standardized guidelines.^103,104 In contrast, artificial intelligence systems use deep learning models to analyze each patient's unique physiological traits, lifestyle habits, and treatment responses, generating highly personalized management plans.^105,106 This personalized decision-making capability is particularly evident in areas such as dietary recommendations, medication selection, and prevention of complications. For example, a computer vision-based dietary analysis system can provide customized nutritional advice based on the patient's specific metabolic characteristics, dietary preferences, and real-time blood glucose levels. These core advantages not only improve the efficiency of diabetes management, but also fundamentally change the quality and mode of medical services, providing a technical foundation for the realization of precision medicine (Table 1).

Table 1.

Evolution of technology-driven paradigms in diabetes prevention and management.

Dimension	The information management stage	The digital intervention stage	The AI-driven stage
Primary objective	Data recording, storage, and sharing	Patient engagement, behavioral interventions, and reminders	Precision prediction, personalized and automated management
Key technology	Electronic health records, hospital information systems	Mobile health applications, wearable devices	Machine learning, deep learning, digital twins
Data characteristics	Structured and in-hospital data	Continuous, dynamic patient-generated data	Multimodal, high-dimensional, integrated data
Interaction pattern	Physician-led and passively recorded	Human-machine collaboration, alerts, and feedback	Intelligence-led, personalized recommendations and decision support
Paradigm shift	Transition from paper-based to electronic	Transition from in-hospital to out-of-hospital	Transition from generalized to precision

Note: This table delineates the conceptual and technological progression across three distinct phases of digital health integration in diabetes care. Each stage is characterized by its primary objective, enabling technologies, data attributes, interaction paradigm, and underlying shift in operational focus, collectively illustrating the trajectory from foundational digitization to intelligent, personalized health management.

Future trends in intelligent diabetes management

Direction of technology integration

In the future, intelligent diabetes management will exhibit significant technological integration, primarily in data collaboration and patient modeling. In data collaboration, the application of federated learning technology will overcome the limitations of traditional medical data silos, enabling collaborative modeling of multicenter medical data. This distributed machine learning framework allows healthcare institutions to jointly train high-precision diabetes prediction models without sharing raw data, thus protecting patient privacy while fully utilizing dispersed medical data resources. Research has shown that blood glucose prediction models built using federated learning can achieve an accuracy rate of over 95% compared to centralized training, providing a viable technical pathway for inter-institutional collaboration.^77,107,108 In the field of patient modeling, the development of digital twin technology will facilitate the creation of virtual patient models. By integrating patients “genomic data, physiological parameters, and lifestyle habits, highly realistic personal digital twins can be constructed, allowing doctors to test the expected outcomes of different treatment plans in a virtual environment.”^109–112 This treatment priori approach not only significantly reduces clinical trial costs but also helps identify the optimal personalized treatment plan for each patient. Preliminary clinical studies have shown that treatment decisions based on digital twins can increase the rate of achieving target blood glucose levels by more than 25% and reduce the incidence of adverse reactions by 30%.^113,114 The development of these integrated technologies will drive diabetes management from the current intelligent stage toward a higher level of precise and predictive management.

Implementation challenges and coping strategies

The clinical implementation of intelligent diabetes management still faces several key challenges. In terms of data privacy and ethics, the integration and application of multisource medical data raise concerns about patient information security, particularly the protection of sensitive information such as genomic data, which urgently needs to be addressed. To tackle this challenge, differential privacy techniques and homomorphic encryption methods can be employed to ensure data availability while strictly protecting patient privacy. The improvement of regulatory frameworks, such as the EU's GDPR, also provides institutional support for the data governance of medical artificial intelligence. Clinical explainability (XAI) is a critical factor affecting doctors trust, and current black box deep learning models struggle to gain full acceptance from clinical physicians. By developing visualization tools based on attention mechanisms and integrating white-box algorithms like decision trees into hybrid models, the transparency of artificial intelligence decision-making can be significantly enhanced. Research indicates that artificial intelligence systems with good explainability can increase doctors’ adoption rate by over 40%.¹¹⁵ In terms of medical accessibility, high technical costs hinder the widespread adoption of artificial intelligence solutions in primary healthcare facilities. Developing lightweight models, optimizing edge computing architectures, and establishing regional artificial intelligence sharing platforms can effectively reduce deployment costs. For instance, model distillation technology can compress deep learning models to 10% of their original size while maintaining over 90% accuracy.¹¹⁵ The systematic implementation of these strategies will pave the way for the large-scale implementation of intelligent diabetes management.

Discussion and prospect

Smart technology is profoundly transforming the traditional model of diabetes management, bringing unprecedented changes to patients and healthcare systems.^19–22 The integration of technologies such as artificial intelligence, the Internet of Things (IoT), and big data has made real-time blood glucose monitoring, personalized treatment recommendations, and remote medical support possible.^21–24 The widespread use of closed-loop insulin pump systems, artificial intelligence-driven decision support tools, and wearable devices has significantly improved the precision and convenience of diabetes management, while also enhancing patients’ self-management capabilities.^23–26. The present paper provides a narrative review of research advances from information management to artificial intelligence in the field of diabetes treatment and management, while also discussing the opportunities and challenges in clinical translation and application. The present review offers a conceptual framework to inform future research and development in intelligent diabetes care. However, its conclusions are inevitably constrained by the rapid evolution of the field and the narrative design's inherent susceptibility to selection bias.

In the future, the widespread adoption and integration of smart technologies will depend critically on sustained interdisciplinary collaboration and robust policy support to facilitate a comprehensive shift from reactive treatment models to proactive, patient-centered health management. Looking ahead, the deep integration of artificial intelligence and telemedicine is poised to become a pivotal research frontier. Remote diagnosis systems based on deep learning, virtual health assistants and tiered diagnosis and treatment mechanisms supported by 5G networks are expected to provide more timely and accurate interventions for diabetic patients. At the same time, it is crucial to establish a health management system that covers the entire life cycle. This includes using multiomics data and artificial intelligence models for early risk prediction, designing tailored intervention plans for different groups, and developing a comprehensive platform for managing complications. These innovations will drive diabetes management toward more personalized and continuous approaches, thereby improving patients’ long-term outcomes and quality of life.

Finally, the widespread application of intelligent technology in diabetes management still faces ethical and regulatory challenges. Ensuring data security, enhancing algorithm transparency, conducting multicenter clinical trials, and optimizing medical resource allocation are key areas for future research. Through ongoing technological innovation and interdisciplinary collaboration, intelligent technology is expected to create a patient-centered smart healthcare ecosystem, offering more efficient and inclusive solutions for global diabetes prevention and control. This transformation will not only improve individual health but also set new standards for chronic disease management.

Footnotes

ORCID iD

Geer Deng

CRediT authorship contribution statement

Geer Deng: project administration and writing–original draft. Wang Chengshi: writing-review and editing and funding acquisition.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Program of Natural Science Foundation of Sichuan (grant number 2022NSFSC1559).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Hruby

. The epidemiology of obesity: a big picture. Pharmacoeconomics 2015; 33: 673–689.

WHO. Diabetes, https://www.who.int/health-topics/diabetes#tab=tab_1 (2025, accessed 8 April 2025).

Jia

. Diabetes care in China: innovations and implications. J Diabetes Investig 2022; 13: 1795–1797.

Cai

Liu

, et al. Effectiveness of quality of care for patients with type 2 diabetes in China: findings from the Shanghai Integration Model (SIM). Front Med 2022; 16: 126–138.

Bergman

Manco

Satman

, et al. International diabetes federation position statement on the 1-hour post-load plasma glucose for the diagnosis of intermediate hyperglycaemia and type 2 diabetes. Diabetes Res Clin Pract 2024; 209: 111589.

Author

. The prevention of diabetes mellitus. JAMA 2021; 325: 190.

Author

. Expanding access to newer medicines for people with type 2 diabetes in low-income and middle-income countries: a cost-effectiveness and price target analysis. Lancet Diabetes Endocrinol 2021; 9: 825–836.

Choi

Winn

Skandari

, et al. First-Line therapy for type 2 diabetes with sodium-glucose cotransporter-2 inhibitors and glucagon-like peptide-1 receptor agonists: a cost-effectiveness study. Ann Intern Med 2022; 175: 1392–1400.

Azizi

Rodriguez

Assimes

. Digital footprints of obesity treatment: GLP-1 receptor agonists and the health equity divide. Circulation 2024; 150: 171–173.

10.

Fuchs

Caserto

Liu

, et al. A glucose-responsive cannula for automated and electronics-free insulin delivery. Adv Mater 2024; 36: e2403594.

11.

Kong

Lim

Yoo

, et al. Asia-Pacific consensus recommendations for application of continuous glucose monitoring in diabetes management. Diabetes Res Clin Pract 2023; 201: 110718.

12.

Baumert

Buchmann

, et al. Use of glucose monitoring devices among adults with diabetes in Germany: results from nationwide surveys conducted in 2017 and 2021/2022. Diabetes Technol Ther 2025; 27: 488–499.

13.

Seidu

Kunutsor

Ajjan

, et al. Efficacy and safety of continuous glucose monitoring and intermittently scanned continuous glucose monitoring in patients with type 2 diabetes: a systematic review and meta-analysis of interventional evidence. Diabetes Care 2024; 47: 169–179.

14.

Kumar

Goel

. Fully automated standalone microfluidic integrated electrochemiluminescence platform for sample-to-answer detection of diabetes complication markers. Lab Chip 2025; 16: 4138–4150.

15.

van Gennip

Schram

Kohler

, et al. Association of type 2 diabetes according to the number of risk factors within the recommended range with incidence of major depression and clinically relevant depressive symptoms: a prospective analysis. Lancet Healthy Longev 2023; 4: e63–e71.

16.

Tomic

Morton

Salim

, et al. Hospitalisation for mental health disorders in Australians with type 1 or type 2 diabetes. Diabetes Res Clin Pract 2023; 196: 110244.

17.

Schiepatti

Maimaris

Nicolardi

, et al. Determinants and trends of adherence to a gluten-free diet in adult celiac patients on a long-term follow-up (2000–2020). Clin Gastroenterol Hepatol 2022; 20: e741–e749.

18.

Henry

Jebb

Aveyard

, et al. Lifestyle advice for hypertension or diabetes: trend analysis from 2002 to 2017 in England. Br J Gen Pract 2022; 72: e269–e275.

19.

Thabit

Schofield

. Technology in the management of diabetes in hospitalised adults. Diabetologia 2024; 67: 2114–2128.

20.

Oikonomou

Khera

. Machine learning in precision diabetes care and cardiovascular risk prediction. Cardiovasc Diabetol 2023; 22: 259.

21.

Neumann

Zghal

Cremona

, et al. A data-driven personalized approach to predict blood glucose levels in type-1 diabetes patients exercising in free-living conditions. Comput Biol Med 2025; 190: 110015.

22.

Lebech

Hasselstrom

Schou

. Development and validation of a machine learning model to predict weekly risk of hypoglycemia in patients with type 1 diabetes based on continuous glucose monitoring. Diabetes Technol Ther 2024; 26: 457–466.

23.

Poalelungi

Musat

Fulga

, et al. Advancing patient care: how artificial intelligence is transforming healthcare. J Pers Med 2023; 13: 1214.

24.

PLOS DHS. Correction: aI-driven healthcare: a review on ensuring fairness and mitigating bias. PLOS Digit Health 2025; 4: e994.

25.

Agarwal

Galindo

Shah

, et al. Diabetes technology in people with type 2 diabetes: novel indications. Curr Diab Rep 2024; 24: 85–95.

26.

Forlenza

Lal

. Current status and emerging options for automated insulin delivery systems. Diabetes Technol Ther 2022; 24: 362–371.

27.

Nomura

Noguchi

Kometani

, et al. Artificial intelligence in current diabetes management and prediction. Curr Diabetes Rep 2021; 21: 61.

28.

Wang

SCY

Nickel

Venkatesh

, et al. AI-based diabetes care: risk prediction models and implementation concerns. npj Digit Med 2024; 7: 36.

29.

Smyth

Jenkins

Dunham

, et al. Systematic review of clinical practice guidelines to identify recommendations for sleep in type 2 diabetes mellitus management. Diabetes Res Clin Pract 2020; 170: 108532.

30.

Zhao

Zheng

Tian

, et al. Objective sleep characteristics and continuous glucose monitoring profiles of type 2 diabetes patients in real-life settings. Diabetes Obes Metab 2023; 25: 823–831.

31.

Franceschi

. Precision medicine in diabetes, current research and future perspectives. J Pers Med 2022; 12: 1233.

32.

Mann

Bar

Lev-Tzion

, et al. Medical concept embedding of real-valued electronic health records with application to inflammatory bowel disease. Artif Intell Med 2023; 145: 102684.

33.

Kim

Lee

Kim

, et al. High-throughput algorithm for discovering new drug indications by utilizing large-scale electronic medical record data. Clin Pharmacol Ther 2020; 108: 1299–1307.

34.

Wang

Liu

Zhang

, et al. Deep fusion learning facilitates anatomical therapeutic chemical recognition in drug repurposing and discovery. Brief Bioinform 2021; 22: bbab289.

35.

Peng

Wang

, et al. Drug repositioning by prediction of drug's anatomical therapeutic chemical code via network-based inference approaches. Brief Bioinform 2021; 22: 2058–2072.

36.

Crema

Verde

Tiraboschi

, et al. Medical information extraction with NLP-powered QABots: a real-world scenario. IEEE J Biomed Health Inform 2024; 28: 6906–6917.

37.

Moqurrab

Ayub

Anjum

, et al. An accurate deep learning model for clinical entity recognition from clinical notes. IEEE J Biomed Health Inform 2021; 25: 3804–3811.

38.

Seki

Kawazoe

Ohe

. Clinical Feature Vector Generation using Unsupervised Graph Representation Learning from Heterogeneous Medical Records. In: AMIA … Annual Symposium proceedings. AMIA Symposium 2024, 2023, pp. 618–623.

39.

Ling

Lau

, et al. Impact of body composition and anemia on accuracy of a real-time continuous glucose monitor in diabetes patients on continuous ambulatory peritoneal dialysis. Diabetes Technol Ther 2024; 26: 70–75.

40.

Shilo

Keshet

Rossman

, et al. Continuous glucose monitoring and intrapersonal variability in fasting glucose. Nat Med 2024; 30: 1424–1431.

41.

Salaami

Kuo

Drake

, et al. Antidiabetic effects of the senolytic agent Dasatinib. Mayo Clin Proc 2021; 96: 3021–3029.

42.

Kukreti

Yeh

, et al. Physiological sensors equipped in wearable devices for management of long COVID persisting symptoms: scoping review. J Med Internet Res 2025; 27: e69506.

43.

Torres-Soto

Ashley

. Multi-task deep learning for cardiac rhythm detection in wearable devices. NPJ Digit Med 2020; 3: 116.

44.

Mizuno

Changolkar

Patel

. Wearable devices to monitor and reduce the risk of cardiovascular disease: evidence and opportunities. Annu Rev Med 2021; 72: 459–471.

45.

Cheng

, et al. Diet- and lifestyle-based prediction models to estimate cancer recurrence and death in patients with stage III colon cancer (CALGB 89803/alliance). J Clin Oncol 2022; 40: 740–751.

46.

Sinaci

Gencturk

Teoman

, et al. A data transformation methodology to create findable, accessible, interoperable, and reusable health data: software design, development, and evaluation study. J Med Internet Res 2023; 25: e42822.

47.

Soares

Schilling

Richardson

, et al. Making science computable using evidence-based medicine on fast healthcare interoperability resources: standards development project. J Med Internet Res 2024; 26: e54265.

48.

Gulden

Macho

Reinecke

, et al. recruIT: a cloud-native clinical trial recruitment support system based on health level 7 fast healthcare interoperability resources (HL7 FHIR) and the observational medical outcomes partnership common data model (OMOP CDM). Comput Biol Med 2024; 174: 108411.

49.

Cichosz

Kronborg

Hangaard

, et al. Assessing the accuracy of continuous glucose monitoring metrics: the role of missing data and imputation strategies. Diabetes Technol Ther 2025; 27: 790–800.

50.

Carlson

Martens

Johnson

, et al. Continuous glucose monitoring integration for remote diabetes management: virtual diabetes care with case studies. Diabetes Technol Ther 2021; 23: S56–S65.

51.

Lin

Juan

, et al. Joint effect of blood pressure and glycemic variation on the risk of cardiovascular morbidity and mortality in persons with type 2 diabetes. Metabolism 2022; 136: 155308.

52.

Zhao

Zhu

Shen

, et al. Chinese Diabetes datasets for data-driven machine learning. Sci Data 2023; 10: 35.

53.

Zhu

Zhong

, et al. A survey on the integration of blockchains and databases. Data Sci Eng 2023; 8: 196–219.

54.

Denison

Zhu

. A survey of few-shot learning for biomedical time series. IEEE Rev Biomed Eng 2025; 18: 192–210.

55.

Ranchal

Bastide

Wang

, et al. Disrupting healthcare silos: addressing data volume, velocity and variety with a cloud-native healthcare data ingestion service. IEEE J Biomed Health Inform 2020; 24: 3182–3188.

56.

Bathelt

Lorenz

Weidner

, et al. Application of modular architectures in the medical domain - a scoping review. J Med Syst 2025; 49: 27.

57.

Belazi

Mabrouk

. A refined sine-derived chaotic map for securing medical image encryption in telemedicine. Comput Biol Med 2025; 196: 110667.

58.

Bai

Wei

Yang

, et al. Data-class-specific all-optical transformations and encryption. Adv Mater 2023; 35: e2212091.

59.

McCoy

Allen

Kopp

, et al. Ethical responsibilities for companies that process personal data. Am J Bioeth 2023; 23: 11–23.

60.

Bahls

Pung

Heinemann

, et al. Designing and piloting a generic research architecture and workflows to unlock German primary care data for secondary use. J Transl Med 2020; 18: 394.

61.

Moreno-Barea

Lopez-Garcia

Mesa

, et al. Named entity recognition for de-identifying spanish electronic health records. Comput Biol Med 2025; 185: 109576.

62.

Hasan

Datta

Namasudra

. Inter-hospital secure healthcare data exchange process by using proxy re-encryption and blockchain technology. Comput Biol Med 2025; 194: 110462.

63.

Lalova-Spinks

Valcke

Ioannidis

, et al. EU-US data transfers: an enduring challenge for health research collaborations. NPJ Digit Med 2024; 7: 215.

64.

Fang

Sun

Almutiq

, et al. Distributed medical data storage mechanism based on proof of retrievability and vector commitment for metaverse services. IEEE J Biomed Health Inform 2024; 28: 6298–6307.

65.

Shalom

Goldstein

Ariel

, et al. Distributed application of guideline-based decision support through mobile devices: implementation and evaluation. Artif Intell Med 2022; 129: 102324.

66.

Hamadi

Lussier

Fantoni

, et al. Data fusion fault tolerant strategy for a quadrotor UAV under sensors and software faults. ISA Trans 2022; 129: 520–539.

67.

Hoffman

Song

Brundage

, et al. Online improvement of condition-based maintenance policy via Monte Carlo tree search. IEEE Trans Autom Sci Eng 2022; 19: 2540–2551.

68.

Liu

Wang

Liu

. Utility of ChatGPT in clinical practice. J Med Internet Res 2023; 25: e48568.

69.

Ying

Fan

Chen

, et al. Real-Time AI-assisted insulin titration system for glucose control in patients with type 2 diabetes: a randomized clinical trial. JAMA Netw Open 2025; 8: e258910.

70.

Wasylewicz

van de Burgt

Manten

, et al. Contextualized drug-drug interaction management improves clinical utility compared with basic drug-drug interaction management in hospitalized patients. Clin Pharmacol Ther 2022; 112: 382–390.

71.

Hua

Lam

, et al. The use of potentially interacting supplement-drug pairs in adults with type 2 diabetes: a large population-based cohort study in the UK biobank. Diabetes Res Clin Pract 2024; 211: 111658.

72.

Wilson

Yamamoto

Martin

, et al. A randomized clinical trial assessing the effect of automated medication-targeted alerts on acute kidney injury outcomes. Nat Commun 2023; 14: 2826.

73.

Chin

Wilson

Trask

, et al. Repurposing clinical decision support system data to measure dosing errors and clinician-level quality of care. J Med Syst 2020; 44: 185.

74.

Nimri

Tirosh

Muller

, et al. Comparison of insulin dose adjustments made by artificial intelligence-based decision support systems and by physicians in people with type 1 diabetes using multiple daily injections therapy. Diabetes Technol Ther 2022; 24: 564–572.

75.

Yoo

Kim

. Advances in continuous glucose monitoring and integrated devices for management of diabetes with insulin-based therapy: improvement in glycemic control. Diabetes Metab J 2023; 47: 27–41.

76.

Lee

Kim

Woo

. Glucose transformer: forecasting glucose level and events of hyperglycemia and hypoglycemia. IEEE J Biomed Health Inform 2023; 27: 1600–1611.

77.

Koutny

Mayo

. Predicting glucose level with an adapted branch predictor. Comput Biol Med 2022; 145: 105388.

78.

Piao

Zhu

Baldeweg

, et al. GARNN: an interpretable graph attentive recurrent neural network for predicting blood glucose levels via multivariate time series. Neural Netw 2025; 185: 107229.

79.

Zhu

Herrero

, et al. GluGAN: generating personalized glucose time series using generative adversarial networks. IEEE J Biomed Health Inform 2023; 27: 5122–5133.

80.

Deng

Aponte

, et al. Deep transfer learning and data augmentation improve glucose levels prediction in type 2 diabetes patients. NPJ Digit Med 2021; 4: 109.

81.

Elhadd

Mall

Bashir

, et al. Artificial intelligence (AI) based machine learning models predict glucose variability and hypoglycaemia risk in patients with type 2 diabetes on a multiple drug regimen who fast during Ramadan (the PROFAST - IT Ramadan study). Diabetes Res Clin Pract 2020; 169: 108388.

82.

Singh

Malhotra

Parikh

. Integrative approach for early detection of Parkinson's disease and atypical Parkinsonian syndromes leveraging hemodynamic parameters, motion data & advanced AI models. Comput Meth Prog Bio 2025; 271: 108989.

83.

Grauslund

. Diabetic retinopathy screening in the emerging era of artificial intelligence. Diabetologia 2022; 65: 1415–1423.

84.

Bora

Balasubramanian

Babenko

, et al. Predicting the risk of developing diabetic retinopathy using deep learning. Lancet Digit Health 2021; 3: e10–e19.

85.

Nderitu

Nunez

DRJ

Webster

, et al. Predicting 1, 2 and 3 year emergent referable diabetic retinopathy and maculopathy using deep learning. Commun Med (Lond) 2024; 4: 167.

86.

Syed

Trucco

Mookiah

, et al. Deep-learning prediction of cardiovascular outcomes from routine retinal images in individuals with type 2 diabetes. Cardiovasc Diabetol 2025; 24: 3.

87.

Hua

Kim

Huynh-The

, et al. Convolutional network with twofold feature augmentation for diabetic retinopathy recognition from multi-modal images. IEEE J Biomed Health Inform 2021; 25: 2686–2697.

88.

Meng

Guan

, et al. Non-invasive biopsy diagnosis of diabetic kidney disease via deep learning applied to retinal images: a population-based study. Lancet Digit Health 2025; 7: 100868.

89.

Sun

Zhang

Lan

, et al. An AI dietitian for type 2 diabetes Mellitus management based on large language and image recognition models: preclinical concept validation study. J Med Internet Res 2023; 25: e51300.

90.

Qiu

, et al. Egocentric image captioning for privacy-preserved passive dietary intake monitoring. IEEE Trans Cybern 2024; 54: 679–692.

91.

Wang

Zhang

, et al. WeChat mini-program, a preliminary applied study of the gestational blood glucose management model for pregnant women with gestational diabetes mellitus. Diabetes Res Clin Pract 2025; 219: 111943.

92.

Suh

Cha

, et al. Explainable deep learning approaches for risk screening of periodontitis. J Dent Res 2025; 104: 45–53.

93.

Huang

Chen

Yang

. AI-enhanced integration of genetic and medical imaging data for risk assessment of Type 2 diabetes. Nat Commun 2024; 15: 4230.

94.

Zhu

Georgiou

. Offline deep reinforcement learning and off-policy evaluation for personalized basal insulin control in type 1 diabetes. IEEE J Biomed Health Inform 2023; 27: 5087–5098.

95.

Cinar

Basu

Bequette

, et al. Metabolic models, in silico trials, and algorithms. Diabetes Technol Ther 2025; 27: S21–S32.

96.

Arrieta

Battelino

Scaramuzza

, et al. Comparison of MiniMed 780G system performance in users aged younger and older than 15 years: evidence from 12 870 real-world users. Diabetes Obes Metab 2022; 24: 1370–1379.

97.

Gros

Amuedo

Bellido

, et al. Effectiveness and safety of an advanced hybrid closed-loop system in adolescents and adults with type 1 diabetes previously treated with continuous subcutaneous insulin infusion and flash glucose monitoring. Diabetes Technol Ther 2023; 25: 151–156.

98.

Elbarbary

Ismail

. Glycemic control during Ramadan fasting in adolescents and young adults with type 1 diabetes on MiniMed 780G advanced hybrid closed-loop system: a randomized controlled trial. Diabetes Res Clin Pract 2022; 191: 110045.

99.

Rossi

Montefusco

Reseghetti

, et al. Daytime hypoglycemic episodes during the use of an advanced hybrid closed loop system. Diabetes Res Clin Pract 2023; 206: 111011.

100.

Halim

Abraham

Manos

, et al. Advances in automated insulin delivery with the Medtronic 780G: the Australian experience. Diabetes Technol Ther 2024; 26: 190–197.

101.

Linseisen

Renner

Gedrich

, et al. Data in personalized nutrition: bridging biomedical, psycho-behavioral, and food environment approaches for population-wide impact. Adv Nutr 2025; 16: 100377.

102.

Mayer

Pilz

, et al. Foodcoach: fully automated diet counseling. IEEE J Biomed Health Inform 2025; 29: 5257–5269.

103.

Zafar

Mahjoub

Mehta

, et al. Artificial intelligence algorithms in diabetic retinopathy screening. Curr Diab Rep 2022; 22: 267–274.

104.

Ruamviboonsuk

Tiwari

Sayres

, et al. Real-time diabetic retinopathy screening by deep learning in a multisite national screening programme: a prospective interventional cohort study. Lancet Digit Health 2022; 4: e235–e244.

105.

Lee

Song

Chun

, et al. Development of a machine learning model for precision prognosis of rapid kidney function decline in people with diabetes and chronic kidney disease. Diabetes Res Clin Pract 2024; 217: 111897.

106.

Tangri

Ferguson

Bamforth

, et al. Machine learning for prediction of chronic kidney disease progression: validation of the Klinrisk model in the CANVAS program and CREDENCE trial. Diabetes Obes Metab 2024; 26: 3371–3380.

107.

Moulahi

Jdey

Moulahi

, et al. A blockchain-based federated learning mechanism for privacy preservation of healthcare IoT data. Comput Biol Med 2023; 167: 107630.

108.

Zhao

Liu

, et al. Medical federated model with mixture of personalized and shared components. IEEE Trans Pattern Anal Mach Intell 2025; 47: 433–449.

109.

Venkatesh

Raza

Kvedar

. Health digital twins as tools for precision medicine: considerations for computation, implementation, and regulation. NPJ Digit Med 2022; 5: 150.

110.

Checcucci

Oing

Amparore

, et al. Digital twins in urological oncology: precise treatment planning via Complex modeling. Eur Urol Oncol 2025; 8: 1–4.

111.

Shen

Liu

, et al. From virtual to reality: innovative practices of digital twins in tumor therapy. J Transl Med 2025; 23: 348.

112.

Ding

Thao

PNM

Peng

, et al. Large language multimodal models for new-onset type 2 diabetes prediction using five-year cohort electronic health records. Sci Rep-Uk 2024; 14: 20774.

113.

Mosquera-Lopez

Jacobs

. Digital twins and artificial intelligence in metabolic disease research. Trends Endocrinol Metab 2024; 35: 549–557

114.

Thangaraj

Benson

Oikonomou

, et al. Cardiovascular care with digital twin technology in the era of generative artificial intelligence. Eur Heart J 2024; 45: 4808–4821.

115.

Dusing

Cimiano

Rehberg

, et al. Integrating federated learning for improved counterfactual explanations in clinical decision support systems for sepsis therapy. Artif Intell Med 2024; 157: 102982.

Intelligent technology-driven diabetes prevention and control: From informatization management to artificial intelligence

Abstract

Keywords

Introduction

Information technology architecture for diabetes management

Data acquisition and standardization

Data storage and management

Application scenarios of informatization

Artificial intelligence-driven intelligent diabetes management

Breakthrough in core technologies

Typical application scenarios

Core advantages of artificial intelligence compared with informatization

Future trends in intelligent diabetes management

Direction of technology integration

Implementation challenges and coping strategies

Discussion and prospect

Footnotes

ORCID iD

CRediT authorship contribution statement

Funding

Declaration of conflicting interests

References