Chronic disease follow-up visit data management method and system based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent medical treatment, in particular to a chronic disease follow-up visit data management method based on artificial intelligence, which comprises the following steps: collecting physiological parameter data, behavior activity data and medication record data of a patient through a medical equipment terminal; performing time sequence standardization processing on the collected data to generate a data set divided according to a preset time period; setting a first abnormal threshold value based on a clinical medicine guide, and setting a second abnormal threshold value based on historical data statistical analysis; when the data exceeds a first abnormal threshold value, generating a red early warning identifier, and when the data exceeds a second abnormal threshold value, generating an orange early warning identifier; and executing grading response according to the early warning grade. According to the method, a double-threshold cooperation mechanism and clinical threshold and statistical boundary cross validation are utilized, personalized chronic disease data management and dynamic adjustment can be carried out on different patients, monitoring of the patients can be accurately adjusted, and the medical diagnosis effect of the chronic disease patients is improved.

Description

Technical Field

[0001] This invention relates to the field of medical data management technology, and in particular to a method and system for managing chronic disease follow-up data based on artificial intelligence. Background Technology

[0002] Chronic disease follow-up data management is an important component of modern medical informatization and health management. Its core objective is to improve the management efficiency and treatment outcomes of chronic disease patients through scientific data collection, analysis, and application. With the rapid development of technologies such as artificial intelligence, big data, and deep learning, chronic disease follow-up management is gradually shifting from a traditional manual model to an intelligent and precise approach.

[0003] Big data technology offers a novel solution for chronic disease follow-up management. By integrating multi-source data such as patients' electronic health records (EHRs), wearable device data, and laboratory test results, big data technology enables comprehensive monitoring and dynamic analysis of patients' health status.

[0004] However, current AI-based chronic disease follow-up data management systems have three core flaws: they rely on fixed clinical thresholds to determine abnormal events, leading to missed detections of statistical fluctuations and a disconnect between intervention measures and data visualization; they lack automatic response mechanisms, requiring manual confirmation of hyperglycemic events, resulting in significant delays; and they lack group collaborative management functions, making it impossible to generate virtual follow-up groups using data from similar patients.

[0005] Therefore, improvements are needed in the intervention and timeliness of abnormal detection and automatic response for chronic disease data. Summary of the Invention

[0006] The purpose of this invention is to overcome the problems in the prior art that rely on fixed clinical thresholds to determine abnormal events, which leads to missed detection of statistical fluctuations and a disconnect between intervention measures and data visualization, and lack of automatic response mechanisms.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for managing chronic disease follow-up data based on artificial intelligence, characterized by comprising the following steps:

[0008] S1. Collect patients' physiological parameter data, behavioral activity data, and medication record data through medical device terminals;

[0009] S2. Perform time-series multimodal standardization on the collected data to generate a dataset divided according to a preset time period, specifically including:

[0010] S3. Based on an artificial intelligence model, set a first anomaly threshold and a second anomaly threshold, and perform cross-combination judgment;

[0011] S4. When real-time data is determined to exceed the first abnormal threshold, a red warning sign is generated; when it is determined to exceed the second abnormal threshold but not the first abnormal threshold, an orange warning sign is generated.

[0012] S5. Implement tiered responses based on the warning level.

[0013] Preferably, the multimodal normalization process specifically includes the following steps:

[0014] S21. Convert the raw data from different medical device terminals into a preset standard data format;

[0015] S22. Convert physiological parameter data recorded in different unit formats into preset standard units of measurement;

[0016] S23. Based on a unified high-precision timestamp, time alignment is performed on multi-source data with different acquisition frequencies, and resampling is performed through linear interpolation or mean filling methods to generate a data sequence with a unified time granularity.

[0017] S24. Aggregate the processed uniform time-granularity data sequence according to a preset period to generate a data set containing key statistical features within each period.

[0018] Preferably, the rule for cross-combining the first abnormal threshold and the second abnormal threshold is as follows:

[0019] S31. First abnormal threshold setting: Based on preset clinical medical guidelines, set the absolute abnormal threshold for key physiological parameters;

[0020] S32. Second abnormal threshold setting: Based on the statistical analysis of the patient's individual historical dataset, calculate the statistical characteristics (such as mean and standard deviation) of key physiological parameters for each preset time period, and use the confidence interval of statistical methods or time series prediction models to set individualized dynamic abnormal thresholds.

[0021] S33. Utilize an artificial intelligence model to perform a dual assessment of real-time data points: whether they exceed the first anomaly threshold; or

[0022] Does it exceed the second abnormal threshold?

[0023] Among them, the key cross-combination states analyzed by the artificial intelligence model include at least: only touching the second threshold, touching the second threshold and deviating from the first threshold, exceeding both thresholds at the same time, and exceeding the lower limit of the second threshold for multiple consecutive cycles.

[0024] S34. Based on the analysis of the multi-dimensional judgment results in S33 by the artificial intelligence model, and combined with the predefined rule logic, the level and specific data points for triggering the warning are finally determined. The combined state of red warning includes at least exceeding both thresholds at the same time or exceeding the lower limit of the second threshold for multiple consecutive periods. The combined state of orange warning includes at least touching only the second threshold or touching the second threshold and deviating from the first threshold.

[0025] Preferably, the graded response is divided into at least the following levels:

[0026] Level 1 Response: Automatically adjusts the monitoring frequency parameters of medical equipment;

[0027] Level 2 response: Create a virtual patient follow-up team to conduct group intervention;

[0028] Level 3 Response: Activate the Emergency Medical Contact Protocol and send the patient's location.

[0029] Preferably, the physiological parameter data in step S1 includes at least one of blood glucose, blood pressure, and heart rate;

[0030] The behavioral activity data includes at least two of the following: exercise duration, number of steps, and calorie consumption; the medication record data includes the drug name, dosage, and time of administration.

[0031] Preferably, the setting of the second abnormal threshold in step S32 includes: calculating the standard deviation of key physiological parameters in historical data of the same time period, and setting the range of N times the standard deviation as the statistical abnormality boundary.

[0032] Preferably, the generation of the virtual patient follow-up group for the secondary response in step S5 includes: calculating the similarity based on the standardized historical health indicator data and the current status, clustering the patient groups into virtual groups, and assigning a person who has achieved blood glucose control target to each group as a mentor member.

[0033] Preferably, the third-level response further includes: sending an encrypted data packet containing the most recent three periodic aggregate values ​​after the standardized processing to the associated medical institution, generating a cross-platform video consultation link, and synchronizing the real-time data stream of the medical device.

[0034] Preferably, the data format conversion in step S21 specifically includes identifying and converting device-specific data fields, removing irrelevant identifiers, and constructing a data structure with unified field names.

[0035] A chronic disease follow-up data management system, comprising:

[0036] Multi-source data acquisition module: used to connect heterogeneous data interfaces of blood glucose monitors, fitness trackers and smart pillboxes;

[0037] Intelligent analysis engine: used to perform dual anomaly threshold detection and generate early warning indicators;

[0038] The intelligent analysis engine includes:

[0039] Clinical rule base: Stores personalized threshold parameters based on chronic disease classification;

[0040] Dynamic statistical unit: The standard deviation calculation model is updated weekly.

[0041] Dynamic Protocol Generator: Used to generate graded response instructions based on the warning level. When executing a level 2 response, the dynamic protocol generator opens an encrypted group chat channel for the virtual follow-up team and automatically generates a group health trend comparison report every week.

[0042] Visualization terminal: Used to display risk heat maps and warning signs on mobile devices.

[0043] A computer-readable storage medium having computer-executable instructions etched on its surface, which are executed by a processor:

[0044] The device driver layer instruction set controls the medical terminal to collect data; calls the data conversion algorithm library to execute standardized processing procedures; loads the artificial intelligence model weight file to determine thresholds; triggers the hierarchical response control protocol; when the instruction detects that the communication load exceeds the preset threshold, it activates the data compression transmission routine and resets the transmission queue priority.

[0045] The beneficial effects of this invention are:

[0046] 1. This invention collects multi-dimensional health data from patients through medical device terminals, uses artificial intelligence technology to standardize and detect anomalies in the data, generates graded early warnings based on dual thresholds, and executes corresponding graded response strategies, thereby realizing intelligent management of chronic disease follow-up data. It effectively solves the problems of untimely data collection, delayed anomaly identification, and single response measures in traditional follow-up. Through multi-source data integration and dynamic analysis, it improves the accuracy and timeliness of data and reduces the cost of manual intervention.

[0047] 2. This invention constructs virtual follow-up groups by clustering health indicators similarity, which promotes mutual assistance management among patients, improves the effectiveness of group health intervention, ensures rapid allocation of medical resources in emergency situations, shortens emergency response time, realizes the transformation of chronic disease follow-up from passive monitoring to active intervention, improves the intelligence level of follow-up management and the effectiveness of patient health management, and provides strong support for the long-term management of chronic diseases. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the process of the chronic disease follow-up data management method based on artificial intelligence of the present invention.

[0049] Figure 2 This is a schematic diagram of the structure of the artificial intelligence-based chronic disease follow-up data management system of the present invention.

[0050] Figure 3 This is a network connection diagram of the chronic disease follow-up data of the artificial intelligence of this invention. Detailed Implementation

[0051] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] like Figure 1 As shown in the embodiments of this application, the chronic disease follow-up data management method includes:

[0055] Step S1. Periodically collect monitoring data for the corresponding patients through multiple medical device terminals, including physiological parameter data, behavioral activity data, and medication record data;

[0056] Specifically, multiple medical device terminals installed on the patient's end synchronously collect the aforementioned physiological parameter data (including at least one of blood glucose, blood pressure, and heart rate), the aforementioned behavioral activity data (including at least two of exercise duration, steps, and calorie consumption), and medication record data (including drug name, dosage, and precise medication time). Through the above technical solution, a complete data source is provided for subsequent multimodal data analysis.

[0057] Step S2. Perform time series multimodal standardization on the collected raw data to generate a dataset divided according to a preset period, specifically including:

[0058] Step S21. Identify the unique data fields of different medical device terminals and remove irrelevant identifiers, convert the heterogeneous raw data into a preset standard data format, and construct a structured dataset containing unified field names;

[0059] Step S22. Convert physiological parameters recorded using different unit systems (such as blood glucose values ​​recorded in mmol / L and mg / dL respectively) into preset international standard units of measurement.

[0060] Step S23. Generate a unified time reference based on a high-precision atomic clock, align the time axis of data streams with different acquisition frequencies (such as blood pressure every 5 minutes and steps every second), supplement missing time point data by linear interpolation, and downsample high-frequency data by mean filling method to generate an equally spaced data sequence with unified time granularity.

[0061] Step S24. Divide and aggregate the standardized time series data according to the preset period (hour / day / week), calculate the key statistical features (including mean, maximum, minimum and standard deviation) in each period, and generate a periodic feature data set;

[0062] The above technical solution solves the compatibility problem of multi-source heterogeneous data and establishes a standardized input for subsequent threshold determination.

[0063] Step S3. Based on the artificial intelligence model, set a first anomaly threshold and a second anomaly threshold, and perform a cross-combination judgment, as follows:

[0064] Step S31. Load the preset clinical medical guideline database, parse the absolute safety boundary values ​​for key physiological parameters (e.g., blood glucose > 16.7 mmol / L is defined as clinical high risk), and generate fixed threshold rules based on evidence-based medicine.

[0065] Step S32. Call the patient's individual historical dataset, calculate the statistical characteristics of physiological parameters (such as the mean ± standard deviation of blood glucose in the same period of the past 30 days) through the statistical analysis module, and use the statistical model (specifically: set the individual historical mean ± twice the standard deviation) to generate a dynamically changing safety boundary, or use the 99% confidence interval of the time series prediction model (such as LSTM neural network) as the individual floating threshold.

[0066] Step S33. Input the real-time data points into the artificial intelligence judgment engine and simultaneously execute the triple logic judgment: whether the current value exceeds the first abnormal threshold (clinical absolute red line); whether the current value exceeds the second abnormal threshold (individualized dynamic boundary); analyze the multi-dimensional combination state (including only touching the second threshold, touching the second threshold and deviating from the first threshold, exceeding both thresholds at the same time, and exceeding the lower limit of the second threshold for three consecutive cycles).

[0067] Step S34. Determine the anomaly level and trigger point: Based on the predefined rule mapping table (such as automatically marking it as the highest risk when both thresholds are exceeded at the same time), and combined with the weight analysis of the combined state by the artificial intelligence model, the warning level and the corresponding anomaly data timestamp are finally output;

[0068] The above technical solutions enable the collaborative analysis of medical rules and individual physical signs, thereby improving the specificity of abnormality identification.

[0069] Step S4. Generate a warning sign based on the judgment result of step S34: when the real-time data exceeds the first abnormal threshold and the second abnormal threshold at the same time, a red warning sign is generated; when the data only exceeds the second abnormal threshold but does not exceed the first abnormal threshold, an orange warning sign is generated.

[0070] Step S5. Implement a tiered response strategy based on the warning level:

[0071] Level 1 Response (corresponding to Orange Alert): Automatically sends instructions to the medical device terminal to increase the monitoring frequency from the basic cycle (e.g., every 30 minutes) to the enhanced cycle (e.g., every 5 minutes).

[0072] Level 2 response (corresponding to the continued existence of the orange alert or a combination of factors indicating a worsening trend): Calculate the group similarity based on the patient's historical health indicators (including blood glucose control stability and medication adherence), dynamically form virtual follow-up groups of similar patients through clustering algorithms, assign a person who has achieved blood glucose control target to each group as a mentor member, and initiate group intervention message push.

[0073] Level 3 Response (corresponding to red alert or high-risk combination judgment): Activate the emergency medical communication protocol, transmit AES-256 encrypted data packets containing the aggregated values ​​of the most recent three periods to the associated medical institutions, simultaneously generate cross-platform video consultation links and establish a real-time data flow channel for medical devices, and call the geolocation service API to push the patient's real-time location coordinates.

[0074] The above technical solutions establish an intelligent linkage mechanism between abnormal events and medical resources.

[0075] Further, step S2 includes:

[0076] The system receives raw data streams from multiple medical device terminals via a distributed data access gateway. It then uses a format parsing engine to identify device-specific codes (such as UUIDs in the Bluetooth protocol), strips off non-standard data headers, and reconstructs them into unified format data packets. A unit converter is used to call an international standard metrology conversion formula library. A sliding time window algorithm is employed to compensate for and align time-offset data. Through these technical solutions, the consistency of multi-source data in the spatiotemporal dimensions is ensured.

[0077] Further, step S3 includes:

[0078] During the initialization phase, when historical patient data is insufficient, a temporary second threshold is generated using the population mean completion method.

[0079] When the AI ​​model detects a new abnormal pattern (such as being consistently below the second lower threshold but not triggering the clinical threshold), it automatically initiates the threshold update process: extracting the feature vector of the abnormal pattern (including the slope of change, duration, and combination of associated physiological parameters), and updating the individualized threshold model parameters through an incremental learning algorithm.

[0080] The above technical solutions enable the system to have continuous optimization capabilities.

[0081] Further, step S5 includes:

[0082] When implementing a Level 2 response, the virtual follow-up team uses a dynamic weighting algorithm to calculate member similarity:

[0083] Similarity = 0.4 × DTW distance of blood glucose curve + 0.3 × medication time matching rate + 0.3 × correlation coefficient of exercise volume;

[0084] The encrypted data packet of the three-level response contains a three-layer structure, where the first layer is a summary of the patient's basic information, the middle layer is an aggregated statistical value matrix, and the bottom layer is the original data hash check code;

[0085] The above technical solutions ensure the integrity and privacy of medical data transmission.

[0086] Furthermore, when the system communication load is ≤50% of the bandwidth threshold, the full patient health status report is pushed to the central monitoring platform in real time; when the communication load is >50% of the bandwidth threshold, a data transmission priority queue is set according to the patient's risk level (red alert > orange alert > routine monitoring), and the complete data packets of high-risk patients are transmitted first.

[0087] The above technical solutions optimize the efficiency of medical network resource allocation.

[0088] Example 2

[0089] This invention also provides a chronic disease follow-up data management system for implementing the above methods, such as... Figure 2 As shown, the system includes:

[0090] Multi-source data acquisition module: This module provides a heterogeneous data interface for connecting blood glucose monitors, fitness trackers, and smart pillboxes. It consists of multiple medical device terminals, including wearable blood glucose meters, smart blood pressure monitors, fitness trackers, and medication record terminals. It periodically uploads raw monitoring data via a hybrid Bluetooth 5.0 and LoRa networking protocol. The multi-source data acquisition module supports three communication protocols: Bluetooth, Wi-Fi, and cellular networks, and can automatically switch communication modes to maintain data continuity.

[0091] Intelligent processing engine: used to perform dual anomaly threshold detection and early warning label generation, with built-in standardized processing module (execution steps S21-S24), dual threshold analysis module (execution steps S31-S34) and early warning generation module (execution step S4).

[0092] The intelligent analysis engine includes a clinical rule base and a dynamic statistical unit. The clinical rule base stores personalized threshold parameters based on chronic disease classification; the dynamic statistical unit is used to update the standard deviation calculation model on a weekly basis.

[0093] Dynamic Protocol Generator: Used to generate graded response instructions based on the warning level, including device control interface (to achieve Level 1 response), virtual team management platform (to achieve Level 2 response) and medical emergency channel gateway (to achieve Level 3 response).

[0094] Among them, the dynamic protocol generator opens an encrypted group chat channel for the virtual follow-up team when executing the secondary response, and automatically generates a group health trend comparison report every week.

[0095] Visualization terminal: Used to display risk heat maps and warning signs on mobile devices, thereby monitoring the system communication load in real time and dynamically adjusting the data transmission strategy through a priority queue manager;

[0096] The present invention also provides a computer-readable storage medium having computer-executable instructions engraved on its surface, which are implemented when executed by a processor:

[0097] The device driver layer instruction set controls the medical terminal to collect data; calls the data conversion algorithm library to execute standardized processing procedures; loads the artificial intelligence model weight file to determine thresholds; triggers the hierarchical response control protocol; when the instruction detects that the communication load exceeds the preset threshold, it activates the data compression transmission routine and resets the transmission queue priority.

[0098] In summary, this invention periodically collects multimodal health data through medical device terminals and uses an artificial intelligence model to apply dual thresholds to standardized time series data. Through cross-analysis, it generates dynamic early warning levels and triggers a tiered response mechanism based on these levels. This method effectively solves the problem of data heterogeneity in chronic disease management, reduces the amount of invalid data transmission, and achieves accurate early warning. Through collaborative analysis of clinical rules and individual models, it intelligently filters key anomalies, reducing computational load. It automatically updates model parameters when new abnormal patterns are detected, enhancing system adaptability. Time series aggregation and geographic information mapping improve the efficiency of health event tracing. Dynamically optimizing transmission strategies based on communication load ensures priority processing of high-risk patient data. Through the synergy of these technical solutions, intelligent matching of medical resource allocation and health risk levels is achieved, improving the precision of chronic disease management and the speed of emergency response, providing reliable technical support for smart healthcare.

[0099] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for managing chronic disease follow-up data based on artificial intelligence, characterized in that... Includes the following steps: S1. Collect patients' physiological parameter data, behavioral activity data, and medication record data through medical device terminals; S2. Perform time-series multimodal standardization on the collected data to generate a dataset divided according to a preset time period, specifically including: S3. Based on an artificial intelligence model, set a first anomaly threshold and a second anomaly threshold, and perform cross-combination judgment; S4. When real-time data is determined to exceed the first abnormal threshold, a red warning sign is generated; when it is determined to exceed the second abnormal threshold but not the first abnormal threshold, an orange warning sign is generated. S5. Implement tiered responses based on the warning level.

2. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The multimodal normalization process specifically includes the following steps: S21. Convert the raw data from different medical device terminals into a preset standard data format; S22. Convert physiological parameter data recorded in different unit formats into preset standard units of measurement; S23. Based on a unified high-precision timestamp, time alignment is performed on multi-source data with different acquisition frequencies, and resampling is performed through linear interpolation or mean filling methods to generate a data sequence with a unified time granularity. S24. Aggregate the processed uniform time-granularity data sequence according to a preset period to generate a data set containing key statistical features within each period.

3. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The rules for cross-combining the first and second abnormal thresholds are as follows: S31. First abnormal threshold setting: Based on preset clinical medical guidelines, set the absolute abnormal threshold for key physiological parameters; S32. Second abnormal threshold setting: Based on the statistical analysis of the patient's individual historical dataset, calculate the statistical characteristics (such as mean and standard deviation) of key physiological parameters for each preset time period, and use the confidence interval of statistical methods or time series prediction models to set individualized dynamic abnormal thresholds. S33. Use an artificial intelligence model to make a dual judgment on real-time data points: whether they exceed the first abnormal threshold; or Does it exceed the second abnormal threshold? Among them, the key cross-combination states analyzed by the artificial intelligence model include at least: only touching the second threshold, touching the second threshold and deviating from the first threshold, exceeding both thresholds at the same time, and exceeding the lower limit of the second threshold for multiple consecutive cycles. S34. Based on the analysis of the multi-dimensional judgment results in S33 by the artificial intelligence model, and combined with the predefined rule logic, the level and specific data points for triggering the warning are finally determined. The combined state of red warning includes at least exceeding both thresholds at the same time or exceeding the lower limit of the second threshold for multiple consecutive periods. The combined state of orange warning includes at least touching only the second threshold or touching the second threshold and deviating from the first threshold.

4. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The graded response is divided into at least the following levels: Level 1 Response: Automatically adjusts the monitoring frequency parameters of medical equipment; Level 2 response: Create a virtual patient follow-up team to conduct group intervention; Level 3 Response: Activate the Emergency Medical Liaison Protocol and send the patient's location.

5. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The physiological parameter data mentioned in step S1 includes at least one of blood glucose, blood pressure, and heart rate; The behavioral activity data includes at least two of the following: exercise duration, number of steps, and calorie consumption; the medication record data includes the drug name, dosage, and time of administration.

6. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The setting of the second abnormal threshold in step S32 includes: calculating the standard deviation of key physiological parameters in historical data of the same time period, and setting the range of N times the standard deviation as the statistical abnormality boundary.

7. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The generation of the virtual patient follow-up group for the secondary response in step S5 includes: calculating the similarity based on the standardized historical health indicator data and the current status, clustering the patient groups into virtual groups, and assigning a person who has achieved the blood glucose control target to each group as a mentor member.

8. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The third-level response also includes: sending an encrypted data packet containing the most recent three periodic aggregate values ​​after the standardized processing to the associated medical institution, generating a cross-platform video consultation link and synchronizing the real-time data stream of the medical device.

9. The method for managing chronic disease follow-up data based on artificial intelligence according to claim 1, characterized in that, The data format conversion in step S21 specifically includes identifying and converting device-specific data fields, removing irrelevant identifiers, and constructing a data structure with unified field names.

10. A chronic disease follow-up data management system for implementing the method of any one of claims 1-10, characterized in that, include: Multi-source data acquisition module: used to connect heterogeneous data interfaces of blood glucose monitors, fitness trackers and smart pillboxes; Intelligent analysis engine: used to perform dual anomaly threshold detection and generate early warning indicators; The intelligent analysis engine includes: Clinical rule base: Stores personalized threshold parameters based on chronic disease classification; Dynamic statistical unit: The standard deviation calculation model is updated weekly. Dynamic Protocol Generator: Used to generate graded response instructions based on the warning level. When executing a level 2 response, the dynamic protocol generator opens an encrypted group chat channel for the virtual follow-up team and automatically generates a group health trend comparison report every week. Visualization terminal: Used to display risk heat maps and warning signs on mobile devices.

11. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the method as described in any one of claims 1-9.

Images