AI-based diabetes care: risk prediction models and implementation concerns
Open accessAbstract
The utilization of artificial intelligence (AI) in diabetes care has focused on early intervention and treatment management. Notably, usage has expanded to predict an individual’s risk for developing type 2 diabetes. A scoping review of 40 studies by Mohsen et al. shows that while most studies used unimodal AI models, multimodal approaches were superior because they integrate multiple types of data. However, creating multimodal models and determining model performance are challenging tasks given the multi-factored nature of diabetes. For both unimodal and multimodal models, there are also concerns of bias with the lack of external validations and representation of race, age, and gender in training data. The barriers in data quality and evaluation standardization are ripe areas for developing new technologies, especially for entrepreneurs and innovators. Collaboration amongst providers, entrepreneurs, and researchers must be prioritized to ensure that AI in diabetes care is providing quality and equitable patient care.
AI snapshot
Written from the paper’s full text.
TL;DR
This commentary reviews AI-based risk prediction for type 2 diabetes, noting that multimodal models often outperform unimodal ones but face implementation barriers in data quality, bias, and evaluation standardization. It calls for diverse datasets, external validation, fairness metrics, and collaboration to ensure equitable diabetes care.
Key findings
- In Mohsen et al.’s scoping review of 40 studies, most models were classical machine learning using electronic health records; 30 were unimodal and 10 were multimodal, with multimodal approaches shown to be superior.
- One multimodal model combining genomic, metabolomic, and clinical risk factors predicted T2DM with an AUC of 0.96, compared with 0.586 for a genomics-only model and 0.798 for a clinical-only model.
- The same XGBoost algorithm used in three unimodal studies produced widely disparate AUC values of 0.91, 0.83, and 0.679–11, reflecting differences in risk predictors and datasets.
- Only five studies provided insights into model calibration and only five conducted external validations; most did not evaluate performance across demographic groups or use calibration and fairness metrics.
Methods
- Design: a commentary/perspective that summarizes and interprets a scoping review by Mohsen et al. of 40 studies on AI-based diabetes risk prediction.
- Data or sample: the 40 studies in the scoping review, including unimodal and multimodal models using EHR, multi-omics, medical imaging, genomic, metabolomic, and clinical data.
- Main analysis: narrative synthesis of model performance (primarily AUC), implementation concerns such as bias and lack of validation, and recommendations for data quality and standardized evaluation.
Limitations
- The paper states that determining the best-performing model is challenging because input risk predictors, datasets, evaluation metrics, and prediction horizons vary, so direct comparisons across studies must be considered carefully.
- The paper states that many studies had demographic imbalance in gender, ethnicity, and age; few addressed calibration or external validation; and most did not assess disparities across demographic groups.
- Evident from the text: this is a commentary rather than a primary systematic review, so its conclusions depend on the cited scoping review and selected examples.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
14 of 16 cited sources checked · 10 supported · 0 flagged · accuracy not yet measured
- Premisecitation
Black and Mexican Americans have an increased prevalence of diabetes compared to their white counterparts.
“As Black and Mexican Americans have an increased prevalence of diabetes compared to their white counterparts, the omission of certain demographics in training models could perpetuate existing health inequities for already at-risk populations and introduce a high probability of bias14.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [14] Fang, L., Sheng, H., Tan, Y. & Zhang, Q. Prevalence of diabetes in the USA from the perspective of demographic characteristics, physical indicators and living habits based on NHANES 2009-2018. Front. Endocrinol. 14, 1088882 (2023). source
- Premisecitation
Most studies in the Mohsen et al. scoping review of 40 studies used unimodal AI models.
“A scoping review of 40 studies by Mohsen et al. shows that while most studies used unimodal AI models, multimodal approaches were superior because they integrate multiple types of data.”
- Also stated in: § Risk prediction · plain
“Although a modest number (n = 10) studies adopted multimodal approaches (combining EHR with multi-omics or medical imaging data), these were shown to be superior to unimodal models (n = 30).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Also stated in: § Risk prediction · plain
- Premisecitation
AI is now being utilized to predict an individual's risk for developing type 2 diabetes (T2DM) and potential complications.
“Beyond intervention and treatment, AI is now being utilized to predict an individual’s risk for developing type 2 diabetes (T2DM) and potential complications.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
The Introduction and Results describe AI models used to predict individuals’ risk of developing T2DM. The Introduction also says precision prognostics aims to predict T2DM and its complications, but the review does not establish that the AI models it examines predict complications. (LLM, conf=0.85)
Experimental — accuracy not yet measured
- Premisecitation
The XGBoost algorithm was used in three unimodal studies.
“For example, the XGBoost algorithm—a machine learning algorithm that uses a sequentially iterative process to learn and correct previous errors—was used in three unimodal studies8.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [8] Chen, T., Guestrin, C. XGBoost: A scalable tree boosting system. In: Proc. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 785–794. https://doi.org/10.1145/2939672.2939785 (2016). source
- Premisecitation
Multimodal AI approaches were superior to unimodal models for diabetes risk prediction in the Mohsen et al. scoping review.
“A scoping review of 40 studies by Mohsen et al. shows that while most studies used unimodal AI models, multimodal approaches were superior because they integrate multiple types of data.”
- Also stated in: § Risk prediction · plain
“Although a modest number (n = 10) studies adopted multimodal approaches (combining EHR with multi-omics or medical imaging data), these were shown to be superior to unimodal models (n = 30).”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Also stated in: § Risk prediction · plain
- Premisecitation
One multimodal approach combining genomic, metabolomic, and clinical risk factors was superior in predicting T2DM (AUC of 0.96) compared to a genomics-only model (AUC of 0.586) and clinical-only model (AUC of 0.798).
“For example, one multimodal approach revealed that a model combining genomic, metabolomic, and clinical risk factors was superior in predicting T2DM (AUC of 0.96) compared to a genomics-only model (AUC of 0.586) and clinical-only model (AUC of 0.798)6.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [6] Liu, J. et al. Metabolic and genetic markers improve prediction of incident type 2 diabetes: a nested case-control study in Chinese. J. Clin. Endocrinol. Metab. 107, 3120–3127 (2022). source
The Results report an AUC of 0.960 for the model combining clinical characteristics, a metabolite score, and a genetic risk score, versus 0.798 for the clinical-only model. Figure 2 names a genetic model, but our copy does not give it an AUC of 0.586, so that numerical comparison is not supported. (LLM, conf=0.86)
Experimental — accuracy not yet measured
- Premisecitation
Only five studies in the Mohsen et al. review conducted external validations.
“Only five studies provided insights into calibration of models and only five conducted external validations.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Premisecitation
Only five studies in the Mohsen et al. review provided insights into calibration of models.
“Only five studies provided insights into calibration of models and only five conducted external validations.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Premisecitation
There are concerns with bias due to variability of demographic representation across models, with many studies showing a pronounced imbalance of gender, ethnicity, and age.
“There are further concerns with bias due to the variability of demographic representation across models, with many in this study showing a pronounced imbalance of gender, ethnicity, and age.”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Premisecitation
A modest number (n = 10) of studies in the Mohsen et al. review adopted multimodal approaches combining EHR with multi-omics or medical imaging data.
“Although a modest number (n = 10) studies adopted multimodal approaches (combining EHR with multi-omics or medical imaging data), these were shown to be superior to unimodal models (n = 30).”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
- Interpretationreasoning
Merging data sources can complicate the understanding of interactions among modalities and the rationale behind predictions, resulting in a scarcity of multimodal AI models for T2DM.
“Merging data sources can complicate the understanding of interactions among modalities and the rationale behind predictions, resulting in a scarcity of multimodal AI models for T2DM.”
Quote in the paperNot a substantial claim - Interpretationreasoning
Determining the best-performing model is challenging because the type and combination of input risk predictors for diabetes can influence the model's performance.
“Determining the best-performing model is challenging, given that the type and combination of input risk predictors for diabetes (e.g., BMI, waist circumference, fasting plasma glucose, age, alcohol intake, blood pressure, etc.) can influence the model’s performance.”
Quote in the paperSubstantial claim - Interpretationreasoning
Improved performance of multimodal models can be attributed to integrating multiple types of data such as clinical, genetic, and biomarker data, providing a comprehensive view of an individual's health status.
“Improved performance of multimodal models can be attributed to integrating multiple types of data such as clinical, genetic, and biomarker data, providing a comprehensive view of an individual’s health status.”
Quote in the paperSubstantial claim
- May combine two claims. 5
- Classified as probably not a substantial claim. 4
May combine two claims.
PremisecitationMost models in the Mohsen et al. review were classical machine learning models with electronic health records as the predominant data source.
“Of these models, most were classical machine learning models with electronic health records as the predominant data source.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationThe performance of the algorithms in the Mohsen et al. review was measured using the area under the curve (AUC) metric in most studies.
“The performance of the algorithms was measured using the area under the curve (AUC) metric in most studies.”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [5] Mohsen, F., Al-Absi, H. R. H., Yousri N. A., Hajj N. E., Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. npj Digit. Med. https://doi.org/10.1038/s41746-023-00933-5 (2023). source
May combine two claims.
PremisereasoningAlthough the same XGBoost algorithm was implemented, each of the three studies used different risk predictors and datasets, resulting in widely disparate AUC values of 0.91, 0.83, and 0.679-11.
“Although the same algorithm was implemented, each study used different risk predictors and datasets, which resulted in widely disparate AUC values of 0.91, 0.83, and 0.679–11.”
Quote in the paperSubstantial claim2 of 2 sources checked- cites [8] Chen, T., Guestrin, C. XGBoost: A scalable tree boosting system. In: Proc. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM; 785–794. https://doi.org/10.1145/2939672.2939785 (2016). source
The Introduction identifies XGBoost as the tree-boosting system, so the paper supports identifying the algorithm. Section 6.2 describes this paper’s own datasets, and Table 3 reports an AUC of 0.8304 for a Higgs-dataset experiment. Neither passage establishes the three cited studies’ predictors, their AUC values of 0.91, 0.83 and 0.679, or the proposed reason those values differ. (LLM, conf=0.83)
Experimental — accuracy not yet measured - cites [11] Wu, Y. et al. Machine learning for predicting the 3-year risk of incident diabetes in Chinese adults. Front. Public Health 9, 626331 (2021). source
The Results (Development of XGBoost Model) confirm that this study used XGBoost, but report XGBoost AUCs of 0.977 and 0.920, not the values listed in the citing sentence. The Results (Performance of the Stepwise Model; External Validation) report 0.910 and 0.830 for a stepwise model based on XGBoost-selected predictors, not for XGBoost itself. This paper does not establish that differences in predictors and datasets caused the disparities across three studies. (LLM, conf=0.86)
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.
Interpretationnone givenDeveloping multimodal models is extremely time-consuming, making it challenging to quickly and easily scale such models.
“However, developing multimodal models is extremely time-consuming and as a result, it is challenging to quickly and easily scale such models.”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationcitationThe omission of certain demographics in training models could perpetuate existing health inequities for already at-risk populations and introduce a high probability of bias.
“As Black and Mexican Americans have an increased prevalence of diabetes compared to their white counterparts, the omission of certain demographics in training models could perpetuate existing health inequities for already at-risk populations and introduce a high probability of bias14.”
Quote in the paperSubstantial claim- Rests on an assumption: T2DM is a multifactorial disease impacted by biological, clinical, and socioeconomic factors.
May combine two claims.
PremisecitationMost studies did not evaluate the algorithm's performance across different demographic groups or use calibration and fairness metrics to assess disparities in predictions across these groups.
“Most studies did not evaluate the algorithm’s performance across different demographic groups or use calibration and fairness (e.g., demographic parity, equal opportunity) metrics to assess disparities in predictions across these groups12,13.”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [12] Barocas, S., Hardt, M., Narayanan, A. Fairness and Machine Learning: Limitations and Opportunities. https://fairmlbook.org/pdf/fairmlbook.pdf (MIT Press., 2023).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites [13] Bella, A., Ferri, C., Hernandez-Orallo, J., Ramirez-Quintana, M. J. Calibration of machine learning models. Handbook of Research on Machine Learning Applications and Trends: Algorithms, Methods, and Techniques. 128–146 (2010). source
Classified as probably not a substantial claim.
InterpretationreasoningDirect comparisons across studies of multimodal models must be considered carefully due to biases such as variations in datasets, evaluation metrics, and prediction horizons.
“Even with multimodal models, direct comparisons across studies must be considered carefully due to biases such as variations in datasets, evaluation metrics, and prediction horizons.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
InterpretationreasoningThere is a growing effort to create more individualized multimodal virtual representations of patients (digital twins) for chronic conditions such as T2DM.
“Still, in light of promising outcomes from multimodal models for chronic conditions such as T2DM, there is a growing effort to create more individualized multimodal virtual representations of patients, also known as digital twins.”
Quote in the paperNot a substantial claim
The paper relies on this without support.
Identifying high-risk individuals and personalizing prevention strategies and targeted treatments could delay or prevent the onset of diabetes and future health complications.
“Identifying high-risk individuals and personalizing prevention strategies and targeted treatments could delay or prevent the onset of diabetes and future health complications.”
The paper relies on this without support.
T2DM is a multifactorial disease impacted by biological, clinical, and socioeconomic factors.
“T2DM is a multifactorial disease impacted by biological, clinical, and socioeconomic factors.”
The paper relies on this without support.
External validation ensures models can generalize beyond specific datasets used for training.
“External validation is also crucial to ensure models can generalize beyond specific datasets used for training.”
The paper relies on this without support.
Comprehensive and diverse data sets to train models are necessary to create accurate tools.
“Comprehensive and diverse data sets to train models are necessary to create accurate tools.”
The paper relies on this without support.
Universal evaluation protocols are necessary to mitigate the propagation of health inequities.
“Universal evaluation protocols are also necessary to mitigate the propagation of health inequities.”
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Machine learning and artificial intelligence approaches have demonstrated growing promise in diabetes risk prediction, complication detection and healthcare utilization forecasting.
“More recently, machine learning and artificial intelligence approaches have demonstrated growing promise in diabetes risk prediction, complication detection and healthcare utilization forecasting [21–28].”
From Machine learning-driven risk prediction for post-hospitalization diabetes case management: Integrating clinical and social determinants of health (2026) · cites this paper as [25]
The Risk prediction section describes AI models for predicting type 2 diabetes, and the Introduction reports an AI system that detects diabetic retinopathy in screening images. The paper does not report healthcare utilization forecasting, so it supports only part of the attributed claim. (LLM, conf=0.88)
Experimental — accuracy not yet measured
Cites
XGBoost: A Scalable Tree Boosting System
Chen, T., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939785 · Cited by 54,569 · see citing works
Multimodal Machine Learning: A Survey and Taxonomy
Baltrusaitis, T., Ahuja, C., & Morency, L.-P. (2019) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2018.2798607 · Cited by 5,018 · see citing works
Sounderajah, V., Ashrafian, H., Rose, S., Shah, N. H., Ghassemi, M., Golub, R., Kahn, C. E., Esteva, A., Karthikesalingam, A., Mateen, B. A., et al. (2021) · Nature Medicine · doi:10.1038/s41591-021-01517-0 · Cited by 277 · see citing works
Nimri, R., Battelino, T., Laffel, L. M., Slover, R. H., Schatz, D., Weinzimer, S. A., Dovc, K., Danne, T., Phillip, M., NextDREAM Consortium, et al. (2020) · Nature Medicine · doi:10.1038/s41591-020-1045-7 · Cited by 242 · see citing works
A scoping review of artificial intelligence-based methods for diabetes risk prediction
Mohsen, F., Al-Absi, H. R. H., Yousri, N. A., El Hajj, N., & Shah, Z. (2023) · npj Digital Medicine · doi:10.1038/s41746-023-00933-5 · Cited by 125 · see citing works
Health Digital Twins in Life Science and Health Care Innovation
Venkatesh, K. P., Brito, G., & Boulos, M. N. K. (2023) · The Annual Review of Pharmacology and Toxicology · doi:10.1146/annurev-pharmtox-022123-022046 · Cited by 109 · see citing works
Self-citationRavaut, M., Harish, V., Sadeghi, H., Leung, K. K., Volkovs, M., Kornas, K., Watson, T., Poutanen, T., & Rosella, L. C. (2021) · JAMA Network Open · doi:10.1001/jamanetworkopen.2021.11315 · Cited by 90 · see citing works
Abramoff, M. D., Whitestone, N., Patnaik, J. L., Rich, E., Ahmed, M., Husain, L., Hassan, M. Y., Tanjil, M. S. H., Weitzman, D., Dai, T., et al. (2023) · npj Digital Medicine · doi:10.1038/s41746-023-00931-7 · Cited by 81 · see citing works
Calibration of Machine Learning Models
Bella, A., Ferri, C., Hernández-Orallo, J., & Ramírez-Quintana, M. J. (2010) · Handbook of Research on Machine Learning Applications and Trends · doi:10.4018/978-1-60566-766-9.ch006 · Cited by 59 · see citing works
Fang, L., Sheng, H., Tan, Y., & Zhang, Q. (2023) · Frontiers in Endocrinology · doi:10.3389/fendo.2023.1088882 · Cited by 52 · see citing works
Machine Learning for Predicting the 3-Year Risk of Incident Diabetes in Chinese Adults
Wu, Y., Hu, H., Cai, J., Chen, R., Zuo, X., Cheng, H., & Yan, D. (2021) · Frontiers in Public Health · doi:10.3389/fpubh.2021.626331 · Cited by 28 · see citing works
Ekanayake, P. S., Juang, P. S., & Kulasa, K. M. (2020) · Current Diabetes Reports · doi:10.1007/s11892-020-01364-2 · Cited by 21 · see citing works
Liu, Q., Zhou, Q. S., He, Y., Zou, J., Guo, Y., & Yan, Y. (2022) · Journal of Personalized Medicine · doi:10.3390/jpm12071055 · Cited by 21 · see citing works
Liu, J., Wang, L., Qian, Y., Shen, Q., Yang, M., Dong, Y., Chen, H., Yang, Z., Liu, Y., Cui, X., et al. (2022) · The Journal of Clinical Endocrinology & Metabolism · doi:10.1210/clinem/dgac487 · Cited by 12 · see citing works
Šoupal, J., Hašková, A., & Prázný, M. (2021) · Diabetes Care · doi:10.2337/dci20-0078 · Cited by 3 · see citing works
Flagged first, then by citations
Research works printed but not linked: 1
- Barocas, S., Hardt, M., Narayanan, A. Fairness and Machine Learning: Limitations and Opportunities. https://fairmlbook.org/pdf/fairmlbook.pdf (MIT Press., 2023).
Cited by
Sorted by citations
Singh, A. P., Shadangi, S., Gupta, P. K., & Rana, S. (2025) · Comprehensive physiology · doi:10.1002/cph4.70003 · Cited by 183 · see citing works
Wang, X., Si, J., Li, Y., Tse, P., Zhang, G., Wang, X., Ren, J., Xu, J., Sun, J., & Yao, X. (2025) · Diabetology & Metabolic Syndrome · doi:10.1186/s13098-025-01819-0 · Cited by 30 · see citing works
Diagnosis, Pathophysiology and Management of Microvascular Dysfunction in Diabetes Mellitus
Jan, Y., Kelhofer, N. C., Tu, T., Mansuri, O., Onyemere, K. U., Dave, S. D., & Pappu, S. (2024) · Diagnostics · doi:10.3390/diagnostics14242830 · Cited by 27 · see citing works
“The emerging of the machine learning and deep learning of artificial intelligence (AI) may assist the clinicians by developing smart algorithms to diagnose microvascular dysfunction [39,91].”
Future horizons in diabetes: integrating AI and personalized care
Zhang, K., Qi, Y. P., Wang, W., Tian, X., Wang, J., Xu, L., & Zhai, X. (2025) · Frontiers in Endocrinology · doi:10.3389/fendo.2025.1583227 · Cited by 26 · see citing works
Artificial Intelligence in Diabetes Care: Applications, Challenges, and Opportunities Ahead
Parab, R., Feeley, J. M., Valero, M., Chadalawada, L., Garcia, G. P., Kar, S. S., Madabhushi, A., Breton, M. D., Li, J., Shao, H., et al. (2025) · Endocrine Practice · doi:10.1016/j.eprac.2025.07.008 · Cited by 32 · see citing works
Baluta, S., Suresh, V., Chmielowska, M., Smeesters, L., & Cabaj, J. (2026) · ACS Omega · doi:10.1021/acsomega.5c06072 · Cited by 21 · see citing works
AI-Derived Blood Biomarkers for Ovarian Cancer Diagnosis: Systematic Review and Meta-Analysis
Xu, H., Li, X.-Y., Jia, M.-Q., Ma, Q.-P., Zhang, Y., Liu, F.-H., Qin, Y., Chen, Y.-H., Li, Y., Chen, X.-Y., et al. (2025) · Journal of Medical Internet Research · doi:10.2196/67922 · Cited by 17 · see citing works
“Studies without external validation might yield results that were hard to generalize owing to factors such as sample selection bias and the characteristics of the research setting [95].”
Artificial intelligence in nephrology: predicting CKD progression and personalizing treatment
Yuan, S., Guo, L., & Xu, F. (2025) · International Urology and Nephrology · doi:10.1007/s11255-025-04878-4 · Cited by 12 · see citing works
Artificial Intelligence Enabled Lifestyle Medicine in Diabetes Care: A Narrative Review
Gonzalez-Rivas, J. P., Seyedi, S. A., & Mechanick, J. I. (2025) · American Journal of Lifestyle Medicine · doi:10.1177/15598276251359185 · Cited by 8 · see citing works
GHRH in diabetes and metabolism
Steenblock, C., & Bornstein, S. R. (2024) · Reviews in Endocrine and Metabolic Disorders · doi:10.1007/s11154-024-09930-9 · Cited by 7 · see citing works
Thanthrige, A. N. W., & Wickramasinghe, N. S. (2025) · JMIR Diabetes · doi:10.2196/80582 · Cited by 7 · see citing works
Rodríguez-Almeida, A. J., Socorro-Marrero, G. V., Acosta, C. B., Zamora-Zamorano, G., Déniz-García, A., Álvarez-Malé, M. L., Årsand, E., Soguero-Ruíz, C., Wägner, A. M., Granja, C., et al. (2025) · Frontiers in Digital Health · doi:10.3389/fdgth.2025.1534830 · Cited by 6 · see citing works
“This evidences the need for performing a subject-oriented evaluation of the DL models when proposing approaches such as a DIY using diabetes-speci fi c metrics (52).”
Jena, T., Suryodai, R., Reddy, D. N., Kumar, K. V., Muniyandy, E., & Kumar, N. V. P. S. (2025) · Intelligence-Based Medicine · doi:10.1016/j.ibmed.2025.100302 · Cited by 5 · see citing works
Integrating wearable devices with AI to explore a disease world spanning time and space
Liu, Y., Wu, Y., Ren, H.-X., Wu, S., Hong, J., Du, J., Sun, M., Lin, Z.-H., & Miao, Y.-B. (2025) · Nano Energy · doi:10.1016/j.nanoen.2025.111481 · Cited by 5 · see citing works
The transformative role of artificial intelligence in diabetes care and research
Canha, D., Bour, C., Barraud, S., Aguayo, G., & Fagherazzi, G. (2024) · Diabetes & Metabolism · doi:10.1016/j.diabet.2024.101565 · Cited by 4 · see citing works
Phan, L. T., Rakkiyappan, R., & Manavalan, B. (2025) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2025.109771 · Cited by 4 · see citing works
Artificial Intelligence in Personalized Medicine for Diabetes Mellitus: A Narrative Review
Ghosh, K., Chandra, S., Ghosh, S., & Ghosh, U. S. (2025) · Cureus · doi:10.7759/cureus.91520 · Cited by 4 · see citing works
“The implementation of AI-driven precision medicine in T2D is hindered by generalizability challenges, as models often overlook the heterogeneity of the population in terms of genetic and socioeconomic determinants, which risks biased outcomes [92].”
Rosella, L. C., Shaw, J., Guha, S., Abejirinde, I. O., Gibson, J., Lipscombe, L. L., Kornas, K., Zaim, R., Chui, V., & Itanyi, I. U. (2025) · Digital Health · doi:10.1177/20552076251358541 · Cited by 4 · see citing works
Alsayed, A. O., Ismail, N. A., Hasan, L., et al. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.2642 · Cited by 3 · see citing works
“Many current approaches experience problems in processing large datasets when temporal and spatial data analysis is needed with various problems related to real-time applicability, scalability, or the reliability of predictions ( Wang et al., 2024 ).”
Yang, L., Du, L., Jiang, L., Zhang, Y., & Fan, Q. (2026) · British Journal of Hospital Medicine · doi:10.31083/bjhm50385 · Cited by 2 · see citing works
Tachkov, K., Somolinos-Simón, F. J., Tapia, J. A. R., Hernando, M. E., García-Sáez, G., Dimitrova, M. J., Kamusheva, M., Mitkova, Z. E., Petykó, Z. I., Németh, B., et al. (2024) · Biotechnology & Biotechnological Equipment · doi:10.1080/13102818.2024.2371354 · Cited by 1 · see citing works
“We do not consider discussing the methods as technical and mathematical approaches but only their usability from the point of view of HTA experts from CEE countries [29].”
Ai for autonomous health care on diabetes diagnostics
Vege, H. K., Yandamuri, S. K., Vennela, J., & Venkat, S. (2025) · South Health and Policy · doi:10.56294/shp2025236 · Cited by 2 · see citing works
Harnessing dysglycaemia heterogeneity for precision type 2 diabetes prevention
Xie, S., & Merino, J. (2025) · Current Opinion in Lipidology · doi:10.1097/mol.0000000000001018 · Cited by 2 · see citing works
Haldar, U., Rahman, M. M., Miah, M. A., & Uddin, M. S. (2025) · International Conference on Electrical Systems & Automation (ICESA) · doi:10.1109/icesa66763.2025.11281187 · Cited by 1 · see citing works
“Wang et al. (2024) note that most AI models in diabetes have been classical machine learning (logistic regression, decision trees, etc.) models trained on EHR-derived features[10].”
The Role of Artificial Intelligence in Chronic Disease Management
Manokaran, S., Muralidharan, A., Ramakrishnan, A., et al. (2026) · Signals and communication technology · doi:10.1007/978-3-032-03554-7_7 · Cited by 1 · see citing works
Young, C. F., & MacLeod, J. (2026) · International Journal of Obesity · doi:10.1038/s41366-026-02119-x · Cited by 1 · see citing works
“Multimodal AI models that integrate electronic health record (EHR) data, genomics, imaging, wearable sensor data, and behavioral inputs deliver better risk strati fi cation performance than unimodal models [10].”
Kim, S. M., & Cho, J. M. (2026) · Diabetology & Metabolic Syndrome · doi:10.1186/s13098-026-02176-2 · Cited by 1 · see citing works
“…scoping review of 40 studies on AI-based methods for type 2 diabetes risk prediction reported that most models used classical machine learning applied to electronic health records, with only five studies reporting external validation and only five providing any assessment of calibration [16, 17].”
Narayan, K. M. V. (2026) · Diabetes Care · doi:10.2337/dci25-0116 · Cited by 1 · see citing works
Thanthrige, A. N. W., & Wickramasinghe, N. S. (2025) · doi:10.2196/preprints.80582
Artificial Intelligence in Obesity and Diabetes
Rafiq, S., Majeed, T., Din, N. M. U., Sabha, S. U., & Assad, A. (2025) · Springer eBooks · doi:10.1007/978-981-96-8176-1_11
Radovanović, J., Gluvic, Z. M., Pajčin, A., et al. (2025) · Expert Review of Molecular Diagnostics · doi:10.1080/14737159.2025.2555867
Equitable Diabetes Diagnosis: Tackling Ethnic and Gender Disparities
Ruback, L., Felix, L., & Teles, A. S. (2025) · Simpósio Brasileiro de Bancos de Dados (SBBD) · doi:10.5753/sbbd.2025.247265
“…step involves selecting the ML algorithm and training the model using the pre-processed data, which is evaluated using performance classification metrics, such as accuracy, recall, precision, F1-score, and the Area Under the Receiver Operating Characteristic curve (AUC-ROC) [Wang et al. 2024].”
Thanthrige, A. N. W., & Wickramasinghe, N. S. (2025) · doi:10.2196/preprints.86515
Deng, G., & Chengshi, W. (2026) · Science Progress · doi:10.1177/00368504261418899 · Cited by 1 · see citing works
Thanthrige, A. N. W., & Wickramasinghe, N. S. (2026) · JMIR Research Protocols · doi:10.2196/86515
“A systematic literature review of 32 studies identified key engagement barriers, including high dropout rates [24,25], inadequate personalization [26,27], environmental constraints [7,28], and cultural or language barriers [29,30].”
Retracted: AI-Based Predictive Models for Effective Diabetes Management
S.Girija, S.Thangamayan, Adiyattil, R., M.Sivadharshini, & Shrivastava, A. (2025) · International Conference on Artificial Intelligence for Innovations in Healthcare Industries (ICAIIHI) · doi:10.1109/icaiihi67124.2025.11403763
“Besides, ethical concerns, data privacy, and accessibility in diverse healthcare settings are posing additional barriers to its widespread implementation [11, 12].”
RetractedJoshi, D. M., Bhatt, H., & Patel, H. K. (2026) · BENTHAM SCIENCE PUBLISHERS eBooks · doi:10.2174/9798898813635126010011
Thanthrige, A. N. W., & Wickramasinghe, N. S. (2026) · Digital Health · doi:10.1177/20552076261436271
Diabetes Risk Prediction Based on Clinical and Lifestyle Data
Mao, T. (2026) · ITM Web of Conferences · doi:10.1051/itmconf/20268401015
“Looking ahead, it remains important for research to balance predictive accuracy with model transparency and practical usability, while also building prediction frameworks that can be validated across different populations and applied at scale [5,7,9].”
Nagar, P., Rashid, M., Arora, S., Alam, S., Agrawal, M., & Basist, P. (2026) · International Journal of Diabetes in Developing Countries · doi:10.1007/s13410-026-01651-w
Precision medicine in low-income settings and small island developing states
Saluja, S., Mannan, F., Paré, G., Anand, S. S., Hanchard, N. A., Langenberg, C., & Anderson, S. G. (2026) · Nature Reviews Endocrinology · doi:10.1038/s41574-026-01250-z · Cited by 1 · see citing works
“Such tools can be applied in a targeted way to groups who are at high risk, for example, young adults with early-onset T2DM or individuals with strong family histories of cardiometabolic disease, to guide intensified lifestyle and pharmacological interventions 69 .”
Diabetes Risk Prediction and Classification using Machine Learning Techniques
Gomathi, N., & Karuppasamy, K. (2025) · International Conference on Sustainable Computing and Smart Systems (ICSCSS) · doi:10.1109/icscss64956.2025.11500636
“Artificial intelligence (AI) technologies [3] are revolutionizing the prediction and management of diabetes across all age groups, including both the young and elderly populations.”
Sun, H., Shi, Z., Xia, Y., Bai, Y., Yu, M., Liu, R., & He, P. (2026) · BMJ Open · doi:10.1136/bmjopen-2025-115755
Gumede, S. B., Venter, W. D. F., Mndebele, N., & Lalla‐Edward, S. T. (2026) · Mayo Clinic Proceedings Digital Health · doi:10.1016/j.mcpdig.2026.100380
Lee, S.-Y., Saleem, M., Land, A. M., DeLaney, E. W., Garretson, A. R., Patel, M., Hall, A. G., Westrick, S., Ovalle, F., Cherrington, A. L., et al. (2026) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2026.106604
Advances in Cardiovascular Disease Risk Prediction in People with Diabetes
Li, Y., & Xue, J. (2026) · Journal of Biosciences and Medicines · doi:10.4236/jbm.2026.149011
Vers une diabétologie augmentée par l’IA : état des lieux et perspectives
Fagherazzi, G. (2026) · EMC - Endocrinologie - Nutrition · doi:10.1016/s1155-1941(25)48886-6
Le, J., Li, X.-X., & Qin, S.-L. (2026) · Digital Health · doi:10.1177/20552076261494727
Multimodal artificial intelligence in diabetes care: A scoping review
Xia, T., Lu, X., You, J., Liu, H., & Xu, R. S. (2026) · Digital Health · doi:10.1177/20552076261495758
- Authors
- Wang, S. C. Y., Nickel, G., Venkatesh, K. P., Raza, M. M., & Kvedar, J. C.
- Affiliations
- Harvard University (US)
- Corresponding author
- Wang, Serena C. Y.
- Venue
- npj Digital Medicine
- Published
- 2024-02-15
- Volume
- 7
- Issue
- 1
- Pages
- 36-36
- ISSN
- 2398-6352 (ISSN-L)
- PMC
- PMC10869708
- OpenAlex
- W4391839209
- Cited by
- 50
- References
- 15
- Citations by year
- 2026: 20 · 2025: 26 · 2024: 4
- FWCI
- 14.87
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- npj Digital Medicine (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Keywords
- type 2 diabetes prediction, multimodal AI, algorithmic bias, external validation, data quality, health equity, implementation barriers, early intervention
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-16.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-16.
Cited sources are checked by gpt-6-sol.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
