Artificial intelligence in diabetes management: Advancements, opportunities, and challenges

Probably not a substantial claim: may not pass S3 (load-bearing).

Diabetes is a significant health challenge worldwide due to increasing prevalence, high avoidable morbidity and mortality, and substantial economic burden.

“The increasing prevalence of diabetes, high avoidable morbidity and mortality due to diabetes and diabetic complications, and related substantial economic burden make diabetes a significant health challenge worldwide.”

• cites [4] Jia W. Diabetes care in China: Innovations and implications J. Diabetes Investig. 13 2022 1795 1797 10.1111/jdi.13908 PMC9623513 36121277 sourcePartly supported

  The opening passage links rising prevalence, avoidable morbidity and mortality, and economic burden to a major health challenge in China. The citing sentence extends that conclusion to the world, which this passage does not establish. (LLM, conf=0.89)

  Experimental — accuracy not yet measured

May combine two claims.

Despite the rapid progress of ML methods, there are several potential flaws, including data bias, overfitting, resource-intensive training, and limited transfer learning.

“Despite the rapid progress of ML methods, there are several potential flaws, including data bias,27 overfitting,28 resource-intensive training,29 and limited transfer learning.30”

• cites [27] Gianfrancesco M.A. Tamang S. Yazdany J. Schmajuk G. Potential Biases in Machine Learning Algorithms Using Electronic Health Record Data JAMA Intern. Med. 178 2018 1544 1547 10.1001/jamainternmed.2018.3763 30128552 PMC6347576 sourceNot checked — no full text of the source
• cites [28] Ying X. An Overview of Overfitting and its Solutions J. Phys, Conf. Ser. 1168 2019 022022 sourceNot checked — no full text of the source
• cites [29] Gómez-Carmona O. Casado-Mansilla D. López-de-Ipiña D. García-Zubia J. Optimizing computational resources for edge intelligence through model cascade strategies IEEE Internet Things J. 9 2022 7404 7417 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

A targeted and reasonable diet can control blood sugar and lipids and supplement protein and other nutrients.

“A targeted and reasonable diet can control blood sugar and lipids and supplement protein and other nutrients.65”

• cites [65] Sami W. Ansari T. Butt N.S. Ab Hamid M.R. Effect of diet on type 2 diabetes mellitus: A review Int J Health Sci-Ijh 11 2017 10.1002/dmrr.2515 PMC5426415 28539866Partly supported

  The sections “Relation between Dietary Practices and Diabetes Complications” and “Dietary Knowledge of Type 2 Diabetics” say that diet affects blood sugar and lipid levels and that nutrition therapy can help manage dyslipidemia. They do not support supplementing protein or other nutrients; the latter section instead notes that protein intake in one population exceeded recommendations. Page numbers are not available in our copy. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

May combine two claims.

ML methods enable the development of AI applications that facilitate the discovery of previously unrecognized patterns in the data without the need to specify decision rules for each specific task or to account for complex interactions among input features. Therefore, ML has become the preferred framework for building AI utilities.

“In summary, ML methods enable the development of AI applications that facilitate the discovery of previously unrecognized patterns in the data without the need to specify decision rules for each specific task or to account for complex interactions among input features. Therefore, ML has become the preferred framework for building AI utilities.34”

• cites [34] Roberts K. Boland M.R. Pruinelli L. Dcruz J. Berry A. Georgsson M. Hazen R. Sarmiento R.F. Backonja U. Yu K.-H. Biomedical informatics advancing the national health agenda: the AMIA 2015 year-in-review in clinical and consumer informatics J. Am. Med. Inf. Assoc. 24 2017 e185 e190 10.1093/jamia/ocw103 PMC6080724 27497798 sourcePartly supported

  Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim. May combine two claims.

With the inputs of SMBG and non-rapid insulin, Fong et al.'s work is worth discussing today because SMBG and non-rapid insulin formulations are still used in many parts of the world.

“With the inputs of SMBG and non-rapid insulin, this work is worth discussing today because SMBG and non-rapid insulin formulations are still used in many parts of the world.”

• cites [82] Fong S. Zhang Y. Fiaidhi J. Mohammed O. Mohammed S. Evaluation of stream mining classifiers for real-time clinical decision support system: A case study of blood glucose prediction in diabetes therapy BioMed Res. Int. 2013 2013 274193 10.1155/2013/274193 PMC3791638 24163813 sourceSupported

May combine two claims.

The prediction horizon range previously explored ranged between 5 and 180 min, with short-term predictions most frequently explored (most studies used prediction horizons below 60 min).

“The prediction horizon range previously explored ranged between 5 and 180 min. Short-term predictions were the most frequently explored, as most studies used prediction horizons below 60 min.71”

• cites [71] Contreras I. Vehi J. Artificial Intelligence for Diabetes Management and Decision Support: Literature Review J. Med. Internet Res. 20 2018 e10775 10.2196/10775 PMC6000484 29848472 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

A proper diet is indispensable in managing diabetic patients by maintaining normal blood sugar levels and reducing the burden of pancreatic islet β cells.

“A proper diet is indispensable in managing diabetic patients by maintaining normal blood sugar levels and reducing the burden of pancreatic islet β cells.64”

• cites [64] Kitagishi Y. Nakanishi A. Minami A. Asai Y. Yasui M. Iwaizako A. Suzuki M. Ono Y. Ogura Y. Matsuda S. Certain Diet and Lifestyle May Contribute to Islet beta-cells Protection in Type-2 Diabetes via the Modulation of Cellular PI3K/AKT Pathway Open Biochem. J. 8 2014 74 82 10.2174/1874091X01408010074 25400709 PMC4231374 sourcePartly supported

  The section “Certain Diet and Lifestyle May Contribute to Islet β-Cells Protection” (page unavailable) says oxidative damage can be reduced by controlling hyperglycemia and calorie intake, and discusses possible dietary contributions to β-cell protection. It does not establish that a proper diet is indispensable for diabetic patients, maintains normal blood sugar levels, or reduces β-cell burden by doing so. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

May combine two claims.

Prior AI models in diabetes management offered several advantages, such as data analysis, pattern recognition, decision support, and patient education, but faced limitations in natural language understanding, personalized treatment, accuracy, and potential biases.

“Prior AI models in diabetes management offered several advantages, such as data analysis, pattern recognition, decision support, and patient education, but faced limitations in natural language understanding, personalized treatment, accuracy, and potential biases.71”

• cites [71] Contreras I. Vehi J. Artificial Intelligence for Diabetes Management and Decision Support: Literature Review J. Med. Internet Res. 20 2018 e10775 10.2196/10775 PMC6000484 29848472 sourcePartly supported

  Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

May combine two claims.

User adherence is crucial to the effectiveness of DHT applications in the real world, which can be affected by convenience, user experience, and true benefits brought by this technology.

“User adherence is crucial to the effectiveness of DHT applications in the real world, which can be affected by convenience, user experience, and true benefits brought by this technology (for example, helping them decide when to contact their provider, being more aware of their symptoms, and feeling more connected to their provider).138,139”

• cites [138] Blakey J.D. Bender B.G. Dima A.L. Weinman J. Safioti G. Costello R.W. Digital technologies and adherence in respiratory diseases: the road ahead Eur. Respir. J. 52 2018 1801147 10.1183/13993003.01147-2018 PMC6364097 30409819 sourcePartly supported

  In ‘Considering the intended users of the interventions’ (page unavailable), the paper says sustained real-world use is critical to digital technologies’ success and discusses user experience, desirable features, and reasons patients disengage. It does not specifically identify convenience as a factor or substantiate the cited sentence’s examples of deciding when to contact a provider, becoming more aware of symptoms, or feeling more connected to a provider. (LLM, conf=0.83)

  Experimental — accuracy not yet measured
• cites [139] Rudin R.S. Fanta C.H. Qureshi N. Duffy E. Edelen M.O. Dalal A.K. Bates D.W. A Clinically Integrated mHealth App and Practice Model for Collecting Patient-Reported Outcomes between Visits for Asthma Patients: Implementation and Feasibility Appl. Clin. Inf. 10 2019 783 793 10.1055/s-0039-1697597 PMC6795530 31618782 sourcePartly supported

  The Patient Themes section (page not available) reports that patients valued the app’s simplicity, symptom awareness, and connection to their provider; the discussion (p. 789) says its core features helped patients decide when to seek care. The Clinical Relevance Statement reports questionnaire adherence, but this single-app feasibility study does not establish that adherence is crucial to the real-world effectiveness of DHT applications generally, or that the listed benefits caused adherence. (LLM, conf=0.83)

  Experimental — accuracy not yet measured

May combine two claims.

Federated computing approaches have been developed, wherein dedicated parameter servers are responsible for aggregating and distributing local learning; however, a remainder of a central structure is kept.

“Federated computing approaches have further been developed, wherein dedicated parameter servers are responsible for aggregating and distributing local learning; however, a remainder of a central structure is kept.136”

• cites [136] Warnat-Herresthal S. Schultze H. Shastry K.L. Manamohan S. Mukherjee S. Garg V. Sarveswara R. Händler K. Pickkers P. Aziz N.A. Swarm Learning for decentralized and confidential clinical machine learning Nature 594 2021 265 270 10.1038/s41586-021-03583-3 34040261 PMC8189907 sourceSupported

May combine two claims.

Participants had interest in having technology to support self-management, but reported a lack of confidence with technology, as well as frustration with design features and navigation.

“Although participants had interest in having technology to support self-management, they reported a lack of confidence with technology, as well as frustration with design features and navigation.130”

• cites [130] Sarkar U. Gourley G.I. Lyles C.R. Tieu L. Clarity C. Newmark L. Singh K. Bates D.W. Usability of Commercially Available Mobile Applications for Diverse Patients J. Gen. Intern. Med. 31 2016 1417 1426 27418347 PMC5130945 sourceSupported

May combine two claims.

These individuals often have the poorest understanding of the disease, low use of maintenance medications, and poor health outcomes.

“These individuals often have the poorest understanding of the disease, low use of maintenance medications, and poor health outcomes.122”

• cites [122] Himes B.E. Weitzman E.R. Innovations in health information technologies for chronic pulmonary diseases Respir. Res. 17 2016 38 10.1186/s12931-016-0354-3 27048618 PMC4822326 sourcePartly supported

  The Background’s health-disparities section (page not provided) identifies health literacy and medication non-adherence as factors in chronic pulmonary disease disparities. Another Background passage links poor adherence to adverse clinical outcomes. The paper does not establish that the unspecified individuals have the poorest disease understanding or specifically low use of maintenance medications. (LLM, conf=0.79)

  Experimental — accuracy not yet measured

May combine two claims.

Application of AI systems in the real world may lead to many unintended outcomes, including alert fatigue, additional burdens for clinicians, disruption of interpersonal communications, and generation of specific threats requiring increased vigilance to recognize.

“Application of AI systems in the real world may lead to many unintended outcomes, including alert fatigue,131 additional burdens for clinicians, disruption of interpersonal communications, and generation of specific threats requiring increased vigilance to recognize.132”

• cites [131] Carspecken C.W. Sharek P.J. Longhurst C. Pageler N.M. A clinical case of electronic health record drug alert fatigue: consequences for patient outcome Pediatrics 131 2013 e1970 e1973 10.1542/peds.2012-3252 23713099 sourceNot checked — no full text of the source
• cites [132] Ash J.S. Berg M. Coiera E. Some unintended consequences of information technology in health care: The nature of patient care information system-related errors J. Am. Med. Inf. Assoc. 11 2004 104 112 10.1197/jamia.M1471 PMC353015 14633936 sourcePartly supported

  The introduction (p. 1) says patient care information systems can add work tasks, disrupt communication routines, and create hazards requiring vigilance to detect. These points support the outcomes attached to citation 132, but the paper discusses patient care information systems rather than AI systems specifically. Alert fatigue is attached to citation 131, not this source. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

May combine two claims.

Warnat-Herresthal et al. introduced a decentralized ML approach named swarm learning, which dispenses with a dedicated server, shares the parameters via the swarm network, and builds the models independently on private data at the individual sites.

“As an alternative, Warnat-Herresthal et al. introduced a decentralized ML approach named swarm learning,136 which dispenses with a dedicated server, shares the parameters via the swarm network, and builds the models independently on private data at the individual sites.”

• cites [136] Warnat-Herresthal S. Schultze H. Shastry K.L. Manamohan S. Mukherjee S. Garg V. Sarveswara R. Händler K. Pickkers P. Aziz N.A. Swarm Learning for decentralized and confidential clinical machine learning Nature 594 2021 265 270 10.1038/s41586-021-03583-3 34040261 PMC8189907 sourceSupported

May combine two claims.

Dai et al. developed an automated, interpretable, and validated system (DeepDR) that performs real-time image quality feedback, retinal lesion detection, and early- to late-stage DR grading with high sensitivity and specificity.

“To overcome this issue, Dai et al. developed an automated, interpretable, and validated system (named DeepDR) that performs real-time image quality feedback, retinal lesion detection, and early- to late-stage DR grading with high sensitivity and specificity.103”

• cites [103] Dai L. Wu L. Li H. Cai C. Wu Q. Kong H. Liu R. Wang X. Hou X. Liu Y. A deep learning system for detecting diabetic retinopathy across the disease spectrum Nat. Commun. 12 2021 3242 10.1038/s41467-021-23458-5 34050158 PMC8163820 sourceSupported

• AssumptionRests on an assumption: Proper intervention at an early stage could significantly delay the progression of DR and even reverse mild non-proliferative DR to a DR-free stage.

May combine two claims.

Two main barriers have been identified that may undermine clinicians' enthusiasm to integrate AI systems into clinical workflows: experts may struggle to develop trust because of numerous inputs and complex data integration; and some evidence suggests AI systems could be perceived as encroaching on clinicians' professional role.

“In addition, two main barriers have been identified that may undermine clinicians’ enthusiasm to integrate AI systems into clinical workflows.133 First, experts may struggle to develop trust with these systems because of the numerous inputs and complex data integration involved, which can make it challenging or impossible to convey the specific logic behind an alert or recommendation. Second, some”

• cites [133] Henry K.E. Kornfield R. Sridharan A. Linton R.C. Groh C. Wang T. Wu A. Mutlu B. Saria S. Human-machine teaming is key to AI adoption: clinicians' experiences with a deployed machine learning system NPJ Digit. Med. 5 2022 97 10.1038/s41746-022-00597-7 35864312 PMC9304371 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

Precision medicine is an emerging disease treatment and prevention approach that considers individual variability in genes, environment, and lifestyle for each person.

“As described by the National Institutes of Health, precision medicine is an emerging disease treatment and prevention approach that considers individual variability in genes, environment, and lifestyle for each person.”

• cites [120] Mesko B. The role of artificial intelligence in precision medicine Expert Review of Precision Medicine and Drug Development 2 2017 239 241 10.1080/23808993.2017.1380516 sourceSupported

May combine two claims.

There exist inherent disadvantages in centralized solutions, including increased data traffic and concerns about data ownership, confidentiality, privacy, and security and creating data monopolies that favor data aggregators.

“There exist inherent disadvantages in centralized solutions, including increased data traffic and concerns about data ownership, confidentiality, privacy, and security and creating data monopolies that favor data aggregators.135”

• cites [135] Kaissis G.A. Makowski M.R. Rückert D. Braren R.F. Secure, privacy-preserving and federated machine learning in medical imaging Nat. Mach. Intell. 2 2020 305 311 10.1038/s42256-020-0186-1 sourcePartly supported

  The section ‘Decentralized data and federated machine learning’ says keeping data with its owner addresses governance and ownership concerns, while the discussion of patient-data protection addresses privacy and confidentiality. Our copy does not say centralized solutions create data monopolies favoring aggregators. It also describes burdensome communication as a limitation of decentralized learning, not as increased traffic caused by centralization; no page numbers are available in the indexed passages. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

May combine two claims.

The first well-established device was IDx-DR, approved by the FDA in 2018 for its high diagnostic performance in clinical trials.

“The first well-established device was IDx-DR, approved by the FDA in 2018 for its high diagnostic performance in clinical trials.104,142”

• cites [104] Abràmoff M.D. Lou Y. Erginay A. Clarida W. Amelon R. Folk J.C. Niemeijer M. Improved Automated Detection of Diabetic Retinopathy on a Publicly Available Dataset Through Integration of Deep Learning Invest. Ophthalmol. Vis. Sci. 57 2016 5200 5206 10.1167/iovs.16-19964 27701631 sourceNot checked — no full text of the source
• cites [142] Abràmoff M.D. Lavin P.T. Birch M. Shah N. Folk J.C. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices NPJ Digit. Med. 1 2018 39 10.1038/s41746-018-0040-6 31304320 PMC6550188 sourceSupported

The quote only partly states this claim.

Dagliati et al. reported that RF, SVM, and LR models using risk factors to predict development of diabetes-related complications provided an accuracy up to 0.838.

“final models, tailored in accordance with the complications, provided an accuracy up to 0.838”

• cites Dagliati et al.97 Dagliati A. Marini S. Sacchi L. Cogni G. Teliti M. Tibollo V. De Cata P. Chiovato L. Bellazzi R. Machine Learning Methods to Predict Diabetes Complications J. Diabetes Sci. Technol. 12 2018 295 302 10.1177/1932296817706375 28494618 PMC5851210 sourcePartly supported

  The Abstract (page not provided) reports complication-specific final models with accuracy up to 0.838 and lists the risk factors used. The Results (page not provided) clarify that the final retained models were LR models: RF and SVM were evaluated, but the paper does not attribute that final-model accuracy to all three methods. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

May combine two claims.

Dai et al. reported that DeepDR (CNN) using retinal images for DR screening achieved AUCs of 0.943, 0.955, 0.960, and 0.972 for grading DR as mild, moderate, severe, and proliferative, respectively, in internal validation, and AUCs from 0.916 to 0.970 in external validations.

“the grading of DR as mild, moderate, severe, and proliferative achieves AUCs of 0.943, 0.955, 0.960, and 0.972, respectively (internal validation); in external validations, AUCs for grading range from 0.916 to 0.970”

• cites Dai et al.103 Dai L. Wu L. Li H. Cai C. Wu Q. Kong H. Liu R. Wang X. Hou X. Liu Y. A deep learning system for detecting diabetic retinopathy across the disease spectrum Nat. Commun. 12 2021 3242 10.1038/s41467-021-23458-5 34050158 PMC8163820 sourcePartly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

May combine two claims.

Ljubic et al. reported that an RNN using hospitalization longitudinal data to predict development of 10 selected complications in T2D achieved prediction accuracy between 73% (myocardial infarction) and 83% (chronic ischemic heart disease), while traditional models achieved between 66% and 76%.

“the prediction accuracy was between 73% (myocardial infarction) and 83% (chronic ischemic heart disease), while the accuracy of traditional models was between 66% and 76%”

• cites Ljubic et al.99 Ljubic B. Hai A.A. Stanojevic M. Diaz W. Polimac D. Pavlovski M. Obradovic Z. Predicting complications of diabetes mellitus using advanced machine learning algorithms J. Am. Med. Inf. Assoc. 27 2020 1343 1351 10.1093/jamia/ocaa120 PMC7647294 32869093 sourceSupported

May combine two claims.

Huang et al. reported that DT, RF, SVM, and BN using clinical and genotyping data for diabetic nephropathy screening yielded accuracy, specificity, and sensitivity of 85.27%, 83.32%, and 85.24% on internal testing, and 78.50%, 80.64%, and 81.40% on a separate testing dataset.

“the proposed method yielded accuracy, specificity, and sensitivity of 85.27%, 83.32, and 85.24%, respectively (internal testing dataset); on a separate testing dataset, the classification accuracy, specificity, and sensitivity were 78.50%, 80.64, and 81.40%, respectively”

• cites Huang et al.112 Huang G.M. Huang K.Y. Lee T.Y. Weng J. An interpretable rule-based diagnostic classification of diabetic nephropathy among type 2 diabetes patients BMC Bioinf. 16 Suppl 1 2015 S5 10.1186/1471-2105-16-S1-S5 PMC4331704 25707942 sourcePartly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

LLMs are built using natural language processing techniques and designed to recognize and understand the structure and meaning of human language, classify texts according to their content or purpose, and generate responses that are appropriate and coherent.

“LLMs are built using natural language processing techniques and designed to recognize and understand the structure and meaning of human language, classify texts according to their content or purpose, and generate responses that are appropriate and coherent.146”

• cites [146] Ali S.R. Dobbs T.D. Hutchings H.A. Whitaker I.S. Using ChatGPT to write patient clinic letters Lancet. Digit. Health 5 2023 e179 e181 10.1016/S2589-7500(23)00048-1 36894409 sourceSupported

May combine two claims.

Ting et al. reported that a CNN using retinal images for DR screening achieved an AUC of 0.936, sensitivity of 90.5%, and specificity of 91.6% for referable DR; and for vision-threatening DR, AUC was 0.958, sensitivity was 100%, and specificity was 91.1%.

“the AUC of the system for referable DR was 0.936, sensitivity was 90.5%, and specificity was 91.6%; for vision-threatening DR, AUC was 0.958, sensitivity was 100%, and specificity was 91.1%”

• cites Ting et al.107 Ting D.S.W. Cheung C.Y.-L. Lim G. Tan G.S.W. Quang N.D. Gan A. Hamzah H. Garcia-Franco R. San Yeo I.Y. Lee S.Y. Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images From Multiethnic Populations With Diabetes JAMA 318 2017 2211 2223 10.1001/jama.2017.18152 29234807 PMC5820739 sourceSupported

May combine two claims.

Bora et al. reported that a CNN using retinal images and risk factors to predict risk of developing DR within 2 years achieved an AUC of 0.79 in the internal validation set and 0.70 in the external validation set with one-field color fundus photographs.

“the three-field DL system had an AUC of 0.79 in the internal validation set; assessment of the external validation set, which contained only one-field color fundus photographs, with the one-field DL system gave an AUC of 0.70”

• cites Bora et al.102 Bora A. Balasubramanian S. Babenko B. Virmani S. Venugopalan S. Mitani A. de Oliveira Marinho G. Cuadros J. Ruamviboonsuk P. Corrado G.S. Predicting the risk of developing diabetic retinopathy using deep learning Lancet. Digit. Health 3 2021 e10 e19 10.1016/S2589-7500(20)30250-8 33735063 sourceSupported

May combine two claims.

The emergence of DHTs, especially AI, may help address obstacles and alleviate disease burden of diabetes in the future because AI-based DHTs could help implement better prevention strategies, manage patients unable to attend appointments, deliver real-time health information, promote self-management, and save time and money.

“The emergence of digital health technologies (DHTs), especially artificial intelligence (AI), may help address these obstacles and alleviate the disease burden of diabetes in the future, because AI-based DHTs in diabetes care could help implement better prevention strategies for high-risk populations, manage diabetic patients who are unable to attend physician appointments in person, deliver real-”

• cites [8] Ashrafzadeh S. Hamdy O. Patient-Driven Diabetes Care of the Future in the Technology Era Cell Metabol. 29 2019 564 575 10.1016/j.cmet.2018.09.005 30269984 sourceNot checked — no full text of the source

• AssumptionRests on an assumption: A shortage of diabetes specialists, uneven distribution of medical resources, low adherence to medications, and improper self-management contribute to poor glycemic control in patients with diabetes.

Probably not a substantial claim: may not pass S3 (load-bearing).

The application of ML in diabetes care and research has been widely explored in basic biomedical research, translational research, and clinical practice.

“The application of ML in diabetes care and research has been widely explored in basic biomedical research,16 translational research,14 and clinical practice17 (Figure 1).”

• cites [16] Check Hayden E. The automated lab Nature 516 2014 131 132 10.1038/516131a 25471888 sourceNot checked — no full text of the source
• cites [14] Yu K.-H. Beam A.L. Kohane I.S. Artificial intelligence in healthcare Nat. Biomed. Eng. 2 2018 719 731 10.1038/s41551-018-0305-z 31015651 sourceNot checked — no full text of the source
• cites [17] Klonoff A.N. Andy Lee W.-A. Xu N.Y. Nguyen K.T. DuBord A. Kerr D. Six Digital Health Technologies That Will Transform Diabetes J. Diabetes Sci. Technol. 17 2023 239 249 10.1177/19322968211043498 34558330 PMC9846384 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Recent advancements in digital health technologies, especially AI, provide a significant opportunity to achieve better efficiency in diabetes care, which may diminish the increase in diabetes-related health-care expenditures.

“Recent advancements in digital health technologies, especially artificial intelligence (AI), provide a significant opportunity to achieve better efficiency in diabetes care, which may diminish the increase in diabetes-related health-care expenditures.”

• AssumptionRests on an assumption: A shortage of diabetes specialists, uneven distribution of medical resources, low adherence to medications, and improper self-management contribute to poor glycemic control in patients with diabetes.

May combine two claims.

Rule-based systems are expensive to build and brittle, and it is challenging to encode higher-order interactions between different pieces of knowledge provided by different specialists, with system effectiveness restricted by comprehensiveness of pre-existing medical knowledge.

“However, rule-based systems are expensive to build and brittle. Furthermore, it is challenging to encode higher-order interactions between different pieces of knowledge provided by different specialists, with system effectiveness restricted by the comprehensiveness of pre-existing medical knowledge.12”

• cites [12] Samuel A.L. Some studies in machine learning using the game of checkers (Reprinted from Journal of Research and Development, vol 3, 1959 IBM J. Res. Dev. 44 2000 206 226 10.1147/rd.441.0206 sourceNot checked — no full text of the source

May combine two claims.

AI can predict the risk of diabetes onset and classify diabetes.

“AI can predict the risk of diabetes onset and classify diabetes.”

May combine two claims.

One explanation for the shortcomings of existing treatment strategies is that the diagnosis of diabetes is based on the measurement of only one metabolite, glucose; however, the disease is heterogeneous in terms of clinical presentation and progression.

“One explanation for these shortcomings is that the diagnosis of diabetes is based on the measurement of only one metabolite, glucose; however, the disease is heterogeneous in terms of clinical presentation and progression.”

• AssumptionRests on an assumption: Existing treatment strategies for diabetes cannot stop the progressive course of the disease and prevent the development of chronic diabetic complications.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

AI algorithms could be quite useful in times when it is difficult to manage diabetes.

“Particularly, it is worth highlighting that AI algorithms could be quite useful in times when it is difficult to manage diabetes.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

AI-based models are less invasive and highly accessible, which could effectively improve the willingness of the general population to screen and provide personalized screening plans for high-risk populations.

“These AI-based models are less invasive and highly accessible, which could effectively improve the willingness of the general population to screen and provide personalized screening plans for high-risk populations.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Health education aims to empower diabetic patients with increased knowledge and awareness of diabetes, which can further facilitate better disease self-management.

“Health education aims to empower diabetic patients with increased knowledge and awareness of diabetes, which can further facilitate better disease self-management.”

May combine two claims.

A refined classification system could provide a powerful tool to individualize treatment regimens and identify individuals with an increased risk of complications at diagnosis.

“A refined classification system could provide a powerful tool to individualize treatment regimens and identify individuals with an increased risk of complications at diagnosis.”

May combine two claims.

Predicting near-future BG on a minute, hour, or overnight timescale enables better diabetes management and therefore remains an active research topic.

“Predicting near-future BG on a minute, hour, or overnight timescale enables better diabetes management and therefore remains an active research topic.”

May combine two claims.

Early stages of T2D are often asymptomatic, so individuals can carry the disease for years undiagnosed.

“Considering that the early stages of T2D are often asymptomatic, individuals can carry the disease for years undiagnosed.48”

• cites [48] American Diabetes Association 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes–2021 Diabetes Care 44 2021 S15 S33 33298413 sourceNot checked — no full text of the source

May combine two claims.

These limitations have motivated the use of AI-based diagnostics that have high classification accuracy and leverage large datasets that are not easily interpreted.

“These limitations have motivated the use of AI-based diagnostics that have high classification accuracy and leverage large datasets (including data from wearable and continuous monitoring devices) that are not easily interpreted.”

• AssumptionRests on an assumption: Late-stage diagnosis can lead to health complications and lower life expectancy.

May combine two claims.

Unlike many other medications, insulin must be titrated, which complicates insulin dosing and requires providers to perform complex calculations to determine the best dose.

“Unlike many other medications, insulin must be titrated, which means that the dose will change in response to real-time indicators such as current BG or based on what the patient has recently eaten. This process complicates insulin dosing and requires providers to perform complex calculations to determine the best dose.”

• AssumptionRests on an assumption: Errors in insulin dosing can be dangerous, as incorrect dosages can lead to serious adverse effects, such as hypoglycemia, which can cause death.

May combine two claims.

It is often poorly implemented that a scientific, personalized, and quantitative exercise prescription can potentially be an essential therapeutic agent for patients with diabetes, as clinicians lack necessary knowledge and skills, while participants have low adherence due to design defects.

“However, it is often poorly implemented, as clinicians lack the necessary knowledge and skills, while participants have low adherence due to design defects (e.g., prescriptions fail to take individual willingness, the appeal of exercise, and complex physical conditions into account).”

Classified as probably not a substantial claim. May combine two claims.

Clinicians provide therapeutic advice beyond insulin dosing guidance, as diabetes is often associated with multiple comorbidities, leading to complex guidelines for personalized treatment, with gaps often present in which no recommendations are available.

“Clinicians provide therapeutic advice beyond insulin dosing guidance (particularly T2D), as diabetes is often associated with multiple comorbidities, leading to complex guidelines for personalized treatment, with gaps often present in which no recommendations are available.”

May combine two claims.

Artificial neural network approaches are the most widely applied methodology for BG prediction, but other ML methodologies such as random forest and support vector machine are being adopted with increasing frequency.

“Artificial neural network approaches are the most widely applied methodology, but other ML methodologies, such as random forest and support vector machine, are being adopted with increasing frequency.”

May combine two claims.

AI can classify different types of diabetes more accurately, but further algorithm upgrading is required to meet clinical treatment needs.

“Although AI can classify different types of diabetes more accurately,51,54,55 further algorithm upgrading is required to meet the clinical treatment needs.”

• cites [51] Maniruzzaman M. Kumar N. Menhazul Abedin M. Shaykhul Islam M. Suri H.S. El-Baz A.S. Suri J.S. Comparative approaches for classification of diabetes mellitus data: Machine learning paradigm Comput. Methods Progr. Biomed. 152 2017 23 34 10.1016/j.cmpb.2017.09.004 29054258 sourceNot checked — no full text of the source
• cites [54] Barakat N.H. Bradley A.P. Barakat M.N.H. Intelligible Support Vector Machines for Diagnosis of Diabetes Mellitus Ieee T Inf Technol B 14 2010 1114 1120 10.1109/Titb.2009.2039485 20071261 sourceNot checked — no full text of the source
• cites [55] Amit kumar D. Pragati A. Classification of Diabetes Mellitus Using Machine Learning Techniques Int. J. Eng. Appl. Sci. 2 2015Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S4 (based).

The ability to anticipate BG excursions could provide an early warning regarding ineffective or poor treatment.

“The ability to anticipate BG excursions could provide an early warning regarding ineffective or poor treatment.”

• AssumptionRests on an assumption: It is usual for people with diabetes to experience BG variation throughout the day owing to carbohydrate ingestion and insulin action.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

The prediction of diabetes onset is a part of pre-emptive medicine, accurately identifying individuals highly likely to develop diabetes at the pre-illness stage. Thus, this technology could eventually decrease the incidence of diabetes by implementing medical interventions for these people at a very early stage.

“The prediction of diabetes onset is a part of pre-emptive medicine, accurately identifying individuals highly likely to develop diabetes at the pre-illness stage. Thus, this technology could eventually decrease the incidence of diabetes by implementing medical interventions for these people at a very early stage.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Previous studies have focused on optimizing the insulin dose for both patients and clinicians.

“Previous studies have focused on optimizing the insulin dose for both patients and clinicians.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Most patients with diabetes do not utilize automated insulin delivery systems on account of personal preferences or relatively high costs.

“This treatment model is still worth discussing today, since most patients with diabetes do not utilize automated insulin delivery systems on account of personal preferences or relatively high costs.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Early AI-enabled decision support systems were generally designed to assist clinicians who were authoritative in the insulin dosing strategies for their patients.

“For clinicians, early AI-enabled decision support systems were generally designed to assist clinicians who were authoritative in the insulin dosing strategies for their patients.”

May combine two claims. The quote only partly states this claim.

The control algorithm acts as a 'brain' in the closed-loop system, analyzing data fed back by CGM and automatically adjusting insulin infusion rate; therefore, an accurate control algorithm is crucial for a closed-loop system.

“The control algorithm acts as a”

Classified as probably not a substantial claim.

AI could be employed in understanding risk factors for diabetes onset owing to human limitations and biases when dealing with large spaces of risk factors.

“AI could be employed in understanding risk factors for diabetes onset owing to human limitations and biases when dealing with large spaces of risk factors.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Glycemic episode detection encompasses a set of tools that address the complexity of effective BG control, enabling detection of glycemic episodes and opportunity to respond promptly.

“As with BG prediction, glycemic episode detection encompasses a set of tools that address the complexity of effective BG control. These tools enable us to detect the occurrence of glycemic episodes and provide us with the opportunity to respond promptly to their effects.”

Classified as probably not a substantial claim. May combine two claims.

Dietary monitoring is critical in patients with diabetes, especially as self-reporting of food intake is inaccurate and often impractical; therefore, it is necessary to develop an automated solution for dietary monitoring.

“Dietary monitoring is critical in patients with diabetes, especially as self-reporting of food intake is inaccurate and often impractical59; therefore, it is necessary to develop an automated solution for dietary monitoring.”

• cites [59] Schoeller D.A. Limitations in the assessment of dietary energy intake by self-report Metabolism 44 1995 18 22 10.1016/0026-0495(95)90204-x 7869932 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Treatment escalation of diabetes is another critical aspect of drug therapy strategies, required as a patient's disease progresses, and multiple studies have aimed to identify optimal escalation points as an aid to clinicians.

“Treatment escalation of diabetes is another critical aspect of drug therapy strategies, which is required as a patient’s disease progresses, and multiple studies have aimed to identify optimal escalation points as an aid to clinicians.”

May combine two claims.

Most reviewed studies on glycemic episode detection focus on detecting hyperglycemia or hypoglycemia in situations where it is not possible to monitor BG effectively, and most dealt with real-time approaches rather than predictions of future events.

“In contrast to BG prediction, most of the reviewed studies on this topic focus on detecting hyperglycemia or hypoglycemia in situations where it is not possible to monitor BG effectively. Therefore, most of these studies dealt with real-time approaches rather than predictions of future events.”

May combine two claims.

Current diagnostic guidelines for diabetes are driven by invasive measurements in clinics, possibly influenced by behavioral and ethnic factors.

“Current diagnostic guidelines for diabetes47 are driven by invasive measurements in clinics, possibly influenced by behavioral and ethnic factors.”

• cites [47] American Diabetes Association 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes-2021 Diabetes Care 44 2021 S15 S33 10.2337/dc21-S002 33298413 sourceNot checked — no full text of the source

Classified as probably not a substantial claim.

A challenging aspect of measuring agreement between human expert- and AI-generated recommendations is that, even when they disagree, both can be correct, because humans often express a wide variety of priorities and risk preferences.

“A challenging aspect of measuring the agreement between human expert- and AI-generated recommendations is that, even when they disagree, both can be correct, because humans (including AI developers) often express a wide variety of priorities and risk preferences.”

Probably not a substantial claim: may not pass S4 (based).

In diabetes management, the prediction of the onset of diabetes and diabetic complications would eventually decrease the incidence of diabetes and diabetic complications by implementing appropriate medical interventions for those at high risk at a very early stage.

“In diabetes management, the prediction of the onset of diabetes and diabetic complications would eventually decrease the incidence of diabetes and diabetic complications by implementing appropriate medical interventions for those at high risk at a very early stage.”

May combine two claims.

The Guardian Connect System is an AI system based on CGM, has an accompanying smartphone application, and was certified by the FDA in 2018.

“The Guardian Connect System (manufactured by Medtronic) is an example of an AI system with this functionality. This system is based on CGM, has an accompanying smartphone application,144 and was certified by the FDA in 2018.”

• cites [144] Soupal J. Haskova A. Prazny M. Real-time CGM Is Superior to Flash Glucose Monitoring for Glucose Control in Type 1 Diabetes: The CORRIDA Randomized Controlled Trial Diabetes Care 43 2021 2744 2750 10.2337/dci20-0078 PMC7576432 32859607 sourceNot checked — no full text of the source

May combine two claims.

The Guardian Connect System uses AI to predict a hypoglycemic attack an hour in advance based on CGM data and alerts the patient.

“It is characterized by using AI to predict a hypoglycemic attack an hour in advance based on the CGM data and alerts the patient.”

May combine two claims.

It is feasible to integrate retinal-image-based prediction algorithms into DR screening programs because retinal examinations are routinely performed for patients with diabetes and the retina is a convenient window into body homeostasis.

“Since retinal examinations are routinely performed for patients with diabetes and the retina is a convenient window into the homeostasis of the body, it is feasible for the integration of such algorithms into DR screening programs.”

• AssumptionRests on an assumption: Retinal examinations are routinely performed for patients with diabetes and the retina is a convenient window into body homeostasis.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Previous studies have shown that applying AI in diabetes management involves all aspects of disease control, including prediction, prevention, screening, diagnosis, and treatment.

“Previous studies have shown that applying AI in diabetes management involves all aspects of disease control, including prediction, prevention, screening, diagnosis, and treatment.”

May combine two claims.

Two important mechanisms of the AI domain can be used to develop decisional systems for medical predictions: a knowledge-based system that involves logic and a system based on probabilistic reasoning.

“Two important mechanisms of the AI domain can be used to develop decisional systems for medical predictions: a knowledge-based system that involves logic and a system based on probabilistic reasoning.125”

• cites [125] Albu A. Stanciu L. Benefits of Using Artificial Intelligence in Medical Predictions 19–21 2015 IEEE 1 4 sourceNot checked — no full text of the source

May combine two claims.

AI technology can increase the productivity of existing DR tele-ophthalmology screening programs and lower unit economic costs within the community.

“For example, with the growing burden of DR, AI technology can increase the productivity of existing DR tele-ophthalmology screening programs and lower unit economic costs within the community.123”

• cites [123] Xie Y. Nguyen Q.D. Hamzah H. Lim G. Bellemo V. Gunasekeran D.V. Yip M.Y.T. Qi Lee X. Hsu W. Li Lee M. Artificial intelligence for teleophthalmology-based diabetic retinopathy screening in a national programme: an economic analysis modelling study Lancet. Digit. Health 2 2020 e240 e249 10.1016/S2589-7500(20)30060-1 33328056 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

With the deployment of automated AI for specific clinical tasks, the credentials needed for diagnostic, therapeutic, supportive, and paramedical tasks need to be updated, and the roles of health-care professionals will continue to evolve as various AI modules are incorporated into the standard of care.

“With the deployment of automated AI for specific clinical tasks, the credentials needed for diagnostic, therapeutic, supportive, and paramedical tasks need to be updated, and the roles of health-care professionals will continue to evolve as various AI modules are incorporated into the standard of care.14”

• cites [14] Yu K.-H. Beam A.L. Kohane I.S. Artificial intelligence in healthcare Nat. Biomed. Eng. 2 2018 719 731 10.1038/s41551-018-0305-z 31015651 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable).

Integrating AI into clinical practice care could shift diabetes care toward precision, penetration, prediction, and personalization.

“Integrating AI into clinical practice care could shift diabetes care toward precision, penetration, prediction, and personalization.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Several obstacles remain in the research and application of AI in diabetes management, including data quality control, poor technology design, privacy concerns, lack of clinical integration, non-adherence, and the establishment of laws and regulations.

“However, several obstacles remain in the research and application of AI in diabetes management, including data quality control, poor technology design, privacy concerns, lack of clinical integration, non-adherence, and the establishment of laws and regulations (Figure 2).”

• AssumptionRests on an assumption: Clinical AI systems' labels and data quality directly determine model performance.
• AssumptionRests on an assumption: AI in medicine inevitably results in legal challenges regarding medical negligence attributed to complex decision-support systems.

May combine two claims.

Armengol et al. reported that a case-based reasoning system using risk factors to estimate risk of complications was 100% correct in determining the kind of risk (progression or development) and the risk of stroke, 90% correct in determining amputation risk, and 72.45% correct in determining the global risk and the risk of infarct.

“the results were 100% correct in determining the kind of risk (progression or development) and the risk of stroke, 90% correct in determining amputation risk, and 72.45% correct in determining the global risk and the risk of infarct”

• cites Armengol et al.96 Armengol E. Palaudàries A. Plaza E. Individual prognosis of diabetes long-term risks: a CBR approach Methods Inf. Med. 40 2001 46 51 11310159 sourceNot checked — no full text of the source

May combine two claims. The quote only partly states this claim.

Lagani et al. reported that an ANN using risk factors to predict diabetes-related complications achieved internal validation C index between 0.66 and 0.833 (all significantly different from 0.5, p ≤ 0.0001), and external validation in 49 T1D patients showed similarly high and statistically significant C indexes (p < 0.05 except hypoglycemia and CVD models with p = 0.0584 and 0.0932), with good T2D

“the results of the internal validation (T1D) reported a C index between 0.66 and 0.833, all statistically significantly different from 0.5 (p ≤ 0.0001); these results were further corroborated by the external validation (49 patients with T1D), where all models obtained similarly high and statistically significant C indexes (p < 0.05 except for the hypoglycemia and CVD model, with p = 0.0584 and 0.”

• cites Lagani et al.94 Lagani V. Chiarugi F. Manousos D. Verma V. Fursse J. Marias K. Tsamardinos I. Realization of a service for the long-term risk assessment of diabetes-related complications J. Diabet. Complicat. 29 2015 691 698 10.1016/j.jdiacomp.2015.03.011 25953402 sourceNot checked — no full text of the source

May combine two claims.

Abramoff et al. reported that IDx-DR (CNN) using retinal images for DR screening exceeded all pre-specified superiority endpoints with sensitivity of 87.2% (>85%), specificity of 90.7% (>82.5%), and imageability rate of 96.1% for detecting more than mild DR.

“the AI system exceeded all pre-specified superiority endpoints at sensitivity of 87.2% (>85%), specificity of 90.7% (>82.5%), and imageability rate of 96.1% for detecting the presence of more than mild DR”

• cites Abramoff et al.104 Abràmoff M.D. Lou Y. Erginay A. Clarida W. Amelon R. Folk J.C. Niemeijer M. Improved Automated Detection of Diabetic Retinopathy on a Publicly Available Dataset Through Integration of Deep Learning Invest. Ophthalmol. Vis. Sci. 57 2016 5200 5206 10.1167/iovs.16-19964 27701631 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

By addressing the challenges and capitalizing on the opportunities, AI could play a pivotal role in transforming diabetes care and improving the lives of millions of people worldwide.

“By addressing the challenges and capitalizing on the opportunities, AI could play a pivotal role in transforming diabetes care and improving the lives of millions of people worldwide.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Despite these risks, LLMs have the potential to revolutionize patient monitoring, treatment personalization, and patient education, leading to improved outcomes and better quality of life for patients with diabetes.

“Despite these risks, LLMs have the potential to revolutionize patient monitoring, treatment personalization, and patient education, leading to improved outcomes and better quality of life for patients with diabetes.149,150”

• cites [149] Stokel-Walker C. Van Noorden R. What ChatGPT and generative AI mean for science Nature 614 2023 214 216 10.1038/d41586-023-00340-6 36747115 sourceNot checked — no full text of the source
• cites [150] Howard A. Hope W. Gerada A. ChatGPT and antimicrobial advice: the end of the consulting infection doctor? Lancet Infect. Dis. 23 2023 405 406 10.1016/S1473-3099(23)00113-5 36822213 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

AI has the potential to optimize diabetes care by providing personalized, precise, and data-driven support to patients and health-care professionals.

“In conclusion, AI has the potential to optimize diabetes care by providing personalized, precise, and data-driven support to patients and health-care professionals.”

May combine two claims.

Acharya et al. reported that a system using morphological image processing and SVM on retinal images could identify different stages of DR with an average accuracy of more than 85%, a sensitivity of 82%, and a specificity of 86%.

“the system could identify different stages of DR with an average accuracy of more than 85%, a sensitivity of 82%, and a specificity of 86%”

• cites Acharya et al.108 Acharya U.R. Lim C.M. Ng E.Y.K. Chee C. Tamura T. Computer-based detection of diabetes retinopathy stages using digital fundus images Proc. Inst. Mech. Eng. H 223 2009 545 553 10.1243/09544119jeim486 19623908 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Integrating AI-based DHTs will probably become increasingly feasible in the future as technology improves, and such integration will enable new models of diabetes care.

“Integrating AI-based DHTs will probably become increasingly feasible in the future as technology improves, and such integration will enable new models of diabetes care.”