Artificial Intelligence-Based Full-Cycle Pathway-Based Chronic Disease Management System and Method

Abstract

This invention discloses a full-cycle path-based chronic disease management system and method based on artificial intelligence, belonging to the field of smart healthcare technology. It includes: proactively identifying, intervening in, and managing target populations through a proactive diagnosis and treatment module based on the results of an AI-powered chronic disease risk prediction model and prescription generation model; providing refined follow-up, prescription, and screening services tailored to the health status and risk characteristics of different individuals through a personalized diagnosis and treatment module; and providing semantic search, indicator commonality identification, and multi-source data integration capabilities for clinical data through an intelligent data governance and decision support module. This invention constructs a new smart healthcare model for proactive, personalized, and intelligent chronic disease management, assisting clinical decision-making, improving management efficiency, contributing to the improvement of chronic disease management levels, promoting the rational flow of medical resources, and driving the development of chronic disease prevention, management, and referral towards full life-cycle management.

Description

Technical Field

[0001] This invention belongs to the field of smart healthcare technology, specifically relating to a full-cycle path-based chronic disease management system and method based on artificial intelligence. Background Technology

[0002] Currently, the treatment and management of chronic diseases has become a key focus of global healthcare systems. With socio-economic development and the improvement of medical informatization, chronic disease management models are gradually shifting from a "treatment-centered" to a "health-centered" approach, emphasizing health maintenance throughout the entire life cycle. However, current chronic disease management practices still face three prominent problems that severely hinder the comprehensive management of chronic diseases throughout their entire life cycle.

[0003] I. Reliance on individual physicians' abilities, with insufficient capacity for standardized and personalized support:

[0004] In the existing healthcare model, disease diagnosis and treatment decision-making heavily relies on the individual capabilities of physicians. While evidence-based clinical guidelines provide a scientific basis for clinical practice, their actual effectiveness is largely constrained by the professional competence and execution ability of individual physicians. Specifically, there are significant differences in physicians' depth of understanding of the guidelines, their willingness to apply them, and their ability to integrate multi-source health information (such as patient history data and current disease information). These individual differences in execution, especially in chronic disease management, directly lead to inconsistent management outcomes, making it difficult to simultaneously address the standardized and personalized needs of treatment. The main problems with the existing model can be summarized as follows:

[0005] Over-reliance on individual physician capabilities: Core aspects of chronic disease management (screening, diagnosis, and treatment) are severely limited by the individual experience and judgment of attending physicians. Different physicians show significant differences in their understanding and adherence to the same guidelines, as well as their attention to individual patient circumstances (such as adequacy of communication and interpretation of past medical history). Under high workloads, physicians struggle to systematically review complete patient histories and accurately assess potential risks, leading to a significant increase in the risk of misdiagnosis and missed diagnosis.

[0006] Low level of homogeneity in diagnosis and treatment: Due to the uneven distribution of medical resources across regions, some primary healthcare physicians struggle to master and effectively apply the latest clinical guidelines. This results in a lack of standardized and pathway-based systematic support for the management of patients with chronic diseases, leading to significant disparities in the quality of diagnosis and treatment among medical institutions at different levels and in different regions.

[0007] Inefficient allocation of medical resources: Inconsistencies or biases in individual physician judgments may lead to inaccurate assessments of the severity of a patient's condition. The consequences are that patients with mild symptoms may receive unnecessary overtreatment, while intervention for severely ill patients may be delayed. This situation not only affects patient prognosis but also results in the underutilization of valuable medical resources.

[0008] II. Primarily reactive, lacking proactive management mechanisms:

[0009] The current chronic disease management system is still characterized by "post-event management," passively responding to the "onset-progression-serious adverse events" of diseases, lacking proactive and forward-looking management interventions. This is mainly manifested in the following ways:

[0010] Delayed medical attention: Patients usually seek medical attention for the first time only after symptoms become obvious or complications occur, thus missing the window of early intervention;

[0011] Delayed referrals: Due to limited clinical experience and disease knowledge, primary care physicians often only recommend referrals after treatment has proven ineffective or the condition has worsened, leading to delays in diagnosis and treatment. Patients at higher-level hospitals then have to receive more complex and costly treatments, further straining healthcare resources.

[0012] Delayed prescription / treatment response: The formulation and adjustment of prescriptions and treatment plans rely heavily on periodic retrospective assessments, making it difficult to respond promptly to changes in the patient's condition. This delayed response may lead to inadequate intervention, increase the risk of complications, and negatively impact patient prognosis.

[0013] III. Weakness in intelligent collaboration and data closed-loop capabilities:

[0014] Current medical information systems in chronic disease management primarily serve the function of information recording, lacking support for proactive and personalized services throughout the entire process.

[0015] Severe information fragmentation: There is a lack of effective data sharing mechanisms between clinical, laboratory, and public health systems; inconsistent data coding standards across different institutions make it difficult to create complete health records for patients. This not only leads to duplicate testing and wasted resources but also limits comprehensive judgment on the progression of the disease.

[0016] Duplicate data entry and redundant processes: Due to the lack of data interoperability between different platforms, doctors need to repeatedly enter follow-up, referral, prescription and other data on multiple platforms, which increases their workload and consumes clinical energy.

[0017] The system only supports a limited range of functions: existing medical information systems are mainly focused on data collection and storage, making it difficult to provide dynamic guidance and services tailored to the specific conditions of patients, and lacking proactive functions such as clinical decision support, risk warning, and follow-up plan development.

[0018] In conclusion, the current chronic disease management system has significant shortcomings in terms of "proactiveness," "personalization," and "intelligence," and needs further optimization and improvement. Summary of the Invention

[0019] In view of the above, the present invention aims to provide a full-cycle path-based chronic disease management system and method based on artificial intelligence. By integrating artificial intelligence technology with regional multi-source medical data, a full-cycle path-based management system covering the prevention of disease onset in high-risk groups and the management of disease progression in chronic disease groups is constructed, promoting three major transformations in chronic disease management: (1) transformation from passive disease response to closed-loop management with proactive early warning and intervention; (2) transformation from decision-making based on individual ability to intelligent-driven personalized path dynamic collaboration; and (3) transformation from rule-based logical judgment to intelligent support system with deep understanding of clinical logic.

[0020] Through the aforementioned transformation, the core objectives of this invention are: to enhance management initiative by establishing a proactive intervention mechanism based on risk prediction to reduce health losses caused by delayed medical visits / referrals; to strengthen the synergy between standardization and personalization by improving the homogeneity of diagnosis and treatment through path-based navigation, while dynamically adjusting plans based on individual health trajectories; to overcome the bottleneck of intelligent support by constructing an AI-assisted decision-making system that deeply understands clinical logic, reducing the cognitive load on doctors and optimizing resource allocation efficiency; and ultimately to achieve a systematic improvement in the effectiveness of chronic disease management, significantly reducing the incidence, disability, and mortality rates of chronic diseases, and increasing the treatment achievement rate and health management coverage.

[0021] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0022] In a first aspect, embodiments of the present invention provide an artificial intelligence-based full-cycle path-based chronic disease management system, comprising:

[0023] The proactive diagnosis and treatment module includes: a proactive visit and follow-up unit that proactively pushes visit reminders, personalized management paths, and follow-up plans based on a chronic disease risk prediction model; a proactive health prescription generation unit that automatically generates structured health prescriptions by combining the chronic disease risk prediction model and the health prescription generation model; and a proactive referral unit that automatically generates referral suggestions based on the chronic disease risk prediction model and tracks referral status.

[0024] Personalized diagnosis and treatment module: The personalized medical visit and follow-up unit is used to automatically generate and present customized follow-up plans and content based on patient data; the personalized health prescription generation unit is used to automatically generate customized health prescriptions through the health prescription generation model; the personalized disease screening unit is used to conduct intelligent screening based on individual characteristics for residents who have not yet been diagnosed with specific chronic diseases.

[0025] Intelligent data governance and decision support module: The indicator intelligent correlation identification unit is used to automatically identify potential correlations of clinical indicators through artificial intelligence models and achieve automatic classification and synchronization; the natural language semantic retrieval unit is used to provide natural language query services with semantic understanding capabilities; the data closed-loop integration management unit is used to access and integrate multi-source heterogeneous medical systems within the region.

[0026] Preferably, the proactive medical visit follow-up unit is used to proactively push medical visit reminders, personalized management paths, and follow-up plans based on a chronic disease risk prediction model, including:

[0027] The system's background periodically runs a chronic disease risk prediction model to assess the risk of disease onset in high-risk groups and the risk of disease progression in chronic disease groups. It automatically filters out individuals with risk levels above the set thresholds for disease onset and progression, generating high-risk disease onset lists and high-risk progression lists respectively. Based on these lists, the system pushes medical appointment reminders to the contracted chronic disease doctors.

[0028] When a patient arrives at a medical institution for treatment, the system automatically matches the corresponding management path and follow-up plan based on their risk label and provides real-time guidance on the doctor's interface. The management path is displayed in a list format, recording the diagnosis and treatment process and task execution status throughout the follow-up process. The follow-up plan is based on the patient's individual characteristics and disease stage, setting the frequency of visits, examination items, and the next follow-up date, and is dynamically updated.

[0029] After the consultation is completed, the system automatically synchronizes the patient's consultation and follow-up records, and supports one-click synchronization to the public health platform for higher-level supervision and quality evaluation. At the same time, it serves as training samples and reference data for the next cycle of chronic disease risk prediction model operation.

[0030] Preferably, the automated health prescription generation unit is used to automatically generate structured health prescriptions by combining a chronic disease risk prediction model and a health prescription generation model, including:

[0031] The chronic disease risk prediction model adopts a modular structure to support parallel prediction of multiple diseases. The multivariate time series prediction submodule based on the Transformer structure processes dense features, the large language model submodule processes sparse features, the filter submodule processes missing patterns and time interval matrices, and the static feature encoding submodule processes static features. The feature vectors output by each submodule are fused by the classification head module to output the risk of onset and progression of multiple chronic diseases, and are compared with the set risk threshold to form risk labels.

[0032] The health prescription generation model generates a health examination summary by summarizing abnormal physical examination results and chronic disease label information, and retrieves the patient's historical doctor advice. It constructs a prompt word containing the physical examination summary, risk label, and historical advice, and calls a large language model to generate a structured health prescription that includes disease risk assessment results, disease progression prediction results, and personalized lifestyle recommendations. After deduplication and compliance review, the prescription text is output.

[0033] Preferably, the proactive referral unit is used to automatically generate referral recommendations based on a chronic disease risk prediction model and track referral status, including:

[0034] Based on the chronic disease incidence risk score and chronic disease progression risk score generated by the chronic disease prediction model, and the referral rules set by the system, the system automatically generates a decision suggestion on whether to refer the patient to another hospital. The suggestion includes key referral information such as the reason for referral, the recommended referral institution, and the recommended referral department. The reason for referral is automatically generated by an AI referral engine based on a large language model. The model is trained by introducing chronic disease-related referral decision samples and combines a thought chain reasoning mechanism to achieve step-by-step decision-making ability. At the same time, a retrieval-enhanced generation technology is used to perform similarity retrieval between the patient assessment data text and the referral rule vector database during the model input stage. The matched referral rules are dynamically injected into the model input as external knowledge context to improve the rationality and standardization of the generated results.

[0035] When a doctor accepts a recommendation, the system will automatically embed the referral suggestion into the health prescription, displaying information such as the reason for the referral, the recommended institution and department; when a doctor refuses a recommendation, he must fill in the reason for refusal. The reason for refusal will be used as an important training sample for model iteration and optimization and sent back to the fine-tuning database to further enhance the model's judgment accuracy.

[0036] In the referral tracking, the system automatically tracks whether patients have completed the recommended referral process and assigns a referral tag, forming a closed-loop management of referral results. At the same time, the system connects to the regional health data platform based on the patient's ID number to monitor changes in the patient's medical records at various levels of medical institutions in real time, thereby judging their referral behavior. In addition, the referral tag will be fed back to the system for subsequent referral effectiveness evaluation, medical behavior analysis, and supplementation of model training data.

[0037] Preferably, the personalized medical follow-up unit is used to automatically generate and present customized follow-up plans and content based on patient data, including:

[0038] The system first automatically identifies or allows doctors to manually select one or more chronic diseases currently suffered by the patient based on the target chronic disease types managed by the system. For each identified or selected chronic disease, the system automatically matches and displays its related core symptoms, common complications, and follow-up content of comorbidities based on the chronic disease complication database.

[0039] Based on individualized risk predictions of disease / progression, past health data, and medical records, personalized test and follow-up plan adjustment recommendations are generated, including:

[0040] (1) Based on the patient's current chronic disease type and the predicted risk of onset / progression, and on the basis of a standardized follow-up pathway generated based on clinical guidelines and local public health service requirements, the follow-up frequency and follow-up items are dynamically adjusted to form a personalized follow-up plan;

[0041] (2) Integrate health data platform information to provide patients with a historical list of past test and examination items across medical institutions, and support hierarchical viewing by time range and specific items to avoid duplicate examinations;

[0042] (3) Based on the personalized follow-up plan and the patient's recent test results, dynamically generate and prompt a list of personalized test items to be completed in the future cycle; for items that have been completed recently, the system automatically identifies and marks the status, and only prompts items to be completed.

[0043] Preferably, the personalized health prescription generation unit is used to automatically generate customized health prescriptions through a health prescription generation model, including:

[0044] Customized health prescriptions include visualization of key chronic disease indicators, prediction results of chronic disease onset / progression risks, disease summaries, and lifestyle-based treatment recommendations, among which:

[0045] (1) Visualization of key indicators for chronic diseases: Based on patients’ cross-institutional health data, identify chronic diseases that require special attention, dynamically extract and visualize the trends of their key physiological indicators; key indicators are determined based on the actual disease status and risk prediction results of patients, and outliers are highlighted.

[0046] (2) Chronic disease incidence / progression risk prediction section: Integrates the results of chronic disease risk prediction models to generate a graded incidence / progression risk assessment report; the risk level determination threshold is configurable;

[0047] (3) Disease summary section: Input patient's medical information and key indicator data, output structured disease summary, covering the test frequency, historical abnormalities, current status and future monitoring focus of important / abnormal indicators;

[0048] (4) Lifestyle-based treatment recommendations: Input multidimensional patient data to generate personalized non-drug intervention plans, including lifestyle recommendations, key points for indicator observation and precautions; prescription weights and levels of detail are dynamically adjusted according to the patient's condition and risk level;

[0049] Meanwhile, the personalized health prescription generation unit supports a mechanism for chronic disease physicians to review and confirm the generated prescriptions. If the content is reasonable, the prescription will be approved by signing it. If the prescription is rejected, a clear reason for rejection must be provided. The system will automatically record the review behavior and incorporate the reason for rejection into the model optimization closed loop for subsequent model training and inference logic correction, thereby continuously improving the clinical adaptability of the model-generated content.

[0050] Preferably, after identifying potentially high-risk individuals through intelligent screening, the personalized disease screening unit also includes:

[0051] For high-risk individuals newly identified during medical visits but who have not yet signed up for chronic disease management services, we will push signing suggestions to doctors, recommend signing up for corresponding chronic disease management services, and authorize primary care physicians to dynamically manage them, thus building a closed-loop service path of screening and follow-up.

[0052] For high-risk individuals who have signed up, a customized disease screening plan is automatically generated, which includes the specific diseases to be screened, screening items, recommended screening institutions, and suggested screening frequency. All screening results will be automatically archived into the resident's electronic health record after completion and will serve as a data source for subsequent model iteration training.

[0053] Preferably, the intelligent correlation identification unit is used to automatically identify potential correlations among clinical indicators through an artificial intelligence model and to achieve automatic classification and synchronization, including:

[0054] A medical knowledge graph is constructed, encompassing seven categories of nodes: diseases, symptoms, drugs, foods, examination items, and departments. These nodes are linked to various relationships, including department-department relationships, disease-avoided food relationships, disease-recommended food relationships, disease-general drug relationships, disease-popular drug relationships, disease-examination relationships, manufacturer-drug relationships, disease-symptom relationships, disease-concurrency relationships, and disease-department relationships. The system writes patients' historical structured data into the graph as event nodes or attributes. Combined with structured health data including physiological indicators, laboratory test results, and diagnostic results, the system uses rule-based reasoning and path query algorithms to dynamically determine potential concurrency relationships and causal connections between diseases. This drives the automatic selection logic of corresponding indicator items in the doctor's interface, returning information on selectable related diseases / indicators / complications and intelligently rendering them as selected.

[0055] Preferably, the natural language semantic retrieval unit is used to provide natural language query services with semantic understanding capabilities, including:

[0056] A natural language semantic retrieval unit supporting multi-turn conversational query mode is constructed based on the integration of large language models and artificial intelligence agent technology. The semantic retrieval process is completed autonomously by the large model, including call path judgment, resource scheduling, parameter organization and semantic generation. At the same time, it supports the fusion retrieval of structured and unstructured information. Doctors can set limited dimensions in the query, including specified time period, medical institution or indicator type, and the system will retrieve information efficiently and in a targeted manner within the limited scope.

[0057] Secondly, embodiments of the present invention also provide an artificial intelligence-based full-cycle path-based chronic disease management method, implemented using the aforementioned artificial intelligence-based full-cycle path-based chronic disease management system, comprising the following steps:

[0058] Based on the proactive diagnosis and treatment module, the proactive medical visit and follow-up unit proactively pushes medical visit reminders, personalized management paths, and follow-up plans based on the chronic disease risk prediction model; the proactive health prescription generation unit automatically generates structured health prescriptions by combining the chronic disease risk prediction model and the health prescription generation model; and the proactive referral unit automatically generates referral suggestions based on the chronic disease risk prediction model and tracks referral status.

[0059] Based on the personalized diagnosis and treatment module, the personalized medical follow-up unit automatically generates and presents customized follow-up plans and content based on patient data; the personalized health prescription generation unit automatically generates customized health prescriptions through a health prescription generation model; and the personalized disease screening unit conducts intelligent screening based on individual characteristics for residents who have not yet been diagnosed with specific chronic diseases.

[0060] Based on the intelligent data governance and decision support module, the intelligent correlation identification unit uses an artificial intelligence model to automatically identify potential correlations among clinical indicators and achieve automatic classification and synchronization; the natural language semantic retrieval unit provides natural language query services with semantic understanding capabilities; and the data closed-loop integration management unit accesses and integrates multi-source heterogeneous medical systems within the region.

[0061] Compared with the prior art, the beneficial effects of the present invention include at least the following:

[0062] (1) Breakthrough in full-cycle proactive management capabilities: Establish a closed-loop management system covering high-risk groups and people with chronic diseases, realize proactive management throughout the entire process of medical visits, follow-up visits, prescription generation, and referral coordination, and support the dual-layer prevention and control goal of "prevention before disease occurs and prevention of disease progression";

[0063] (2) Individualized and standardized collaborative optimization: Provides system support for personalized path collaboration in chronic disease management, achieves precise guidance in the consultation and follow-up, prescription generation and referral stages, and allows doctors to make the final decision, thereby improving the efficiency of diagnosis and treatment services;

[0064] (3) Upgrade of intelligent decision-making and data governance: realize the interconnection of regional medical data, realize the intelligent association of clinical follow-up indicators and natural language query, and reduce the clinical workload;

[0065] (4) Dual-effect improvement of clinical resource optimization and medical insurance cost control: Through intelligent analysis and accurate prediction, the allocation of clinical resources is optimized to ensure the efficient use of medical resources; at the same time, the diagnosis and treatment process is guided to promote the rational flow of medical insurance funds, avoid resource waste, effectively control the overall medical cost, and achieve dual optimization management of medical resources and medical insurance costs.

[0066] (5) Significant achievements in the homogenization of primary healthcare and prevention and control of chronic diseases: By combining standardized proactive management processes with personalized service plans, the homogenization level of primary healthcare services is improved, the incidence and mortality of chronic diseases are reduced, the quality of life of patients is significantly improved, and the average life expectancy is extended.

[0067] (6) Regionalized and systematic promotion and strengthening of public health capacity: Support the systematic and modular promotion and deployment on a regional basis, and improve the overall health level of the region by strengthening public health management capacity and primary medical service capacity, so as to provide a strong guarantee for building a healthy and harmonious social environment. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0069] Figure 1 This is a schematic diagram of the structure of the AI-based full-cycle path-based chronic disease management system provided in the embodiment;

[0070] Figure 2 This is an overview schematic diagram of the automated diagnosis and treatment module provided in the embodiment;

[0071] Figure 3 This is a system interface diagram of the automated medical follow-up unit provided in the embodiment;

[0072] Figure 4 This is a system interface diagram of the automated / personalized health prescription generation unit provided in the embodiment;

[0073] Figure 5 This is a system interface diagram of the automated referral unit provided in the embodiment;

[0074] Figure 6 This is an overview diagram of the personalized diagnosis and treatment module provided in the embodiment;

[0075] Figure 7 This is a system interface diagram of the follow-up content in the personalized medical follow-up unit provided in the embodiment;

[0076] Figure 8 This is a diagram of the test and examination checklist system interface in the personalized medical follow-up unit provided in the embodiment;

[0077] Figure 9 This is a system interface diagram of the personalized disease screening unit provided in the embodiment;

[0078] Figure 10This is an overview diagram of the intelligent data governance and decision support module provided in the embodiment;

[0079] Figure 11 This is a system interface diagram of the indicator intelligent association recognition unit provided in the embodiment;

[0080] Figure 12 This is a system interface diagram of the natural language semantic retrieval unit provided in the embodiment. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0082] The inventive concept of this invention is as follows: Addressing key issues in existing chronic disease management models, such as insufficient homogenization of diagnosis and treatment, reliance on individual physician capabilities, weak personalized support, and inadequate passive response and system intelligence support, this invention provides an artificial intelligence-based, full-cycle, path-based chronic disease management system and method. Based on AI-powered risk prediction and prescription generation technologies, it effectively manages the onset of chronic diseases in high-risk populations and the progression of chronic diseases in existing chronic disease populations. While improving the homogenization of clinical diagnosis and treatment, it possesses three core characteristics: proactive management, personalized path-based collaborative management adapted to individual health trajectories, and human-machine collaborative intelligent support with a deep understanding of clinical logic. Under the premise of reducing clinical burden, it fully leverages the role of primary healthcare institutions to construct a full-cycle proactive chronic disease management system covering the health-high-risk-onset-progression stages, achieving a closed-loop intelligent chronic disease management system from risk prediction, intervention suggestion generation, treatment behavior recommendation to data closed-loop governance.

[0083] Furthermore, the system provided in this embodiment of the invention is applicable to the continuous, dynamic, and precise comprehensive management of various chronic diseases on a resident-by-resident basis, including but not limited to major chronic disease types such as hypertension, diabetes, coronary heart disease, stroke, chronic obstructive pulmonary disease (COPD), and chronic kidney disease (CKD). It is primarily deployed in primary healthcare institutions such as community health service centers, community health service stations, and village clinics (hereinafter collectively referred to as the "doctor's end"). The service targets cover all residents within the region who are at high risk of developing chronic diseases or already suffering from chronic diseases and require screening, initial diagnosis, follow-up visits, and long-term follow-up management. Simultaneously, the system supports management services for chronic disease control statistics and performance evaluation functions for medical institutions and government departments, as well as mobile services for patients, realizing a multi-dimensional, interconnected chronic disease management architecture involving "medical care, management, and patient." The system provided in this embodiment of the invention is applicable to the management of chronic disease onset in high-risk populations and the management of chronic disease progression in individuals with chronic diseases.

[0084] Specifically, high-risk individuals refer to those who have not yet been diagnosed with a chronic disease in the electronic medical system, but whose risk score for a certain chronic disease exceeds a specific threshold (default value 0.8) after assessment by the AI ​​chronic disease risk prediction model embedded in the system. This threshold can be flexibly adjusted by the local health department based on the region's public health management capabilities. Although this population has not been diagnosed, they have a significant risk of developing the disease and should be included in this invention to receive proactive contract signing, follow-up intervention, and health education to achieve early identification, early intervention, and early management.

[0085] Chronic disease morbidity refers to a resident meeting the systematic diagnostic criteria or having a chronic disease diagnosis recorded in the electronic medical system. The systematic diagnostic criteria are based on clinical guidelines and public health management needs for the chronic disease, adapted to the application of this invention, and evaluated for feasibility by a team of clinical experts. For example, the systematic diagnostic criteria for diabetes are: the presence of typical symptoms of diabetes and a fasting blood glucose level ≥11.1 mmol / L, a random blood glucose level ≥11.1 mmol / L, or a 2-hour post-oral glucose tolerance test blood glucose level ≥11.1 mmol / L.

[0086] The chronic disease population refers to individuals who, from a certain point in time, have met the criteria for the onset of a specific chronic disease as defined by the system. These criteria are developed by a team of clinical experts, taking into account national clinical guidelines, public health management requirements, and the actual implementation capabilities at the grassroots level, and are tailored to specific application scenarios. All individuals meeting the diagnostic criteria will be included in the chronic disease population managed throughout the entire lifecycle of this invention. The system further supports a tiered management mechanism for chronic diseases based on clinical guidelines, allowing adjustments to the management level according to changes in the patient's condition. Simultaneously, the system automatically records and identifies chronic disease progression events, including the development of other chronic diseases (such as hypertension combined with diabetes), progression to a higher disease stage, or the occurrence of adverse clinical events such as new hospitalizations or death, promptly triggering high-risk warnings and referral recommendations, supporting closed-loop dynamic management.

[0087] Chronic disease progression refers to the development of other chronic diseases in a patient with a chronic disease, the progression of a chronic disease to a new systemic clinical classification, or adverse events such as new hospitalizations or death. For example, a patient with stage I hypertension may develop diabetes, progress to stage II diabetes, or experience new hospitalizations or death. Systemic clinical classifications are based on clinical guidelines for the chronic disease and public health management needs, adapted to the application of this invention, and are assessed by a team of clinical experts to determine feasible clinical classifications. For example, the systemic clinical classifications for chronic kidney disease are shown in Table 1.

[0088] Table 1 Systemic Clinical Classification of Chronic Kidney Disease

[0089]

[0090] The system of this invention manages a population (including high-risk groups and people with chronic diseases) and establishes a one-to-one contract with a primary care physician (hereinafter referred to as a chronic disease physician). The chronic disease physician is the patient's primary responsible physician and uses the system of this invention to manage the patient.

[0091] like Figure 1 As shown in the embodiment, this invention provides a full-cycle, pathway-based chronic disease management system based on artificial intelligence, including a proactive diagnosis and treatment module, a personalized diagnosis and treatment module, and an intelligent data governance and decision support module. This system integrates multi-source regional health data and, through artificial intelligence modeling and modular services, forms a complete system encompassing proactive identification, risk prediction, follow-up intervention, prescription recommendations, intelligent referral, and closed-loop data governance.

[0092] 1. Automated diagnosis and treatment module.

[0093] like Figure 2 As shown, this module is used to proactively identify, intervene in, and manage target populations based on the results of artificial intelligence chronic disease risk prediction models and prescription generation models, enabling proactive follow-up visits, health prescription generation, and referrals. Specifically, it includes:

[0094] (1) The proactive medical visit and follow-up unit is used to scan and predict the risks of contracted high-risk groups and chronic disease groups in real time based on the AI ​​chronic disease risk prediction model. It automatically identifies high-risk individuals with the risk of disease onset or disease progression and pushes medical visit reminders, personalized management paths and follow-up plans to their corresponding chronic disease contracted doctors, so as to realize an intelligent and closed-loop follow-up management mechanism.

[0095] Specifically, the system's background runs an AI-powered chronic disease risk prediction model periodically (every 10 minutes) to assess the risk of disease onset in high-risk individuals and the risk of disease progression (adverse events such as graded progression, complications, hospitalization, or death) in chronic disease patients. The system automatically filters individuals with risk thresholds exceeding these thresholds (both set to a default value of 0.8), generating "High-Risk Onset List" and "High-Risk Progression List," respectively. These risk thresholds can be flexibly adjusted by local health authorities based on local medical capabilities. Based on these lists, the system sends follow-up reminders to contracted chronic disease physicians through various channels (including generating tags under the patient's name in the system's physician list and sending SMS notifications). For example, tags such as "High Risk of Hypertension Onset" and "High Risk of Hypertension Progression" will be generated under the patient's name in the patient list. Following the system's prompts, chronic disease physicians can contact patients via phone, SMS, or other means to arrange follow-up appointments at medical institutions.

[0096] Once a patient arrives at a medical institution for treatment, the system will automatically match the appropriate management path and follow-up plan based on their risk tags, and provide real-time guidance on the doctor's interface. Specifically, for example... Figure 3 As shown, the management path is displayed in a list format, recording the diagnosis and treatment steps and task execution status throughout the follow-up process. The system marks the completion status of completed steps and specifies the content to be executed. For example, on the doctor's side of the system, it shows that the current patient has completed the pre-visit steps of "collecting vital signs," "inquiring about the condition," and "lifestyle habits," the in-visit steps of "pre-visit information," "medication status," and "health assessment," and the post-visit steps of "confirming and submitting to community services," but has not yet completed the in-visit steps of "health prescription and annual assessment." The follow-up plan is based on the patient's individual characteristics and disease stage, setting the frequency of visits, examination items, and the next follow-up date, and is dynamically updated. This process ensures that doctors have clear execution guidance and visual support for the path during clinical practice, improving the standardization and operational efficiency of chronic disease management.

[0097] After the consultation is completed, the system will automatically synchronize the patient's consultation and follow-up records to the data platform, and support one-click synchronization to the public health platform for government supervision and quality evaluation, while also serving as training samples and reference data for the next cycle of AI model operation.

[0098] (2) The proactive health prescription generation unit is used to automatically generate structured health prescriptions, including disease risk assessment, disease progression prediction and personalized lifestyle suggestions, based on the patient's historical health data and current medical data (including changes in disease symptoms, disease complications and comorbidities, prescription changes, etc.), combined with the artificial intelligence chronic disease risk prediction model and the health prescription generation model. These prescriptions are then reviewed and implemented by chronic disease doctors, thereby enhancing patients' forward-looking health awareness and proactive participation.

[0099] Specifically, such as Figure 4 As shown, the health prescription includes a chronic disease risk prediction section (risk grading for future chronic disease onset / progression, risk warnings and suggestions, and health guidance) and a lifestyle prescription generation section (personalized non-drug intervention plans, including lifestyle suggestions such as exercise, diet, and weight, key points for indicator observation, and precautions). For example, the prescription might display, "Based on your current lifestyle and medication, the estimated risk of developing diabetes in the future is 95% (high). You can reduce your risk factors through a healthy lifestyle and strict adherence to your doctor's advice." In the health prescription, the risk of onset and progression are divided into two levels: "high" and "low." The system defaults to a risk >0.8 for high-risk patients, and this value can be adjusted by the local health department based on regional medical capabilities.

[0100] To achieve the function of predicting the risk of disease onset and progression, this invention presents a chronic disease risk prediction model composed of multiple sub-modules for six types of chronic diseases. The model can automatically assess the risk of onset of all chronic diseases for a user based on input. To address the problem of unequal follow-up intervals and different test items in real clinical data, resulting in unequal lengths and intervals of time-series clinical data, a novel artificial intelligence-based chronic disease risk prediction model method is proposed.

[0101] Specifically, the input to the chronic disease risk prediction model includes four types of key feature data:

[0102] (a) Dense features: The top 50 dense features with the fewest missing data selected from the test and examination items by the missing rate, combined with several important feature data proposed by the clinical expert team based on prior medical knowledge;

[0103] (b) Sparse features: Sparse feature data other than the selected dense feature data in the inspection and testing items;

[0104] (c) Missing pattern and time interval matrix: The missing pattern of the user's dense input data and the user's time interval matrix (recording the time interval between each examination and the last examination with data) are used to characterize the temporal frequency and observation missing features of the patient's physical examination data;

[0105] (d) Static characteristics: such as the user's gender, age, and other demographic information;

[0106] The chronic disease risk prediction model consists of four sub-modules: a multivariate time series prediction sub-module based on the Transformer structure, responsible for handling dense input; a large language model sub-module, which handles sparse input and unstructured data; a filter sub-module, which is used to extract missing patterns and time frequency information; and a static feature encoding sub-module, which handles auxiliary information.

[0107] In actual prediction, the model first obtains all user input data: 1. Basic user information, including name, gender, date of birth, and ID number (unique identifier); 2. Physical examination data, including examination date, examination / test name, test result value, and unit of measurement; 3. Historical medical information, i.e., historical medical labels formed from multi-source data such as outpatient, inpatient, and medical records; 4. Other features, including age and examination time interval (the time interval between two consecutive physical examinations). Then, data processing is performed, including standardizing and unifying the names and units of examination / test names, and organizing all features into a data format suitable for the prediction model input. This is divided into four parts: dense input data, sparse input data, missing patterns, and other static features, and then loaded into the prediction model.

[0108] During the model inference phase, the prepared data is input into the corresponding submodule. Dense input data is treated as a vector for multivariate time series interpolation, and then input into a Transformer-based prediction model suitable for time series analysis tasks. After passing through multiple Transformer decoder layers, a feature vector of dense data is obtained. Sparse input data is treated as JSONB format data and input into the large language model submodule for processing to obtain a feature vector of sparse data. The large language model used here is also a large model fine-tuned according to its own dataset in this embodiment of the invention.

[0109] Furthermore, the user's missing patterns and time interval matrix are input as two important sets of information into a filter submodule for processing time-frequency information, resulting in a feature vector for the missing patterns. This invention addresses the widespread missing and irregular sampling phenomena in clinical time series data by treating the missing pattern matrix and time interval matrix as inputs with important temporal dynamic properties, and designs a separate time-frequency filtering extraction structure. A dedicated filter bank is introduced to extract time-frequency features across multiple scales and frequency bands. This operation significantly improves the model's sensitivity to complex time-frequency patterns such as periodicity, persistence, and abrupt changes in missing events; for time interval heterogeneity, it can model its actual impact on observation dynamics and causal structure. Features preprocessed by the filter, whether periodic missing, sparsely distributed, or time-varying in intervals, are efficiently encoded, thus better reflecting the structural uncertainty and potential physiological rhythms of the sequence when fused with the main data features, providing more discriminative supplementary information for the downstream prediction layer.

[0110] In addition to the inputs mentioned above, the model also uses other important static features of the user as auxiliary reference information for one-hot encoding to obtain static feature vectors. Finally, a classification head submodule is used to integrate the four feature vectors for classification prediction. Inputting the four feature vectors, the classification head outputs the risk of developing all of the user's chronic diseases (between 0 and 1). These probabilities are compared with a system-set threshold (default 0.8). If the predicted probability of a certain chronic disease is higher than the corresponding threshold, the risk label for that chronic disease is set to "high," otherwise it is set to "low." Multiple labels can be output simultaneously for multiple diseases. After prediction, the system generates a health summary of the chronic disease prediction (e.g., "You have no abnormal indicators, please continue to maintain this."). Finally, the disease probability, risk label, and prediction summary obtained from the chronic disease prediction algorithm are written into the database.

[0111] In summary, this embodiment introduces a time series analysis model based on the Transformer architecture, optimizing it for time series analysis tasks. Combined with learnable positional encoding, it better adapts to multivariate time series analysis tasks. The prediction model based on the Transformer architecture consists of multiple Transformer Decoder layers. Specifically, the prediction model design of this invention adopts a hierarchical feature extraction and top-level dynamic fusion structure. Dense observations, missing patterns, and time interval information are processed through independent feature extraction channels. Dedicated filters are introduced for the latter two to achieve multi-scale time-frequency feature modeling. The higher-order features extracted from the aforementioned auxiliary information are not directly merged with the main data at the input end, but rather integrated within a dynamic fusion unit at the top layer of the prediction network, primarily using attention and gating mechanisms. Theoretical derivation and experimental verification show that this method significantly outperforms traditional input layer mixing methods, more accurately capturing missing mechanisms, sampling heterogeneity, and potential time dependencies in time series data, greatly improving the model's prediction performance and generalization ability for complex, multivariate clinical time series data. This method, after derivation and testing, has been found to effectively improve prediction performance.

[0112] Furthermore, dense input data requires multivariate time series imputation when entering the model for analysis. This invention, when imputing dense input data, utilizes the missing pattern matrix as a feature, abandoning the traditional "imputation before modeling" approach. Instead, it adopts a joint modeling and imputation approach, jointly optimizing the imputation process with the target task, using the missing values ​​themselves as an information source to achieve end-to-end learning. In practical application, this invention adds two auxiliary training objectives to the main model training objective (prediction loss function): mask imputation loss and observation reconstruction loss. Through these two objectives, the model learns how to better fill in missing values ​​during training, demonstrating significant advantages for time series imputation.

[0113] Furthermore, after the prediction is completed, the system will generate a health risk summary text (e.g., "Your current risk is high; it is recommended to adjust your diet") and write the prediction results, tags, and summary into the database for the health prescription generation model to use. The health prescription generation model is used to generate dietary, medication, symptom and clinical indicator precautions, and other precautions for future onset and chronic disease progression risks, thereby enhancing patients' proactive awareness and ability to participate in health management.

[0114] The health prescription generation model process includes the following input data: 1. Basic user information, including age and gender; 2. Physical examination results data, including examination items, results, and abnormalities; 3. Chronic disease risk tags, output by a chronic disease risk prediction algorithm; 4. Doctor's advice (optional), derived from historical treatment opinions and notes related to the user in a medical database. When developing a health guidance plan, a brief summary of the physical examination is first generated based on the results data. Then, chronic disease tags are integrated. If historical doctor's advice is available, it is retrieved to extract content relevant to the current chronic disease, serving as a reference for generating health advice. A prompt is constructed by organizing user age, physical examination summary, chronic disease tags, and historical advice into input text. This text is then processed using a finely tuned large language model to generate 3-5 multi-faceted, structured health suggestions in a doctor's voice, covering lifestyle, diet, and exercise, avoiding specific prescriptions and medication recommendations. Finally, the generated suggestions are deduplicated, and content inconsistent with the prompts is removed for compliance processing, ensuring the suggestions are compliant, concise, and unambiguous. The final output is personalized health advice in HTML format.

[0115] Health Prescription Generation Model Training: First, based on existing patient prescription data within the system, a structured and high-quality prescription dataset is constructed, incorporating patient basic information, diagnostic results, test indicators, disease type, allergy history, complications, historical treatment plans, drug usage, and subsequent efficacy feedback. To protect privacy and ensure model generalization ability, the dataset undergoes rigorous anonymization, noise filtering, and label consistency checks. Simultaneously, domain knowledge is used for data standardization, feature annotation, and label refinement to ensure the data used for model training is comprehensive, accurate, and possesses medical professionalism. Based on this, the currently mainstream open-source large language pre-trained model Deepseek-r1 is selected as the base model. The parameter-efficient fine-tuning algorithm "LoRA (Low-Rank Adaptation)" is employed for secondary incremental training of the base model. LoRA uses a low-rank matrix to fine-tune some key parameters, enabling the model to converge more quickly to the specialized knowledge distribution for health management and medical advice generation while maintaining the original knowledge system. This improves the input-output ratio and algorithm maintainability, thereby enhancing the reliability and professionalism of the large language model's health advice solutions, making it more suitable for the service population of this invention. After deployment, the model can be continuously fine-tuned by combining doctor feedback and patient follow-up data to dynamically optimize its accuracy and professional intelligence level.

[0116] Finally, the system automatically generates a health prescription and submits it to the patient's contracted chronic disease physician for review. If the physician agrees with the prescription, they sign it; if they reject any part of the prescription, they must provide the reason for rejection. The health prescription is provided to the patient via paper printout, mobile push notification, etc. The system automatically records the chronic disease physician's approval or rejection of the health prescription, and the reasons for rejection will be used for model optimization after evaluation.

[0117] (3) The proactive referral unit is used to automatically generate referral suggestions and push them to the doctor based on the model's assessment of the patient's condition (considering factors such as disease severity, risk of complications, and the processing capacity of lower-level medical institutions); the doctor can refer to the risk warnings provided by the system to make a decision signature; the referral results can be fed back to the system (for iterative optimization of model parameters) to realize referral tracking.

[0118] Specifically, such as Figure 5 As shown, in the referral recommendation generation process, based on the chronic disease incidence risk score and chronic disease progression risk score generated by the chronic disease prediction model, and the referral rules set by the system, a decision-making suggestion on whether to refer is automatically generated. This suggestion includes key referral information such as the reason for referral, the recommended receiving institution, and the recommended receiving department. The reason for referral is automatically generated by an AI referral engine based on a large language model. The receiving institution and receiving department are dynamically determined according to the standardized referral path provided by the local health authorities. Depending on regional differences, the referral path supports a three-tiered structure: "Community Health Service Station / Village Clinic → Community Health Service Center → District Hospital," with clearly defined upstream and downstream institutions and departments at each level. In some areas, due to the high degree of integration between community health service stations / village clinics and community health service centers, the referral path only includes a two-tiered structure: "Primary Community Health Service Institution → District Hospital."

[0119] The AI ​​referral engine is built upon a finely tuned large language model. This model is trained using referral decision samples related to chronic diseases and incorporates a Chain of Thought (CoT) reasoning mechanism to achieve step-by-step decision-making capabilities. Simultaneously, to ensure that referral decisions follow standardized referral paths and rules, the system employs Retrieval Augmentation (RAG) technology: during the model input phase, similarity searches are performed between patient assessment data text and a referral rule vector database. Matching referral rules are dynamically injected into the model input as external knowledge context, thereby improving the rationality and standardization of the generated results.

[0120] During execution, the system integrates the patient's basic information, chronic disease onset and progression risk scores, and examination data into standardized structured input, which is then fed into a large language model. Based on contextual knowledge and prompt words, the large language model generates output including a referral decision and its reasons. After standardization, the output is saved as part of a referral recommendation report and simultaneously pushed to the patient's contracted physician for review.

[0121] For example, if a patient's chronic disease progression risk score is "high," and considering the patient's hospital examination results and treatment history, the large model will, according to referral rules, progressively execute the above steps to determine that the patient needs referral. Based on the score and other input data, it will provide what it considers to be the scientifically sound reason for referral. Subsequently, the standardized referral judgment and reasons will be added to the database as part of the referral recommendation and written into a referral recommendation report.

[0122] In referral tracking, the system backend automatically searches for the patient's medical records at various medical institutions in the regional health data platform (including historical medical information from the aforementioned medical institutions) based on the patient's ID number. It automatically tracks whether the patient has completed the recommended referral process by monitoring changes in the medical institutions visited, assigning referral tags and forming a closed-loop management system for referral results. The system connects to the regional health data platform based on the patient's ID number, monitoring changes in the patient's medical records at various medical institutions in real time to determine their referral behavior.

[0123] Specifically, the system will determine the referral result based on the change of the medical institution and assign the following three referral tags:

[0124] (1) "Transferred out": This means that the patient has been transferred to a designated institution according to the referral recommendation;

[0125] (2) "Not referred as recommended": This means that the patient was referred, but not to the recommended institution or department;

[0126] (3) "No": This means that the patient was not referred.

[0127] The referral tag results will be fed back to the system for subsequent referral effectiveness evaluation, medical behavior analysis, and supplementation of AI model training data.

[0128] 2. Personalized diagnosis and treatment module.

[0129] like Figure 6 As shown, it is used to provide personalized medical follow-up, prescription, and screening services based on the health status and risk characteristics of different individuals, specifically including:

[0130] (1) Personalized follow-up unit, which is used to automatically generate and present customized follow-up plans and content on the doctor's end based on the individual's disease risk label (generated by the disease risk prediction model or the progression risk prediction model), historical health data and current diagnosis. The follow-up plan includes follow-up time, frequency, specific follow-up items and recommended test and examination list to improve the efficiency and accuracy of follow-up and reduce the workload of doctors' follow-up work.

[0131] Specifically, such as Figure 7 As shown, the system first automatically identifies or allows doctors to manually select one or more chronic diseases currently suffered by a patient based on the target chronic disease categories managed by the system. Personalized follow-up content is then provided on the doctor's end of the system, accurately matching the patient's actual condition, reducing irrelevant interference, and alleviating the workload of follow-up. For each identified or selected chronic disease, based on the chronic disease complication database, the system automatically matches and displays its related core symptoms, common complications, and comorbidities. For example, for a patient with both hypertension and diabetes, the system will display hypertension and diabetes follow-up content on the interface, such as common complications of hypertension including "coronary atherosclerotic heart disease," "hypertrophic cardiomyopathy," "hypertensive retinopathy," "hypertensive nephropathy," and "hypertensive encephalopathy," and common complications of diabetes including "diabetic nephropathy," "diabetic foot," "diabetic retinopathy," "diabetic peripheral neuropathy," and "coronary atherosclerotic heart disease." Common complications of a particular chronic disease are assessed and provided by a clinical expert team. Simultaneously, doctors are allowed to perform fuzzy searches for diseases in the chronic disease complication database using disease names. This database is constructed based on the International Classification of Diseases (ICD) Medicare version coding and combined with the consensus assessment of the clinical expert team.

[0132] Furthermore, such as Figure 8 As shown, in the personalized medical follow-up unit, based on the patient's individualized risk prediction results for disease / progression, past health data, and medical records, a personalized list of laboratory tests and suggestions for adjusting the follow-up plan are generated, including:

[0133] (a) Based on the patient’s current chronic disease type and the predicted risk of onset / progression, and on the basis of a standardized follow-up pathway generated based on clinical guidelines and local public health service requirements, the follow-up frequency and follow-up items are dynamically adjusted to form a personalized follow-up plan;

[0134] (b) Integrate health data platform information to provide patients with a historical list of past test and examination items across medical institutions. It supports hierarchical viewing by time range (e.g., the past 90 days) and specific items (e.g., complete blood count - hemoglobin) to avoid duplicate tests.

[0135] (c) Based on the individualized follow-up plan and the patient's recent test and examination completion status, dynamically generate and prompt a list of individualized test and examination items recommended to be completed in the future cycle (e.g., within the next 90 days); for items that have been completed recently, the system automatically identifies and marks the status (e.g., "completed"), and only prompts items that need to be completed.

[0136] For example, specifically, based on information in the health data platform, a list of past laboratory tests can be provided by searching the patient's ID number. If a patient has undergone multiple blood routine tests at different medical institutions within the past 90 days, the doctor can click on "Past 90 Days Laboratory Tests - Blood Routine - Hemoglobin" to view hemoglobin data across institutions within the data platform. Specifically, based on the disease / progression risk prediction results and the patient's follow-up plan, a personalized list of recommended future laboratory tests can be provided. For example, for a patient with both hypertension and diabetes, according to the follow-up plan, they should complete tests such as "blood lipids," "fasting blood glucose," and "ambulatory blood pressure" within the next 90 days. Since an "ambulatory blood pressure" test has recently been completed, the system automatically indicates that the "ambulatory blood pressure" test has been completed and recommends that tests such as "blood lipids" and "fasting blood glucose" be completed within the next 90 days.

[0137] (2) Personalized health prescription generation unit, which is used to integrate the patient's multidimensional health data (including current and past health data, medical records, and chronic disease onset / progression risk prediction results) after the screening or follow-up visit process is completed, and automatically generate a structured customized health prescription through an artificial intelligence prescription generation model. The health prescription includes at least visualization of key chronic disease indicators, chronic disease onset / progression risk prediction results, disease summary and lifestyle treatment suggestions; this unit also supports chronic disease doctors to review, modify and electronically sign the generated prescription content, and use the review feedback information for model optimization.

[0138] Specifically, after screening and follow-up visits using this system, a health prescription is generated by the AI ​​chronic disease risk prediction model and the AI ​​prescription generation model. This health prescription includes the following four parts:

[0139] (a) Visualization of Key Chronic Disease Indicators: Based on patients' cross-institutional health data, the system identifies chronic diseases requiring special attention, dynamically extracts and visualizes the trends of their key physiological indicators. Key indicators are determined based on the patient's actual disease status and risk prediction results, with outliers highlighted. For example, a line chart is used to display the trend of a certain indicator across medical institutions within a region. Specifically, the system defaults to displaying "triglycerides," "total cholesterol," "fasting blood glucose," "postprandial blood glucose," "systolic blood pressure," and "diastolic blood pressure." Based on the patient's actual disease status and the risk of developing / progressing chronic diseases, the system generates key chronic diseases requiring special attention for the patient, and generates key indicators for those diseases, outputting key indicators and abnormal indicators. For example, a patient with diabetes and dyslipidemia will display "fasting blood glucose," "postprandial blood glucose," "glycated hemoglobin," "triglycerides," and "total cholesterol."

[0140] (b) Chronic Disease Onset / Progression Risk Prediction Section: Integrates the results of chronic disease risk prediction models to generate a graded (high / low) onset / progression risk assessment report; the risk level determination threshold is configurable. For example, the prescription will display "Based on current lifestyle and medication, the estimated future risk of developing diabetes is 95% (high). You can reduce your risk factors through a healthy lifestyle and strict adherence to medical advice." In the health prescription, the risk of onset and progression are divided into two levels: "high" and "low." The system defaults to patients with a risk >0.8 as high, and this value can be adjusted by the local health department based on regional medical capabilities.

[0141] (c) Disease summary section: Input patient's medical information and key indicator data, and output a structured disease summary through AI health prescription generation model, covering the test frequency, historical abnormalities, current status and future monitoring focus of important / abnormal indicators;

[0142] (d) Lifestyle-based treatment recommendations: Input multidimensional patient data and generate personalized non-drug intervention plans through the AI ​​health prescription generation model, including lifestyle recommendations (exercise, diet, weight, etc.), key points for indicator observation and precautions; the prescription weight and level of detail are dynamically adjusted according to the condition and risk level.

[0143] Furthermore, the personalized health prescription generation unit supports a mechanism for chronic disease physicians to review and confirm generated prescriptions. Specifically, chronic disease management physicians can review the content of generated prescriptions; if the content is reasonable, they will sign it off; if some suggestions are rejected, a clear reason for rejection must be provided. The system will automatically record this review behavior and incorporate the reasons for rejection into the model optimization loop for subsequent model training and inference logic correction, continuously improving the clinical adaptability of AI-generated content.

[0144] The final health prescription can be delivered to patients in various forms, such as paper printing, SMS / WeChat push, and App push, to achieve continuity and accessibility of health management after follow-up visits.

[0145] (3) Personalized disease screening unit, which is used to conduct intelligent screening based on individual characteristics for residents who have not yet been diagnosed with a specific chronic disease, in order to identify potential high-risk groups and thus achieve the goal of early screening, early prevention and early intervention in chronic disease management.

[0146] Specifically, the personalized disease screening unit can assess the risk of developing a specific chronic disease in any primary healthcare institution where the system is deployed, based on residents' past health data (including high-risk factors, lifestyle, medical history, family history, and laboratory test data) and combined with an AI-powered chronic disease risk prediction model (disease incidence prediction model). If a resident's chronic disease risk exceeds a preset threshold, the system automatically includes them in high-risk population management and initiates a targeted screening process.

[0147] For newly identified high-risk individuals during medical visits who have not yet signed up for chronic disease management services, a recommendation to sign up for corresponding chronic disease management services will be sent to the doctor's end. The doctor will be authorized to dynamically manage these individuals based on their specific circumstances. Figure 9 The content shown allows for the inquiry of medical conditions, establishing a closed-loop service path of screening and follow-up.

[0148] For high-risk individuals who have already signed up, the personalized disease screening unit can automatically generate customized disease screening plans, including the specific diseases to be screened, screening items, recommended screening institutions, and suggested screening frequency. For example, for high-risk individuals with chronic kidney disease, it is required to complete routine urine tests, urine albumin-to-creatinine ratio (UACR) and serum creatinine tests annually at a medical institution at the level of their chronic disease physician, and receive CKD prevention and treatment education.

[0149] All screening results will be automatically archived in the resident's electronic health record upon completion and can be used as a data source for subsequent model iteration training to continuously improve the accuracy of the risk prediction model and the adaptability of individualized screening recommendations.

[0150] Through the deployment of this unit, the system can achieve intelligent identification and early intervention for high-risk individuals with chronic diseases in the region, improve the disease prevention capabilities at the grassroots level, and help move the prevention and control of chronic diseases forward and achieve precise control.

[0151] 3. Intelligent data governance and decision support module.

[0152] like Figure 10 As shown, it provides semantic search, indicator commonality identification, and multi-source data integration capabilities for clinical data, specifically including:

[0153] (1) Intelligent correlation identification unit for indicators, which is used to automatically identify potential correlations between different clinical indicators (such as key indicators shared by multiple cardiovascular diseases) through artificial intelligence models (based on knowledge graph reasoning mechanism), and realize the automatic classification and synchronous operation (such as selection) of clinical indicators (symptoms, complications, comorbidities, smoking and drinking lifestyle) on the follow-up interface and screening interface, so as to reduce the repetitive work of doctors in actual follow-up and data entry.

[0154] Specifically, the intelligent association and recognition unit for indicators is based on a self-built medical knowledge graph and combines structured health data (such as physiological indicators, laboratory test results, and diagnostic results). Through rule-based reasoning and path query algorithms, it can dynamically judge the potential concurrent relationships and causal connections between diseases, thereby driving the automatic selection logic of corresponding indicator items in the doctor's interface, effectively reducing redundant data entry operations.

[0155] For example, if a patient is identified by the system as having "diabetes," "hypertension," and "coronary heart disease," then among the types of diabetic complications and hypertension complications, "coronary atherosclerotic heart disease" will be automatically selected (e.g., Figure 11 (As shown).

[0156] For example, a male patient's waist circumference was measured at the nurse station's all-in-one machine before his visit and was 100cm, meeting the male obesity standard (waist circumference > 90cm). This satisfies the logical inference of "obesity". When the chronic disease doctor imports the data from the all-in-one machine during the follow-up visit, the option related to "obesity" will be automatically checked in the doctor's interface of this invention for screening related diseases.

[0157] For example, in a patient's historical test results, the low-density cholesterol level in the blood biochemistry test was 4.0 mmol / L, indicating that the patient met the clinical criteria for hyperlipidemia. In the doctor's interface, the option for "hyperlipidemia" will be automatically checked to remind the patient to screen for coronary heart disease and stroke.

[0158] To support the aforementioned reasoning mechanism, this invention, through combining publicly available open-source medical knowledge and supplementation by a clinical expert team, constructs a dedicated knowledge graph database based on Neo4j. When constructing the knowledge graph database, this invention defines entity nodes, relationship modeling, and attribute definitions. Subsequently, this invention integrates structured data (patient basic data) and simultaneously writes or synchronizes it to a data table structure available in the knowledge graph. Rules are defined in the knowledge graph, for example: "Diabetes" and "Hypertension" → if also suffering from "Coronary Heart Disease" → automatically select "Coronary Heart Disease" in either "Diabetic Complication Type" or "Hypertension Complication Type". Neo4j supports the Cypher query language, enabling complex path retrieval and conditional judgment. It detects the status of relevant indicators through nodes such as diseases and examination results, and outputs selection results. Data is automatically synchronized in the system background, writing patient examination data as attributes or event nodes into Neo4j. A series of Cypher statements are pre-written for each type of indicator requiring automatic selection. When a doctor opens the patient interface, the system backend queries Neo4j in real time to retrieve all selectable indicators / diseases / complications for that patient, and returns the results to the frontend via an interface. The frontend automatically renders the selected status. The database contains seven types of nodes: medicines, food, examinations, departments, drug categories, diseases, and symptoms. Each disease has its own disease information. It includes node entity relationships such as: department-department relationship, disease-avoided food relationship, disease-recommended food relationship, disease-general medicine relationship, disease-popular medicine relationship, disease-examination relationship, manufacturer-drug relationship, disease symptom relationship, disease concurrency relationship, and disease-department relationship. Compared to traditional if-else rules, knowledge graphs can use nodes and the relationships between nodes to reason about and identify the logical relationships between disease indicators, achieving intelligent association. Relationship adjustments, updates, and visual reasoning are all very intuitive and dynamic.

[0159] (2) Natural language semantic retrieval units, such as Figure 12 As shown, it is used to provide natural language query services with semantic understanding capabilities based on large language models and artificial intelligence agent (AI Agent) technology, realize the structured extraction and semantic hierarchical scheduling query of medical information, assist doctors to conduct multi-round interactive retrieval through natural language, and significantly improve the accessibility of medical record data and the efficiency of clinical auxiliary decision-making.

[0160] This unit allows doctors to input natural language questions (including but not limited to "Patient Zhang San's blood sugar control last year" and "the trend of kidney function changes in the past 6 months"). The system outputs relevant historical medical records, test data, diagnostic conclusions and indicator trends through semantic parsing and knowledge-driven matching mechanisms, and presents them in a structured manner.

[0161] Specifically, the natural language semantic retrieval unit is built based on the integration of the large language model and the Model Context Protocol (MCP) technology. Its internal implementation process is as follows: (1) The MCP server pre-sets query tools, resources and execution templates, and listens for connections; (2) When a doctor initiates a natural language query request on the client interface, the system starts a subprocess to connect to the server and obtain available query components; (3) The model determines whether to call a tool based on the input content. If it needs to be called, the system automatically decides which tool to call and which resources to use, and constructs the input parameters; (4) After the tool is executed, the model can determine whether it needs to call other tools in a loop based on the returned results until the final response is obtained; (5) The large language model merges all intermediate results with the generated text to generate a comprehensive query result output that can be read by the doctor. The above semantic retrieval process is completed autonomously by the large model, including the judgment of the call path, resource scheduling, parameter organization and semantic generation, and has high adaptability and context memory capabilities. Furthermore, this unit supports the fusion retrieval of structured and unstructured information. Doctors can set limited dimensions in the query, such as specifying a time period, medical institution, or indicator type. The system will retrieve information within the limited scope, improving the targeting and efficiency of the retrieval.

[0162] Furthermore, the system supports a multi-turn conversational query mode. After a doctor enters the natural language search interface, the system automatically records the current conversation state and remembers the context content in each round of querying. As the doctor continues to ask questions, the system maintains conversational coherence and optimizes the current output based on historical interactions. When the doctor closes the query window or enters the "exit" command, the system automatically ends the session and clears the context state.

[0163] (3) Data closed-loop integration management unit, which is used to access and integrate multi-source heterogeneous medical systems in the region (based on the regional health data middleware architecture of the present invention), including data sources such as electronic medical record system, public health platform, and medical insurance settlement system. Through unified identity identification, standardized data processing and interface integration, it realizes the structured governance and cross-system closed-loop sharing of residents' health information, and improves the efficiency of data interoperability and the synergy of clinical services throughout the entire process.

[0164] Specifically, the system database uses the patient's ID number as a unique identifier to uniformly aggregate historical health data from different institutions, ensuring long-term continuous management and vertical alignment of various medical records, follow-up information, and test results. The data mapping engine supports automatic cleaning and standardization of test data from various heterogeneous systems, including the following dimensions: variable name mapping, variable unit conversion, reference value standardization, and error value filtering. Variable names are standardized to unify the names of certain test items from different institutions. For example, "creatinine," "creatinine (CREA)," "creatinine (enzymatic method)," and "creatinine measurement" are standardized to "serum creatinine."

[0165] The cleaned and standardized data supports multi-institutional data visualization and interactive query. Doctors can view the historical records of a certain indicator in multiple institutions in the form of lists, trend line charts, etc., on the platform to help judge the continuity of long-term chronic disease management and cross-hospital treatment.

[0166] Furthermore, this unit also supports regional information sharing and performance evaluation integration. After a doctor completes a follow-up visit, issues a prescription, or sets a referral plan, the system supports a one-click "synchronize to regional platform" operation. For example, by clicking the "synchronize to community service" button, the follow-up information can be pushed to the regional public health service platform in real time, automatically completing the submission of performance evaluation data, greatly simplifying the doctor's operation process and reducing the burden of repetitive manual data entry.

[0167] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An artificial intelligence-based full-cycle pathogenic chronic disease management system, characterized by, include: The proactive diagnosis and treatment module: The proactive medical visit and follow-up unit is used to proactively push medical visit reminders, personalized management paths, and follow-up plans based on the chronic disease risk prediction model; The proactive health prescription generation unit automatically generates structured health prescriptions by combining a chronic disease risk prediction model and a health prescription generation model. The inputs of the chronic disease risk prediction model include dense features, sparse features, missing pattern matrix and time interval matrix, and static features. Among them, the missing pattern matrix and time interval matrix are used to characterize the observational gaps and temporal frequency features of the patient's physical examination data. Through a filter bank, multi-scale and multi-band time-frequency features are extracted from the missing pattern matrix of the dense input data and the time interval matrix that records the time interval between each examination and the last examination with data, so as to reflect the structural uncertainty and potential physiological rhythm of the sequence. Finally, the classification head integrates the four types of features to perform classification prediction and output the incidence risk and risk label of all chronic diseases of the user. The proactive referral unit is used to automatically generate referral suggestions based on the chronic disease risk prediction model and track the referral status. Personalized diagnosis and treatment module: The personalized medical visit and follow-up unit is used to automatically generate and present customized follow-up plans and content based on patient data; the personalized health prescription generation unit is used to automatically generate customized health prescriptions through the health prescription generation model; the personalized disease screening unit is used to conduct intelligent screening based on individual characteristics for residents who have not yet been diagnosed with specific chronic diseases. The intelligent data governance and decision support module includes: an intelligent indicator association identification unit that automatically identifies potential associations among clinical indicators using an artificial intelligence model and achieves automatic classification and synchronization; a unit based on medical knowledge graphs and structured health data that uses rule-based reasoning and path query algorithms to dynamically determine potential concurrent relationships and causal connections between diseases, thereby driving the automatic selection logic of corresponding indicator items in the doctor's interface; a natural language semantic retrieval unit that provides natural language query services with semantic understanding capabilities; and a data closed-loop integration and management unit that connects and integrates multi-source heterogeneous medical systems within the region. 2.The artificial intelligence-based whole-cycle path-based chronic disease management system according to claim 1, wherein, The proactive medical visit and follow-up unit is used to proactively push medical visit reminders, personalized management paths, and follow-up plans based on a chronic disease risk prediction model, including: The system's background periodically runs a chronic disease risk prediction model to assess the risk of disease onset in high-risk groups and the risk of disease progression in chronic disease groups. It automatically filters out individuals with risk levels above the set thresholds for disease onset and progression, generating high-risk disease onset lists and high-risk progression lists respectively. Based on these lists, the system pushes medical appointment reminders to the contracted chronic disease doctors. When a patient arrives at a medical institution for treatment, the system automatically matches the corresponding management path and follow-up plan based on their risk label and provides real-time guidance on the doctor's interface. The management path is displayed in a list format, recording the diagnosis and treatment process and task execution status throughout the follow-up process. The follow-up plan is based on the patient's individual characteristics and disease stage, setting the frequency of visits, examination items, and the next follow-up date, and is dynamically updated. After the consultation is completed, the system automatically synchronizes the patient's consultation and follow-up records, and supports one-click synchronization to the public health platform for higher-level supervision and quality evaluation. At the same time, it serves as training samples and reference data for the next cycle of chronic disease risk prediction model operation. 3.The artificial intelligence-based whole cycle path management system for chronic diseases according to claim 1, wherein, The automated health prescription generation unit is used to automatically generate structured health prescriptions by combining a chronic disease risk prediction model and a health prescription generation model, including: The chronic disease risk prediction model adopts a modular structure to support parallel prediction of multiple diseases. The multivariate time series prediction submodule based on the Transformer structure processes dense features, the large language model submodule processes sparse features, the filter submodule processes missing patterns and time interval matrices, and the static feature encoding submodule processes static features. The feature vectors output by each submodule are fused by the classification head module to output the risk of onset and progression of multiple chronic diseases, and are compared with the set risk threshold to form risk labels. The health prescription generation model generates a health examination summary by summarizing abnormal physical examination results and chronic disease label information, and retrieves the patient's historical doctor advice. It constructs a prompt word containing the physical examination summary, risk label, and historical advice, and calls a large language model to generate a structured health prescription that includes disease risk assessment results, disease progression prediction results, and personalized lifestyle recommendations. After deduplication and compliance review, the prescription text is output.

4. The artificial intelligence-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, The automated referral unit is used to automatically generate referral suggestions based on a chronic disease risk prediction model and track referral status, including: Based on the chronic disease incidence risk score and chronic disease progression risk score generated by the chronic disease prediction model, and the referral rules set by the system, the system automatically generates a decision suggestion on whether to refer the patient to another hospital. The suggestion includes key referral information such as the reason for referral, the recommended referral institution, and the recommended referral department. The reason for referral is automatically generated by an AI referral engine based on a large language model. The model is trained by introducing chronic disease-related referral decision samples and combines a thought chain reasoning mechanism to achieve step-by-step decision-making ability. At the same time, a retrieval-enhanced generation technology is used to perform similarity retrieval between the patient assessment data text and the referral rule vector database during the model input stage. The matched referral rules are dynamically injected into the model input as external knowledge context to improve the rationality and standardization of the generated results. When a doctor accepts a recommendation, the system will automatically embed the referral suggestion into the health prescription, displaying information including the reason for the referral, the recommended institution, and the department. When a doctor refuses a recommendation, they must fill in the reason for refusal. The reason for refusal will be used as an important training sample for model iteration and optimization and will be sent back to the fine-tuning database to further enhance the model's judgment accuracy. In the referral tracking, the system automatically tracks whether patients have completed the recommended referral process and assigns a referral tag, forming a closed-loop management of referral results. At the same time, the system connects to the regional health data platform based on the patient's ID number to monitor changes in the patient's medical records at various levels of medical institutions in real time, thereby judging their referral behavior. In addition, the referral tag will be fed back to the system for subsequent referral effectiveness evaluation, medical behavior analysis, and supplementation of model training data.

5. The artificial intelligence-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, The personalized medical follow-up unit is used to automatically generate and present customized follow-up plans and content based on patient data, including: The system first automatically identifies or allows doctors to manually select one or more chronic diseases currently suffered by the patient based on the target chronic disease types managed by the system. For each identified or selected chronic disease, the system automatically matches and displays its related core symptoms, common complications, and follow-up content of comorbidities based on the chronic disease complication database. Based on individualized risk predictions of disease / progression, past health data, and medical records, personalized test and follow-up plan adjustment recommendations are generated, including: (1) Based on the patient's current chronic disease type and the predicted risk of onset / progression, and on the basis of a standardized follow-up pathway generated based on clinical guidelines and local public health service requirements, the follow-up frequency and follow-up items are dynamically adjusted to form a personalized follow-up plan; (2) Integrate health data platform information to provide patients with a historical list of past test and examination items across medical institutions, and support hierarchical viewing by time range and specific items to avoid duplicate examinations; (3) Based on the personalized follow-up plan and the patient's recent test results, dynamically generate and prompt a list of personalized test items to be completed in the future cycle; for items that have been completed recently, the system automatically identifies and marks the status, and only prompts items to be completed.

6. The artificial intelligence-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, The personalized health prescription generation unit is used to automatically generate customized health prescriptions through a health prescription generation model, including: Customized health prescriptions include visualization of key chronic disease indicators, prediction results of chronic disease onset / progression risks, disease summaries, and lifestyle-based treatment recommendations, among which: (1) Visualization of key indicators for chronic diseases: Based on patients’ cross-institutional health data, identify chronic diseases that require special attention, dynamically extract and visualize the trends of their key physiological indicators; key indicators are determined based on the actual disease status and risk prediction results of patients, and outliers are highlighted. (2) Chronic disease incidence / progression risk prediction section: Integrates the results of chronic disease risk prediction models to generate a graded incidence / progression risk assessment report; the risk level determination threshold is configurable; (3) Disease summary section: Input patient's medical information and key indicator data, output structured disease summary, covering the test frequency, historical abnormalities, current status and future monitoring focus of important / abnormal indicators; (4) Lifestyle-based treatment recommendations: Input multidimensional patient data to generate personalized non-drug intervention plans, including lifestyle recommendations, key points for indicator observation and precautions; prescription weights and levels of detail are dynamically adjusted according to the patient's condition and risk level; Meanwhile, the personalized health prescription generation unit supports a mechanism for chronic disease physicians to review and confirm the generated prescriptions. If the content is reasonable, the prescription will be approved by signing it. If the prescription is rejected, a clear reason for rejection must be provided. The system will automatically record the review behavior and incorporate the reason for rejection into the model optimization closed loop for subsequent model training and inference logic correction, thereby continuously improving the clinical adaptability of the model-generated content.

7. The AI-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, After identifying potentially high-risk individuals through intelligent screening, the personalized disease screening unit also includes: For high-risk individuals newly identified during medical visits but who have not yet signed up for chronic disease management services, we will push signing suggestions to doctors, recommend signing up for corresponding chronic disease management services, and authorize primary care physicians to dynamically manage them, thus building a closed-loop service path of screening and follow-up. For high-risk individuals who have signed up, a customized disease screening plan is automatically generated, which includes the specific diseases to be screened, screening items, recommended screening institutions, and suggested screening frequency. All screening results will be automatically archived into the resident's electronic health record after completion and will serve as a data source for subsequent model iteration training.

8. The artificial intelligence-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, The intelligent correlation identification unit is used to automatically identify potential correlations among clinical indicators through an artificial intelligence model and to achieve automatic classification and synchronization, including: A medical knowledge graph is constructed, encompassing seven categories of nodes: diseases, symptoms, drugs, foods, examination items, and departments. These nodes are linked to various relationships, including department-department relationships, disease-avoided food relationships, disease-recommended food relationships, disease-general drug relationships, disease-popular drug relationships, disease-examination relationships, manufacturer-drug relationships, disease-symptom relationships, disease-concurrency relationships, and disease-department relationships. The system writes patients' historical structured data into the graph as event nodes or attributes. Combined with structured health data including physiological indicators, laboratory test results, and diagnostic results, the system uses rule-based reasoning and path query algorithms to dynamically determine potential concurrency relationships and causal connections between diseases. This drives the automatic selection logic of corresponding indicator items in the doctor's interface, returning information on selectable related diseases / indicators / complications and intelligently rendering them as selected.

9. The artificial intelligence-based full-cycle path-based chronic disease management system according to claim 1, characterized in that, The natural language semantic retrieval unit is used to provide natural language query services with semantic understanding capabilities, including: A natural language semantic retrieval unit supporting multi-turn conversational query mode is constructed based on the integration of large language models and artificial intelligence agent technology. The semantic retrieval process is completed autonomously by the large model, including call path judgment, resource scheduling, parameter organization and semantic generation. At the same time, it supports the fusion retrieval of structured and unstructured information. Doctors can set limited dimensions in the query, including specified time period, medical institution or indicator type, and the system will retrieve information efficiently and in a targeted manner within the limited scope.

10. A method for full-cycle path-based chronic disease management based on artificial intelligence, implemented using the full-cycle path-based chronic disease management system based on artificial intelligence as described in any one of claims 1 to 9, characterized in that, Includes the following steps: Based on the proactive diagnosis and treatment module, the proactive medical visit and follow-up unit proactively pushes medical visit reminders, personalized management paths, and follow-up plans based on the chronic disease risk prediction model; An automated health prescription generation unit, combined with a chronic disease risk prediction model and a health prescription generation model, automatically generates structured health prescriptions. The inputs to the chronic disease risk prediction model include dense features, sparse features, missing patterns and time interval matrices, and static features. Among them, the missing patterns and time interval matrices are used to characterize the observational gaps and temporal frequency characteristics of the patient's physical examination data. A filter bank is used to extract multi-scale, multi-band time-frequency features from the missing patterns of the dense input data and the time interval matrix that records the time interval between each examination and the last examination with data, in order to reflect the structural uncertainty and potential physiological rhythms of the sequence. Finally, a classification head is used to integrate the four types of features for classification prediction and output the incidence risk and risk labels of all chronic diseases of the user. An automated referral unit is used to automatically generate referral suggestions based on the chronic disease risk prediction model and track the referral status. Based on the personalized diagnosis and treatment module, the personalized medical follow-up unit automatically generates and presents customized follow-up plans and content based on patient data; the personalized health prescription generation unit automatically generates customized health prescriptions through a health prescription generation model; and the personalized disease screening unit conducts intelligent screening based on individual characteristics for residents who have not yet been diagnosed with specific chronic diseases. Based on the intelligent data governance and decision support module, the intelligent indicator association identification unit automatically identifies potential correlations among clinical indicators through artificial intelligence models and achieves automatic classification and synchronization. Based on medical knowledge graphs and combined with structured health data, rule reasoning and path query algorithms are used to dynamically judge potential concurrent relationships and etiological connections between diseases, thereby driving the automatic selection logic of corresponding indicator items in the doctor's interface. The natural language semantic retrieval unit provides natural language query services with semantic understanding capabilities. The data closed-loop integration management unit accesses and integrates multi-source heterogeneous medical systems within the region.

Images