Sage Journals: Discover world-class research

Abstract

Introduction:

Prediabetes is a prevalent condition in which early detection and lifestyle interventions can prevent or delay progression to diabetes. Artificial intelligence (AI) and machine learning (ML) offer enhanced tools for diagnosis, risk stratification, and scalable delivery of lifestyle interventions. This review synthesizes current applications of AI/ML in patients with prediabetes.

Methods:

We conducted a scoping review using PubMed, EMBASE, and Web of Science (through May 2025) to identify original studies applying AI/ML to prediabetes prediction or management. Population-level forecasting and models combining prediabetes with other conditions were excluded. Data were extracted via structured REDCap instruments and validated through secondary review. Descriptive statistics summarized findings.

Results:

Of 2072 records screened, 149 studies met criteria: 118 prediction model studies, 20 intervention studies, and 11 miscellaneous. Machine learning models primarily targeted prediction of prediabetes, progression to diabetes, diabetic complications, and glucose metrics. Overall model performance was favorable (mean C-statistic 0.81), with random forests, neural networks, and support vector machines showing better performance. Only 20 studies reported external validation, few compared ML to standard risk tools, and data/code availability was limited. Six AI-based diabetes prevention programs showed positive clinical outcomes, though randomized controlled trial (RCT) evidence was limited. Three personalized nutrition interventions showed mixed efficacy.

Conclusion:

Most AI/ML research in prediabetes focused on predictive modeling, which shows promise but limited translation to real-world settings. Artificial intelligence-based interventions may scale behavioral change support but need further evaluation versus standard care. Future efforts should prioritize external validation, assess added value over standard tools, and address barriers to integration into care.

Keywords

artificial intelligence diabetes lifestyle intervention machine learning impaired glucose tolerance prediabetes

Introduction

Prediabetes is characterized by elevated blood glucose levels below diabetes thresholds and is diagnosed through impaired fasting glucose (100-125 mg/dL), impaired glucose tolerance (140-199 mg/dL at 2 hours post-75g glucose load), or A1C levels of 5.7% to 6.4%. Its global prevalence is projected to reach 1 billion by 2045.¹

Approximately 10% of individuals with prediabetes progress to type 2 diabetes (T2D) each year.² Prediabetes also poses health risks, including macrovascular complications such as stroke, myocardial infarction, heart failure, and peripheral vascular,^3,4 disease. Despite its prevalence, prediabetes often remains asymptomatic (81% of adults with prediabetes are unaware of their condition),⁵ leading to inadequate screening and care,⁶ further compounded by underutilization of risk assessment tools and lifestyle interventions.⁷ As such, there is an urgent need for improved prevention, early detection, and management strategies for prediabetes.

In recent years, the use of artificial intelligence (AI) and machine learning (ML) has expanded to chronic disease prevention and management,⁸ including prediabetes. Leveraging diverse data sources, such as large electronic health records (EHR), continuous glucose monitoring (CGM), and wearable devices, these tools can predict disease onset,^9,10 identify key risk factors,¹¹ and track progression to T2D or diabetes-related complications. By detecting patterns potentially missed by conventional tests like fasting glucose and A1C, AI can enhance early detection through glucose trend analysis, personalized risk stratification, and image-based diagnostics.^12-15

Beyond predictive analytics, AI can enhance lifestyle interventions for prediabetes by offering scalable and cost-effective methods to support lifestyle changes, which may address challenges like low participation and limited access to in-person diabetes prevention programs (DPPs).^16,17 These tools can emulate the personalization of human coaching through feedback loops, goal setting, behavioral tracking, and reminders.^18-20 More recently, an increasing number of direct-to-consumer digital products have emerged leveraging CGM data,^21-23 or photo-based meal scanning,^23,24 to identify and personalize lifestyle choices.

The rapid growth of AI/ML in prediabetes underscores the need for a comprehensive review detailing its applications. As such, this scoping review aims to describe how AI/ML are being applied in the context of prediabetes.

Methods

Definitions of AI and ML

We broadly defined AI as computational systems that simulate human intelligence processes such as learning, reasoning, and decision-making. ML, a subfield of AI, was defined as the use of data-driven algorithms that identify patterns and generate predictions or classifications. Within the context of prediabetes, AI/ML applications include predictive analytics to support clinical decision-making and patient-specific risk assessment (i.e., ML models), as well as for the personalized delivery of behavioral lifestyle change support (i.e., AI-based interventions).

Study Identification and Selection Process

This review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines.²⁵ As this was a scoping review and not a systematic review, our review protocol was not prospectively registered.

We conducted a systematic literature search in PubMed, Web of Science, and Embase from database inception to May 1, 2025. Search queries incorporated terms related to AI/ML and prediabetes. Studies representing original articles employing AI/ML for the clinical care of patients with prediabetes were included. Further details, including the complete search strategy, are provided in the Supplemental Methods.

Data Extraction and Validation

Data extraction was performed using REDCap electronic data capture tools hosted at Johns Hopkins University.^28,29 Studies were categorized into predictive modeling, AI-based interventions, or miscellaneous. Further details, including extraction methods, data validation, and REDCap data extraction instruments, are provided in the Supplemental Methods.

Data Analysis

Descriptive statistics were used to summarize extracted data, including counts of predicted outcomes, ML techniques, performance metrics, study designs, and validation methods. Top-performing ML techniques were identified based on internal validation C-statistics or accuracy. For studies reporting prediabetes-specific metrics, model performance was aggregated by technique, with forest plots generated when sufficient data were available. Intervention-related studies were summarized using counts and frequencies of platforms, data sources, and study designs.

Results

Included Studies

Figure 1 shows the PRISMA flow diagram and study selection process. A total of 2072 records were identified through database searching: 845 from PubMed, 626 from Embase, and 601 from Web of Science. After removing duplicates, 1359 records remained for title and abstract screening. Of these, 1190 were excluded for the following reasons: not an original article (n = 270), not meeting study definitions for AI or ML (n = 579), or not relevant to prediabetes (n = 341). The remaining 169 articles underwent full-text review, of which 149 met the eligibility criteria and were included in the final review. These comprised 118 prediction model studies, 20 intervention studies, and 11 miscellaneous articles.

Figure 1.

PRISMA flow diagram.

Figure 2 shows the distribution of included studies by year and application type, with most studies published within the past 5 years. Figure 3 highlights the primary areas of AI/ML use in prediabetes: lifestyle interventions and predictive analytics. Artificial intelligence-based interventions include personalized nutrition and automated DPPs, while ML-driven prediction models target transitions between glycemic states (e.g., normal glucose tolerance, prediabetes, diabetes, complications) and key metabolic metrics, including glucose levels, glycemic response to interventions, A1C trajectories, and insulin resistance.

Figure 2.

Annual number of publications describing AI-based interventions (purple) and ML-based predictive models (blue) focused on prediabetes from 2013 to 2025. Data reflect publication counts from the final dataset, which included 20 intervention studies and 118 predictive modeling studies.

Figure 3.

Study focus areas.

Predictive Modeling Using ML

A total of 145 ML models were developed across 118 unique studies. Table 1 categorizes these studies by predicted outcomes, with Figure 4a illustrating their distribution. As displayed in Figure 4b, most models used k-fold cross-validation (n = 71, 49.0%) or split-sample validation (n = 52, 35.9%), but only 20 models (13.8%) underwent external validation. Among the 17 models that reported both internal and external validation C-statistics, the mean internal C-statistic was 0.81 (SD: 0.08), while the mean external C-statistic was 0.77 (SD: 0.08).

Table 1.

Classification of Prediction Models by Outcomes.

Prediabetes		Progression to diabetes	Diabetic complications	Glucose values
N = 68		N = 40	N = 14	N = 10
Acciaroli et al¹³	Ahamed et al²⁸	Anderson et al²⁹	Chauhan et al³⁰	Ahmed et al³¹
Anderson et al²⁹	Alanis et al³²	Aoki et al³³	Hayashi³⁴	Bent et al³⁵
Babbar et al³⁶	Alqahtani et al³⁷	Cahn et al³⁸	Hu et al³⁹	van Doorn et al⁴⁰
Byeon⁴¹	Altuve et al⁴²	Cardozo et al⁴³	Huang et al⁴⁴	Hammour and Mandic⁴⁵
Cardozo et al⁴³	Cichosz et al⁴⁶	Carrasco-Zanini et al⁴⁷	Lai et al⁴⁸	Kashif et al⁴⁹
Carrasco-Zanini et al⁵⁰	Console et al⁹	Chen et al⁵¹	Schallmoser et al⁵²	Zeevi et al⁵³
Chen et al⁵¹	Cui et al⁵⁴	Edlitz and Segal⁵⁵	Wang et al⁵⁶	Karim et al⁵⁷
Choi et al¹⁴	Deng et al⁵⁸	Guevara et al⁵⁹	Schork et al⁶⁰	Li et al⁶¹
Chung et al⁶²	Emam et al⁶³	Han et al⁶⁴
Deberneh and Kim⁶⁵	Goodrich et al⁶⁶	Kent et al⁶⁷		Changes in A1C
De Silva et al¹¹	Igbe et al⁶⁸	Li et al⁶⁹		N = 3
Dietz et al⁷⁰	Li et al⁶¹	Li et al⁷¹		Aasmets et al⁷²
Gollapalli et al⁷³	Li et al⁷⁴	Liu et al⁷⁵		Bent et al⁷⁶
Hu et al⁷⁷	Mukhtar and Azwari Sana⁷⁸	Maeta et al⁷⁹		Patel et al⁸⁰
Jain et al⁸¹	Nguyen et al⁸²	Morgan-Benita et al⁸³
Javidi et al⁸⁴	Ray et al⁸⁵	Ou et al⁸⁶
Joglekar et al⁸⁷	Sadek et al⁸⁸	Sahid et al⁸⁹
Kamble et al⁹⁰	Smit et al⁹¹	Varga et al⁹²		Insulin resistance
Kushwaha et al⁹³	Tanim et al⁹⁴	Yashar et al⁹⁵		N = 2
Li et al⁹⁶	Tatli et al⁹⁷	Zeng et al⁹⁸		Khan et al⁹⁹
Li et al⁷¹	Tobore et al¹⁰⁰	Zhu et al¹⁰¹		Lukic et al¹⁰²
Liu et al¹⁰	Wang et al¹⁰³	Zou et al¹⁰⁴
Longato et al¹⁰⁵	Xu et al¹⁰⁶	Zueger et al¹⁰⁷
Longato et al¹⁰⁸	Yang et al¹⁰⁹	Shin et al¹¹⁰
Michel et al¹¹¹	Yuk et al¹¹²	Yuk et al¹¹²		Efficacy of Intervention
Morgan-Benita et al⁸³	Zhang et al¹¹³	Zhang et al¹¹⁴		N = 8
Muhli et al¹¹⁵	Zhong et al¹¹⁶	Han et al¹¹⁷		Diaz-Canestro et al¹¹⁸
Neri-Rosario et al¹¹⁹		Zheng 2023¹²³		Joblin-Mills et al¹²¹
Oh et al¹²²				Liu et al¹²³
Ou et al⁸⁶		Frank et al¹²⁴		Murphree et al¹²⁵
Parkhi et al¹²⁶		Ren et al¹²⁷		Sabatini et al¹²⁸
Prabha et al¹²⁹		Li et al¹³⁰		Foppiani et al¹³¹
Schwartz et al¹³²		Okere et al¹³³		Jiang et al¹³⁴
Vîrgolici and Vîrgolici¹³⁵		Li et al¹³⁶
Wang et al¹³⁷		Jaeger et al¹³⁸
Yacamán Méndez et al¹³⁹
Zeng et al⁹⁸
Zhao et al¹⁴⁰

N refers to the number of machine learning models developed for each outcome category included in this review.

Figure 4.

Characterization of ML-based predictive models: (a) Distribution of prediction model outcomes. (b) Distribution of internal validation methods used in models. (c) Reporting frequency of internal validation performance metrics. (d) Frequency of top-performing machine learning models.

Internal validation metrics (Figure 4c) were most frequently reported as C-statistic (n = 98, 67.6%), sensitivity (n = 74, 51.0%), and accuracy (n = 65, 44.8%). Sixty-five percent of studies developed multiple ML models using different techniques to predict the same outcome, typically to compare their relative performance; on average, three models were constructed per study. Figure 4d shows that the most common top-performing techniques were random forest (RF; n = 25, 17.2%), neural networks (NN; n = 21, 14.5%), and support vector machines (SVM; n = 19, 13.1%). Derivation datasets and ML model or code resources were available in 27.1% (n = 32) and 17.8% (n = 21) of studies, respectively.

Diagnosing or predicting prediabetes

Supplementary Table 1 summarizes 68 ML models predicting real-time prediabetes diagnosis (diagnostic models, n = 48, 70.6%), future prediabetes development (prognostic models, n = 19, 27.9%), and EHR-based prediabetes discussions (n = 1). Prediabetes was predicted as a binary outcome in 35 studies (51.5%) and as part of a multi-class outcome (e.g., normal, prediabetes, diabetes) in 32 studies (47.1%). Study designs included retrospective cohort (n = 26, 38.2%), prospective cohort (n = 16, 23.5%), and cross-sectional (n = 26, 38.2%), with sample sizes ranging from 18 to 381,077 (median = 1,461).

The most common top-performing methods were NN (n = 12), RF (n = 9), and SVM (n = 10). Internal validation C-statistics were reported in 43 studies; mean was 0.81 (SD: 0.10; Figure 5a). Diagnostic models (n = 27) demonstrated higher performance than prognostic models (n = 16), with a mean C-statistic of 0.84 (SD: 0.09) compared to 0.75 (SD: 0.10), respectively. Figures 5b and 5c show RF and decision tree (DT) performing best for prognostic models, and extreme gradient boosting (XGB) and ensemble methods (EM) for diagnostic models. External validation was reported in 10 studies (14.7%).

Figure 5.

Performance of ML models predicting prediabetes. (a) C-statistic distribution for prediabetes prediction models. (b) Performance of prognostic prediabetes models by ML technique. (c) Performance of diagnostic prediabetes models by ML technique. Error bars represent confidence intervals (if reported) for panel a and standard deviation for panels b and c. All plotted values reflect model performance in prediabetes-specific cohorts or subgroups within mixed cohorts.

Most studies used ADA criteria to define prediabetes, with some applying World Health Organization (WHO) criteria.^17,25 Common predictors included age, body mass index (BMI), sex, waist circumference, fasting plasma glucose, and A1C. Unconventional predictors included microRNAs,⁸⁷ tongue images and aquaphotomics,⁷¹ biochemical metabolites,⁸³ and gut microbiome profiling.¹¹⁹

Diagnosing or predicting progression to diabetes

Supplemental Table 2 summarizes 40 models predicting diabetes progression in patients with prediabetes. Most were prognostic (n = 32), with the rest being diagnostic (n = 8). Study designs included retrospective cohort (n = 25, 62.5%), prospective cohort (n = 7, 17.5%), cross-sectional (n = 5, 12.5%), and randomized clinical trial (RCT; n = 3, 7.5%). Derivation cohort sizes ranged from 80 to 1.89 million (median = 2615). Models were derived from prediabetes-only cohorts (n = 22, 55.0%) or mixed cohorts (n = 18, 45.0%). Overall mean C-statistic in studies reporting performance in prediabetes-specific populations was 0.81 (SD: 0.08), with diagnostic models performing better than prognostic models (Figure 6a). The most common top-performing ML techniques were RF (n = 7), XGB (n = 7), SVM (n = 6), and gradient boosting machine (GBM; n = 5). Figure 6b shows RF and LogR models achieving the highest C-statistics for this category. External validation was reported for 7 models (17.5%).

Figure 6.

Performance of ML models predicting diabetes, diabetic complications, and glucose values. (a) C-statistic distribution for diabetes prediction models. (b) Performance of diabetes models by ML technique. (c) C-statistic distribution for diabetic complication models predicting microvascular and macrovascular outcomes. (d) RMSE distribution for glucose value prediction models. Error bars represent confidence intervals (if reported) for panels a, c, and d, and standard deviation for panels b. All plotted values reflect model performance in prediabetes-specific cohorts or subgroups within mixed cohorts.

Frequent predictors included triglycerides, age, high-density lipoprotein (HDL) cholesterol, and A1C. Some studies used uncommon features in their ML models, such as pectoral muscle mass,⁹⁵ microbiota analysis,⁹⁸ Raman spectroscopy on the skin,⁵⁹ plasma proteins,¹⁰¹ high-performance liquid chromatography,¹⁰¹ and Ultra-Performance Liquid Chromatography.¹⁰¹

Predicting diabetic complications

Supplementary Table 3 summarizes 14 ML models predicting diabetic complications in patients with prediabetes. Common top-performing ML techniques included DT (n = 7) and NN (n = 2). Complications predicted were microvascular (neuropathy [n = 4], diabetic kidney disease [n = 3], retinopathy [n = 1]) and macrovascular (peripheral artery disease [n = 3], cerebrovascular disease, cardiovascular disease, heart failure [each n = 1]). As displayed in Figure 6c, the mean C-statistic for seven models was 0.79 (SD: 0.1), with it higher in microvascular (0.81, SD: 0.1) versus macrovascular models (0.77, SD: 0.1).

Predicting glucose values, glycemia, and insulin resistance

Supplementary Table 4 summarizes 23 ML models glucose values (n = 10), intervention efficacy (n = 8), A1C changes (n = 3), and insulin resistance (n = 2). Most used prospective cohort designs (n = 15, 65.2%) or RCTs (n = 4, 17.4%). Cohort sizes ranged from 1 to 12,147 (median = 38).

The most common top-performing techniques for predicting glucose values were NN (n = 4) and RF (n = 2). Among models reporting prediabetes-specific root mean square error (RMSE), the mean was 13.3 mg/dL (SD: 10; Figure 6d). Prediction intervals varied: 2-hour postprandial glucose (RMSE: 31.87 mg/dL;⁴⁸ R = 0.68),⁵³ 5-minute,³⁵ 15-minute, and 60-minute,⁴⁰ and 3-hour interstitial glucose (R² = 0.99) using physiological signals.⁴⁹ One study evaluated an in-ear photoplethysmography-based monitor for glucose prediction.⁴⁵

Eight ML models predicted the efficacy of interventions in patients with prediabetes, including metformin response,¹²⁵ exercise-induced changes in insulin sensitivity,¹²³ changes in insulin resistance and fasting insulin following high-intensity interval training,¹¹⁸ and metabolic responses to whey protein intake.¹²¹ Three studies predicted A1C changes using noninvasive wearables for activity patterns,⁸⁰ physiological metrics (skin temperature, heart rate),⁷⁶ and gut microbiome data.⁷² Two models predicted insulin sensitivity/resistance: RF for insulin resistance in obese Hispanic adolescents⁹⁹ and NN for hyperinsulinemia in school-age adolescents.¹⁰²

AI/ML-Based Interventions

Interventions directed toward patients with diabetes fell into two broader categories: automated DPPs that aim to replicate human coaching for behavior change (e.g., weight loss, increased physical activity) using AI; and personalized nutrition platforms that use AI to generate individualized dietary guidance. Table 2 presents nine AI-based interventions: six automated DPPs and three personalized nutrition programs. These interventions utilize various AI methodologies, including reinforcement learning to adapt interventions in real time, natural language processing for interactive, conversational support, and deep learning (e.g., recurrent neural network and long-short-term memory network) to predict future blood glucose levels and tailor nutritional recommendations.

Table 2.

Identified AI-Based Interventions.

Intervention category	Intervention name	Intervention description	Platform(s) used	Data source(s) used
Automated diabetes prevention program	AlivePD^141-143	Fully automated, algorithm-driven web-based intervention providing personalized weekly goal-setting and feedback on diet and physical activity to promote weight loss and increase physical activity	Web-based Application, Smartphone App, Emails, Phone Calls	User-entered Data
	January AI^144,145	Fully automated, app-based intervention that uses an RNN with LSTM and dense layers to predict future blood glucose levels, which inform personalized dietary and exercise recommendations. V2 includes a “digital twin” model trained on 5 days of CGM and heart rate data, enabling continued personalized recommendations without ongoing CGM use.	Smartphone App	CGM, Wearable Device, User-entered Data, Engagement Data
	Lark^146-148	CDC-recognized DPP powered by conversational AI that emulates human coaching and is monitored by human lifestyle coaches	Smartphone App	Wearable Device, Digital Body Weight Scale, Engagement Data, User-entered Data
	SMS text messaging program by Sapre et al¹⁴⁹	Conversational AI delivered two personalized daily messages to provide additional support and motivation to youth in a diabetes prevention workshop using natural language processing for interactive exchanges	Text Message	User-entered Data, Engagement Data
	Sweetch^17,150	A fully automated DPP utilizing a reinforcement learning algorithm that integrates multi-modal data to deliver personalized, just-in-time adaptive interventions	Smartphone App	Wearable Device, Smartphone Data, Digital Body Weight Scale, User-entered Data, Engagement Data
	Tess¹⁵¹	A behavioral coaching chatbot that provides personalized, evidence-based interventions through text messages, using ML to adapt its responses based on user input and emotional analysis	Text Message	Engagement Data
Personalized Nutrition	DayTwo^12,152-154	A gradient-boosting regression-based personalized PPT diet aimed at lowering PPGRs with real-time feedback through a smartphone app	Smartphone App	CGM, Biological/laboratory Samples, User-entered Data
	The Low Carb Program^155,156	NHS-approved app that promotes adoption of a low-carbohydrate diet via AI-driven recipe recommendations	Web-based Application, Smartphone App, Smartwatch App, Smart Speaker, Smart Assistant	User-entered Data, Engagement Data
	Web-based AI-driven nutrition platform¹⁵⁷	Web platform that utilizes deep learning and natural language processing to personalize recipe suggestions, meal planning, and shopping lists	Web-based Application	Smartphone Data, User-entered Data, Engagement Data

Abbreviations: AI, artificial intelligence; CGM, continuous glucose monitoring; CDC, Centers for Disease Control and Prevention; DPP, diabetes prevention program; LSTM, long short-term memory (a type of recurrent neural network); ML, machine learning; NHS, National Health Service; PPT, postprandial targeting; PPGRs, postprandial glucose responses; RNN, recurrent neural network.

Platforms for diabetes prevention (Figure 7a) primarily included smartphone apps, followed by web applications and text messaging. Less common platforms were emails, phone calls, and smart devices. Most interventions used multimodal data (Figure 7b), with data sources mainly being user-entered data (e.g., food or activity logs or health goals manually inputted by users), engagement data (e.g., app usage patterns used to optimize behavioral nudges through reinforcement learning), wearables, CGM, and body weight scales. Occasionally, smartphone data and biological samples were incorporated.

Figure 7.

Characterization of AI-based interventions: (a) distribution of platforms utilized and (b) distribution of data sources used by AI.

Supplementary Tables 5 and 6 summarize studies evaluating the performance of nine interventions in cohorts that included, at least partially, patients with prediabetes. Of the 19 studies identified, three were published RCT protocols. Because two of these protocols had completed RCTs with published results, the protocol papers were not included in the data extraction tables; instead, only the publications presenting the corresponding results were included.^141,153

Automated DPPs

Supplementary Table 5 summarizes 11 publications (10 clinical studies) evaluating six automated DPPs. Most used prospective (n = 4) or retrospective cohorts (n = 4), with only two RCTs—one for Sweetch and one for Alive-PD. The comparator arms differed: Sweetch was tested against a human coach-based DPP,¹⁷ while Alive-PD was compared to usual care.¹⁴² Study phases were evenly split across pilot, postimplementation, and full-scale evaluations. Average study duration was 7 months (most lasting 12 months), with cohorts ranging from 13 to 16 327 participants, primarily prediabetes-only, except January AI, which included mixed glucose-control populations.^144,145 Two chatbot interventions targeted youth.^149,151

Automated DPPs consistently improved clinical outcomes, including glycemic control, weight management, and physical activity. January AI significantly improved glucose levels, glycemic variability, body weight (−2.5 lbs), dietary intake (calories, carbohydrates, sugars), and physical activity (+73 min/day), with fewer hyperglycemic episodes at 28 days.¹⁴⁴ A follow-up retrospective study of January V2—an updated version of the intervention that used a 5-day CGM and heart rate-based training period to generate a digital twin simulation model—reported greater time-in-range and weight loss of 4 lbs over ~1 month among users with prediabetes, with more favorable results in more engaged users.¹⁴⁵

Lark DPP demonstrated sustained 12-month weight loss (−4.96%), with effectiveness consistent across rural, urban, and health care shortage areas.^146-148 Alive-PD showed greater reductions than usual care in A1C (−0.26%), fasting glucose, refined carbohydrate intake, and increased aerobic activity (+1.21 days/week) at 6 months in an RCT.^141-143 Sweetch pilot results (12 weeks) indicated high usability, retention, weight loss (−1.6 kg), improved physical activity (+2.8 MET-hours/week), and reduced A1C (−0.1%),^146,147 a larger Sweetch RCT has concluded with published results not yet available.¹⁷

Personalized nutrition

Supplementary Table 6 summarizes seven publications evaluating three personalized nutrition interventions across five clinical studies (three full-scale, two pilot/feasibility). Designs included prospective cohorts (n = 3) and RCTs (n = 2, both of which tested the DayTwo intervention). Study durations were typically 12 months, except one 2-month trial.¹⁵⁷ Cohorts ranged from 45 to 269 participants (average: 153), featuring mostly mixed populations except one prediabetes-only RCT.¹²

The DayTwo intervention, applying an ML model developed by Zeevi et al⁵³ predicting postprandial glucose response (PPGR), significantly reduced PPGR and improved microbiome alpha-diversity compared to a Mediterranean diet in one RCT.¹² However, another RCT found no difference in weight loss, glycemic variability, or A1C change compared to standard calorie restriction.¹⁵⁴ In addition, the web-based AI-driven nutrition platform evaluated by Bul et al¹⁵⁷ achieved significant short-term weight loss (−4.5 kg at 8 weeks), though participant engagement and usability were limited. Finally, the Low Carb Program, assessed in two prospective cohorts, significantly improved weight (−2.82 kg) and A1C (−2.35 mmol/mol) at 12 months, with outcomes closely linked to program completion.^155,156

Two interventions reported A1C outcomes: the Low Carb Program consistently reduced A1C,^155,156 whereas DayTwo showed mixed results (−0.16% vs. no significant change) across two separate RCTs.^12,158 Weight-loss results varied similarly: the Low Carb Program demonstrated meaningful weight loss,^155,156 DayTwo did not differ significantly from controls,¹⁵⁴ and the web-based AI platform showed substantial short-term weight loss.¹⁵⁷

Miscellaneous Studies

Supplemental Table 7 summarizes the 11 articles that were classified as Miscellaneous. Studies employed diverse methodologies, such as latent Markov models and cluster analysis, to classify individuals into risk categories based on obesity indices, psychosocial factors, metabolic phenotypes, and risk groups, while AI-based techniques were used to identify metabolic biomarkers, clinical predictors, and retinal vascular features, to enhance understanding of prediabetes and its implications for diabetes and cardiovascular risk.

Discussion

Most AI/ML applications identified in this review focused on predictive modeling. In contrast, relatively few studies evaluated AI-based lifestyle interventions, despite growing consumer and commercial interest in digital approaches for diabetes prevention.¹⁵⁹ However, clinical integration remains limited in both areas. Predictive models often lacked external validation and data/code accessibility, limiting their generalizability and clinical utility. Meanwhile, AI-driven lifestyle interventions were hindered by a lack of rigorous evaluations, with only a few studies directly comparing them to standard care. Together, these gaps suggest that while AI is creating promising tools for prediabetes prediction and management, their integration into routine clinical care is still in its early stages.

Across the 145 predictive models reviewed, outcome selection was skewed toward glycemic-state transitions—normoglycemia to prediabetes (n = 68) or prediabetes to diabetes (n = 40)—while fewer studies predicted continuous metrics such as glucose values, A1C, or insulin resistance. This emphasis on disease progression and diagnosis is likely more clinically actionable, as it aligns with established diagnostic thresholds used in treatment decisions and risk-based screening. In contrast, predicting surrogate markers such as HOMA-IR may have limited clinical utility, as these metrics are not direct endpoints. For example, Khan et al⁹⁹ used a breath-based random forest model to predict HOMA-IR in adolescents with prediabetes, but its relevance remains unclear without evidence of impact on diagnosis, monitoring, or intervention decisions.

One promising application of ML to improve clinical care in prediabetes is through more personalized and efficient risk screening. This review identified 40 models that predict progression of diabetes in patients with prediabetes. When embedded into EHRs, such models could operate behind the scenes to flag at-risk individuals based on routinely collected data—helping to close screening gaps (only 50%-60% of U.S. adults receive recommended A1C screening)¹⁶⁰ without adding burden to clinician workflows. By facilitating earlier diagnosis and timely intervention, these tools have the potential to improve patient outcomes and reduce progression of diabetes. In addition, ML algorithms predicting patient-specific health metrics are also being applied to tailor interventions; for instance, Zeevi et al⁵³ developed a model predicting postprandial glucose levels based on meal content, activity, CGM data, microbiome, and lifestyle habits, demonstrating benefits over the Mediterranean diet in one RCT.^12,154

Overall, diagnostic models in both the prediabetes and diabetes categories demonstrated better C-statistic performance than their prognostic counterparts, likely due to stronger contemporaneous associations between predictors and disease status. Commonly top-performing ML techniques—RF, NN, SVM, Lasso, DTs, LightGBM, and XGB—align with findings from diabetes and chronic disease prediction literature.^161-164

In our review, we observed that less than 15% of studies performed external validation on their ML models. Several factors may explain this shortfall: (1) the field is still nascent, and many research groups remain in an algorithm-development phase; (2) truly independent, well-phenotyped prediabetes cohorts are scarce, partly because prediabetes is often undocumented in routine care; (3) privacy regulations and fragmented EHR infrastructures complicate cross-institutional data sharing; and (4) many journals do not require external validation for publication. Without testing in diverse populations, even technically strong models risk overfitting to local practice patterns and will likely fail to gain regulatory approval given the Food and Drug Administration (FDA)’s requirement for external validation within its regulatory framework for AI/ML-based medical devices.¹⁶⁵

In prediction-focused studies, we observed considerable variability in the reporting of performance metrics, with over 20 distinct metrics used across studies. This heterogeneity hampers cross-study comparisons and limits benchmarking of model performance in clinically meaningful ways. The TRIPOD-AI statement provides guidance on reporting standard metrics for ML models predicting binary outcomes, including AUC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).¹⁶⁶ In the context of prediabetes diagnosis, where disease prevalence is high, concerns about overstated performance due to class imbalance are reduced. Although considered optional, positive and negative likelihood ratios can offer valuable diagnostic insight. Calibration metrics (e.g., Brier scores), which assess the agreement between predicted probabilities and observed outcomes, are often underreported but strongly recommended.

The selection of performance metrics should align with the model’s intended use. Diagnostic models may prioritize sensitivity to reduce false negatives, particularly in screening contexts, while specificity may be emphasized when false positives carry risks of overtreatment or unnecessary testing. Prognostic models, aimed at forecasting disease progression, benefit from metrics assessing both discrimination (e.g., C-statistic) and calibration for effective risk stratification. Models predicting outcomes along the prediabetes continuum should also be benchmarked against existing clinical tools to clarify their added value—an approach largely absent from the literature. In rare cases where this has been done, ML models demonstrated improved reclassification over A1C and outperformed tools like the Finnish Diabetes Risk Score, highlighting the importance of direct comparisons for clinical relevance.^55,130 The Net Reclassification Index (NRI) can further quantify these improvements over reference models.¹⁶⁷

Beyond the lack of external validation and benchmarking, many ML models face integration barriers due to reliance on impractical predictors—such as tongue images, gut microbiome data, or mass spectrometry—that are not routinely collected in clinical care.¹⁶⁸ This limits their real-world applicability. For ML tools to be useful in clinical workflow, they must be practical and operate in harmony with existing approaches. Models should therefore prioritize accessible inputs, rather than introducing new burdens.

Furthermore, the black-box nature of many ML models has been cited as a barrier to clinical adoption, particularly when providers cannot understand how predictions are made.^169,170 This challenge is compounded by the frequent lack of publicly available code, data, or trained models, which limits independent validation and prevents other researchers from replicating or building upon the work. For tools intended to support frontline decision-making, models must be understandable and interpretable. Yet this review highlights a lack of model transparency, potentially limiting clinician trust and real-world adoption.

This review also highlights several findings regarding AI-based interventions for prediabetes. We found that AI interventions are understudied as most are evaluated in observational studies without comparison to standard care (i.e., the human-coach-based DPP).¹⁷¹ The lack of RCT data in this space may be driven by underlying regulatory procedures. Most AI-driven lifestyle interventions fall outside FDA oversight, as the agency generally exempts general wellness tools from regulation.¹⁷² Instead, these programs pursue validation through Centers for Disease Control and Prevention (CDC) recognition, which relies on observational data and a per-protocol evaluation in program completers—an approach exemplified by Lark’s recognition based on retrospective data alone.¹⁵⁹ The broader literature has highlighted the need for more RCTs to assess AI tools in clinical care,¹⁷³ and comparisons to human-delivered interventions are especially important given that many AI systems, including the interventions in this review, aim to replicate functions traditionally performed by health professionals. However, a large-scale RCT evaluating Sweetch, which concluded in late 2024 (results not yet available), is the first to compare an AI-based intervention directly to the standard-of-care DPP. Demonstrating comparable performance between AI and human coaching could unlock scalable delivery of lifestyle interventions and help address the widespread underuse of preventive care in prediabetes.¹⁷⁴

While these interventions are generally well-received and show potential for promoting lifestyle changes, their effectiveness depends on user engagement, which is inconsistently measured.¹⁷⁵ For instance, the Lark DPP study defined engagement via lesson completion, coaching exchanges, and weigh-in frequency—each independently linked to ≥5% weight loss—whereas the January V2 study used food logging frequency to identify “power users,” who achieved greater improvements in time in range and modestly greater weight loss.^145,146 This engagement-outcome association is analogous to that observed in human-coach DPPs,¹⁷⁶ suggesting that future work should focus on identifying strategies that reliably sustain user engagement in AI-based interventions.

In addition, our findings reveal that AI-based interventions for diabetes prevention are integrating diverse data inputs like CGM, photo-based meal analysis, and data from wearables to provide hyper-personalized recommendations. This trend fits well with the growing availability of over-the-counter CGMs, which are making real-time glucose monitoring more accessible to individuals with prediabetes as well as healthy adults seeking wellness insights.¹⁷⁷ As these devices become more widely used in this population, they open new opportunities for AI tools to integrate dynamic glucose data into personalized feedback loops. Interventions like January AI have already begun leveraging CGM data to guide individualized dietary and activity recommendations.

The most common delivery modalities for AI-based interventions were mobile applications, web-based platforms, and text messaging. Each offers distinct advantages and limitations. Mobile apps may be the most promising due to their portability, bidirectional communication capabilities, and potential for real-time, context-aware feedback. Web platforms support structured content delivery and are conducive to in-depth education but may be less accessible for users seeking on-the-go support. Text messaging is broadly accessible and low-cost yet limited in interactivity and capacity for personalized engagement.

Integration of AI-based tools for diabetes prevention faces several implementation barriers. First, the lack of rigorous effectiveness data compared to standard-of-care lifestyle programs raises questions about whether these tools should be adopted at scale or if resources would be better spent expanding the CDC’s established National DPP. Second, although asynchronous delivery can lower geographic barriers, real-world uptake still depends on reliable internet, smartphone access, and digital literacy—factors that can vary widely by age, income, and rurality. Consequently, future work must conduct head-to-head trials with appropriate endpoints linked to diabetes incidence (e.g., weight loss, A1C reduction) and pragmatic studies of access, engagement, and adherence to determine where and for whom AI interventions add the greatest value.

Strengths of our review include a comprehensive multidatabase literature search, clear inclusion and exclusion criteria that promoted the relevance of included studies, and a rigorous multireviewer process that reduced bias and errors in data extraction. Structured data extraction further supported consistency and reliability in reporting. However, our study also had limitations. Relying on published studies may have introduced publication bias, as negative results are often underreported, and some performance data may remain in company white papers rather than peer-reviewed literature. Gray literature and conference proceedings were not reviewed, and no non-English articles were identified or included. In addition, by focusing only on top-performing ML models within studies, we may have overlooked potentially useful algorithms.

Future research should focus on strengthening the clinical relevance and implementation potential of AI/ML tools for prediabetes. For predictive models, priorities include conducting external validation in diverse populations, improving model transparency through open code and interpretable features, and benchmarking performance against existing risk scores to clarify added value. A systematic synthesis of top-performing predictors across outcome categories could enhance interpretability and guide the selection of clinically practical variables that consistently contribute to model performance. For AI-based interventions, randomized trials comparing digital programs to human-coach DPPs are needed, alongside implementation studies that address disparities in access, digital literacy, and sustained engagement.

Conclusions

This scoping review provides an overview of the applications of AI/ML for prediabetes. The majority of research focuses on predictive analytics, with fewer studies evaluating AI-based interventions. While ML-based prediction models demonstrated good discrimination across various outcomes, their clinical utility and practicality remain largely unvalidated. For interventions, AI-driven lifestyle programs show promising results; however, their real-world effectiveness remains uncertain.

Supplemental Material

sj-docx-1-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-docx-1-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-2-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-2-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-3-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-3-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-4-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-4-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-5-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-5-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-6-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-6-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Supplemental Material

sj-xlsx-7-dst-10.1177_19322968251351995 – Supplemental material for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review

Supplemental material, sj-xlsx-7-dst-10.1177_19322968251351995 for Applications of Artificial Intelligence and Machine Learning in Prediabetes: A Scoping Review by Benjamin Lalani, Rohan Herur, Daniel Zade, Grace Collins, Devin M. Dishong, Setu Mehta, Jalene Shim, Yllka Valdez and Nestoras Mathioudakis in Journal of Diabetes Science and Technology

Footnotes

Abbreviations

A1C, hemoglobin A1C; ADA, American Diabetes Association; AI, artificial intelligence; BMI, body mass index; CatBoost, categorical boosting; CDC, Centers for Disease Control and Prevention; CGM, continuous glucose monitoring; Cox, Cox proportional hazards model; DPP, diabetes prevention program; DT, decision tree; EHR, electronic health record; EM, ensemble methods; FDA, Food and Drug Administration; GBM, gradient boosting machine; GMI, glucose management indicator; ITT, intention-to-treat; Lasso, least absolute shrinkage and selection operator; LGBM, light gradient boosting machine; LogR, logistic regression; MET, metabolic equivalent of task; ML, machine learning; NB, naïve bayes; NN, neural network; PPGR, postprandial glucose response; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT, randomized controlled trial; RF, random forest; RMSE, root mean squared error; SVM, support vector machine; T2D, type 2 diabetes; WHO, World Health Organization; XGB, eXtreme gradient boosting.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was funded by a grant (grant no. R01DK125780) from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institutes of Health (NIH; healthinfo@niddk.nih.gov; USA; 1-800-860-8747).

ORCID iDs

Benjamin Lalani

Rohan Herur

Daniel Zade

Grace Collins

Devin M. Dishong

Setu Mehta

Jalene Shim

Yllka Valdez

Nestoras Mathioudakis

Supplemental Material

Supplemental material for this article is available online.

References

Rooney

Fang

Ogurtsova

, et al. Global prevalence of prediabetes. Diabetes Care. 2023;46:1388-1394.

Echouffo-Tcheugui

Perreault

Dagogo-Jack

. Diagnosis and management of prediabetes: a review. JAMA. 2023;329:1206-1216.

Zand

Ibrahim

Patham

. Prediabetes: why should we care? Methodist Debakey Cardiovasc J. 2018;14(4):289-297.

Brannick

Dagogo-Jack

. Prediabetes and cardiovascular disease: pathophysiology and interventions for prevention and risk reduction. Endocrinol Metab Clin North Am. 2018;47(1):33-50.

National Diabetes Statistics Report . Atlanta, GA: Centers for Disease Control and Prevention; 2024.

Karve

Hayward

. Prevalence, diagnosis, and treatment of impaired fasting glucose and impaired glucose tolerance in nondiabetic U.S. adults. Diabetes Care. 2010;33(11):2355-2359.

Dhippayom

Chaiyakunapruk

Krass

. How diabetes risk assessment tools are implemented in practice: a systematic review. Diabetes Res Clin Pract. 2014;104(3):329-342.

Singareddy

Jaramillo

, et al. Artificial intelligence and its role in the management of chronic medical conditions: a systematic review. Cureus. 2023;15(9):e46066.

Console

Lenatti

Simeone

, et al. Exploring prediabetes pathways using explainable AI on data from electronic medical records. Stud Health Technol Inform. 2024;316:736-740.

10.

Liu

Chang

Chen

, et al. Machine learning prediction of prediabetes in a young male Chinese cohort with 5.8-year follow-up. Diagnostics (Basel). 2024;14:979.

11.

De Silva

Jönsson

Demmer

. A combined strategy of feature selection and machine learning to identify predictors of prediabetes. J Am Med Inform Assoc. 2020;27:396-406.

12.

Ben-Yacov

Godneva

Rein

, et al. Personalized postprandial glucose response-targeting diet versus Mediterranean diet for glycemic control in prediabetes. Diabetes Care. 2021;44(9):1980-1991.

13.

Acciaroli

Sparacino

Hakaste

, et al. Diabetes and prediabetes classification using glycemic variability indices from continuous glucose monitoring data. J Diabetes Sci Technol. 2018;12:105-113.

14.

Choi

Kim

Yoo

, et al. Screening for prediabetes using machine learning models. Comput Math Methods Med. 2014;2014:618976.

15.

Zee

Lee

Lai

, et al. Digital solution for detection of undiagnosed diabetes using machine learning-based retinal image analysis. BMJ Open Diabetes Res Care. 2022;10(6):e002914.

16.

Venkataramani

Pollack

Yeh

Maruthur

. Prevalence and correlates of diabetes prevention program referral and participation. Am J Prev Med. 2019;56(3):452-457.

17.

Abusamaan

Ballreich

Dobs

, et al. Effectiveness of artificial intelligence vs. human coaching in diabetes prevention: a study protocol for a randomized controlled trial. Trials. 2024;25:325.

18.

Auster-Gussman

Lockwood

Graham

Pitter

Branch

. Engagement in digital health app-based prevention programs is associated with weight loss among adults age 65. Front Digit Health. 2022;4:886783.

19.

Joiner

Nam

Whittemore

. Lifestyle interventions based on the diabetes prevention program delivered via eHealth: a systematic review and meta-analysis. Prev Med. 2017;100:194-207.

20.

Sepah

Jiang

Ellis

McDermott

Peters

. Engagement and outcomes in a digital diabetes prevention program: 3-year update. BMJ Open Diabetes Res Care. 2017;5(1):e000422.

21.

Veri. Your Personalized Metabolic Health Program. Helsinki, Finland: Veri; 2024.

22.

Signos. How It Works. Signos; 2024.

23.

Rxfood. Rxfood for Patients. RxFood; 2024.

24.

January AI. How It Works. Menlo Park, CA: January AI; 2024.

25.

Tricco

Lillie

Zarin

, et al. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Ann Intern Med. 2018;169:467-473.

26.

Harris

Taylor

Thielke

, et al. Conde, Research electronic data capture (REDCap) - A metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377-381.

27.

Harris

Taylor

Minor

, et al. The REDCap consortium: Building an international community of software partners. J Biomed Inform. 2019. doi:10.1016/j.jbi.2019.103208.

28.

Ahamed

Arya

Sangeetha

SKB

Auxilia Osvin

. Diabetes mellitus disease prediction and type classification involving predictive modeling using machine learning techniques and classifiers. Appl Comput Intell Soft Comput. 2022;2022:7899364.

29.

Anderson

Parikh

Shenfeld

, et al. Reverse engineering and evaluation of prediction models for progression to type 2 diabetes: an application of machine learning using electronic health records. J Diabetes Sci Technol. 2015;10:6-18.

30.

Chauhan

Varre

Izuora

Trabia

Dufek

. Prediction of diabetes mellitus progression using supervised machine learning. Sensors (Basel). 2023;23.

31.

Ahmed

Ali

Masud

Azad

Naznin

. After-meal blood glucose level prediction for type-2 diabetic patients. Heliyon. 2024;10:e28855.

32.

Alanis

Sanchez

Vaca-González

Rangel-Heras

. Intelligent classification and diagnosis of diabetes and impaired glucose tolerance using deep neural networks. Mathematics. 2023;11:4065.

33.

Aoki

Khalid

Kaya

Nagymanyoki

Hussong

Salama

. Progression from prediabetes to diabetes in a diverse U.S. population: a machine learning model. Diabetes Technol Ther. 2024;26:748-753.

34.

Hayashi

. Detection of lower albuminuria levels and early development of diabetic kidney disease using an artificial intelligence-based rule extraction approach. Diagnostics (Basel). 2019;9:133.

35.

Bent

Cho

Henriquez

, et al. Engineering digital biomarkers of interstitial glucose from noninvasive smartwatches. NPJ Digit Med. 2021;4:89.

36.

Babbar

Heni

Peter

, et al. Prediction of glucose tolerance without an oral glucose tolerance test. Front Endocrinol (Lausanne). 2018;9:82.

37.

Alqahtani

SAM

Alobaid

Alshammari

, et al. Feature importance and model performance for prediabetes prediction: a comparative study. J King Saud Univ Sci. 2024;36:103583.

38.

Cahn

Shoshan

Sagiv

, et al. Prediction of progression from pre-diabetes to diabetes: development and validation of a machine learning model. Diabetes Metab Res Rev. 2020;36(2):e3252.

39.

Reaven

Saremi

, et al. Machine learning to predict rapid progression of carotid atherosclerosis in patients with impaired glucose tolerance. EURASIP J Bioinform Syst Biol. 2016;2016(1):14.

40.

van Doorn

WPTM

Foreman

Schaper

, et al. Machine learning-based glucose prediction with use of continuous glucose and physical activity monitoring data: the Maastricht Study. PLoS ONE. 2021;16(6):e0253125.

41.

Byeon

. Exploring the risk factors of impaired fasting glucose in middle-aged population living in South Korean communities by using categorical boosting machine. Front Endocrinol (Lausanne). 2022;13:1013162.

42.

Altuve

Alvarez

Severeyn

. Multiclass classification of metabolic conditions using fasting plasma levels of glucose and insulin. Health Technol. 2021;11:953-962.

43.

Cardozo

Pintarelli

Andreis

Lopes

ACW

Marques

JLB

. Use of machine learning and routine laboratory tests for diabetes mellitus screening. Biomed Res Int. 2022;2022:8114049.

44.

Huang

Huth

Covic

, et al. Machine learning approaches reveal metabolic signatures of incident chronic kidney disease in individuals with prediabetes and type 2 diabetes. Diabetes. 2020;69(12):2756-2765.

45.

Hammour

Mandic

. An in-ear PPG-based blood glucose monitor: a proof-of-concept study. Sensors (Basel). 2023;23:3319.

46.

Cichosz

Kronborg

Laugesen

, et al. From stability to variability: classification of healthy individuals, prediabetes, and type 2 diabetes using glycemic variability indices from continuous glucose monitoring data. Diabetes Technol Ther. 2025;27(1):34-44.

47.

Carrasco-Zanini

Pietzner

Wheeler

Kerrison

Langenberg

Wareham

. Multi-omic prediction of incident type 2 diabetes. Diabetologia. 2024;67(1):102-112.

48.

Lai

Chiu

Huang

, et al. Longitudinal artificial intelligence-based deep learning models for diagnosis and prediction of the future occurrence of polyneuropathy in diabetes and prediabetes. Neurophysiol Clin. 2024;54(4):102982.

49.

Kashif

Flesca

Veltri

. GLSTM: on using LSTM for glucose level prediction. Stud Health Technol Inform. 2024;314:103-107.

50.

Carrasco-Zanini

Pietzner

Lindbohm

, et al. Proteomic signatures for identification of impaired glucose tolerance. Nat Med. 2022;28(11):2293-2300.

51.

Chen

Zhou

Yang

, et al. Risk prediction of diabetes progression using big data mining with multifarious physical examination indicators. Diabetes Metab Syndr Obes. 2024;17:1249-1265.

52.

Schallmoser

Zueger

Kraus

, et al. Machine learning for predicting micro- and macrovascular complications in individuals with prediabetes or diabetes: retrospective cohort study. J Med Internet Res. 2023;25:e42181.

53.

Zeevi

Korem

Zmora

, et al. Personalized nutrition by prediction of glycemic responses. Cell. 2015;163:1079-1094.

54.

Cui

, et al. Study on risk factors of impaired fasting glucose and development of a prediction model based on extreme gradient boosting algorithm. Front Endocrinol (Lausanne). 2024;15:1368225.

55.

Edlitz

Segal

. Prediction of type 2 diabetes mellitus onset using logistic regression-based scorecards. eLife. 2022;11:e71862.

56.

Wang

Hou

, et al. Development and validation of a prediction model based on machine learning algorithms for predicting the risk of heart failure in middle-aged and older US people with prediabetes or diabetes. Clin Cardiol. 2023;46:1234-1243.

57.

Karim

Anderson

Currie

Bui

Klyve

Somers

. A glycemic status classification model using a radiofrequency noninvasive blood glucose monitor. Diabetes Technol Ther. 2024;26(12):979-983.

58.

Deng

Dai

Liu

, et al. Application of tongue image characteristics and oral-gut microbiota in predicting pre-diabetes and type 2 diabetes with machine learning. Front Cell Infect Microbiol. 2024;14:1477638.

59.

Guevara

Torres-Galván

González

, et al. Feasibility of Raman spectroscopy as a potential in vivo tool to screen for pre-diabetes and diabetes. J Biophotonics. 2022;15:e202200055.

60.

Schork

Fritsche

Schleicher

, et al. Differential risk assessment in persons at risk of type 2 diabetes using urinary peptidomics. Metabolism. 2025;167:156174.

61.

Tobore

Liu

Kandwal

Wang

Nie

. Non-invasive monitoring of three glucose ranges based on ECG by using DBSCAN-CNN. IEEE J Biomed Health Inform. 2021;25(9):3340-3350.

62.

Chung

Kim

Choi

Park

Kim

. Screening for pre-diabetes using support vector machine model. Annu Int Conf IEEE Eng Med Biol Soc. 2014;2014:2472-2475.

63.

Emam

Tarek

Soudy

, et al. Comparative evaluation of multiomics integration tools for the study of prediabetes: insights into the earliest stages of type 2 diabetes mellitus. Netw Model Anal Health Inform Bioinform. 2024;13:8.

64.

Han

Yang

Huang

, et al. Risk prediction of diabetes and pre-diabetes based on physical examination data. Math Biosci Eng. 2022;19:3597-3608.

65.

Deberneh

Kim

. Prediction of type 2 diabetes based on machine learning algorithm. Int J Environ Res Public Health. 2021;18:3317.

66.

Goodrich

Wang

Walker

, et al. Postprandial metabolite profiles and risk of prediabetes in young people: a longitudinal multicohort study. Diabetes Care. 2024;47:151-159.

67.

Kent

Nelson

Pittas

, et al. An electronic health record-compatible model to predict personalized treatment effects from the diabetes prevention program: a cross-evidence synthesis approach using clinical trial and real-world data. Mayo Clin Proc. 2022;97(4):703-715.

68.

Igbe

Kandwal

, et al. An absolute magnitude deviation of HRV for the prediction of prediabetes with combined artificial neural network and regression tree methods. Artif Intell Rev. 2022;55:2221-2244.

69.

Yuan

, et al. A tongue features fusion approach to predicting prediabetes and diabetes with machine learning. J Biomed Inform. 2021;115:103693.

70.

Dietz

Machann

Agrawal

, et al. Detection of diabetes from whole-body MRI using deep learning. JCI Insight. 2021;6:e146999.

71.

Chen

, et al. Establishment of noninvasive diabetes risk prediction model based on tongue features and machine learning techniques. Int J Med Inform. 2021;149:104429.

72.

Aasmets

Lüll

Lang

, et al. Machine learning reveals time-varying microbial predictors with complex effects on glucose regulation. Msystems. 2021;6:10.1128/msystems.01191-20.

73.

Gollapalli

Alansari

Alkhorasani

, et al. A novel stacking ensemble for detecting three types of diabetes mellitus using a Saudi Arabian dataset: pre-diabetes, T1DM, and T2DM. Comput Biol Med. 2022;147:105757.

74.

Ding

Zhang

, et al. Interpretable machine learning method to predict the risk of pre-diabetes using a national-wide cross-sectional data: evidence from CHNS. BMC Public Health. 2025;25:1145.

75.

Liu

Zhou

, et al. Predicting the 2-year risk of progression from prediabetes to diabetes using machine learning among Chinese elderly adults. J Pers Med. 2022;12:1055.

76.

Bent

Cho

Wittmann

, et al. Non-invasive wearables for remote monitoring of HbA1c and glucose variability: proof of concept. BMJ Open Diabetes Res Care. 2021;9(1):e002027.

77.

Han

Liu

, et al. A nomogram model for predicting 5-year risk of prediabetes in Chinese adults. Sci Rep. 2023;13:22523.

78.

Mukhtar

Azwari Sana

. Investigating non-laboratory variables to predict diabetic and prediabetic patients from electronic medical records using machine learning. Int J Comput Sci Netw Secur. 2021;21:19-30.

79.

Maeta

Nishiyama

Fujibayashi

, et al. Prediction of glucose metabolism disorder risk using a machine learning algorithm: pilot study. JMIR Diabetes. 2018;3:e10212.

80.

Patel

Polsky

Small

, et al. Predicting changes in glycemic control among adults with prediabetes from activity patterns collected by wearable devices. NPJ Digit Med. 2021;4:172.

81.

Jain

Patel

Hanawal

, et al. Machine learning for predicting diabetic metabolism in the Indian population using polar metabolomic and lipidomic features. Metabolomics. 2023;20:1.

82.

Nguyen

Choi

Byeon

. An explainable hybrid deep learning model for prediabetes prediction in men aged 30 and above. J Mens Health. 2024;20:52-72.

83.

Morgan-Benita

Sánchez-Reyna

Espino-Salinas

, et al. Metabolomic selection in the progression of type 2 diabetes mellitus: a genetic algorithm approach. Diagnostics (Basel). 2022;12:2803.

84.

Javidi

Mariam

Alkhaled

, et al. An interpretable predictive deep learning platform for pediatric metabolic diseases. J Am Med Inform Assoc. 2024;31:1227-1238.

85.

Ray

Sangavi

Vishwakarma

, et al. Breath analysis using quartz tuning forks for predicting blood glucose levels using artificial neural networks. ACS Sens. 2024;9:5468-5478.

86.

Jin

Lin

Chen

Quan

. LASSO-based machine learning algorithm to predict the incidence of diabetes in different stages. Aging Male. 2023;26(1):2205510.

87.

Joglekar

Kunte

Wong

WKM

, et al. Circulating microRNAs from early childhood and adolescence are associated with pre-diabetes at 18 years of age in women from the PMNS cohort. J Dev Orig Health Dis. 2022;13(6):806-811.

88.

Sadek

Abdelhafez

Al-Hashimi

, et al. Screening for diabetes and impaired glucose metabolism in Qatar: models’ development and validation. Prim Care Diabetes. 2022;16(1):69-77.

89.

Sahid

Babar

MUH

Uddin

. Predictive modeling of multi-class diabetes mellitus using machine learning and filtering Iraqi diabetes data dynamics. PLoS ONE. 2024;19(5):e0300785.

90.

Kamble

Collins

Harvey

, et al. Understanding prediabetes in a Medicare advantage population using data adaptive techniques. Popul Health Manag. 2018;21(6):477-485.

91.

Smit

Botterblom

Mulder

. Skin autofluorescence based decision tree in detection of impaired glucose tolerance and diabetes. PLoS ONE. 2013;8(6):e65592.

92.

Varga

Liu

Goldberg

, et al. Predictive utilities of lipid traits, lipoprotein subfractions and other risk factors for incident diabetes: a machine learning approach in the diabetes prevention program. BMJ Open Diabetes Res Care. 2021;9(1):e001953.

93.

Kushwaha

Srivastava

Jain

, et al. Harnessing machine learning models for non-invasive pre-diabetes screening in children and adolescents. Comput Methods Programs Biomed. 2022;226:107180.

94.

Tanim

Aurnob

Shrestha

, et al. Explainable deep learning for diabetes diagnosis with DeepNetX2. Biomed Signal Process Control. 2025;99:106902.

95.

Yashar

Izci

Gungoren

, et al. Can artificial intelligence detect type 2 diabetes in women by evaluating the pectoral muscle on tomosynthesis: diagnostic study. Insights Imaging. 2024;15:99.

96.

Guo

, et al. Early diagnosis of type 2 diabetes based on near-infrared spectroscopy combined with machine learning and aquaphotomics. Front Chem. 2020;8:580489.

97.

Tatli

Papapanagiotou

Liakos

, et al. Prediabetes detection in unconstrained conditions using wearable sensors. Clin Nutr Open Sci. 2024;58:163-174.

98.

Zeng

Zhao

Wang

, et al. Integrating choline and specific intestinal microbiota to classify type 2 diabetes in adults: a machine learning based metagenomics study. Front Endocrinol (Lausanne). 2022;13:906310.

99.

Khan

Cuda

Karere

Cox

Bishop

. Breath biomarkers of insulin resistance in pre-diabetic Hispanic adolescents with obesity. Sci Rep. 2022;12:339.

100.

Tobore

Kandwal

, et al. Towards adequate prediction of prediabetes using spatiotemporal ECG and EEG feature analysis and weight-based multi-model approach. Knowl Based Syst. 2020;209:106464.

101.

Zhu

Huang

, et al. Serological phenotyping analysis uncovers a unique metabolomic pattern associated with early onset of type 2 diabetes mellitus. Front Mol Biosci. 2022;9:841209.

102.

Lukic

Savic

Simic

, et al. Risk assessment and determination of factors that cause the development of hyperinsulinemia in school-age adolescents. Medicina (Kaunas). 2021;58:9.

103.

Wang

Zhang

Jia

Zhao

. Construction and validation of a model for predicting impaired fasting glucose based on more than 4000 general population. Int J Gen Med. 2023;16:1415-1428.

104.

Zou

Luo

Huang

, et al. Differential effect of interventions in patients with prediabetes stratified by a machine learning-based diabetes progression prediction model. Diabetes Obes Metab. 2024;26(1):97-107.

105.

Longato

Acciaroli

Facchinetti

, et al. Glycaemic variability-based classification of impaired glucose tolerance vs. type 2 diabetes using continuous glucose monitoring data. Comput Biol Med. 2018;96:141-146.

106.

Sun

Wang

, et al. Risk factor assessment of prediabetes and diabetes based on epidemic characteristics in new urban areas: a retrospective and a machine learning study. Sci Rep. 2025;15:3792.

107.

Zueger

Schallmoser

Kraus

Saar-Tsechansky

Feuerriegel

Stettler

. Machine learning for predicting the risk of transition from prediabetes to diabetes. Diabetes Technol Ther. 2022;24(11):842-847.

108.

Longato

Acciaroli

Facchinetti

Maran

Sparacino

. Simple linear support vector machine classifier can distinguish impaired glucose tolerance versus type 2 diabetes using a reduced set of CGM-based glycemic variability indices. J Diabetes Sci Technol. 2020;14(2):297-302.

109.

Yang

Fang

, et al. Pre-diabetes diagnosis based on ATR-FTIR spectroscopy combined with CART and XGBoots. Optik. 2019;180:189-198.

110.

Shin

Kim

Lee

, et al. Development of various diabetes prediction models using machine learning techniques. Diabetes Metab J. 2022;46(4):650-657.

111.

Michel

Rospleszcz

Reisert

, et al. Deep learning to estimate impaired glucose metabolism from Magnetic Resonance Imaging of the liver: an opportunistic population screening approach. PLoS Digit Health. 2024;3(1):e0000429.

112.

Yuk

Juhui

Kee

Jaesuk

Heo

T-Y

. Artificial intelligence-based prediction of diabetes and prediabetes using health checkup data in Korea. Appl Artif Intell. 2022;36:2145644.

113.

Zhang

Yao

Wang

Zhang

. Development and validation of machine learning models for identifying prediabetes and diabetes in normoglycemia. Diabetes Metab Res Rev. 2024;40(8):e70003.

114.

Zhang

Luo

, et al. Microbiome and metabolome dysbiosis analysis in impaired glucose tolerance for the prediction of progression to diabetes mellitus. J Genet Genomics. 2024;51(1):75-86.

115.

Muhli

Benchraka

Lotankar

, et al. Aberrations in the early pregnancy serum metabolic profile in women with prediabetes at two years postpartum. Metabolomics. 2023;19:20.

116.

Zhong

Fang

Fan

, et al. Lipidomic profiling reveals distinct differences in plasma lipid composition in healthy, prediabetic, and type 2 diabetic individuals. Gigascience. 2017;6:1-12.

117.

Han

Liu

Wang

Liu

. Nomogram model and risk score to predict 5-year risk of progression from prediabetes to diabetes in Chinese adults: development and validation of a novel model. Diabetes Obes Metab. 2023;25(3):675-687.

118.

Diaz-Canestro

Chen

Liu

, et al. A machine-learning algorithm integrating baseline serum proteomic signatures predicts exercise responsiveness in overweight males with prediabetes. Cell Rep Med. 2023;4:100944.

119.

Neri-Rosario

Martínez-López

Esquivel-Hernández

, et al. Dysbiosis signatures of gut microbiota and the progression of type 2 diabetes: a machine learning approach in a Mexican cohort. Front Endocrinol (Lausanne). 2023;14:1170459.

120.

Zheng

Chen

, et al. Development and validation of risk prediction models for new-onset type 2 diabetes in adults with impaired fasting glucose. Diabetes Res Clin Pract. 2023;197:110571.

121.

Joblin-Mills

Fraser

, et al. The impact of ethnicity and intra-pancreatic fat on the postprandial metabolome response to whey protein in overweight Asian Chinese and European Caucasian women with prediabetes. Front Clin Diabetes Healthc. 2022;3:980856.

122.

Lee

Pak

. An interactive online app for predicting diabetes via machine learning from environment-polluting chemical exposure data. Int J Environ Res Public Health. 2022;19:5800.

123.

Liu

Wang

, et al. Gut microbiome fermentation determines the efficacy of exercise for diabetes prevention. Cell Metab. 2020;31:77-91.e5.

124.

Frank

Hames

Jbaily

Park

Stroyeck

Price

. Feasibility of using a factory-calibrated continuous glucose monitoring system to diagnose type 2 diabetes. Diabetes Technol Ther. 2022;24(12):907-914.

125.

Murphree

Arabmakki

Ngufor

Storlie

McCoy

. Stacked classifiers for individualized prediction of glycemic control following initiation of metformin therapy in type 2 diabetes. Comput Biol Med. 2018;103:109-115.

126.

Parkhi

Periyathambi

Ghebremichael-Weldeselassie

, et al. Prediction of postpartum prediabetes by machine learning methods in women with gestational diabetes mellitus. iScience. 2023;26:107846.

127.

Ren

Lin

Liu

, et al. Potential novel serum metabolic markers associated with progression of prediabetes to overt diabetes in a Chinese population. Front Endocrinol (Lausanne). 2021;12:745214.

128.

Sabatini

Nolan

O’Donoghue

, et al. Baseline phenotypes with preserved beta-cell function and high insulin concentrations have the best improvements in glucose tolerance after weight loss: results from the prospective DEXLIFE and EGIR-RISC studies. Metabolism. 2024;155:155910.

129.

Prabha

Yadav

Rani

Singh

. Design of intelligent diabetes mellitus detection system using hybrid feature selection based XGBoost classifier. Comput Biol Med. 2021;136:104664.

130.

Sun

, et al. Nuclear magnetic resonance-based metabolomics with machine learning for predicting progression from prediabetes to diabetes. eLife. 2024;13:RP98709.

131.

Foppiani

De Amicis

Leone

, et al. Predicting non-responders to lifestyle intervention in prediabetes: a machine learning approach. Eur J Clin Nutr. 2025;79(2):142-147.

132.

Schwartz

Tseng

Maruthur

Rouhizadeh

. Identification of prediabetes discussions in unstructured clinical documentation: validation of a natural language processing algorithm. JMIR Med Inform. 2022;10:e29803.

133.

Nkemdirim Okere

Theran

Nyasani

Ali

. Evaluation of factors predicting transition from prediabetes to diabetes among patients residing in underserved communities in the United States—a machine learning approach. Comput Biol Med. 2025;187:109824.

134.

Jiang

Lai

Muirhead

, et al. Deep serum lipidomics identifies evaluative and predictive biomarkers for individualized glycemic responses following low-energy diet-induced weight loss: a PREVention of diabetes through lifestyle intervention and population studies in Europe and around the World (PREVIEW) substudy. Am J Clin Nutr. 2024;120(4):864-878.

135.

Vîrgolici

. Predicting prediabetes using simple a multi-layer perceptron neural network model. Stud Health Technol Inform. 2023;305:168-171.

136.

Mao

, et al. Machine learning models integrating dietary indicators improve the prediction of progression from prediabetes to type 2 diabetes mellitus. Nutrients. 2025;17:947.

137.

Wang

Zhao

Wang

Che

. IGRNet: a deep learning model for non-invasive, real-time diagnosis of prediabetes through electrocardiograms. Sensors (Basel). 2020;20:2556.

138.

Jaeger

Casanova

Demesie

, et al. Development and validation of a diabetes risk prediction model with individualized preventive intervention effects. J Clin Endocrinol Metab. 2025. doi:10.1210/clinem/dgaf250.

139.

Yacamán Méndez

Zhou

Trolle Lagerros

, et al. Characterization of data-driven clusters in diabetes-free adults and their utility for risk stratification of type 2 diabetes. BMC Med. 2022;20:356.

140.

Zhao

Getmanenko

Zhang

Yao

. A formula based on autonomic test using EZSCAN and anthropometric data for diagnosis of DM in China. Sci Rep. 2020;10:4870.

141.

Block

Azar

Block

, et al. A fully automated diabetes prevention program, Alive-PD: program design and randomized controlled trial protocol. JMIR Res Protoc. 2015;4:e3.

142.

Block

Azar

Romanelli

, et al. Diabetes prevention and weight loss with a fully automated behavioral intervention by email, web, and mobile phone: a randomized controlled trial among persons with prediabetes. J Med Internet Res. 2015;17:e240.

143.

Block

Azar

Romanelli

, et al. Improving diet, activity and wellness in adults at risk of diabetes: randomized controlled trial. Nutr Diabetes. 2016;6:e231.

144.

Zahedani

McLaughlin

Veluvali

, et al. Digital health application integrating wearable data and behavioral patterns improves metabolic health. NPJ Digit Med. 2023;6:216.

145.

Veluvali

Dehghani Zahedani

Hosseinian

, et al. Impact of digital health interventions on glycemic control and weight management. NPJ Digit Med. 2025;8:20.

146.

Graham

Pitter

Hori

Stein

Branch

. Weight loss in a digital app-based diabetes prevention program powered by artificial intelligence. Digit Health. 2022;8:20552076221130619.

147.

Graham

Auster-Gussman

Lockwood

Branch

. Weight loss in a digital diabetes prevention program for people in health professional shortage and rural areas. Popul Health Manag. 2023;26(3):149-156.

148.

Auster-Gussman

Lockwood

Graham

Stein

Branch

. Reach of a fully digital diabetes prevention program in health professional shortage areas. Popul Health Manag. 2022;25(4):441-448.

149.

Sapre

Elaiho

Brar Prayaga

, et al. The development of a text messaging platform to enhance a youth diabetes prevention program: observational process study. JMIR Form Res. 2024;8:e45561.

150.

Everett

Kane

Yoo

Dobs

Mathioudakis

. A novel approach for fully automated, personalized health coaching for adults with prediabetes: pilot clinical trial. J Med Internet Res. 2018;20:e72.

151.

Stephens

Joerin

Rauws

Werk

. Feasibility of pediatric obesity and prediabetes treatment support through Tess, the AI behavioral coaching chatbot. Transl Behav Med. 2019;9:440-447.

152.

Ben-Yacov

Godneva

Rein

, et al. Gut microbiome modulates the effects of a personalised postprandial-targeting (PPT) diet on cardiometabolic markers: a diet intervention in pre-diabetes. Gut. 2023;72(8):1486-1496.

153.

Popp

St-Jules

, et al. The rationale and design of the personal diet study, a randomized clinical trial evaluating a personalized approach to weight loss in individuals with pre-diabetes and early-stage type 2 diabetes. Contemp Clin Trials. 2019;79:80-88.

154.

Popp

Kharmats

, et al. Effect of a personalized diet to reduce postprandial glycemic response vs a low-fat diet on weight loss in adults with abnormal glucose metabolism and obesity: a randomized clinical trial. JAMA Netw Open. 2022;5:e2233760.

155.

Summers

Tobin

Unwin

. Evaluation of the Low Carb Program digital intervention for the self-management of type 2 diabetes and prediabetes in an NHS England general practice: single-arm prospective study. JMIR Diabetes. 2021;6:e25751.

156.

Scott

Shehata

Panesar

Summers

Dale

. The Low Carb Program for people with type 2 diabetes and pre-diabetes: a mixed methods feasibility study of signposting from general practice. BJGP Open. 2022;6(1):BJGPO.2021.0137.

157.

Bul

Holliday

Bhuiyan

MRA

, et al. Usability and preliminary efficacy of an artificial intelligence-driven platform supporting dietary management in diabetes: mixed methods study. JMIR Hum Factors. 2023;10:e43959.

158.

Kharmats

Popp

, et al. A randomized clinical trial comparing low-fat with precision nutrition-based diets for weight loss: impact on glycemic variability and HbA1c. Am J Clin Nutr. 2023;118(2):443-451.

159.

Lalani

Shim

Vadini

Valdez

Zade

Mathioudakis

. Consumer-oriented review of digital diabetes prevention programs: insights from the CDC’s diabetes prevention recognition program. Front Clin Diabetes Healthc. 2025;6:1562108.

160.

O’Brien

Bailey

Gregory

, et al. Screening for prediabetes and diabetes in a national network of federally qualified health centers: an observational study. J Gen Intern Med. 2023;38(16):3541-3548.

161.

Ebrahim

Derbew

. Application of supervised machine learning algorithms for classification and prediction of type-2 diabetes disease status in Afar regional state, Northeastern Ethiopia 2021. Sci Rep. 2023;13:7779.

162.

Perreault

Pan

Schroeder

, et al. Regression from prediabetes to normal glucose regulation and prevalence of microvascular disease in the Diabetes Prevention Program Outcomes Study (DPPOS). Diabetes Care. 2019;42(9):1809-1815.

163.

Kanbour

Harris

Lalani

, et al. Machine learning models for prediction of diabetic microvascular complications. J Diabetes Sci Technol. 2024;18(2):273-286.

164.

Tan

Seng

JJB

Kwan

, et al. Evaluation of machine learning methods developed for prediction of diabetes complications: a systematic review. J Diabetes Sci Technol. 2023;17(2):474-489.

165.

Warraich

Tazbaz

Califf

. FDA perspective on the regulation of artificial intelligence in health care and biomedicine. JAMA. 2025;333:241-247.

166.

Collins

Moons

KGM

Dhiman

, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378.

167.

Leening

Vedder

Witteman

, et al. Net reclassification improvement: computation, interpretation, and controversies: a literature review and clinician’s guide. Ann Intern Med. 2014;160:122-131.

168.

Yang

Thompson

. Regulatory framework for clinical decision support software: present uncertainty and prospective proposition. J Am Coll Radiol. 2015;12(7):672-675.

169.

Murdoch

Singh

Kumbier

Abbasi-Asl

. Definitions, methods, and applications in interpretable machine learning. Proc Natl Acad Sci U S A. 2019;116:22071-22080.

170.

Rasheed

Qayyum

Ghaly

Al-Fuqaha

Razi

Qadir

. Explainable, trustworthy, and ethical machine learning for healthcare: a survey. Comput Biol Med. 2022;149:106043.

171.

Knowler

Barrett-Connor

Fowler

, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002;346:393-403.

172.

U.S. Food & Drug Administration. Step 1: Is the Software Function Intended for a Medical Purpose? Silver Spring, MD: U.S. Food & Drug Administration.

173.

Ouyang

Hogan

. We need more randomized clinical trials of AI. NEJM AI. 2024;1:AIe2400881.

174.

Alva

Chakkalakal

Moin

Galaviz

. The diabetes prevention gap and opportunities to increase participation in effective interventions. Health Aff (Millwood). 2022;41(7):971-979.

175.

Eaton

Vallejo

McDonald

, et al. User engagement with mHealth interventions to promote treatment adherence and self-management in people with chronic health conditions: systematic review. J Med Internet Res. 2024;26:e50508.

176.

Herman

Villatoro

Joiner

McEwen

. Retention and outcomes of National Diabetes Prevention Program enrollees and non-enrollees with prediabetes: the University of Michigan experience. J Diabetes Complications. 2023;37(8):108527.

177.

Oganesova

Pemberton

Brown

. Innovative solution or cause for concern? the use of continuous glucose monitors in people not living with diabetes: a narrative review. Diabet Med. 2024;41:e15369.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.88 MB

0.03 MB

0.02 MB