Digital health technologies for diabetes-centered five-condition co-management in China: Theoretical foundations, practical experience, and technical challenges

Abstract

Type 2 diabetes mellitus in China is increasingly accompanied by a clustering of metabolic abnormalities, hypertension, dyslipidemia, hyperuricemia, and excess body weight/obesity, forming a mutually reinforcing network that amplifies cardiometabolic risk and accelerates target-organ damage. Traditional single-disease management and specialty silos leave substantial residual risk and are difficult to scale in primary care. This review delineates the rationale and evolving practice of diabetes-centered five-condition co-management in China, integrating shared mechanisms of glucotoxicity, lip toxicity, urate dysmetabolism, and obesity-related inflammation with relevant policy and academic consensus. We synthesize real-world digital health experience and emerging artificial intelligence applications across screening, multimodal monitoring, risk stratification, individualized target setting, and precision lifestyle and pharmacologic interventions, and discuss data-driven collaborative networks linking tertiary hospitals, primary care, and community management to enable closed-loop follow-up. Finally, we critically examine technical, ethical, and regulatory challenges, particularly data interoperability and governance, privacy protection, algorithmic bias and interpretability, and workflow integration, and propose research priorities, including multicenter evaluations with clinically meaningful endpoints and scalable, primary-care-appropriate solutions.

Key Words: Artificial intelligence; Cardiometabolic risk; Chronic disease management; Digital health; Dyslipidemia; Hypertension; Hyperuricemia; Obesity; Precision medicine; Type 2 diabetes mellitus

Core Tip: Diabetes-centered five-condition co-management integrates glycemic control with concurrent management of hypertension, dyslipidemia, hyperuricemia, and obesity to address shared cardiometabolic pathophysiology and synergistic risk amplification. This review synthesizes the epidemiologic rationale, mechanistic links, and evidence from digital health-enabled care models, highlighting how continuous monitoring, personalized decision support, and multidisciplinary workflows can improve target attainment and adherence. We propose a practical implementation framework with measurable endpoints, risk stratification, and governance requirements for artificial intelligence tools, while outlining barriers such as interoperability, equity, and long-term safety and effectiveness evaluation in real-world settings.

INTRODUCTION

China’s prevention and control of chronic diseases is entering a phase of intensified pressure[1]. Non-communicable diseases have become the leading threat to population health, accounting for approximately 91% of all deaths, while associated medical spending constitutes more than 70% of total health expenditure[1,2]. Within the broad non-communicable disease spectrum, a clustered set of metabolic risk factors, hypertension, hyperlipidemia, hyperglycemia, hyperuricemia, and excess body weight (overweight/obesity), is increasingly recognized as a high-impact “risk-factor constellation” (hereafter referred to as the “five-condition” cluster)[3,4]. Rather than acting independently, these abnormalities interact through overlapping pathophysiological pathways to form a mutually reinforcing “metabolic disorder network”, thereby accelerating the onset and progression of major cardiometabolic outcomes, including cardiovascular and cerebrovascular disease and diabetes-related complications[3,5].

However, contemporary chronic disease management remains largely organized around single diseases, specialty silos, and episodic care, which is poorly aligned with the reality of multimorbidity[6,7]. In China, the control rate of hypertension among adults is reported to be only 18%, and the treatment rate of diabetes is 32.9%[8-10]. Such fragmented care models struggle to address interdependent risk factors and their multiplicative harm[6,11]. For example, individuals with coexisting diabetes and hypertension have a significantly higher risk of stroke than those with hypertension alone; similarly, hypertension co-occurring with hyperlipidemia is associated with a markedly increased stroke risk and evidence of risk amplification consistent with interaction (supra-additive effects)[11-13]. These observations underscore that “risk-factor clustering” is not merely additive; it can amplify downstream event risk and widen residual cardiometabolic risk even when one indicator is partially controlled[11,13].

The traditional “three-condition co-management” (hyperlipidemia, hypertension, hyperglycemia) is insufficient, as overweight/obesity (affecting > 200 million Chinese adults) and hyperuricemia (approximately 170 million adults)[14-16] are key metabolic risk drivers linked to insulin resistance and chronic inflammation[16-18]. Extending to a diabetes-centered five-condition co-management framework, treating the five factors as an interconnected risk cluster, shifts focus from single-disease targets to integrated cardiometabolic risk reduction[15,19].

In this review, five-condition co-management is defined as an integrated, risk-cluster-oriented paradigm with three core tenets: Multidimensional risk assessment, personalized intervention bundles, and data-driven iterative adjustment. Notably, all supporting evidence to date is indirect, derived from metabolic syndrome, multimorbidity, and digital diabetes interventions, no dedicated studies explicitly test the five-condition framework. We therefore distinguish its theoretical plausibility (based on shared pathophysiology) from unvalidated clinical outcomes throughout this review.

Accordingly, we synthesize the theoretical origins and practical evolution of five-condition co-management, focus on diabetes-centered digital health implementation pathways, appraise key barriers (data interoperability, governance, primary care capacity), and propose research priorities for a China-adapted chronic disease management model.

THEORETICAL FOUNDATIONS OF THE “FIVE-CONDITION CO-MANAGEMENT” CONCEPT

Conceptual origins: Background and core breakthroughs

The “five-condition co-management” concept emerged from two converging developments: Growing recognition of the limitations of the traditional “three-condition co-management” model and accumulating epidemiologic evidence on the escalating harm of clustered metabolic risk factors. Historically, chronic disease management has centered on hypertension, hyperglycemia, and hyperlipidemia[20-23]. However, the rapid rise in overweight/obesity and hyperuricemia, and their increasingly prominent contributions to cardiometabolic risk, has made it progressively difficult for the conventional paradigm to comprehensively capture today’s dominant drivers of cardiovascular and cerebrovascular burden[24-26].

Epidemiologic data further underscore the urgency of broadening the management scope. In China, the prevalence of overweight among young adults aged 18-24 years has reached 28.5%, adult obesity has exceeded 16%, and the absolute number of individuals with obesity has surpassed 200 million[27-30]. Meanwhile, the incidence of hyperuricemia is increasing at an estimated rate of 9.7% per year and exhibits a notable trend toward younger onset (“rejuvenation”)[16,31]. The affected population is now close to 170 million, approximately one in eight adults[32]. Importantly, overweight/obesity and hyperuricemia do not simply add risk on top of the traditional “three conditions”; rather, they interact with hyperglycemia, hypertension, and dyslipidemia through intertwined pathophysiological pathways, thereby amplifying atherosclerotic cardiovascular disease risk in a manner that can exceed the arithmetic sum of individual factors[4,33,34].

Against this backdrop, and building on practical experience from “three-condition co-management” (hypertension, hyperglycemia, and hyperlipidemia) alongside established approaches to weight control and hyperuricemia management, we propose an integrated digital health framework and a diabetes-centered, five-dimensional collaborative workflow. The core breakthrough of the “five-condition” concept is a reframing of these abnormalities from independent diseases or isolated risk factors into a single, interconnected risk-cluster target for systematic prevention and control, aimed at comprehensive upstream risk reduction for cardiovascular and cerebrovascular disease.

Pathophysiological mechanisms: Scientific basis for cluster prevention and control

Accumulating evidence indicates that the “five-condition” risk factors share convergent pathophysiological pathways, providing a mechanistic rationale for cluster-based prevention and control at the molecular level (Figure 1). A central unifying axis is insulin resistance, which can be induced or exacerbated by hypertension, hyperglycemia, dyslipidemia, and obesity/overweight through distinct yet overlapping mechanisms[35,36]. Insulin resistance not only impairs vascular endothelial function directly but also promotes renin-angiotensin system activation, enhances pro-inflammatory signaling, and accelerates atherogenesis, thereby linking metabolic dysregulation to downstream cardiovascular risk[37-39].

Figure 1 Shared pathophysiological links underpinning diabetes-centered five-condition co-management and cardiovascular risk. A: Insulin resistance serves as a central hub linking hyperglycemia/diabetes, hypertension, dyslipidemia, and obesity, thereby promoting vascular endothelial dysfunction, activation of the renin-angiotensin system, and atherogenesis; B: A chronic low-grade inflammatory network, reflected by elevated cytokines and inflammatory biomarkers (e.g., tumor necrosis factor-α, interleukin-1, interleukin-6, and C-reactive protein), contributes to endothelial injury and plaque initiation and progression; C: Oxidative stress and endothelial dysfunction arise from increased reactive oxygen species generation and reduced antioxidant capacity, resulting in decreased nitric oxide bioavailability, impaired vasomotor regulation, and a procoagulant milieu; D: Metabolite accumulation-associated toxicity (including urate- and lipid/glucose-related signals) can activate the NOD-like receptor family pyrin domain containing 3 inflammasome, amplifying systemic inflammation and reinforcing the interconnected progression of cardiometabolic abnormalities. CRP: C-reactive protein; IL: Interleukin; NLRP3: NOD-like receptor family pyrin domain containing 3; TNF-α: Tumor necrosis factor-α.

A second shared hallmark is chronic low-grade inflammation. Immune activation in the five-condition state is accompanied by elevated inflammatory mediators, including tumor necrosis factor-α, interleukin-6, and C-reactive protein[40-42]. Persistent inflammatory signaling drives endothelial injury and vascular remodeling, creates a permissive milieu for lipid retention, and facilitates plaque initiation and progression[40-42].

Third, the five-condition cluster jointly aggravates oxidative stress and endothelial dysfunction. Increased production of reactive oxygen species, together with reduced antioxidant capacity, diminishes nitric oxide bioavailability, disrupts vasomotor homeostasis, and promotes a pro-thrombotic state, manifesting as impaired vasodilation, heightened vascular tone, and enhanced coagulation potential[43-46]. These interconnected processes establish a mechanistic bridge from metabolic risk clustering to atherosclerotic cardiovascular disease.

Finally, metabolite-related cumulative toxicity provides an additional layer of synergistic harm. Hyperuricemia, a downstream product of purine metabolism, contributes to vascular injury through crystal deposition-related endothelial damage and activation of the NOD-like receptor family pyrin domain containing 3 inflammasome[47,48]. Importantly, hyperuricemia can interact with the other four conditions to amplify vascular inflammation and injury beyond isolated effects[49,50]. Collectively, these shared pathways support the view that the five-condition factors should be managed as an integrated risk cluster rather than as separate targets. This mechanistic convergence provides key scientific justification for the development and implementation of “five-condition co-management”.

Policy and academic consensus: Dual guarantees for concept implementation

The emergence of “five-condition co-management” reflects a bidirectional reinforcement between policy direction and academic evidence. From a policy perspective, China’s National Health Commission launched the 2024-2026 three-year “weight management year” initiative, which advocates comprehensive prevention and control of metabolism-related risk factors and encourages a shift beyond single-disease management. This orientation effectively extends the conventional “three-condition” focus by emphasizing the growing public health relevance of excess body weight and hyperuricemia, thereby supporting a transition from “single risk-factor control” toward “multi-factor collaborative management”. Collectively, these policy signals provide an institutional and governance backdrop for integrating clustered metabolic risks and advancing the implementation of “five-condition co-management”.

In parallel, academic research has established a progressively stronger evidentiary basis for “five-condition co-management”. A relatively mature expert consensus already exists for “three-condition co-management”, and multiple studies have shown that the coexistence of hypertension, hyperglycemia, and dyslipidemia is associated with a disproportionate escalation in cardiovascular event risk, whereas comprehensive, integrated management yields greater benefits than separate control of individual factors[51,52]. With advancing research, the mechanistic and epidemiologic links between hyperuricemia, overweight/obesity, and the traditional “three conditions” have become clearer: These abnormalities converge on key pathways such as insulin resistance and chronic low-grade inflammation, accelerate metabolic derangement, and collectively heighten the risk of cardiovascular and cerebrovascular complications. Notably, data from the China Kadoorie Biobank indicate that individuals concurrently exposed to three or more “five-condition” factors experience a 5.2-fold higher risk of cardiovascular and cerebrovascular disease and a 2.8-fold higher risk of all-cause mortality compared with those without these risk factors[5,53]. Together, these findings strengthen the rationale for positioning “five-condition co-management” as a core framework within chronic disease prevention and control, facilitating its translation from a conceptual proposal into more standardized, actionable clinical practice.

Core features of five-condition co-management

The core features of five-condition co-management include integration, synergy, and individualization: It treats the five conditions as an interconnected cardiometabolic risk cluster rather than isolated diseases managed in parallel; it designs intervention bundles that explicitly account for pathophysiological and therapeutic interactions among multiple metabolic abnormalities to mitigate amplified (supra-additive) risk; and it delivers risk-stratified, person-centered management plans with tailored intensity, targets, and follow-up frequency based on baseline cardiometabolic/renal risk, comorbidity patterns, and dynamic responses to therapy.

THE PRACTICAL NECESSITY AND PUBLIC HEALTH VALUE OF “FIVE-CONDITION CO-MANAGEMENT”

Epidemiological urgency: The multiplier effect of comorbidity risk

From an epidemiological standpoint, major metabolic chronic conditions in China, hypertension, hyperglycemia/diabetes, dyslipidemia, hyperuricemia, and obesity/overweight, have evolved into a large-scale, overlapping disease burden that poses a substantial public health challenge. Current estimates indicate approximately 245 million individuals living with hypertension, about 600 million people affected by overweight/obesity, and nearly 170 million patients with hyperuricemia[54-58]. Importantly, these conditions rarely occur in isolation; instead, they cluster within the same individuals, creating a multimorbidity pattern that can magnify downstream cardiometabolic risk. For example, evidence suggests that among patients with diabetes, the coexistence of additional metabolic abnormalities such as hypertension, dyslipidemia, and obesity is associated with a 6.16-fold higher risk of myocardial infarction, highlighting the synergistic and potentially supra-additive harm of “five-condition” clustering[59].

This compounded disease burden not only accelerates individual-level morbidity and premature mortality but also exerts sustained pressure on health systems through increased service demand, long-term medication use, and the management of complications. Collectively, these epidemiologic realities underscore the practical urgency of shifting from single-disease prevention and control toward integrated, multidimensional collaborative management.

Limitations of current management systems: Inefficiency and resource waste

The inefficiency inherent in single-disease, fragmented management further reinforces the practical need for “five-condition co-management”. In routine practice, conventional chronic disease care is often organized around separate conditions and specialties, leading to discontinuous follow-up, inconsistent target setting, and delayed treatment intensification. As a result, simultaneous attainment of targets for hypertension, dyslipidemia, and hyperglycemia remains extremely low (reported at approximately 5%), suggesting that siloed approaches are poorly suited to controlling interdependent metabolic disorders in the real world[60,61].

Moreover, delayed or inadequate long-term management can drive avoidable resource consumption and downstream costs[60,61]. For instance, compared with patients identified and managed earlier, individuals with unrecognized advanced chronic kidney disease incur substantially higher medical expenditures, total costs are reported to be 26% higher, and kidney disease-related costs 19% higher[62]. Late-stage recognition and delayed management are also associated with suboptimal dialysis initiation and greater demand for acute care services[62]. Collectively, these findings imply that gaps in proactive, longitudinal management can accelerate progression to severe complications, thereby amplifying end-stage treatment needs and overall health system burden.

Policy and social value: Alignment with healthy China strategic goals

From the perspective of policy alignment and social value, five-condition co-management is highly consistent with the core objectives of the Healthy China Action (2019-2030) and provides a pragmatic pathway to operationalize the chronic disease prevention and control principle of “three-early” (early screening, early assessment, and early intervention). By integrating multidisciplinary resources and establishing a closed-loop management mechanism that links risk screening, comprehensive evaluation, and personalized intervention within primary care settings, this model can enhance the precision and continuity of chronic disease management while facilitating the downward extension of high-quality services to community-level facilities.

Internationally, similar trends emphasize integrated multimorbidity care, e.g., the World Health Organisation’s Integrated Care for Chronic Conditions framework and the United States Patient-Centered Medical Home, sharing core tenets of person-centeredness and care coordination, but differing in implementation due to China’s unique primary care capacity and policy context.

In the context of population ageing and persistent inequities in the distribution of medical resources, five-condition co-management has the potential to strengthen the service capacity of primary healthcare institutions, reduce avoidable referrals, and limit redundant testing through standardized pathways and coordinated follow-up. With process optimization and technology enablement, it also supports more efficient resource allocation and offers an implementable route toward a more functional hierarchical medical system.

Overall, the value proposition of five-condition co-management is coherent and cumulative: Epidemiologic evidence highlights the multiplier effect of clustered risk; real-world performance gaps demonstrate the need for a model shift; and national policy objectives clarify the strategic significance of an integrated, closed-loop practice pathway. Taken together, these elements position five-condition co-management not merely as a tactical response to the growing burden of metabolic disease, but as a strategic approach to advance China’s chronic disease system from reactive treatment toward proactive, population-oriented health management, with public health benefits that are expected to expand as implementation deepens.

EVIDENCE MAP OF DIGITAL HEALTH INTERVENTIONS IN FIVE-CONDITION CO-MANAGEMENT

Implementation of five-condition co-management is workflow- and data-dependent, requiring digital health/artificial intelligence (AI) to translate the concept into practice. These technologies address three core barriers: Multimodal monitoring (early risk warning), precision intervention (personalized decision support), and collaborative management (cross-institution care coordination).

This is a narrative review. Literature search: PubMed, CNKI (2018-2024) using keywords: “digital health”, “AI”, “T2DM”, “hypertension”, “dyslipidemia”, “hyperuricemia”, “obesity”, “co-management”. Inclusion: Type 2 diabetes mellitus (T2DM) with ≥ 2 metabolic abnormalities, digital interventions, quantitative outcomes (English/Chinese). Exclusion: Case reports, reviews, animal studies. Evidence graded by GRADE system.

High-quality clinical evidence is essential for adoption. We summarize evidence from Chinese and international studies on digital health for T2DM with multiple metabolic abnormalities (Table 1), mapping study characteristics and outcomes to identify benefits and gaps. These studies provide only indirect evidence for AI-enabled five-condition co-management, as none explicitly test this model; observed benefits of digital metabolic risk management support the framework’s theoretical plausibility, not validated clinical outcomes. Notably, existing evidence often lacks long-term clinical endpoints (e.g., cardiovascular events) and robust cost-effectiveness data, while algorithm bias (e.g., age/gender disparities in prediction accuracy) and real-world implementation failures (e.g., low adherence due to poor user experience) remain underreported.

Table 1 Evidence map of digital interventions in the management of diabetes complicated with multiple metabolic abnormalities in China and internationally.

Research region	Study design	Sample size	Population characteristics	Intervention form	Follow-up duration	Primary outcome	Risk of bias/Limitations	Ref.
China	Quasi-experiment (pre-post control design)	189	T2DM patients complicated with hypertension	AI-powered remote monitoring system + personalized medical guidance	9 months	HbA1c decreased by 0.6% on average, blood pressure control rate increased by 36%, fasting plasma glucose reduced by 43.8 mg/dL	(1) No randomization (potential selection bias); (2) Loss to follow-up (15% attrition rate); and (3) Unblinded design (performance bias)	[118]
China	Quasi-experiment (single-arm intervention with historical control)	156	Primary T2DM patients (with DR risk)	Multimodal AI system (fundus images + clinical data) + referral management	6 months	DR screening accuracy reached the level of professional ophthalmologists, referral compliance was significantly improved, HbA1c decreased by 0.8% on average	(1) Historical control may introduce confounding bias; (2) Limited to primary care settings; and (3) No assessment of DR progression (only screening accuracy)	[119]
International	Parallel-group randomized clinical trial	246	T2DM combined with metabolic syndrome patients	APP-based self-management (blood glucose/diet/exercise recording) + healthy behavior rewards	12 months	HbA1c decreased by 0.4% (mean difference vs usual care), body weight decreased by 3.0 kg, LDL-C decreased by 0.38 mmol/L, intervention compliance increased by 32%	(1) Selection bias (strict inclusion/exclusion criteria); (2) Reward mechanism may introduce performance bias; and (3) Lack of subgroup analysis by age/gender	[120]
International	Quasi-experiment (non-randomized controlled trial)	112	T2DM patients complicated with proatherogenic dyslipidemia	Multimodal remote monitoring (blood glucose + lipid + postprandial glucose) + telehealth consultation	10 months	HbA1c decreased by 0.7% on average, LDL-C reduced by 0.45 mmol/L, postprandial glucose variability decreased by 52.6 mg/dL	(1) Small sample size (limited statistical power); (2) Single-center design (geographic bias); and (3) No blinding of outcome assessors	[121]
International	Real-world observational study	418	T2DM patients with multiple metabolic abnormalities	AI-integrated management platform (medication + diet + exercise + metabolic prediction)	15 months	HbA1c decreased by 1.3% on average, body weight reduced by 5.1 kg, metabolic index compliance rate increased by 48%, insulin sensitivity improved	(1) Selection bias (voluntary participation); (2) Lack of control group (cannot rule out temporal trends); and (3) Technical threshold for platform use (excludes elderly/illiterate patients)	[122]

T2DM: Type 2 diabetes mellitus; AI: Artificial intelligence; HbA1c: Glycated hemoglobin; DR: Diabetic retinopathy; LDL-C: Low-density lipoprotein cholesterol.

Dynamic monitoring innovation: Early identification and accurate early warning of multi-dimensional risks

A key contribution of AI to monitoring metabolism-related risks is the shift from “single-indicator detection” toward multidimensional risk early warning[63], an approach that aligns closely with the early screening and early intervention demands of five-condition co-management. Emerging evidence suggests that prediction models integrating multi-source data can improve the efficiency and timeliness of identifying metabolic abnormalities, particularly in real-world settings where risk evolves dynamically and single measurements may be misleading[64-67].

For example, a Super Learner ensemble model developed by a Zhejiang University team using approximately 460000 community health examination records incorporated 10 core predictors (e.g., body mass index, age, and liver function indicators) to construct a five-level risk scoring system[68]. The model demonstrated good discrimination, with an area under the curve (AUC) of 0.816 in the training set and 0.810 in the validation set, outperforming traditional rule-based diagnostic criteria[68]. Its translational relevance lies in the ability to flag “subthreshold” or “invisible” high-risk individuals, those who may not meet conventional diagnostic cutoffs yet already exhibit metabolic imbalance, thereby offering a practical tool to support early detection of five-condition-related risk profiles in community-based screening pathways.

Complementing these findings, a SMART-MR analysis in 1232 patients with arterial disease reported that associations between metabolic syndrome (encompassing core components overlapping with the five-condition cluster) and abnormal brain tissue structure could be detected even before overt diabetes develops[69]. This observation reinforces the rationale for AI-enabled early monitoring as a means to prevent metabolic-related complications and supports the construction of an “early screening-early warning” system within the broader five-condition co-management framework.

Precision intervention enhancement: Intelligent optimization and verification of individualized schemes

AI’s capability to integrate and interpret multi-source, heterogeneous data has been increasingly validated in settings such as metabolic syndrome and multimorbidity management, and this technical logic aligns closely with the individualized, risk-stratified core of five-condition co-management. A growing body of studies indicates that AI-driven, personalized intervention strategies can improve the control of key metabolic indicators[70,71]. International research in cardiometabolic disease further suggests that digital health platforms combining electronic medical records, continuous glucose monitoring data, lifestyle records, and other patient-generated information can support dynamic optimization of glucose-lowering, lipid-lowering, and weight-management plans, while improving treatment adherence and target attainment rates[63].

Domestic studies have reported similar findings. For example, a metabolic syndrome prediction model developed using data mining and machine learning achieved strong discrimination (AUC 0.947) after integrating multipoint health examination data, and it was able to identify actionable determinants such as dietary patterns and exercise frequency, thereby providing concrete targets for individualized intervention design[72].

At the primary-care level, a randomized controlled trial evaluating a digital, AI-assisted clinical workflow (built on large-model-enabled decision support) showed that the system could help frontline clinicians synthesize multidimensional data from patients with hypertension and diabetes and iteratively adjust management plans. The resulting closed loop, “data integration → intelligent decision-making → scheme optimization”, was associated with improved precision in chronic disease management in community settings, offering a potentially replicable technical framework for individualized implementation of five-condition co-management at the grassroots level[73].

Collaborative management systems: Multi-factor linkage and cross-scenario integration in practice

A core requirement of five-condition co-management is to move beyond single-disease silos and enable coordinated control of multiple metabolic indicators through integrated workflows. Evidence from multimorbidity co-management programs has already established a workable logic for collaboration, anchored in data sharing, stratified pathways, and multidisciplinary coordination, which offers direct implementation lessons for the five-condition framework.

For example, a community-based program in Chengyang District, Qingdao, applying a model of “three-condition co-management and six diseases co-prevention”, provided four years of continuous management for approximately 34500 older adults. By integrating regional medical data and embedding hierarchical management processes, the program improved coordinated control of blood pressure, glycemia, and lipids. Reported outcomes showed that the lipid control rate increased from 55.20% in 2019 to 60.95% in 2022, while the attainment rates for blood glucose and blood pressure also rose steadily over time[74]. The key contribution of this study is its validation of a scalable operating model characterized by “data integration → tiered management → multidisciplinary collaboration”, a framework that can be readily adapted to the development of five-condition co-management systems.

Internationally, the IMPACT system further advances collaborative management toward greater technical sophistication. Leveraging interpretable AI and multimodal large language model approaches, it applies a non-dominated sorting genetic algorithm to jointly optimize risk reduction across multiple disease domains and to generate personalized recommendations for conditions such as cardiovascular and cerebrovascular disease and diabetes[75]. This concept of multi-disease risk co-optimization is closely aligned with the synergistic, multi-factor hazards inherent to the five-condition cluster and provides a forward-looking technical reference for cross-condition, collaborative intervention under five-condition co-management.

Clinical implementation framework of diabetes-centered five-dimensional co-management

To translate the concept of five-condition co-management into routine care, we propose a standardized, diabetes-centered workflow using T2DM as the practical entry point. This framework spans risk screening and stratification, implementation of individualized intervention bundles, and dynamic, data-driven adjustment, thereby operationalizing a closed-loop process across the care continuum (Figure 2). In parallel, explicit specification of monitoring indicators, early-warning thresholds, and role-based intervention responsibilities is essential to ensure accountability and continuity, and to prevent the pathway from devolving into parallel, uncoordinated target pursuit (Table 2).

Figure 2 Diabetes-centered clinical implementation framework for five-condition co-management. At the initial visit, patients undergo a baseline metabolic assessment covering glycemia (glycated hemoglobin and/or fasting plasma glucose), blood pressure, lipids, serum uric acid, and body weight/waist circumference. Risk stratification is then performed according to overall target attainment and complication status, yielding three tiers: High risk (≥ 2 targets unmet with complications), managed by a multidisciplinary team (endocrinologist lead, nutritionist, exercise therapist, pharmacist, and health manager); intermediate risk (1-2 targets unmet), managed with endocrinologist guidance and structured health-manager follow-up; and low risk (all targets met), managed primarily in the community by a community physician with patient self-management. Digital enablement is embedded across tiers: High/intermediate-risk patients receive comprehensive digital intervention (smart monitoring devices, artificial intelligence decision support, and medication reminders), while low-risk patients use digital self-management (app logging and regular data synchronization). Follow-up intensity is tiered (high risk: Every 1-3 months; intermediate risk: Every 3-6 months; low risk: Every 6-12 months), with periodic re-evaluation of targets to maintain the regimen when controlled or to adjust/intensify strategies when targets are not met, forming a closed-loop management pathway. HbA1c: Glycated hemoglobin; FPG: Fasting plasma glucose; BP: Blood pressure; WC: Waist circumference; AI: Artificial intelligence; MDT: Multidisciplinary team.

Table 2 Digital health monitorable indicators.

Core monitoring indicators	Trigger thresholds (refer to 2024 CDS guidelines)	Intervention actions	Responsible subjects
HbA1c	≥ 7.0% or increase ≥ 0.5% within 3 months	(1) Adjust hypoglycemic regimen (prioritize GLP-1RA); (2) Strengthen diet/exercise intervention; and (3) Increase monitoring frequency to ≥ 3 times/week	Endocrinologist + dietitian
Systolic blood pressure/diastolic blood pressure	≥ 130/80 mmHg	(1) Initiate or adjust antihypertensive drugs (synergistic with hypoglycemic drugs); (2) Restrict sodium intake (< 5 g/day); and (3) Recommend aerobic exercise	Physician + health management nurse
LDL-C	High risk ≥ 1.8 mmol/L; medium-low risk ≥ 2.6 mmol/L	(1) Initiate statins; (2) Adjust diet structure (reduce saturated fat); and (3) Monitor liver function	Physician + dietitian
Uric acid	Male ≥ 420 μmol/L; female ≥ 360 μmol/L	(1) Restrict high-purine foods in diet; (2) Increase water intake (≥ 2000 mL/day); and (3) Initiate uric acid-lowering drugs if necessary	Physician + nurse
Body weight/BMI	BMI ≥ 28 kg/m² or increase ≥ 5% within 3 months	(1) Use weight loss-oriented hypoglycemic drugs (GLP-1RA); (2) Formulate individualized exercise prescription (150 minutes moderate-intensity exercise/week); and (3) Diet calorie control	Physician + exercise therapist + dietitian
Waist circumference	Male ≥ 90 cm; female ≥ 85 cm	(1) Core strength training; and (2) Reduce risk factors related to abdominal obesity (sedentary/high-sugar diet)	Exercise therapist + dietitian
Urinary microalbumin/creatinine ratio	≥ 30 mg/g	(1) Prefer SGLT2i/GLP-1RA; (2) Control blood pressure < 130/80 mmHg; and (3) Recheck every 3 months	Physician + nurse

Trigger thresholds strictly follow the requirements of the “China Guidelines for the Prevention and Treatment of Diabetes (2024 Edition)” and the Healthy China Action Plan. The responsible subjects clarify the division of labor within the diabetes center to ensure the intervention closed loop. HbA1c: Glycated hemoglobin; GLP-1RA: Glucagon-like peptide-1 receptor agonist; LDL-C: Low-density lipoprotein cholesterol; BMI: Body mass index; SGLT2i: Sodium-glucose cotransporter-2 inhibitor; CDS: Chinese Diabetes Society.

Table 1 (evidence map) collates included studies with risk of bias assessed per GRADE criteria. Overall, formal investigations of AI-enabled approaches in metabolic syndrome and multimorbidity management have converged on an increasingly coherent technology-enabled pathway that integrates early monitoring, precision intervention, and collaborative management. Collectively, these studies provide a practical scientific and technological rationale for five-condition co-management and, through accumulating real-world implementation experience, support the feasibility of moving this model from a conceptual framework toward standardized clinical practice.

CHALLENGES AND FUTURE DIRECTIONS

Although AI-enabled approaches show substantial promise for advancing five-condition co-management, translation into routine care remains constrained by barriers spanning technical performance, data interoperability, and ethical/governance implementation. Addressing these challenges requires a systematic, end-to-end pathway that couples clinically reliable devices and standardized data pipelines with transparent governance, safety evaluation, and implementation science.

Technical bottlenecks: Dual constraints of device accuracy and data integration

A fundamental barrier is the measurement validity of consumer- and home-based monitoring devices. For several household devices, reported error rates can reach 15%-20%, which may be insufficient for treatment titration or other clinical decision-making[76-79]. These limitations are often attributable to the technical trade-offs of low-cost sensors and the absence of unified calibration, quality-control, and performance certification standards, undermining the credibility of key indicators such as blood pressure and glucose obtained outside clinical environments.

In parallel, data fragmentation remains pervasive. “Data islands” across institutions are reinforced by heterogeneity in hospital information systems (e.g., HIS/LIS), inconsistent interface standards, incompatible data formats, and institutional barriers around data governance and asset protection, collectively impeding cross-institution and cross-regional integration of five-condition–relevant data streams[65,80-83]. Without reliable interoperability, it is difficult to construct a closed-loop workflow that connects home monitoring, community follow-up, and hospital-based escalation in a timely, standardized manner.

A feasible breakthrough strategy should therefore be two-pronged. First, device-side improvements are needed to accelerate the development and deployment of affordable, clinically reliable monitoring tools (including ultra-low-cost devices, e.g., around RMB 100, where appropriate), supported by technological innovations such as microfluidic chips and optoelectronic sensors, and anchored by a national-level performance certification and calibration system. Second, system-side reforms should prioritize a unified multi-source data integration standard, covering data schemas, interface protocols, terminology mappings, and data-sharing specifications, to enable end-to-end interoperability. Building a full-chain data network spanning “home-community-hospital” would provide the infrastructural basis for scalable, real-world five-condition co-management.

Ethical dilemmas: Balancing data sharing and privacy protection

AI-enabled five-condition co-management depends on access to large-scale, high-quality data for model development and validation. However, an inherent tension exists between the need for data sharing and the imperative to protect patient privacy. Conventional centralized data storage and aggregation can increase exposure to security breaches and misuse, which may erode patient trust and reduce willingness to participate in data sharing. In parallel, healthcare institutions may be hesitant to open or exchange data due to compliance uncertainty and perceived legal liabilities[84-87].

Federated learning offers a potentially practical pathway to mitigate this dilemma[88-91]. Under a decentralized architecture, models are trained locally within participating institutions, and only model updates (e.g., parameters or gradients) are exchanged, rather than raw patient-level data, enabling a privacy-preserving workflow often summarized as “data remain local while models move”. A multicenter chronic disease management study conducted by the Medical School of Tsinghua University reported that a federated learning-based AI model achieved an AUC of 0.89 while maintaining data privacy protections, with performance comparable to centralized training[92]. These findings support the feasibility of federated learning for privacy-preserving modeling in five-condition co-management scenarios.

Going forward, broader implementation will require strengthening both technical and governance safeguards. Key priorities include secure and encrypted transmission of model updates, node qualification and access control, and traceability/audit mechanisms for model training and outputs. In parallel, a clearer legal and regulatory framework is needed to define data rights, responsibilities, accountability, and permissible data-use boundaries across stakeholders to enable responsible data sharing at scale.

Implementation barriers: Structural mismatches between primary-care capacity and resource allocation

Primary healthcare institutions constitute the main delivery setting for five-condition co-management, yet persistent shortages of qualified personnel and uneven resource distribution can constrain the real-world deployment and clinical translation of AI-enabled tools. Available evidence indicates a substantial gap in the supply of general practitioners within China’s primary care system[93,94]. Although the number of primary health personnel increased to 5.26 million from 2020 to 2024, the proportion of “compound” professionals capable of managing multimorbidity (“one patient with multiple diseases”) remains insufficient, and only about 30% of community health service centers are staffed with dedicated chronic disease management teams[95].

These constraints create two practical bottlenecks. First, frontline staff may lack the training and time required to operate AI systems and to interpret multidimensional reports generated for comprehensive five-condition assessment. Second, some facilities remain under-resourced in basic monitoring equipment and digital infrastructure, limiting the feasibility of implementing technology-enabled workflows even when decision-support tools are available.

Overcoming these implementation barriers requires a coordinated strategy across policy, technology, and service delivery. At the policy level, payment reform is essential, particularly the establishment of medical insurance mechanisms that recognize and reimburse AI-assisted monitoring, remote follow-up, and teleconsultation as part of chronic disease management. At the technology level, tools should be designed for primary-care realities, including age-friendly interfaces and simplified operational pathways that translate complex outputs into intuitive, visualized reports. At the service level, workflows should be redesigned to support a standardized, full-cycle closed loop spanning screening, assessment, intervention, and follow-up.

Key elements of a feasible breakthrough pathway include: (1) Policy-driven linkage with medical insurance, incorporating AI-assisted monitoring and remote services into reimbursement catalogues; (2) Technology optimization, developing age-friendly and streamlined AI tools that convert complex risk reports into visual, actionable formats; and (3) Service reengineering, building a collaborative care team model of “specialist-general practitioner-health manager”, supported by standardized processes to improve the efficiency and continuity of primary-care delivery.

China-specific data governance system: Ensuring compliant implementation

The real-world deployment of digital health technologies for five-condition co-management must align with China’s regulatory requirements and the operational realities of the healthcare system. A pragmatic approach is to establish a three-dimensional governance model encompassing legal compliance, technical safeguards, and standards-based interoperability:

Legal basis: Ground governance in China’s data-related regulatory framework, using the Data Security Law of the People’s Republic of China as a key reference. Apply classified and tiered management for medical data; treat high-risk items such as glycemic data and complication-related records as sensitive health information, and enforce full lifecycle protection across collection, storage, processing, sharing, and disposal.

Technical support: Deploy privacy-enhancing technologies (e.g., differential privacy, encryption mechanisms such as fully homomorphic encryption where feasible) to reduce leakage risk. In parallel, use distributed/federated learning architectures to enable cross-institutional collaboration while minimizing raw data movement, thereby mitigating “data island” constraints without indiscriminate data pooling.

Standard specifications: Promote interoperability through standardized data schemas and interface protocols to support structured integration of heterogeneous sources (electronic medical records, laboratory systems, wearable devices, and patient-reported data), consistent with broader national health informatization initiatives and the Healthy China strategy. In addition to the governance framework, compliant implementation requires clearly defined accountability across stakeholders to ensure clinical safety and trustworthy deployment.

Medical institutions: As primary stewards of clinical data governance, institutions should ensure the authenticity, completeness, and standardization of data capture and incorporate data quality into internal management and evaluation processes. A dynamic informed-consent mechanism should be established to strengthen patient agency and clarify data-use boundaries.

Digital platforms: Platforms should assume responsibility for algorithmic security and clinical robustness, including fairness, interpretability, and bias mitigation. Where applicable, systems should undergo relevant ethical review and governance procedures, and provide transparent disclosure of decision logic, uncertainty, and risk alerts to support safe clinical use.

Device manufacturers: Manufacturers should ensure medical-grade measurement performance and transmission security for wearable and home-monitoring devices. For clinically used devices (e.g., cuffless/wearable blood pressure monitors), compliance with applicable medical device certification and quality management requirements is essential, alongside secure, real-time data transmission suitable for clinical workflows.

Cross-border data flows require particular caution. For international multicenter collaborations, data transfer should follow a risk-based and compliance-first principle, including appropriate de-identification, security assessments, and regulatory procedures under China’s cross-border data governance requirements, while also considering interoperability and compliance expectations in partner jurisdictions (e.g., potential alignment needs when collaborating with European Union institutions under General Data Protection Regulation-related constraints).

Overall, the challenges facing AI-enabled five-condition co-management are systemic, spanning technology, ethics/governance, and service delivery. Progress will depend on coordinated advances in device reliability, privacy-preserving data collaboration, and workflow redesign. While existing studies provide feasible directions across these domains, future work should prioritize multicenter clinical validation and implementation evaluation within five-condition co-management pathways to ensure that technical capability translates into measurable, real-world clinical benefit.

DISCUSSION

The “five-condition co-management” framework presented in this review should be understood as a systems-level correction to the prevailing logic of chronic disease care, rather than a simple “add two more conditions” expansion. Single-disease pathways and specialty silos may be workable when comorbidity is occasional; they become increasingly fragile when multimorbidity is routine. In this setting, the central challenge is no longer the formulation of separate targets for five indicators, but the capacity of clinical systems to continuously coordinate risk assessment, intervention, and follow-up across hypertension, dyslipidemia, hyperglycemia/diabetes, hyperuricemia, and overweight/obesity, without reducing care into disconnected checklists. Whether the model becomes scalable practice depends on three tightly linked dimensions: Decision-grade data and technology, feasibility within primary care, and a China-adapted integration of traditional Chinese Medicine (TCM) and Western medicine that remains evidence-aligned and workflow-compatible.

A persistent obstacle in routine care is the fragmentation of patient information. Even within a single case, relevant signals are typically scattered across home blood pressure logs, intermittent glucose records, laboratory results, imaging reports, and medication histories. When these streams remain disconnected, clinical decisions are often made on partial snapshots, sometimes outdated, sometimes incomplete, while cardiometabolic risk continues to evolve between visits. Five-condition co-management therefore requires a different operating logic in which multi-source data integration translates dispersed measurements into a coherent, longitudinal risk profile. Wearable monitoring (e.g., ambulatory blood pressure and glucose trajectories) becomes clinically meaningful in this context, particularly when linked to organ-specific risk signals captured through imaging and structured assessment. Deep-learning-assisted diabetic kidney disease pathways, for example, suggest that retinal imaging may contain kidney-risk-relevant information, creating a realistic opportunity for “single-encounter imaging with multi-disease screening”[92,96-99]. Experience from real-world digital programs further indicates that the decisive step is not simply increasing data volume, but improving synthesis: Once physiological trajectories, imaging phenotypes, and clinical context are fused, AI models can surface comorbidity patterns with direct clinical utility, for instance, identifying profiles in which hyperglycemia, hyperuricemia, and obesity jointly signal accelerated renal risk, thereby supporting earlier warning and more precise adjustment. However, such AI models are not without limitations: Algorithm bias may arise from unrepresentative training data (e.g., underinclusion of elderly or rural populations), leading to suboptimal performance in vulnerable groups; real-world implementation has reported failures due to poor workflow integration and low health literacy, while long-term cost-effectiveness remains unproven given the high upfront investment in digital infrastructure. In this sense, integration has value only when it improves decision timing and decision quality, pushing management upstream.

Translation also depends on how well the framework performs outside tertiary centers. Programs that look effective in well-resourced hospitals may underperform in community settings, often because of implementation mismatch: Tools designed for stable connectivity, ample staff time, and standardized devices are introduced into environments where those assumptions do not hold. Given the vast regional disparities in economic development and medical resource allocation across China, cost-effectiveness has become a critical practical constraint for the primary care implementation of AI-enabled five-condition co-management. In resource-constrained rural and remote areas, the high economic threshold of deploying full-set AI decision support systems and commercial wearable monitoring devices is a realistic barrier that cannot be ignored, and blind promotion of high-cost technical solutions will lead to low utilization and resource waste due to poor affordability. For five-condition co-management to deliver population-level benefit, it must function in the settings where most patients receive care, primary care, rural clinics, and resource-constrained communities. This is where “appropriate technology” matters more than technical sophistication. Low-cost alternative solutions are thus essential: For example, using low-priced universal monitoring devices (≤ 100 RMB) instead of high-end wearables, simplifying AI algorithms into offline usable mobile terminal programs to reduce network and hardware costs, and leveraging community health management teams to carry out group intervention and health education to lower the per capita management cost. Noninvasive, low-cost monitoring tools can reduce barriers to routine data capture and patient self-management; ultra-low-cost portable devices (e.g., around RMB 100) may lower equipment thresholds for primary institutions and decrease travel burden for patients who otherwise require repeated facility-based checks[100-102]. When home-based monitoring (including blood pressure and, where applicable, uric acid-related indicators) can be synchronized to community clinicians, follow-up becomes more continuous and less episodic, helping patients remain in stable management rather than cycling between neglect and acute escalation[103]. Meanwhile, the medical insurance payment system plays a pivotal supporting role in improving cost-effectiveness: Incorporating low-cost monitoring services, AI-assisted clinical decision-making, and remote follow-up into the medical insurance reimbursement catalogue can reduce the out-of-pocket expenses of patients and primary medical institutions, effectively lowering the economic threshold for implementation. Policy-driven insurance linkage is thus a key guarantee for balancing technical application and cost control in primary care. The reliability of this pathway ultimately rests on whether device performance, calibration, and clinical workflows are aligned through standards and governance, rather than on devices alone.

A China-adapted pathway also benefits from structured integration of TCM and Western medicine, and this integration can be effectively empowered and standardized by digital health and AI technologies, forming a synergistic model with modern biomedical management. TCM has long been used in chronic disease care and may provide complementary value in symptom control, constitution-based regulation, and long-term conditioning[104-109]. In many outpatient settings, multimodal care already occurs informally; the priority is to integrate it in a way that is transparent, standardized, and clinically accountable. AI integrates TCM syndrome differentiation/tongue-pulse signs with biomedical metabolic data to standardize syndrome differentiation, embedding TCM principles, particularly syndrome differentiation, into AI-assisted decision support can broaden intervention options while maintaining a structured decision pathway[110,111]. For patients characterized by a “phlegm-dampness constitution”, for example, an AI system could recommend adjunctive acupuncture or dietary therapy alongside guideline-concordant pharmacotherapy, producing a plan tailored to biomedical targets and constitution features[112-114]. Meanwhile, digital tools (e.g., dedicated TCM intervention record modules in the five-condition co-management platform) can quantify and longitudinally track the clinical effects of TCM interventions (e.g., herbal medicine, acupuncture, constitution-regulated diet), and synchronize these data with the dynamic changes of metabolic indicators, enabling objective evaluation of TCM intervention efficacy and forming a closed-loop of integrated TCM-Western medicine management based on digital tracking and iterative optimization. Such coordination preserves the precision of modern medicine for indicator control while allowing TCM-based regulation to address overall metabolic state and, in some cases, reduce medication-related adverse effects. A standardized and auditable TCM-Western medicine pathway may also improve acceptability and patient-centeredness, strengthening engagement over long follow-up intervals[115-117].

Scaling these directions requires incentives and infrastructure to move in parallel. Healthy China 2030 emphasizes healthcare big data and unified health information platforms; when translated into interoperable standards and usable pipelines, these initiatives can accelerate data flow and enable cross-setting collaboration for five-condition co-management. Market innovation is equally important, but it must begin from primary-care realities: Simplified products that older adults can operate, tools that tolerate rural connectivity constraints, and interfaces that convert complex outputs into clear, actionable steps for community clinicians. When policy guidance and product development converge, service delivery becomes more plausible, technically feasible, economically sustainable, and socially acceptable, allowing the model to extend beyond pilots and embed into routine workflows.

Notably, the current field still lacks large-scale, high-quality multicenter randomized clinical trials based on composite hard endpoints such as cardiovascular event mortality (rather than single glycated hemoglobin), which is the core direction for validating the clinical effectiveness of the five-condition co-management model in the future.

Ultimately, five-condition co-management will be evaluated by its ability to sustain a dynamic balance between technology empowerment and system adaptation. Algorithms and devices are necessary but insufficient; workflow redesign, payment policy, workforce capacity, and patient health literacy determine whether technical promise becomes measurable benefit. When these elements co-evolve, so that data are trustworthy, outputs are actionable, and follow-up is continuous, five-condition co-management can meaningfully enhance the quality and efficiency of chronic disease prevention and control and provide a scalable pathway toward proactive, integrated cardiometabolic health management in China.

CONCLUSION

Diabetes-centered five-condition co-management, integrating hypertension, hyperglycemia/diabetes, dyslipidemia, hyperuricemia, and excess body weight/obesity, provides a practical framework for confronting China’s rapidly expanding burden of cardiometabolic multimorbidity. By reframing management from parallel, single-disease targets to risk-cluster-oriented care, this approach is better aligned with shared pathophysiology and the synergistic harms generated when metabolic abnormalities co-occur, and it supports a shift from episodic, reactive treatment toward proactive, longitudinal risk reduction.

Digital health and AI operationalize this framework via multimodal monitoring, risk stratification, clinical decision support, and cross-institution care coordination, enabling closed-loop iterative optimization of personalized interventions. Crucially, the clinical effectiveness of five-condition co-management and its AI-enabled systems remains unvalidated; current support is limited to indirect evidence, and its theoretical plausibility has not been confirmed by targeted clinical research. Key uncertainties persist, including unaddressed algorithm bias, inconsistent real-world adherence, and lack of data on long-term clinical endpoints (e.g., cardiovascular mortality) and cost-effectiveness across diverse healthcare settings.

Looking ahead, multicenter real-world studies that incorporate clinically meaningful composite hard endpoints (e.g., cardiovascular event mortality) beyond single glycemic metrics such as glycated hemoglobin alongside implementation outcomes (e.g., adherence, equity, feasibility, and cost-effectiveness) are needed to establish effectiveness and generalizability. Cross-country comparisons with mature digital chronic disease management systems can further refine implementation strategies for diverse healthcare contexts. If these requirements are met, AI-enabled diabetes-centered five-condition co-management could evolve into a scalable China-adapted chronic disease management model, offering transferable lessons for health systems facing similar cardiometabolic risk clustering globally.