Health monitoring and hierarchical care method and system for elderly people in medical and nursing combination

Abstract

The application discloses a kind of health monitoring and hierarchical care methods and systems of medical care combination of old people, by collecting the health data of old people in hospital, community old-age care institutions, using medical information standardization agreement to carry out format conversion and semantic alignment, fusion generates multidimensional health data set;Distributed ledger is constructed using blockchain technology, ensure the safe record of data access authority and change log, and according to the authorization data sharing of privacy rules;Extract key features such as blood pressure fluctuation value and activity frequency index, risk classification is carried out through support vector machine and random forest model, combine rule engine matching care template, generate personalized scheme and real-time feedback to relevant institutions.The application significantly improves the accuracy and response efficiency of medical care combination service through the organic combination of data standardization, safe sharing and intelligent analysis, and provides comprehensive support for health management of old people.

Description

Technical Field

[0001] This invention relates to the field of health management technology for the elderly, and in particular discloses a method and system for health monitoring and graded care of the elderly that integrates medical and elderly care. Background Technology

[0002] With the increasing aging of the population, health management for the elderly has become a major social concern, its importance being self-evident and directly related to the quality of life of the elderly and the rational allocation of social resources. Integrated medical and elderly care, as an emerging model combining elderly care and medical services, aims to provide comprehensive health protection and daily care for the elderly. However, current practices face numerous challenges and urgently need to overcome existing bottlenecks to promote more efficient and precise service models.

[0003] Existing methods have significant shortcomings in integrating resources and responding to needs. Many solutions remain at a superficial level of collaboration, lacking in-depth information sharing and coordination mechanisms among hospitals, communities, and elderly care institutions. This results in fragmented health data for the elderly, making it impossible to form a unified assessment basis. Furthermore, the matching of care services often relies on manual judgment, lacking systematic risk stratification and personalized plan design, making it difficult to meet the diverse needs of elderly people with different health conditions. This fragmented resource situation and delayed response greatly limit the actual effectiveness of integrated medical and elderly care.

[0004] At the technical level, the core challenges lie in achieving cross-institutional data interoperability and accurate risk classification based on multi-dimensional health information. First, obstacles to data interoperability stem from differing technical standards and privacy protection requirements among different institutions, creating information silos and compromising the integrity and timeliness of health data. Second, because the health status of the elderly is influenced by various factors, such as past medical history, daily monitoring data, and lifestyle habits, relying solely on one type of information is insufficient to accurately determine risk levels, thus affecting the targetedness and timeliness of subsequent care plans. These two issues are interconnected; the ineffective integration of data directly leads to a decline in the accuracy of risk assessment, while inaccurate assessments further exacerbate the inefficiency of resource allocation.

[0005] Therefore, how to break down information barriers between multiple institutions while ensuring data security, and how to achieve accurate risk classification and care matching through comprehensive analysis of the multi-dimensional health information of the elderly, has become a key issue that urgently needs to be addressed in the field of integrated medical and elderly care. Summary of the Invention

[0006] This invention provides a method and system for health monitoring and graded care of the elderly that integrates medical and elderly care, aiming to provide stronger support for the health management of the elderly.

[0007] One aspect of this invention relates to a method for health monitoring and tiered care of the elderly that integrates medical and elderly care, comprising the following steps: S100. Collect health data of the elderly from hospitals, communities and elderly care institutions, and use medical information standardization protocols to convert the format and align the semantics of health data from different sources, and merge them to generate a unified multi-dimensional health dataset, in which health data includes past medical history data, daily monitoring data and lifestyle record data. S200: Based on a multi-dimensional health dataset, a distributed ledger is constructed using blockchain technology. The distributed ledger records the data access permissions and change logs of each institution. Data access requests are authorized according to preset privacy rules, and a secure and interconnected data sharing flow is established. S300. Extract key features from the secure and interconnected data sharing stream, normalize the key features to obtain a standardized feature vector group, where the key features include blood pressure fluctuation values ​​and activity frequency indicators. S400. Input the standardized feature vector group into the support vector machine model for classification. Mark the high-risk category according to whether the blood pressure fluctuation value exceeds the preset threshold to obtain the preliminary risk level distribution. S500: Combining the preliminary risk level distribution and activity frequency index, the multi-dimensional features are input into the random forest model for integrated analysis. The risk level is adjusted according to whether the activity frequency index is lower than the preset threshold to determine the refined risk classification result. S600: Based on the refined risk classification results, the system uses a rule engine to match medical intervention plans and life guidance plans in the preset care template library to generate a personalized care matching plan. S700 generates feedback reports based on personalized care matching plans and distributes them to relevant institutions. At the same time, it adjusts the secure and interconnected data sharing flow through a real-time update mechanism to optimize the response of integrated medical and elderly care services.

[0008] Further, step S100 includes: S110. Obtain multi-source heterogeneous health data packets of elderly health data from hospitals, communities and elderly care institutions, extract features and reassemble them using a medical information standardization protocol parser to obtain intermediate structured records. S120. Perform medical entity recognition and semantic mapping on the intermediate structured records, replace heterogeneous descriptions with standard codes, and obtain standardized data entries; S130. Based on standardized data entries, perform time-series alignment and anomaly verification, and utilize the features of the multi-dimensional data fusion matrix to generate a unified multi-dimensional health dataset.

[0009] Further, step S200 includes: S210. Obtain a multi-dimensional health dataset, generate an institutional node index, and initialize the distributed ledger; S220. Configure access permission tables for the distributed ledger and generate an access permission state tree containing the latest access permission status. S230: Receive a data access request packet, read the permission status tree for verification, and generate a temporary authorization token; S240. Based on the temporary authorization token, record the change log and establish a point-to-point data sharing channel; S250: Transmit multi-dimensional health datasets through a data sharing channel to complete the construction of a secure and interconnected data sharing stream.

[0010] Further, step S300 includes: S310. Capture raw data streams containing blood pressure fluctuation values ​​and activity frequency indicators in real time from secure and interconnected data sharing streams; S320: The raw data stream is divided by a sliding time window. The standard deviation of the blood pressure reading is calculated as the blood pressure fluctuation value in each window, and the number of times the accelerometer exceeds the preset threshold is counted as the activity frequency index. S330. Based on the maximum and minimum values ​​of blood pressure fluctuations in historical data, perform maximum-minimum normalization on the blood pressure fluctuation values ​​calculated in the current window. S340. Using the 95th percentile of the activity frequency index in historical data as a benchmark, perform quantile normalization on the activity frequency index of the current window statistics. S350. Combine the normalized blood pressure fluctuation values ​​with the activity frequency index in chronological order to form a standardized feature vector group with timestamps.

[0011] Further, step S400 includes: S410. Obtain the standardized feature vector set and input it into the support vector machine model. Construct the optimal hyperplane decision boundary using the radial basis kernel function. S420. Calculate the normal distance from the feature vector point to the decision boundary of the optimal hyperplane, and determine the initial binary classification label based on the normal distance; S430. If the blood pressure fluctuation value in the standardized feature vector group exceeds the preset threshold, ignore the initial binary classification label and mark the sample as a high-risk category. S440. By combining high-risk categories with low- and medium-risk levels mapped based on normal distance, a risk level identifier sequence is generated to obtain a preliminary risk level distribution.

[0012] Further, step S500 includes: S510. Obtain the preliminary risk level distribution and activity frequency indicators, and construct a multi-dimensional feature matrix containing numerical risk weight vectors and activity frequency indicators. S520. Input the multi-dimensional feature matrix into the random forest model, and obtain the aggregated classification probability value through decision tree branching and ensemble voting. S530. Determine whether the activity frequency index is lower than the preset threshold. If it is lower, generate a risk level gain value and weight the aggregated classification probability value. S540. Based on the revised aggregated classification probability values, redefine the risk boundaries and determine the refined risk classification results.

[0013] Furthermore, the medical intervention plan includes medical intervention items, the lifestyle guidance plan includes lifestyle guidance items, and step S600 includes: S610. Obtain the intervention intensity corresponding to the refined risk classification results, and retrieve medical intervention items and life guidance items based on the intervention intensity; S620. Compare medical intervention items with lifestyle guidance items against the disease contraindication map to obtain a conflict detection matrix; S630. If the conflict detection matrix contains mutually exclusive markers, then remove mutually exclusive entries to obtain a set of candidate care entries. S640. Assign execution priority values ​​to the candidate care item set, and assemble the execution priority values ​​with the candidate care item set to generate a personalized care matching scheme.

[0014] Further, step S700 includes: S710. Obtain a personalized care matching plan, generate a encrypted feedback report based on the personalized care matching plan, and push it to the target receiving institution node; S720: Receive an acknowledgment receipt with a node status code. The acknowledgment receipt is generated by the target receiving node after receiving the encrypted feedback report. S730: Adjust the initial data sharing stream according to the node status code to obtain the updated data sharing stream; S740: Adopt the updated data sharing stream to push dynamic monitoring data and optimize the response of integrated medical and elderly care services.

[0015] Another aspect of the present invention relates to a health monitoring and graded care system for the elderly that integrates medical and elderly care, and a method for implementing health monitoring and graded care for the elderly that integrates medical and elderly care, comprising: The multi-dimensional health dataset generation module is used to collect health data of the elderly from hospitals, communities and elderly care institutions. It adopts medical information standardization protocol to convert the format and align the semantics of health data from different sources, and integrates them to generate a unified multi-dimensional health dataset, which includes past medical history data, daily monitoring data and lifestyle record data. The data sharing flow establishment module is used to build a distributed ledger based on a multi-dimensional health dataset using blockchain technology. The distributed ledger records the data access permissions and change logs of each institution, authorizes data access requests according to preset privacy rules, and establishes a secure and interconnected data sharing flow. The standardized feature vector group acquisition module is used to extract key features from the secure and interoperable data sharing stream, normalize the key features, and obtain a standardized feature vector group, in which the key features include blood pressure fluctuation values ​​and activity frequency indicators. The preliminary risk level distribution acquisition module is used to input standardized feature vector groups into the support vector machine model for classification, and to mark high-risk categories according to whether blood pressure fluctuation values ​​exceed preset thresholds, thereby obtaining the preliminary risk level distribution. The refined risk classification result determination module is used to combine the preliminary risk level distribution and activity frequency index, input multi-dimensional features into the random forest model for integrated analysis, adjust the risk level according to whether the activity frequency index is lower than the preset threshold, and determine the refined risk classification result. The personalized care matching solution generation module is used to generate personalized care matching solutions by matching medical intervention solutions and life guidance solutions in the preset care template library with the results of refined risk classification through a rule engine. The data sharing flow synchronization adjustment module is used to generate feedback reports based on personalized care matching plans and distribute them to relevant institutions. At the same time, it synchronizes and adjusts the secure and interconnected data sharing flow through a real-time update mechanism to optimize the response of integrated medical and elderly care services.

[0016] The beneficial effects achieved by this invention are as follows: 1. The method and system for health monitoring and graded care of the elderly that integrates medical and elderly care provided by this invention addresses the unique business scenario of health data management and integrated medical and elderly care services for the elderly, namely, how to achieve standardized integration, secure sharing and refined risk assessment of health data among multiple institutions, and generate personalized care plans, and proposes an integrated solution.

[0017] 2. This invention collects health data of the elderly in hospitals and community elderly care institutions, uses standardized medical information protocols for format conversion and semantic alignment, and integrates them to generate a multi-dimensional health dataset; it uses blockchain technology to construct a distributed ledger to ensure secure recording of data access permissions and change logs, and authorizes data sharing according to privacy rules; it extracts key features such as blood pressure fluctuation values ​​and activity frequency indicators, performs risk classification through support vector machines and random forest models, combines a rule engine to match care templates, generates personalized plans, and provides real-time feedback to relevant institutions.

[0018] 3. This invention significantly improves the accuracy and response efficiency of integrated medical and elderly care services by organically combining data standardization, secure sharing, and intelligent analysis, providing comprehensive support for the health management of the elderly. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an embodiment of the method for health monitoring and graded care of the elderly that integrates medical and elderly care according to the present invention. Figure 2 This is a functional block diagram of an embodiment of the medical and elderly health monitoring and graded care system of the present invention.

[0020] Explanation of icon numbers: 10. Multi-dimensional health dataset generation module; 20. Data sharing stream establishment module; 30. Standardized feature vector group acquisition module; 40. Preliminary risk level distribution acquisition module; 50. Refined risk grading result determination module; 60. Personalized care matching plan generation module; 70. Data sharing stream synchronous adjustment module. Detailed Implementation

[0021] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0022] like Figure 1 As shown, the first embodiment of the present invention proposes a method for health monitoring and graded care of the elderly that integrates medical and elderly care, including the following steps: Step S100: Collect health data of the elderly from hospitals, communities and elderly care institutions. Use medical information standardization protocols to convert the format and align the semantics of health data from different sources, and merge them to generate a unified multi-dimensional health dataset. The health data includes past medical history data, daily monitoring data and lifestyle record data.

[0023] This step involves the collection and standardized fusion of multi-source health data. It integrates data from multiple sources, including hospital electronic medical record systems, community health service platforms, and elderly care institution care record systems, to collect comprehensive health data from the elderly. Specifically, this includes: past medical history data (such as chronic disease diagnoses, surgical history, and allergy history), daily monitoring data (such as real-time / periodic monitoring values ​​of blood pressure, blood sugar, heart rate, and sleep duration), and lifestyle habit records (such as dietary structure, exercise frequency, and medication adherence). Using standardized medical information protocols such as HL7 FHIR (Fast Healthcare Interoperability Resources) and CDA (Clinical Document Architecture), the multi-source health data from different institutions and in different formats (structured / unstructured) undergo format conversion, field mapping, and semantic alignment to eliminate data heterogeneity. After data cleaning to remove invalid, duplicate, and abnormal data, a unified and standardized multi-dimensional health dataset is generated, providing a complete data foundation for subsequent risk analysis.

[0024] Step S200: Based on a multi-dimensional health dataset, a distributed ledger is constructed using blockchain technology. The distributed ledger records the data access permissions and change logs of each institution. Data access requests are authorized according to preset privacy rules to establish a secure and interconnected data sharing flow.

[0025] This step is the construction phase of the data security sharing system. Based on the multi-dimensional health dataset generated in step S100, a decentralized distributed ledger is built using blockchain technology. This ledger permanently records data access permissions (e.g., read-only / modifiable / uploadable) and all data change logs (including operator, operation time, and operation content) for participating parties such as hospitals, communities, and elderly care institutions, ensuring data traceability. Pre-set privacy protection rules (e.g., anonymized access, tiered authorization, and informed consent verification) are used to verify permissions and determine compliance for data access requests initiated by various institutions, authorizing only legitimate requests to access data within the corresponding scope. Ultimately, a secure, traceable, and tamper-proof data sharing flow is established across institutions.

[0026] Data sharing flow refers to the orderly, secure, and traceable end-to-end data transmission and interaction link of health data from the data holder to the authorized access party in a cross-institutional sharing scenario of multi-dimensional health datasets. It is supported by a distributed ledger built with blockchain technology, authorized by preset privacy rules, and with participating institutions (hospitals, communities, elderly care institutions, etc.) as the main data interaction subjects. It is the core carrier for cross-institutional data security interoperability.

[0027] Step S300: Extract key features from the secure and interconnected data sharing stream, normalize the key features to obtain a standardized feature vector group, where the key features include blood pressure fluctuation values ​​and activity frequency indicators.

[0028] This step involves key feature extraction and standardization. From a secure and interconnected data sharing stream, key features with a core impact on assessing health risks in the elderly are selected. These include, but are not limited to, blood pressure fluctuation values ​​(such as daily systolic / diastolic blood pressure fluctuation range and weekly mean deviation) and activity frequency indicators (such as daily duration of independent activity and weekly frequency of outings). Auxiliary features such as blood glucose control level, heart rate variability, and medication patterns are also included. The extracted key features are then normalized (e.g., Min-Max standardization and Z-score standardization) to eliminate analytical biases caused by differences in feature dimensions, generating standardized feature vector sets with uniform numerical ranges that can be directly used for model calculations.

[0029] Step S400: Input the standardized feature vector group into the support vector machine model for classification, and mark the high-risk category according to whether the blood pressure fluctuation value exceeds the preset threshold to obtain the preliminary risk level distribution.

[0030] This step is the preliminary risk level classification stage. The standardized feature vector group obtained in step S300 is input into the pre-trained support vector machine (SVM) classification model. The SVM model is trained based on the health risk grading standard, with blood pressure fluctuation value as the core judgment dimension: if the blood pressure fluctuation value exceeds the preset clinical threshold (e.g., daily systolic blood pressure fluctuation ≥20mmHg), the elderly person is marked as high-risk. Combined with other features, multi-classification calculation is completed, and the output is a preliminary risk level distribution covering three basic levels: low, medium, and high, realizing the preliminary screening of health risks.

[0031] Step S500: Combining the preliminary risk level distribution and activity frequency index, input the multi-dimensional features into the random forest model for integrated analysis, adjust the risk level according to whether the activity frequency index is lower than the preset threshold, and determine the refined risk classification result.

[0032] This step involves fine-tuning the risk level. Integrating the initial risk level distribution with activity frequency indicators, standardized feature vector groups, initial risk labels, and activity frequency indicators are input into a random forest model for integrated analysis. The activity frequency indicator serves as the core adjustment criterion: if the activity frequency indicator is below a preset threshold (e.g., daily self-directed activity time < 1 hour), the corresponding risk level is raised (e.g., medium risk is raised to high risk); conversely, the level is maintained or lowered based on the actual situation. Through multi-feature cross-validation and weight calculation, a refined risk classification result comprising five levels—very low, low, medium, high, and very high—is finally determined, improving the accuracy of risk assessment. For example, a score above 0.8 indicates very high risk, requiring emergency medical intervention; 0.6-0.8 indicates high risk, requiring high-frequency monitoring and professional care; 0.4-0.6 indicates medium risk, requiring urgent regular check-ups and lifestyle interventions; 0.2-0.4 indicates low risk, requiring health guidance and routine monitoring; and below 0.2 indicates very low risk, requiring health maintenance and regular follow-up.

[0033] Step S600: Based on the refined risk classification results, the medical intervention plan and life guidance plan in the preset care template library are matched through the rule engine to generate a personalized care matching plan.

[0034] This step is the personalized care plan generation stage. Based on the refined risk classification results, a preset rule engine is invoked. The rule engine has a built-in care template library (covering medical intervention plans and lifestyle guidance plans corresponding to different risk levels): for high / very high-risk groups, medication adjustment suggestions, high-frequency health monitoring plans, and professional medical care intervention plans are matched; for medium-risk groups, dietary / exercise intervention plans and regular check-up reminders are matched; for low / very low-risk groups, healthy lifestyle guidance plans are matched. The rule engine personalizes the preset care templates according to the specific key characteristics of the elderly (such as blood pressure fluctuation type and activity level) (such as adjusting exercise intensity and optimizing dietary restrictions), and finally generates a personalized care matching plan.

[0035] Step S700: Generate a feedback report based on the personalized care matching plan and distribute it to the corresponding institutions. At the same time, adjust the secure and interconnected data sharing flow through a real-time update mechanism to optimize the response of integrated medical and elderly care services.

[0036] This step involves the implementation of the solution and the optimization of the data closed loop. A structured feedback report is generated from the personalized care matching plan and distributed according to permissions to the corresponding hospitals (for medical intervention), community health service centers (for daily monitoring), and elderly care institutions (for daily living care). Simultaneously, a real-time update mechanism is activated to synchronize the implementation status of the personalized care matching plan and new changes in the elderly's health data to the blockchain distributed ledger, dynamically adjusting the feature data in the secure and interconnected data sharing flow. Through real-time data updates and model iterations, the responsiveness and adaptability of integrated medical and elderly care services are continuously optimized, forming a closed-loop service system of "data collection - risk classification - care intervention - data update".

[0037] Furthermore, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S100 as follows: Step S110: Obtain multi-source heterogeneous health data packets of elderly health data from hospitals, communities and elderly care institutions, extract features and reorganize them using a medical information standardization protocol parser to obtain intermediate structured records.

[0038] Multi-source heterogeneous health data packets are derived using the following formula: (1) In formula (1), This indicates a multi-source heterogeneous health data packet. This represents the health data of the elderly in the hospital. This represents health data of the elderly in the community. This indicates the health data of elderly people in nursing homes. This indicates the acquisition operation. The control logic of formula (1) is to aggregate and integrate the health data of the elderly from three sources—hospitals, communities, and elderly care institutions—through the acquisition operation to form a unified multi-source heterogeneous health data package, thereby realizing the centralized collection and fusion of health data from multiple channels.

[0039] The intermediate structured record is derived using the following formula: (2) In formula (2), This represents an intermediate state structured record. Indicates feature extraction, This indicates a recombination operation. The control logic of formula (2) is to perform feature extraction on multi-source heterogeneous health data, and then reassemble the extraction results to obtain intermediate structured records.

[0040] The extracted features are obtained using the following formula: (3) In formula (3), Indicates feature extraction, This represents multi-source heterogeneous health data. This refers to a standardized protocol for medical information (standard specifications / protocol template). This indicates the use of a parser for extraction. The control logic of formula (3) follows the medical information standard protocol to standardize the feature extraction of multi-source heterogeneous health data, unify the data format, eliminate heterogeneous differences, and provide standardized feature input for subsequent structured recombination and health classification.

[0041] The core of processing multi-source heterogeneous data lies in eliminating data barriers between different institutions and achieving lossless information extraction. Consider an elderly person with chronic cardiovascular disease. Their inpatient medical record at a top-tier hospital might be transmitted based on the HL7V3 standard, containing a complex nested XML structure, while follow-up records at a community health service center are only in simple JSON format, or even just a CSV file exported from a smart mattress provided by a nursing home. A medical information standardization protocol parser, as a key component, can automatically identify these differentiated header information and payload structures, uniformly extracting the hospital's systolic blood pressure of 145 mmHg and the nursing home's systolic blood pressure of 145 mmHg into a key-value pair intermediate record. This masks the differences in underlying technology implementations and provides a consistent input basis for subsequent processing. For example, in the medical entity recognition and semantic mapping stages, elderly health monitoring and tiered care systems need to address the problem of diverse descriptions to achieve semantic uniformity.

[0042] Step S120: Perform medical entity recognition and semantic mapping on the intermediate structured records, replace the heterogeneous descriptions with standard codes, and obtain standardized data entries.

[0043] Standardized data entries are derived using the following formula: (4) In formula (4), Represents standardized data entries. This indicates a replacement function (performing a mapping from heterogeneous description to standard encoding). This indicates a heterogeneous description, a non-standard, multi-source, and inconsistently formatted original description text / field. The standard code is represented. The control logic of formula (4) is to unify multi-source, heterogeneous, and unstructured medical descriptions into a standard code form through standardization replacement, so as to realize data standardization, exchangeability, and computability, and improve data quality and cross-system interoperability.

[0044] The semantic mapping result is obtained through the following formula: (5) In formula (5), This represents the semantic mapping result. Represents a semantic mapping function. Indicates a medical entity. The semantic vocabulary is represented. The control logic of formula (5) is to associate medical entities with the standard semantic vocabulary through semantic mapping, thereby completing the standardization and unification of medical concepts and improving the semantic consistency and understandability of data.

[0045] Medical Entities This can be derived from the following formula: (6) In formula (6), Indicates the identified medical entities, This represents a medical entity recognition function. This represents an intermediate structured record. The control logic of formula (6) takes the intermediate structured record as input, processes it through a medical entity recognition function, automatically identifies and extracts the medical entities contained therein, and finally outputs the set of identified medical entities.

[0046] In the aforementioned intermediate records, the hospital record diagnosed "primary hypertension," while the community doctor's handwritten entry stated "hypertension," even containing spelling errors or abbreviations. Natural language processing (NLP) technology identified "hypertension" as the core entity and used a semantic mapping engine to uniformly map it to the ICD-10 standard code I10. Simultaneously, drug names, such as enteric-coated aspirin and Bayer aspirin, were uniformly mapped to the ATC code B01AC06. This process of converting unstructured natural language into a uniquely standard code that computers can understand eliminates semantic ambiguity, ensures precise equivalence of cross-institutional data in medical meaning, and greatly improves data usability.

[0047] Step S130: Perform time-series alignment and anomaly verification based on standardized data entries, and generate a unified multi-dimensional health dataset by aggregating features using a multi-dimensional data fusion matrix.

[0048] A unified, multi-dimensional health dataset is generated using the following formula: (7) In formula (7), This represents a multi-dimensional health dataset. Represents a multidimensional data fusion matrix. This represents the multidimensional aggregated feature matrix. The control logic of formula (7) is to perform feature aggregation on the multidimensional aggregated feature matrix through the multidimensional data fusion matrix to obtain a standardized and unified multidimensional health dataset.

[0049] Time-series alignment and multidimensional fusion based on standardized data entries are the ultimate steps in constructing a high-quality dataset. Because elderly health data exhibits multi-frequency characteristics—for example, blood pressure may be monitored daily, while biochemical indicators may be monitored monthly—direct merging can lead to dimensional misalignment. The elderly health monitoring and tiered care system performs time-series alignment, constructing a timeline at the daily or hourly granularity, and using linear interpolation or forward padding to fill gaps in low-frequency data. Simultaneously, anomaly detection is performed; if a record shows a body temperature of 30 degrees Celsius or a heart rate of 200 beats / min, the system, considering the context, determines it as sensor detachment or data entry error and removes or marks it. Finally, using a multidimensional data fusion matrix, physiological parameters, medication records, and test results are aggregated into a sparse matrix, where each row represents a time point and each column represents a standardized health feature, generating a unified multidimensional health dataset. This not only provides a solid data foundation for subsequent disease progression prediction but also effectively improves the accuracy of multimodal data analysis.

[0050] Preferably, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S200: Step S210: Obtain a multi-dimensional health dataset, generate an institutional node index, and initialize the distributed ledger.

[0051] The organization node index is generated using the following formula: (8) In formula (8), Indicates the first Institutional node indexes are index values ​​used in a distributed ledger to identify and locate institutional nodes. This represents a hash function, a cryptographic hash operation function that maps inputs to fixed-length hash values. Indicates the first Each organization identifier corresponds to a unique identity information of the organization. The control logic of formula (8) is to use the uniqueness, fixed length output and anti-collision characteristics of the hash function to map the organization identifier to a unique, fixed-length organization node index, so as to realize the unique identification and fast addressing of the organization node in the distributed ledger.

[0052] The initial distributed ledger is obtained using the following formula: (9) In formula (9), Represents the initial distributed ledger, a distributed ledger instance that has been initialized, and a ledger carrier used to store basic data. This represents the genesis function, a dedicated mapping function that initializes the distributed ledger and completes the generation of the initial ledger structure. The root hash value represents the hash digest of the initial state of the ledger, used to anchor the initial data characteristics. The control logic of formula (9) is to use the root hash value to anchor the initial state of the ledger, ensuring the uniqueness, integrity and anti-tampering foundation of the initial ledger; to establish the initial trust foundation of the distributed ledger, and to support data consistency management in a multi-node collaborative environment.

[0053] After constructing a comprehensive multi-dimensional health dataset, the elderly health monitoring and tiered care system first needs to address the issues of data trust and identity anchoring across institutions. Assuming a municipal hospital, community rehabilitation center, and several nursing homes act as independent data holders, the system generates a unique institutional node index for each participant; for example, the municipal hospital is labeled as node A, and the community rehabilitation center as node B. A distributed ledger is then initialized based on a hash algorithm. This initial distributed ledger does not store massive amounts of raw medical data, but rather stores the data's metadata fingerprints and access policies, ensuring data asset ownership is established without compromising patient privacy.

[0054] Step S220: Configure access permission tables for the distributed ledger and generate an access permission state tree containing the latest access permission status.

[0055] The permission status tree is derived using the following formula: (10) In formula (10), Represents the permission status tree. This represents the total number of permission entries in the distributed ledger. Indicates the first Access permission table, Indicates the first Latest permission status of the item Representing the union and aggregation operations of tree structures. The formula (10) represents the state binding operation, which is used to generate a permission state tree containing the latest permission state. The control logic of formula (10) is to bind the scattered permission tables with their respective latest states, and then aggregate them into a tree structure to form a unified, complete, and traceable permission state tree, thereby realizing the structured management and fast query of permission states.

[0056] The latest permission status is obtained using the following formula: (11) In formula (11), Indicates the latest permission status. Represents the timestamp of a distributed ledger. Indicates time Access permission table, Indicates time Distributed ledger data, This indicates that the maximum effective state is selected. The formula (11) represents state fusion and is used to update the input of the permission state tree. The control logic of formula (11) combines dynamic data in the time dimension to complete the entire process calculation of permission association, fusion processing and optimal state selection, and generates the latest and effective permission state in real time, providing core data support for the dynamic update and maintenance of the permission state tree.

[0057] Configuring access permission tables for distributed ledgers is a core step in ensuring secure data flow. The elderly health monitoring and tiered care system constructs an access permission state tree, a data structure similar to a Merkle tree, where each leaf node represents a specific access permission state. For example, node B is only allowed to access anonymized data related to hypertension management in the elderly from node A, but not to view specific genetic testing reports. When node A modifies its sharing policy, such as adding sharing authorization for a certain biochemical indicator, the root hash value of the access permission state tree is updated immediately, ensuring that all nodes in the network are synchronously aware of the latest access permission state. This mechanism effectively prevents the risk of access being tampered with or expiring.

[0058] Step S230: Receive the data access request packet, read the permission status tree for verification, and generate a temporary authorization token.

[0059] The following formula is used to generate a temporary authorization token: (12) In formula (12), This indicates a temporary authorization token. Indicates the authorization status. The generation operation is indicated. The control logic of formula (12) is based on the verification result of the permission state tree to generate a temporary authorization token that is bound to the authorization state and has timeliness and permission constraints, so as to realize secure and fine-grained data access control in the distributed ledger environment and ensure the legality and controllability of medical and elderly care data sharing.

[0060] When a research institution initiates a data access request, the elderly health monitoring and tiered care system does not directly transmit the data. Instead, it first reads the permission state tree for verification. For example, if the research institution requests historical heart rate data from an elderly person, the system verifies that the digital signature matches the current permission tree state and generates a temporary authorization token with a time limit. This temporary authorization token contains the permitted data range, its validity period (e.g., 30 minutes), and a one-time encryption key.

[0061] Step S240: Based on the temporary authorization token, record the change log and establish a point-to-point data sharing channel.

[0062] Record change logs using the following formula: (13) In formula (13), This indicates the change log. Indicates the first Temporary authorization token, Indicates the first Each single data change This represents a chained XOR operation. Indicates the join operation. The formula represents the number of changes. It is used to record multiple change logs based on the temporary authorization token to form a complete log chain. The control logic of formula (13) is to bind the authorization token and the change data through a connection operation, and then form a log chain through a chain-like XOR operation to realize the full-link traceability and anti-tampering audit of medical and elderly care data sharing operations, and ensure the compliance and accountability of data access.

[0063] The point-to-point data sharing channel is derived using the following formula: (14) In formula (14), Indicates a point-to-point data sharing channel. and This represents two peer nodes. This represents a digital signature function. This indicates a temporary authorization token. The formula (14) represents the change log and is used to establish a secure point-to-point data sharing channel based on the temporary authorization token and the change log. The control logic of the formula (14) is based on a triple security mechanism of temporary authorization + digital signature + change log to establish a secure point-to-point data sharing channel in the medical and elderly care scenario: 1. It establishes a secure point-to-point data sharing channel in the medical and elderly care scenario through a temporary authorization token. 1. Strictly limit the scope and duration of sharing to avoid the risk of long-term access leakage; 2. Also, use digital signatures. 3. Ensure the authenticity and integrity of data and authorizations to prevent forgery and tampering; 4. Simultaneously, through change logs. It enables full-chain traceability of operations, meets the security and compliance requirements of medical and health data, and perfectly adapts to the multi-party collaborative data sharing needs in the context of integrated medical and elderly care (such as health data exchange between institutions and access to care plans for families).

[0064] Establishing a peer-to-peer data sharing channel based on a temporary authorization token is key to achieving efficient transmission. The elderly health monitoring and tiered care system records change logs based on the token, immutably writing the access behavior into a distributed ledger, and then establishing a direct P2P (Peer-to-Peer) encrypted channel between the requester and the data holder.

[0065] Step S250: Transmit multi-dimensional health datasets through the data sharing channel to complete the construction of a secure and interconnected data sharing stream.

[0066] The data sharing flow is constructed using the following formula: (15) In formula (15), Indicates a data sharing stream. Indicates time A multidimensional health dataset, Indicates time Point-to-point data sharing channel, This indicates the process of constructing a secure and interconnected data sharing flow. The control logic of formula (15) is to continuously integrate and accumulate the multi-dimensional health dataset and the point-to-point sharing channel in the time dimension through time-domain integration, thereby constructing a continuous, traceable, secure and controllable data sharing flow and realizing the orderly, efficient and secure interconnection of medical and health data in cross-institutional scenarios.

[0067] Transmitting multi-dimensional health datasets through this P2P encrypted channel avoids the risk of data leakage that may occur when data passes through third-party relay servers. At the same time, by utilizing the non-repudiation of distributed ledgers, it enables full-process auditing and tracking of every medical data call, thereby completing the construction of a secure and interconnected data sharing flow.

[0068] Furthermore, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S300 as follows: Step S310: Capture the raw data stream containing blood pressure fluctuation values ​​and activity frequency indicators in real time from the secure and interconnected data sharing stream.

[0069] Based on the established secure data channel, the elderly health monitoring and tiered care system begins processing the continuous stream of physiological signals from the secure and interconnected data sharing flow. For example, when analyzing an elderly person with chronic hypertension, the system does not analyze each millisecond-level raw data point in isolation. Instead, it employs a sliding time window mechanism to capture the dynamic changes in the data. For instance, a 30-minute time window is set, sliding forward in 5-minute increments. This means that data from each moment is included in multiple overlapping analysis periods, ensuring continuous capture of sudden health events.

[0070] Step S320: The raw data stream is divided using a sliding time window. The standard deviation of the blood pressure readings is calculated as the blood pressure fluctuation value within each window, and the number of times the accelerometer exceeds the preset threshold is counted as the activity frequency index.

[0071] Blood pressure fluctuation values ​​are calculated using the following formula: (16) In formula (16), Indicates the first Blood pressure fluctuation values ​​within a sliding time window Indicates the first The number of blood pressure readings within each window. Indicates the first The first window One blood pressure reading, Indicates the first The average value of blood pressure readings within a window. The control logic of formula (16) uses a sliding time window as the analysis unit and quantifies the dispersion of blood pressure readings in a short period of time through standard deviation, objectively reflecting the dynamic fluctuation level of blood pressure in the elderly, and providing quantifiable core indicators for blood pressure abnormality monitoring and health risk early warning.

[0072] The activity frequency index is derived using the following formula: (17) In formula (17), Indicates the first Activity frequency index within a sliding time window Indicates the first Number of accelerometer data points within a window Indicates the first The first window One accelerometer reading, Indicates the preset threshold. The indicator function is 1 when the condition is met and 0 otherwise. The control logic of formula (17) uses a sliding time window as the analysis unit and uses threshold filtering + indicator function counting to quantify the number of effective accelerations generated by the elderly in a short period of time, objectively reflecting their activity intensity and frequency, and providing quantifiable behavioral characteristic indicators for health risk assessment (such as the risk of prolonged sitting and early warning of abnormally intense activities).

[0073] Within each sliding time window, the elderly health monitoring and tiered care system first focuses on the stability analysis of blood pressure data. Unlike simply focusing on the absolute values ​​of systolic or diastolic blood pressure, the system calculates the standard deviation of all blood pressure readings within that window. This blood pressure fluctuation value can sensitively reflect the severity of short-term fluctuations in the elderly person's blood pressure, such as oscillations after medication or during emotional excitement. Simultaneously, regarding activity frequency, the system monitors the data stream from the accelerometer, counting the number of times its value exceeds a preset threshold. This effectively filters out minute vibrations during sleep or at rest, recording only meaningful physical activities such as walking and standing.

[0074] Step S330: Based on the maximum and minimum values ​​of blood pressure fluctuations in historical data, perform maximum-minimum normalization on the blood pressure fluctuation values ​​calculated in the current window.

[0075] The normalized blood pressure fluctuation value is obtained by the following formula: (18) In formula (18), This represents the normalized blood pressure fluctuation value after processing. Indicates the first Blood pressure fluctuation values ​​within a sliding time window This represents the minimum historical blood pressure fluctuation value. This represents the maximum value of historical blood pressure fluctuations. The control logic of formula (18) is to linearly map the blood pressure fluctuation values ​​of different time windows and different individuals to the unified interval [0, 1] through min-max scaling, thereby eliminating the difference in dimensions and numerical ranges and realizing the standardized expression of the degree of blood pressure fluctuation, which is convenient for subsequent health risk classification, model training and cross-individual / cross-time comparative analysis.

[0076] To eliminate the influence of dimensional differences caused by individual variations among elderly individuals, normalization is an essential step. For blood pressure fluctuation values, the elderly health monitoring and tiered care system retrieves extreme value records from the elderly person's historical database and uses a max-min normalization algorithm to map the current fluctuation standard deviation to the range of 0 to 1. This processing method ensures that the fluctuation amplitude is comparable across individuals, regardless of whether their baseline blood pressure is high or low.

[0077] Step S340: Using the 95th percentile of the activity frequency index in historical data as a benchmark, perform quantile normalization processing on the activity frequency index of the current window statistics.

[0078] The normalized activity frequency index is obtained using the following formula: (19) In formula (19), This represents the current window activity frequency index after quantile normalization. This indicates the activity frequency metric for the current window. The 95th percentile of the activity frequency index in historical data is used as the benchmark. The control logic of formula (19) is to normalize the activity frequency of the current window to the proportion of the relatively high frequency level based on the 95th percentile of the historical activity frequency. This not only preserves the relative fluctuation characteristics of the activity frequency, but also intuitively reflects whether the current activity exceeds the historical common high frequency range, which is suitable for the activity monitoring scenario of the elderly (such as identifying abnormally intense activities or the risk of sedentary low activity), while eliminating the difference in the numerical range of activity frequency of different individuals and different time periods.

[0079] 95th percentile benchmark of activity frequency indicators in historical data This can be derived from the following formula: (20) In formula (20), The 95th percentile benchmark represents the frequency of activity in historical data. Indicates the candidate threshold. This indicates the number of samples for an activity frequency indicator in historical data. Indicates the first in historical data Each activity frequency index value, Indicates an indicator function, Represents the 95th percentile. This indicates finding the minimum value that satisfies the condition. The control logic of formula (20) accurately calculates the 95th quantile of the historical activity frequency index through empirical distribution function + infimum operation. As a robust benchmark for subsequent activity frequency normalization: ① It retains the distribution characteristics of historical data while avoiding interference from extreme outliers (such as occasional violent activities); ② The 95th percentile represents "the frequency of 95% of historical activities does not exceed this value", which can be used as a critical level to judge whether the current activity belongs to an abnormally high frequency.

[0080] When processing activity frequency, considering that elderly individuals may occasionally experience extreme abnormalities such as falls or violent shaking, directly using the maximum value for normalization could lead to excessive compression of normal activity data. Therefore, the elderly health monitoring and tiered care system selects the 95th percentile of the activity frequency index in historical data as the benchmark. Activity frequencies below this 95th percentile are linearly mapped, while extreme data above this 95th percentile are truncated or specially labeled, thus ensuring that the feature vector accurately represents daily activity patterns.

[0081] Step S350: Combine the normalized blood pressure fluctuation value and the activity frequency index in chronological order to form a standardized feature vector group with timestamps.

[0082] The standardized feature vector set with timestamps is obtained by the following formula: (twenty one) In formula (21), Represents a standardized set of feature vectors. Indicates the first Normalized blood pressure fluctuation values ​​over time. Indicates the first Activity frequency index at any time Indicates the first Normalized blood pressure fluctuation values ​​over time. Indicates the first Activity frequency index at any time. The control logic of formula (21) is to integrate the time-series physiological characteristics (blood pressure fluctuations) and behavioral characteristics (activity frequency) into a structured matrix, which not only preserves the continuity of the time dimension, but also eliminates the difference in dimensions through normalization, realizing the unified expression of multi-dimensional health information of "time-physiology-behavior", and providing a standardized input carrier for the time-series analysis and intelligent early warning of health risks of the elderly.

[0083] Ultimately, these meticulously processed feature values ​​are aligned by timestamps and encapsulated into standardized feature vector sets, providing high-quality input data for subsequent health risk prediction models.

[0084] Preferably, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S400: Step S410: Obtain the standardized feature vector set and input it into the support vector machine model. Construct the optimal hyperplane decision boundary using the radial basis kernel function.

[0085] The optimal hyperplane decision boundary constructed using the radial basis kernel is defined by the following formula: (twenty two) In formula (22), Represents the decision function. Represents a standardized set of feature vectors. Indicates the first A standardized set of feature vectors Indicates the number of support vectors. Indicates the first Each support vector weight Indicates the first Support vector labels Represents the radial basis kernel function. The formula (22) defines the optimal hyperplane decision boundary constructed using the radial basis kernel, with the zero point representing the classification boundary. The control logic of formula (22) is a nonlinear decision model based on support vector machine (SVM) + radial basis kernel (RBF), which maps standardized time-series health characteristics (blood pressure fluctuation + activity frequency) to a high-dimensional space, constructs the optimal hyperplane decision boundary, and achieves accurate identification of health risks in the elderly.

[0086] After obtaining the normalized and standardized feature vector set, the elderly health monitoring and tiered care system inputs it into a support vector machine (SVM) model for deep analysis. Considering the complex nonlinear relationship between human physiological parameters and health risks, simple linear classifiers struggle to accurately capture the subtle boundaries. Therefore, the elderly health monitoring and tiered care system uses a radial basis function (RBF) kernel to map the low-dimensional feature data to a high-dimensional space. In this high-dimensional space, the SVM model constructs an optimal hyperplane decision boundary that maximizes the class margin, thereby distinguishing between healthy states and potential risk states.

[0087] Step S420: Calculate the normal distance from the feature vector point to the decision boundary of the optimal hyperplane, and determine the initial binary classification label based on the normal distance.

[0088] The normal distance from the eigenvector point to the decision boundary of the optimal hyperplane is obtained by the following formula: (twenty three) In formula (23), Indicates the normal distance. Represents the hyperplane normal vector. This represents the transpose of the hyperplane normal vector. Representing feature vector points, Indicates hyperplane offset. The normal vector magnitude is represented. The control logic of formula (23) realizes the geometric distance quantization from the feature vector point to the decision boundary of the optimal hyperplane: 1. It completes the binary classification direction determination through the sign of the molecule, and also through the normalized distance. 1. Measuring the distance of a sample from the decision boundary provides a dual output of category label and confidence level for health risk classification; 2. The geometric distance expression is intuitive and easy to understand, providing an interpretable quantitative basis for risk grading and manual review in medical and elderly care scenarios.

[0089] The initial binary classification labels determined based on the normal distance are obtained using the following formula: (twenty four) In formula (24), Indicates the initial binary classification label. Indicates the normal distance. The decision function symbol is represented by a positive distance corresponding to label 1 and a negative distance corresponding to label -1. The control logic of formula (24) is that the formula combines the magnitude (confidence) of the normal distance with the symbol (classification direction) of the decision function to realize the mapping from geometric distance to binary classification label: 1. It clarifies the risk category of the sample through the symbol and implies the classification confidence through the product magnitude (the larger the magnitude, the more certain the risk judgment); 2. It provides a clear and interpretable binary classification result for the health risk classification of the elderly in the medical and elderly care integration scenario, which is convenient for the formulation of subsequent care strategies.

[0090] To quantify the certainty of risk, the elderly health monitoring and tiered care system does not simply output a binary result, but further calculates the normal distance from each feature vector point to the decision boundary of the optimal hyperplane. The magnitude of this normal distance directly reflects the reliability of the classification and the severity of the risk. For example, when a feature vector falls on the healthy side of the decision boundary and is far away, it means that the elderly person's physiological state is very stable at that moment, and the elderly health monitoring and tiered care system classifies it as low risk. Conversely, when a feature vector is on the risk side but close to the boundary, it suggests a sub-healthy or borderline state, and the elderly health monitoring and tiered care system maps it to a medium-risk level. This fine-grained classification based on distance makes health monitoring no longer a black-and-white judgment, but rather possesses the ability to perceive continuous risk gradients.

[0091] Step S430: If the blood pressure fluctuation value in the standardized feature vector group exceeds the preset threshold, then ignore the initial binary classification label and mark the sample as a high-risk category.

[0092] Relying solely on statistical patterns from machine learning models can lead to misjudgments in extreme cases, particularly for elderly patients with hypertension, where drastic fluctuations in physiological indicators are often more dangerous than overall trends. Therefore, the elderly health monitoring and tiered care system introduces a rule-based circuit breaker mechanism. Assuming that within a certain analysis period, even if the comprehensive feature vector is determined to be within a safe zone in the SVM model (e.g., although blood pressure fluctuates greatly, activity frequency is normal, resulting in the vector point still being within a low-risk range from the decision boundary), if the blood pressure fluctuation value alone exceeds a preset extreme value threshold, the elderly health monitoring and tiered care system will forcibly ignore the initial binary classification label output by the SVM. This logic ensures absolute sensitivity to sudden hypertensive crises, directly labeling such samples as high-risk, thus preventing the algorithm from missing crucial warning signals due to "averaging" features.

[0093] Step S440: Combine the high-risk category with the low- and medium-risk levels mapped by normal distance to generate a risk level identifier sequence to obtain a preliminary risk level distribution.

[0094] The risk level identifier sequence is generated using the following formula: (25) In formula (25), This indicates a sequence of risk level identifiers. Indicates a high-risk category. This represents the mapping function for low to medium risk levels. Indicates the normal distance. The formula represents the set union operation. It combines high-risk categories and low-to-medium risk levels to generate a risk level identifier sequence to obtain a preliminary risk level distribution. The control logic of formula (25) combines binary high-risk identification with low-to-medium risk fine classification driven by normal distance. It breaks through the limitations of traditional binary classification and constructs a three-level risk level sequence of low / medium / high: 1. It retains the accurate identification of high-risk samples and ensures timely warning of emergency risks; 2. It also performs fine classification of non-high-risk samples, distinguishing between medium-risk that needs to be strengthened and low-risk that is routinely monitored, perfectly adapting to the graded care needs in the medical and elderly care integration scenario.

[0095] The final risk assessment result is not an isolated point, but rather a continuous sequence of risk level identifiers generated chronologically by combining the aforementioned mandatory high-risk markers with low- and medium-risk levels mapped based on normal distance. By observing this sequence of risk level identifiers, nursing staff can not only see the current risk status, but also trace back how the risk gradually evolved from low risk to medium risk, ultimately triggering a high-risk alarm, thus obtaining a preliminary and complete view of the risk level distribution.

[0096] Furthermore, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S500: Step S510: Obtain the preliminary risk level distribution and activity frequency indicators, and construct a multi-dimensional feature matrix containing numerical risk weight vectors and activity frequency indicators.

[0097] The elderly health monitoring and tiered care system receives preliminary risk level distribution data from previous stages and transforms it into numerical risk weight vectors. For example, low risk, medium risk, and high risk are mapped to weight values ​​of 0.2, 0.5, and 0.9, respectively. Simultaneously, the system extracts activity frequency indicators for the elderly within the current monitoring period, such as the number of steps per hour or the number of positional changes. These numerical risk weight vectors are then concatenated with the activity frequency indicators to construct a multi-dimensional feature matrix.

[0098] Step S520: Input the multi-dimensional feature matrix into the random forest model, and obtain the aggregated classification probability value through decision tree branching and ensemble voting.

[0099] The aggregated classification probability value is obtained using the following formula: (26) In formula (26), This represents the aggregated classification probability value. This represents the number of decision trees in a random forest. Indicates the first The classification probability output of each decision tree. The formula (26) describes the aggregated classification probability obtained by the ensemble voting operation, which represents the multi-dimensional feature matrix. The control logic of the formula (26) is to aggregate the independent probability outputs of multiple decision trees into a robust final classification probability through the ensemble average voting mechanism of random forest: 1. It retains the fitting ability of a single decision tree to nonlinear health features, and eliminates the overfitting and random fluctuations of a single tree through multi-tree averaging, thereby improving the generalization ability of the model; 2. The output probability vector can be directly used to quantify the risk confidence (e.g., the higher the high-risk probability, the more certain the model's risk judgment of the sample), providing a quantifiable basis for graded care in the medical and elderly care scenario.

[0100] Subsequently, this multi-dimensional feature matrix is ​​input into a random forest model. The random forest model contains multiple independent decision trees, each branching based on a different subset of features. For example, one decision tree might focus on analyzing the fluctuation trend of the risk weight vector, while another tree might focus on analyzing the distribution characteristics of activity frequency. After each decision tree independently outputs its classification results, the elderly health monitoring and tiered care system uses an integrated voting operation to statistically analyze the predictive tendencies of all trees, thereby calculating an aggregated classification probability value between 0 and 1. This aggregated classification probability value reflects the comprehensive probability that the elderly person is currently in a higher-risk state.

[0101] Step S530: Determine whether the activity frequency index is lower than the preset threshold. If it is lower, generate a risk level gain value and perform weighted correction on the aggregated classification probability value.

[0102] The risk level gain value is generated using the following formula: (27) In formula (27), Indicator of activity frequency Indicates the preset threshold. This indicates the risk level gain value. denoted by scaling factor, this formula is used to determine the proportional risk level gain value when the activity frequency index is lower than a preset threshold. The control logic of formula (27) is a refined gain mechanism for the low activity (sedentary / bedridden) risk of the elderly: 1. Proportional risk gain is generated only when the activity frequency is lower than a preset threshold (indicating low activity states such as sedentary or bedridden state) - the less activity and the more it is lower than the threshold, the greater the gain, which accurately reflects the potential health risks (such as thrombosis, muscle atrophy, fall risk, etc.) brought about by low activity; 2. The gain is 0 during normal activity, which does not interfere with the risk judgment of normal samples, which not only makes up for the problem of insufficient identification of low activity risk by the general model, but also ensures the robustness and rationality of risk judgment.

[0103] The corrected aggregate classification probability value is obtained using the following formula: (28) In formula (28), This represents the aggregated classification probability value. This represents the corrected aggregate classification probability value. Indicator of activity frequency Indicates the preset threshold. The formula directly adjusts the aggregated classification probability value proportionally based on the activity frequency index being lower than the preset threshold. The control logic of formula (28) is a refined gain mechanism for the risk of low activity (sedentary / bedridden) in the elderly: 1. Proportional risk gain is generated only when the activity frequency is lower than the preset threshold (indicating low activity states such as sedentary or bedridden). The less activity and the more it is lower than the threshold, the greater the gain, which accurately reflects the potential health risks (such as thrombosis, muscle atrophy, and fall risk) brought about by low activity; 2. The gain is 0 during normal activity, which does not interfere with the risk judgment of normal samples. This not only makes up for the problem of insufficient identification of low activity risk by the general model, but also ensures the robustness and rationality of risk judgment.

[0104] Considering the unique physiological characteristics of the elderly, an abnormal decrease in activity frequency is often a hidden precursor to changes in health conditions. The elderly health monitoring and tiered care system continuously assesses whether the current activity frequency falls below a preset threshold. Assuming the preset threshold is 15 postural changes per hour, if an elderly person's activity frequency is detected to be only 5 times per hour, the system determines that it is below the threshold and generates a risk level gain value, for example, 0.15. The elderly health monitoring and tiered care system then adds this gain value to the aforementioned aggregated classification probability value for weighted correction. For example, if the original aggregated classification probability value is 0.60, after weighted correction, it will become 0.75.

[0105] Step S540: Redefine the risk boundaries based on the corrected aggregated classification probability values ​​to determine the refined risk classification results.

[0106] The refined risk classification results are determined using the following formula: (29) In formula (29), This indicates the results of refined risk classification. This represents the corrected aggregate classification probability value. Indicates the first New risk thresholds Indicates the total number of risk levels. This represents the boundary level coefficient. The control logic of formula (29) is to adjust the aggregated classification probability value. Mapped to discrete multi-level risk levels This enables refined stratification of health risks: 1. Breaking through the limitations of traditional binary classification (high / normal), it subdivides the risk level into multiple levels (such as low / medium / high / extremely high), which better meets the differentiated care needs in medical and elderly care scenarios; 2. The stratification logic is intuitive and explainable: the higher the probability, the higher the matching threshold level, and the higher the final risk level, which makes it easier for medical staff to quickly understand the risk level and formulate precise care strategies.

[0107] No. New risk threshold This can be derived from the following formula: (30) In formula (30), Indicates the first New risk thresholds The mean of the aggregation probability is represented. The standard deviation represents the aggregation probability. This represents the threshold level coefficient. The control logic of formula (30) is that the formula dynamically generates multi-level risk thresholds based on the statistical characteristics (mean + standard deviation) of the probability distribution, replacing the traditional method of manually fixing thresholds: 1. It ensures that the thresholds conform to the actual distribution of current health data, avoiding the subjectivity of manual thresholds; 2. It also... The linear offset allows the risk thresholds at each level to increase in an orderly manner with the risk level, forming a clear risk gradient and providing a scientific and adaptive threshold basis for refined risk classification.

[0108] The elderly health monitoring and tiered care system redefines risk boundaries based on the revised aggregated classification probability values. Before revision, a probability of 0.60 corresponded to a moderate risk range, but after revision to 0.75, this aggregated classification probability value crossed the 0.70 high-risk threshold set in the refined tiering. Therefore, the elderly health monitoring and tiered care system determined the elderly person's final condition to be at the refined high-risk level. Through this processing logic, the elderly health monitoring and tiered care system completes the adjustment of risk assessment scales and outputs refined risk tiering results even when physiological indicators such as blood pressure have not shown drastic fluctuations, but behavioral activity exhibits slowness.

[0109] Preferably, the integrated medical and elderly health monitoring and graded care method provided in this embodiment includes a medical intervention plan with medical intervention items and a lifestyle guidance plan with lifestyle guidance items. Step S600 includes: Step S610: Obtain the intervention intensity corresponding to the refined risk classification results, and retrieve medical intervention items and life guidance items based on the intervention intensity.

[0110] The retrieved medical intervention entries are derived using the following formula: (31) In formula (31), This indicates the total number of medical intervention items. Indicates the intensity of intervention. Indicates the first Matching factors for each medical intervention item This represents medical intervention items. The control logic of formula (31) is to transform the abstract risk classification result into an executable medical intervention item matching quantification index through the weighted average calculation of intervention intensity and item matching factor, thereby achieving: 1. Strong binding between intervention intensity and risk level: higher risk → stronger intervention. 1. Larger weighted matching value indicates a greater bias towards high-intensity intervention items; 2. Item suitability quantification: Matching factor 3. Accurately characterize the fit between items and intervention needs, avoiding the subjectivity of manual matching; 4. Overall matching degree assessment: average matching degree It can be directly used to guide the selection of intervention items, providing a scientific basis for precise intervention in medical and elderly care scenarios.

[0111] The retrieved lifestyle guidance entries are derived from the following formula: (32) In formula (32), This indicates the total number of life guidance items. Indicates the intensity of intervention. Indicates the first The guidance coefficient of each life guidance item. This represents the lifestyle guidance items. The control logic of formula (32) complements that of formula (31) for medical intervention items. By using a weighted average of the intervention intensity and the matching factor of the lifestyle guidance items, the abstract risk classification result is transformed into an executable quantifiable indicator for lifestyle guidance matching, achieving: 1. Strong binding between intervention intensity and risk level: the higher the risk → the stronger the intervention. The larger the weighted matching value, the more it favors high-intensity lifestyle guidance items (such as recommending stricter diet / exercise management for high-risk situations); 2. Item suitability quantification: guidance coefficient 3. Accurately characterize the fit between life guidance items and intervention needs, avoiding the subjectivity of manual matching; It can be directly used to guide the selection of life guidance items, providing a scientific basis for refined care in medical and elderly care scenarios.

[0112] The elderly health monitoring and tiered care system first determines the high-risk level based on the refined results of the previous steps, and then obtains the corresponding strong intervention intensity level through preset mapping rules. Assuming the current risk level is determined to be Level 1 high risk, the system converts this into a corresponding intervention intensity label, such as intensity level A, and retrieves matching intervention strategies from the database accordingly. The search results include medical intervention items such as "take fast-acting antihypertensive medication immediately" and lifestyle guidance items such as "perform 30 minutes of assisted standing training."

[0113] Step S620: Compare the medical intervention items with the lifestyle guidance items against the disease contraindication map to obtain the conflict detection matrix.

[0114] The corresponding elements of the collision detection matrix are obtained using the following formula: (33) In formula (33), Represents the corresponding element of the collision detection matrix. This indicates the index number of the medical intervention entry. Indicates the index number of the life guidance item. This indicates the total number of entries in the disease contraindication map, and the total number of all contraindications included in the disease contraindication map (such as drug contraindications, dietary contraindications, exercise contraindications, etc.). The diagram shows the medical intervention items and contraindications. The comparison value of the item, the first The medical intervention item and the first The degree of association between the taboo items is calculated; a positive value indicates association (e.g., 1 / 0.5), while a value of 0 indicates no association. This section represents the entries for life guidance and taboo charts. The comparison value of the item, the first The first life guidance item and the first The degree of association between the taboo items is calculated; a positive value indicates association (e.g., 1 / 0.5), while a value of 0 indicates no association. The disease contraindication chart is shown in section number 1. The taboo coefficient of the item, the first The severity weight of each contraindication item is as follows: high contraindication → 1, medium contraindication → 0.7, low contraindication → 0.3, no contraindication → 0. The control logic of formula (33) is to accurately quantify the potential conflict between medical intervention items and life guidance items by comparing and weighting the disease contraindication map: 1. Conflict will only occur when both items touch the same contraindication item at the same time, avoiding misjudgment across contraindications; 2. Contraindication coefficient The severity of different contraindications is differentiated, making the conflict contribution of high-risk contraindications more prominent, which is in line with the priority of medical safety; 3. The final conflict matrix can be directly used to avoid conflicts in intervention programs (such as replacing conflict items and adjusting the intensity of intervention), ensuring the safety and rationality of medical and elderly care interventions.

[0115] The elderly health monitoring and tiered care system uses a disease contraindication graph to perform deep security checks on the retrieved entries. This disease contraindication graph stores the associations between various diseases and interventions, especially contraindications, in the form of a knowledge graph. If an elderly person's health record includes a history of "orthostatic hypotension," a strong contraindication link exists between "assisted standing training" and "orthostatic hypotension" in the disease contraindication graph. After comparing the retrieved entries with the disease contraindication graph data, the elderly health monitoring and tiered care system constructs a conflict detection matrix. If a contraindication link is found, the corresponding position in the conflict detection matrix is ​​marked as mutually exclusive, for example, assigned a value of 1.

[0116] Step S630: If there are mutually exclusive markers in the conflict detection matrix, remove the mutually exclusive entries to obtain a set of candidate care entries.

[0117] The candidate care item set is derived using the following formula: (34) In formula (34), This represents the set of candidate care items. Represents a set of items. Represents a single entry. Represents the collision detection matrix. Indicates a mutual exclusion flag. The empty set is represented. The control logic of formula (34) is based on the intersection operation of set theory, which realizes the accurate screening from the initial set of intervention items to the set of candidate care items without mutual exclusion conflicts: 1. Using the conflict detection matrix Based on this, mutually exclusive and conflicting items are precisely eliminated to avoid contradictory combinations between medical intervention and lifestyle guidance (such as recommending both prohibited drugs and prohibited foods at the same time); 2. Safe items without conflicts are retained to provide a compliant and safe basis for the subsequent final care plan. This is a key safety verification step from "recommendation" to "implementation" of medical and elderly care intervention plans.

[0118] Once a conflict detection matrix contains mutually exclusive markers, the elderly health monitoring and tiered care system immediately executes elimination logic, removing "assisted standing training" from the list and retaining the non-conflicting items as the candidate care item set. Subsequently, the system assigns a specific execution priority value to each candidate item based on its clinical urgency and contribution to risk mitigation. For example, "immediate administration of fast-acting antihypertensive drugs," which is directly related to stabilizing vital signs, is assigned a high priority value of 0.95, while "bedridden head elevation care," as an alternative gentle intervention, is assigned a priority value of 0.70.

[0119] Step S640: Assign execution priority values ​​to the candidate care item set, and assemble the execution priority values ​​and the candidate care item set to generate a personalized care matching scheme.

[0120] Personalized care matching plans are generated using the following formula: (35) In formula (35), Indicates the number of elements in the candidate care item set. Indicates the first The execution priority value of each entry. Indicates the first The attribute values ​​of each candidate care item. This represents the personalized care matching plan (or its comprehensive quantitative score), characterizing the overall priority and suitability of the plan, and can also generate the execution order by sorting the comprehensive scores of the items. The control logic of formula (35) is to realize the transformation from a safe candidate item set to an executable personalized care plan by weighted summation of execution priority (urgency) + item attributes (suitability): 1. It ensures that high-risk and high-urgency intervention items are executed first, while taking into account the actual feasibility of the elderly, avoiding the recommendation of high-intensity interventions that cannot be implemented; 2. Quantitative score It allows for an intuitive assessment of the overall suitability of the plan, while the sorted order of items provides caregivers with clear implementation guidelines, improving the efficiency and effectiveness of medical and elderly care interventions.

[0121] Ultimately, the elderly health monitoring and tiered care system sequentially assembles these priority-valued items to generate a personalized care matching plan. In this personalized care matching plan, high-priority medical interventions are placed at the top of the execution sequence, ensuring that caregivers address key risks first, thereby effectively mitigating medical risks caused by individual differences while ensuring the effectiveness of interventions.

[0122] Furthermore, the method for health monitoring and graded care of the elderly integrating medical and elderly care provided in this embodiment includes step S700 as follows: Step S710: Obtain a personalized care matching plan, generate a encrypted feedback report based on the personalized care matching plan, and push it to the target receiving institution node.

[0123] The encrypted feedback report is derived using the following formula: (36) In formula (36), This indicates that the encrypted personalized care matching plan data in the encrypted feedback report is used for secure transmission. The cryptographic base (primitive element) is a public parameter of a public-key cryptosystem and is modulo... The generators of the multiplicative group are used to construct cryptographic operations. This represents the scheme feature data, which is the plaintext core features of the personalized care matching scheme to be encrypted (such as scheme ID, priority, item summary, etc.). It represents the encryption modulus, large prime numbers, public parameters of a public-key encryption system, defines the finite field of encryption operations, and ensures encryption security. This represents the modulo operation and the remainder operation. Limited to the model Within a finite field, ciphertext is formed. The control logic of formula (36) is based on the computational difficulty of the discrete logarithm problem to achieve secure encrypted transmission of personalized care matching schemes: 1. One-way guarantee of modular exponentiation + modular operation: Given , , It is difficult to deduce the plaintext from the reverse. 1. This protects the confidentiality of sensitive medical and elderly care data; 2. Lightweight and efficient encryption logic adapts to resource-constrained scenarios such as edge devices and elderly care gateways, and can quickly generate encrypted feedback reports to meet the real-time push requirements of medical and elderly care intervention plans; 3. Supports secure data sharing across institutions, which can only be decrypted by authorized recipients, balancing data privacy and collaboration efficiency, and meeting the compliance requirements for medical and health data security.

[0124] After receiving the pre-generated personalized care matching plan, the elderly health monitoring and tiered care system extracts the core nursing instructions and medication lists. It then uses a pre-set asymmetric encryption algorithm to obfuscate and encrypt this sensitive information, generating a encrypted feedback report. Subsequently, the system pushes this encrypted feedback report to the designated target receiving institution node via a dedicated medical network, such as the data center of the corresponding community health service center.

[0125] Step S720: Receive an acknowledgment receipt with a node status code. The acknowledgment receipt is generated by the target receiving node after receiving the encrypted feedback report.

[0126] The confirmation receipt is generated using the following formula: (37) In formula (37), This indicates a confirmation receipt, generated by the target node, containing status information, used to provide feedback on the result of receiving and processing the encrypted message. This indicates the target receiving institution node, the node receiving the encrypted feedback report, such as a hospital information system node, a nursing home management platform, or a family terminal device. This indicates a encrypted feedback report containing data on encrypted personalized care matching solutions pending confirmation. Indicates the node status code, the target node pair The result encoding of receiving, decrypting, and verifying. The control logic of formula (37) realizes a closed-loop confirmation mechanism for the transmission of sensitive medical and elderly care data: 1. Taking the processing status of the target node as the core, the initiator can keep track of the entire process status of receiving, decrypting, and verifying the encrypted feedback report in real time through the confirmation receipt with status code, avoiding the black box problem of "sent but not knowing whether it has been received"; 2. The node status code accurately locates the cause of the abnormality (such as data corruption, insufficient permissions), which facilitates quick troubleshooting and improves the reliability of the medical and elderly care intervention plan push; 3. The audit metadata (time stamp, node identifier, encrypted summary) in the receipt meets the medical data compliance audit requirements and realizes full-link traceability of data transmission.

[0127] After attempting to receive the encrypted feedback report, the target receiving node will return an acknowledgment containing a specific node status code to the sender. This node status code directly reflects the receiver's current processing capacity and network environment. For example, if the node status code in the acknowledgment is 200, it means the receiving node is idle and receiving normally; if the node status code is 503, it indicates that the receiving node is currently experiencing excessive concurrent processing, is under high load, or is in a state of network congestion.

[0128] Step S730: Adjust the initial data sharing stream according to the node status code to obtain the updated data sharing stream.

[0129] The updated data sharing stream is derived using the following formula: (38) In formula (38), This indicates the updated data sharing stream and the adjusted medical and elderly care data transmission strategy, including rate, priority, retry mechanism, etc. This represents the initial data sharing stream and the baseline transmission strategy before adjustment. This represents a coefficient used to adjust weights and control the magnitude of traffic adjustments, preventing sudden changes in traffic. This indicates the node status code, the processing status identifier returned by the target node (e.g., 200 / 400 / 403 / 500). The reference status code represents the preset "normal state" baseline (e.g., 200). This formula obtains the updated data sharing flow by adjusting the initial data sharing flow through additive deviation. The control logic of formula (38) is to achieve dynamic adaptive optimization of the medical and elderly care data sharing flow through additive adjustment driven by state deviation: 1. Using the node status code as a feedback signal, the transmission strategy is corrected in real time to avoid continuous invalid transmission under abnormal conditions; 2. Additive adjustment ensures smooth flow changes and avoids drastic fluctuations affecting other medical and elderly care services; 3. Weight Sensitivity can be flexibly controlled and adjusted to adapt to different scenarios (such as prioritizing transmission in emergency care plans and appropriately reducing the speed for daily monitoring).

[0130] After the elderly health monitoring and tiered care system parses the node status code in the aforementioned confirmation receipt, it triggers a flow control adjustment mechanism to reshape the transmission parameters of the initial data sharing stream. Assuming the initial data sharing stream is configured to synchronize the elderly person's real-time heart rate and blood pressure data to the receiving institution once per second, when the elderly health monitoring and tiered care system identifies status code 503, it determines that the current transmission frequency exceeds the receiving end's capacity threshold. At this point, the elderly health monitoring and tiered care system will reduce the transmission frequency of the initial data sharing stream, for example, adjusting it to send the average vital signs once every 60 seconds, or switching the transmission channel from the backbone link to a backup low-bandwidth link, thereby obtaining an updated data sharing stream.

[0131] Step S740: Use the updated data sharing stream to push dynamic monitoring data and optimize the response of integrated medical and elderly care services.

[0132] The optimized response time for integrated medical and elderly care services is derived using the following formula: (39) In formula (39), This indicates the optimized response time of integrated medical and elderly care services. After the data sharing flow is optimized, the actual response time of medical and elderly care services (such as alarm delay and plan push delay) is represented. The smaller the value, the faster the response. This represents the baseline response time, which is the baseline response time for medical and elderly care services before optimization and serves as the starting point for optimization calculations. This represents the optimization weight, a coefficient that controls the proportion of efficiency improvement to response time optimization, preventing over-optimization from causing system fluctuations. The value of efficiency improvement in integrated medical and elderly care services is a quantitative value of performance gain brought about by dynamic data sharing flow optimization (such as increased transmission rate and reduced latency). The control logic of formula (39) is to directly convert the performance gain brought about by data sharing flow optimization into a reduction in the response time of medical and elderly care services through the multiplicative decay mechanism of efficiency improvement value: 1. The core logic is "the higher the efficiency improvement → the more obvious the response time compression", which perfectly matches the high real-time requirements of health monitoring scenarios for the elderly (such as abnormal alarms and emergency care plan push); 2. Optimize weight It balances response speed and system stability, avoiding service fluctuations caused by extreme optimization; 3. The structure is intuitive and explainable, making it easy for medical staff and maintenance personnel to understand the source and extent of response optimization, thus improving the credibility of the medical and elderly care system.

[0133] The elderly health monitoring and tiered care system uses the updated data sharing stream described above to continuously push the elderly person's dynamic monitoring data to the target receiving institution node. During periods of high load at the receiving end, the elderly health monitoring and tiered care system pushes dynamic monitoring data such as body temperature and blood oxygen saturation at a reduced frequency of once every 60 seconds to maintain basic medical and elderly care business data interaction.

[0134] Please see Figure 2This embodiment provides a system for health monitoring and tiered care of the elderly that integrates medical and elderly care, and a method for implementing health monitoring and tiered care of the elderly that integrates medical and elderly care. It includes a multi-dimensional health dataset generation module 10, a data sharing flow establishment module 20, a standardized feature vector group acquisition module 30, a preliminary risk level distribution acquisition module 40, a refined risk grading result determination module 50, a personalized care matching scheme generation module 60, and a data sharing flow synchronization adjustment module 70. The multi-dimensional health dataset generation module 10 collects health data of the elderly from hospitals, communities, and elderly care institutions. It uses a medical information standardization protocol to convert the format and align the semantics of health data from different sources, fusing them to generate a unified multi-dimensional health dataset. The health data includes past medical history data, daily monitoring data, and lifestyle habit records. The data sharing flow establishment module 20 uses blockchain technology to construct a distributed ledger based on the multi-dimensional health dataset. This distributed ledger records data access permissions and change logs for each institution, authorizes data access requests according to preset privacy rules, and establishes a secure and interconnected data sharing flow. The standardized feature vector group acquisition module... Module 30 extracts key features from the secure and interconnected data sharing stream, normalizes these features to obtain a standardized feature vector set, including blood pressure fluctuation values ​​and activity frequency indicators. Module 40, for obtaining the preliminary risk level distribution, inputs the standardized feature vector set into a support vector machine model for classification, marking high-risk categories based on whether blood pressure fluctuation values ​​exceed a preset threshold, thus obtaining a preliminary risk level distribution. Module 50, for determining the refined risk grading result, combines the preliminary risk level distribution with the activity frequency indicator, inputs multi-dimensional features into a random forest model for integrated analysis, adjusts the risk level based on whether the activity frequency indicator is below a preset threshold, and determines the refined risk grading result. Module 60, for generating personalized care matching plans, uses a rule engine to match medical intervention plans and lifestyle guidance plans from a preset care template library based on the refined risk grading result, generating personalized care matching plans. Module 70, for synchronously adjusting the data sharing stream based on the personalized care matching plan, generates feedback reports and distributes them to relevant institutions, while simultaneously adjusting the secure and interconnected data sharing stream through a real-time update mechanism to optimize the response of integrated medical and elderly care services.

[0135] The method and system for health monitoring and tiered care of the elderly integrating medical and elderly care provided in this embodiment, compared with existing technologies, collects health data of the elderly from hospitals and community elderly care institutions, performs format conversion and semantic alignment using a standardized medical information protocol, and integrates the data to generate a multi-dimensional health dataset; it utilizes blockchain technology to construct a distributed ledger to ensure secure recording of data access permissions and change logs, and authorizes data sharing according to privacy rules; it extracts key features such as blood pressure fluctuation values ​​and activity frequency indicators, performs risk classification using support vector machines and random forest models, and combines a rule engine to match care templates, generating personalized plans and providing real-time feedback to relevant institutions. This embodiment significantly improves the accuracy and response efficiency of integrated medical and elderly care services through the organic combination of data standardization, secure sharing, and intelligent analysis, providing comprehensive support for the health management of the elderly.

[0136] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for health monitoring and graded care of the elderly that integrates medical and elderly care, characterized in that, Includes the following steps: S100. Collect health data of the elderly from hospitals, communities and elderly care institutions, and use medical information standardization protocols to convert the format and align the semantics of health data from different sources, and merge them to generate a unified multi-dimensional health dataset, wherein the health data includes past medical history data, daily monitoring data and lifestyle record data. S200. Based on the multi-dimensional health dataset, a distributed ledger is constructed using blockchain technology. The distributed ledger records the data access permissions and change logs of each institution. Data access requests are authorized according to preset privacy rules to establish a secure and interconnected data sharing flow. S300. Extract key features from the secure and interconnected data sharing stream, normalize the key features to obtain a standardized feature vector group, wherein the key features include blood pressure fluctuation values ​​and activity frequency indicators. S400. Input the standardized feature vector group into the support vector machine model for classification, and mark the high-risk category according to whether the blood pressure fluctuation value exceeds the preset threshold to obtain the preliminary risk level distribution. S500. Combining the preliminary risk level distribution with the activity frequency index, the multi-dimensional features are input into the random forest model for integrated analysis. The risk level is adjusted according to whether the activity frequency index is lower than a preset threshold to determine the refined risk classification result. S600. Based on the refined risk classification results, a personalized care matching plan is generated by matching medical intervention plans and life guidance plans in the preset care template library through a rule engine. S700: Generate a feedback report based on the personalized care matching plan and distribute it to the corresponding institutions. At the same time, adjust the secure and interconnected data sharing flow through a real-time update mechanism to optimize the response of integrated medical and elderly care services.

2. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 1, characterized in that, Step S100 includes: S110. Obtain multi-source heterogeneous health data packets of elderly health data from hospitals, communities and elderly care institutions, extract features and reassemble them using a medical information standardization protocol parser to obtain intermediate structured records. S120. Perform medical entity recognition and semantic mapping on the intermediate structured record, replace the heterogeneous description with standard encoding, and obtain standardized data entries; S130. Based on the standardized data entries, perform time-series alignment and anomaly verification, and use the multi-dimensional data fusion matrix to aggregate features to generate a unified multi-dimensional health dataset.

3. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 1, characterized in that, Step S200 includes: S210. Obtain a multi-dimensional health dataset, generate an institutional node index, and initialize the distributed ledger; S220. Configure an access permission table for the distributed ledger and generate an access permission status tree containing the latest access permission status. S230: Receive a data access request packet, read the permission status tree for verification to generate a temporary authorization token; S240. Based on the temporary authorization token, record the change log and establish a point-to-point data sharing channel; S250. Transmit the multi-dimensional health dataset through the data sharing channel to complete the construction of a secure and interconnected data sharing stream.

4. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 3, characterized in that, Step S300 includes: S310. Capture the raw data stream containing blood pressure fluctuation values ​​and activity frequency indicators in real time from the secure and interconnected data sharing stream; S320. The original data stream is divided using a sliding time window. The standard deviation of the blood pressure readings is calculated as the blood pressure fluctuation value in each window, and the number of times the accelerometer exceeds a preset threshold is counted as the activity frequency index. S330. Based on the maximum and minimum values ​​of blood pressure fluctuations in historical data, perform maximum-minimum normalization on the blood pressure fluctuation values ​​calculated in the current window. S340. Using the 95th percentile of the activity frequency index in the historical data as a benchmark, perform quantile normalization on the activity frequency index statistically analyzed in the current window. S350. The normalized blood pressure fluctuation value and the activity frequency index are combined in chronological order to form a standardized feature vector group with timestamps.

5. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 1, characterized in that, Step S400 includes: S410. Obtain the standardized feature vector set and input it into the support vector machine model. Construct the optimal hyperplane decision boundary using the radial basis kernel function. S420. Calculate the normal distance from the feature vector point to the decision boundary of the optimal hyperplane, and determine the initial binary classification label based on the normal distance; S430. If the blood pressure fluctuation value in the standardized feature vector group exceeds a preset threshold, then the initial binary classification label is ignored and the sample is marked as a high-risk category. S440. Combining the high-risk category with the low- and medium-risk levels mapped according to the normal distance, a risk level identifier sequence is generated to obtain a preliminary risk level distribution.

6. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 5, characterized in that, Step S500 includes: S510. Obtain the preliminary risk level distribution and activity frequency index, and construct a multi-dimensional feature matrix containing a numerical risk weight vector and the activity frequency index. S520. Input the multi-dimensional feature matrix into the random forest model, and obtain the aggregated classification probability value through decision tree branching and ensemble voting. S530. Determine whether the activity frequency index is lower than a preset threshold. If it is lower, generate a risk level gain value and perform weighted correction on the aggregated classification probability value. S540. Based on the revised aggregated classification probability values, the risk boundaries are redefined to determine the refined risk classification results.

7. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 1, characterized in that, The medical intervention plan includes medical intervention items, the lifestyle guidance plan includes lifestyle guidance items, and step S600 includes: S610. Obtain the intervention intensity corresponding to the refined risk classification results, and retrieve medical intervention items and lifestyle guidance items based on the intervention intensity; S620. Compare the medical intervention items with the lifestyle guidance items against the disease contraindication map to obtain a conflict detection matrix; S630. If the conflict detection matrix contains mutually exclusive markers, then mutually exclusive entries are removed to obtain a set of candidate care entries. S640. Assign execution priority values ​​according to the candidate care item set, and assemble the execution priority values ​​with the candidate care item set to generate a personalized care matching scheme.

8. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 1, characterized in that, Step S700 includes: S710. Obtain a personalized care matching plan, generate a encrypted feedback report based on the personalized care matching plan, and push it to the target receiving institution node; The encrypted feedback report is derived using the following formula: ； in, This indicates a encrypted feedback report. Indicates the encryption base. Represents the characteristic data of the scheme. Indicates the encryption modulus. Represents modulo operation; S720. Receive an acknowledgment receipt with a node status code, wherein the acknowledgment receipt is generated by the target receiving node after receiving the encrypted feedback report; The confirmation receipt is generated using the following formula: ； in, This indicates a confirmation receipt. Indicates the target receiving agency node. This indicates a encrypted feedback report. Indicates the node status code; S730. Adjust the initial data sharing stream according to the node status code to obtain the updated data sharing stream; The updated data sharing stream is derived using the following formula: ； in, This represents the updated data sharing stream. Indicates the initial data sharing stream, This indicates an adjustment of the weights. Indicates the node status code. Indicates the reference status code; S740: Adopt the updated data sharing stream to push dynamic monitoring data and optimize the response of integrated medical and elderly care services.

9. The method for health monitoring and graded care of the elderly integrating medical and elderly care according to claim 8, characterized in that, In step S740, the optimized response time for integrated medical and elderly care services is obtained using the following formula: ； in, This indicates the optimized response time for integrated medical and elderly care services. Indicates the base response time. Indicates the optimization weights, This indicates the efficiency improvement value of integrated medical and elderly care services.

10. A system for monitoring and tiered care of elderly health integrating medical and elderly care, used to execute the method for monitoring and tiered care of elderly health integrating medical and elderly care as described in any one of claims 1 to 9, characterized in that, include: The multidimensional health dataset generation module (10) is used to collect health data of the elderly from hospitals, communities and elderly care institutions. It adopts the medical information standardization protocol to convert the format and align the semantics of health data from different sources and integrates them to generate a unified multidimensional health dataset. The health data includes past medical history data, daily monitoring data and lifestyle record data. The data sharing flow establishment module (20) is used to construct a distributed ledger based on the multi-dimensional health dataset using blockchain technology, record the data access permissions and change logs of each institution in the distributed ledger, authorize data access requests according to preset privacy rules, and establish a secure and interconnected data sharing flow. The standardized feature vector group acquisition module (30) is used to extract key features from the secure and interconnected data sharing stream, normalize the key features, and obtain a standardized feature vector group, wherein the key features include blood pressure fluctuation value and activity frequency index. The preliminary risk level distribution acquisition module (40) is used to input the standardized feature vector group into the support vector machine model for classification, and to mark the high-risk category according to whether the blood pressure fluctuation value exceeds the preset threshold, so as to obtain the preliminary risk level distribution. The refined risk classification result determination module (50) is used to combine the preliminary risk level distribution with the activity frequency index, input multi-dimensional features into the random forest model for integrated analysis, adjust the risk level according to whether the activity frequency index is lower than a preset threshold, and determine the refined risk classification result. The personalized care matching solution generation module (60) is used to generate a personalized care matching solution by matching medical intervention solutions and life guidance solutions in the preset care template library through the rule engine based on the refined risk classification results. The data sharing flow synchronization adjustment module (70) is used to generate feedback reports based on the personalized care matching scheme and distribute them to the corresponding institutions. At the same time, it synchronizes and adjusts the secure and interconnected data sharing flow through a real-time update mechanism to optimize the response of medical and elderly care services.

Images