Health data management service system and method

Abstract

The invention relates to the technical field of data management, in particular to a health data management service system and method. The method comprises the following steps: acquiring a historical multi-source user data set, wherein the historical multi-source user data set comprises physiological index data and behavior state identifiers; analyzing the physiological index data, and determining physiological anomaly features; analyzing the behavior state identifier, and determining a triggering condition of the physiological anomaly feature; acquiring a real-time multi-source user data set of a to-be-managed user, analyzing the real-time multi-source user data set, and determining user individual characteristics of the to-be-managed user; and based on the triggering condition, determining a personalized intervention scheme according to the individual characteristics of the user. Individualized threshold adjustment under privacy protection is realized, and the problem that a group rule cannot adapt to an individual risk coefficient is eliminated. On the basis of the triggering condition, according to the individual characteristics of the user, a personalized intervention scheme is determined, intervention urgency evaluation errors are reduced, and the user response rate of a data complementary set task is improved.

Description

Technical Field

[0001] This application relates to the technical field of data management, and in particular, to a health data management service system and method. Background Art

[0002] With the increasing number of chronic diseases and stress-related health problems, data-driven health management technology has become the core research direction in the field of medical health. Traditional health monitoring systems usually use static threshold methods to judge physiological abnormalities, but there are still significant defects.

[0003] However, the determination of abnormal trigger conditions in traditional health monitoring systems is mostly based on single-dimensional matching. Due to individual risk coefficient differences, its clinical significance is small, it is difficult to meet the actual health management needs, and it is impossible to provide targeted management measures in a timely manner. Summary of the Invention

[0004] This application provides a health data management service system and method to solve the above problems.

[0005] In a first aspect, this application provides a health data management service method, the method includes: Obtain a historical multi-source user data set, the historical multi-source user data set includes physiological index data and behavior status identifiers; Analyze the physiological index data to determine physiological abnormality characteristics; Analyze the behavior status identifiers to determine the trigger conditions for the physiological abnormality characteristics; Obtain the real-time multi-source user data set of the user to be managed, analyze the real-time multi-source user data set, and determine the user individual characteristics of the user to be managed; Based on the trigger conditions, according to the user individual characteristics, determine a personalized intervention plan.

[0006] Through this solution, obtain a historical multi-source user data set, avoid the lack of behavior status data caused by manual entry, and eliminate the problem of broken attribute associations caused by global anonymization. Analyze the physiological index data to determine physiological abnormality characteristics, reduce the misjudgment rate, and improve the detection sensitivity of hidden abnormalities. Analyze the behavior status identifiers to determine the trigger conditions for the physiological abnormality characteristics, which helps to avoid the overfitting problem caused by a single statistic and improve the accuracy of stress source derivation. Obtain the real-time multi-source user data set of the user to be managed, analyze the real-time multi-source user data set, and determine the user individual characteristics of the user to be managed, which helps to achieve individualized threshold adjustment under privacy protection and eliminate the problem that group rules cannot adapt to individual risk coefficients. Based on the trigger conditions, according to the user individual characteristics, determine a personalized intervention plan, reduce the error in the assessment of intervention urgency, and improve the user response rate of the data complement task.

[0007] Optionally, the obtaining of the historical multi-source user dataset includes: Obtain data attributes; analyze the data attributes to determine sensitive data types; According to the sensitive data types, perform dynamic desensitization processing on the original data to obtain desensitized original data; Use the desensitized original data as the historical multi-source user dataset.

[0008] Through this solution, obtaining data attributes provides a complete metadata basis for sensitive data identification and ensures cross-source data correlation. Analyzing data attributes to determine sensitive data types helps eliminate the problem of loss of individual attribute correlation caused by global desensitization and avoids the inability to adapt to individual characteristics after complete anonymization. According to sensitive data types, performing dynamic desensitization processing on the original data to obtain desensitized original data helps balance data utility and privacy security, maintain the availability of temporal sequence characteristics, and avoid desensitization from damaging short-term variation detections such as premature ventricular contractions. Using the desensitized original data as the historical multi-source user dataset helps eliminate the spatio-temporal offset error caused by device differences.

[0009] Optionally, the analyzing of the physiological index data to determine physiological abnormality characteristics includes: Analyze the physiological index data to determine historical heart rate data; Analyze the historical heart rate data to determine the heart rate period distribution characteristics; Analyze the heart rate period distribution characteristics to determine a dynamic analysis threshold; Based on the dynamic analysis threshold, analyze the physiological index data to determine physiological abnormality characteristics.

[0010] Through this solution, analyzing physiological index data to determine historical heart rate data helps eliminate the sampling frequency difference between devices. Analyzing historical heart rate data to determine the heart rate period distribution characteristics helps eliminate the problem of ignoring the period fluctuation law and avoid misjudgment of the resting heart rate of elderly patients caused by group general rules. Analyzing the heart rate period distribution characteristics to determine a dynamic analysis threshold helps eliminate the problems of over-sensitivity or missed detection caused by a single interval. Based on the dynamic analysis threshold, analyzing physiological index data to determine physiological abnormality characteristics helps eliminate the problem of being insensitive to progressive abnormalities.

[0011] Optionally, the analyzing of the behavior status identifier to determine the triggering condition of the physiological abnormality characteristics includes: Based on the dynamic analysis threshold, analyze the physiological abnormality characteristics to determine the abnormal deviation degree; According to the abnormal deviation degree, determine the physiological abnormality level; Match the physiological abnormality level with the behavior status identifier to obtain a matching degree; Determine the triggering conditions of the physiological abnormality features according to the matching degree.

[0012] Through this solution, based on the dynamic analysis threshold, analyze the physiological abnormality features to determine the abnormal deviation degree, ensure the comparability of abnormalities across time periods, and overcome the problem of undifferentiated evaluation of the same deviation degree in young and elderly patients. According to the abnormal deviation degree, determine the physiological abnormality level, which helps to break through the limitation of not constructing a multi-layer model of abnormality level - individual attributes and provides a basis for differential intervention. Match the physiological abnormality level with the behavior status identifier to obtain the matching degree, which reflects the consistency between the abnormal event and the current behavior pattern and eliminates the problem that the speculation of the stress source stays on the surface level. According to the matching degree, determine the triggering conditions of the physiological abnormality features and achieve a gradient response from low-urgency recording to high-urgency intervention.

[0013] Optionally, the real-time multi-source user dataset includes life records. Based on the triggering conditions and according to the user individual characteristics, determine a personalized intervention plan, including: Analyze the life records to determine the user's living habits; Based on the triggering conditions, analyze the user individual characteristics and the physiological abnormality features to determine the mapping relationship between features and abnormalities; Speculate the stress source according to the mapping relationship; Determine a personalized intervention plan according to the stress source.

[0014] Through this solution, analyzing the life records to determine the user's living habits helps to eliminate the problem of missing benchmark data caused by relying on manual entry. Based on the triggering conditions, analyzing the user individual characteristics and physiological abnormality features to determine the mapping relationship between features and abnormalities replaces the traditional single-index matching mechanism and improves the accuracy of abnormal attribution. Speculating the stress source according to the mapping relationship helps to avoid misjudgment of the stress source caused by shallow statistics. Speculating the stress source according to the mapping relationship; determining a personalized intervention plan according to the stress source helps to correct the misalignment of the intervention strategy caused by the fixed return visit template and improve the executability of the plan.

[0015] Optionally, based on the triggering conditions, analyze the user individual characteristics and the physiological abnormality features to determine the mapping relationship between features and abnormalities, including: Based on the triggering conditions, analyze the user individual characteristics to determine the event-related features; Analyze the event-related features to determine the co-occurrence frequency with the physiological abnormality features; Determine the set of strong association rules according to the co-occurrence frequency; Determine the mapping relationship between features and abnormalities according to the set of strong association rules.

[0016] Through this solution, based on the triggering conditions, analyze the individual characteristics of users, determine the event-related characteristics, realize the multi-level binding of abnormal events and individual characteristics, and avoid the waste of computing power caused by global feature scanning. Analyze the event-related characteristics, determine the co-occurrence frequency with physiological abnormal characteristics, and solve the problem of lack of causal basis in shallow statistics. According to the co-occurrence frequency, determine the set of strong association rules, which helps to eliminate the problem of insufficient timeliness in the analysis of historical multi-source user datasets. According to the set of strong association rules, determine the feature-abnormality mapping relationship, which helps to eliminate the problem that group rules cannot adapt to individuals.

[0017] Optionally, the historical multi-source user dataset includes user attribute data; the analyzing the heart rate period distribution characteristics and determining the dynamic analysis threshold includes: Through a data statistical algorithm, cluster the heart rate period distribution characteristics according to a preset continuous period to obtain several sets of heart rate data for the same period of consecutive N days; Calculate the standard deviation and mean of the set of heart rate data; Parse the user attribute data to determine the age group and physical examination data to which the user belongs; Determine the user's disease history according to the physical examination data; Determine the threshold relationship index according to the age group to which the user belongs and the user's disease history; Determine the dynamic analysis threshold according to the standard deviation, the mean and the threshold relationship index.

[0018] Through this solution, through a data statistical algorithm, cluster the heart rate period distribution characteristics according to a preset continuous period to obtain several sets of heart rate data for the same period of consecutive N days, which helps to eliminate the interference of single-day accidental fluctuations, ensure data stability, and avoid statistical deviations caused by cross-period data mixing. Calculate the standard deviation and mean of the set of heart rate data, replace the fixed heart rate interval, and provide a basis for differential calculation of the dynamic threshold for each period. Parse the user attribute data to determine the age group and physical examination data to which the user belongs, and avoid the loss of individual characteristics caused by global anonymization. Determine the user's disease history according to the physical examination data to distinguish the clinical significance of the same deviation events between young healthy people and elderly patients. Determine the threshold relationship index according to the age group to which the user belongs and the user's disease history, which helps to replace the coarse-grained correction mode of a single dimension. Determine the dynamic analysis threshold according to the standard deviation, the mean and the threshold relationship index, which helps to replace the fixed percentage deviation determination and realize the matching of risk grading and individual tolerance.

[0019] Optionally, the determining the threshold relationship index according to the age group to which the user belongs and the user's disease history includes: Establish a feature vector including age, BMI index, and cardiovascular disease markers; Input the feature vector into a pre-trained risk coefficient prediction model to output a threshold relationship metric.

[0020] Through this solution, a feature vector including age, BMI index, and cardiovascular disease markers is established, eliminating the problem that individual attributes cannot be associated after data desensitization of the age layer, quantifying the potential impact of body fat percentage on heart rate variability, corresponding to the problem of not establishing the association between physiological abnormalities and lifestyle habits, providing a basis for quantifying obesity-related risks, and at the same time eliminating the defect of not differentiating and evaluating individual risk coefficients. Inputting the feature vector into a pre-trained risk coefficient prediction model to output a threshold relationship metric helps to dynamically adjust the heart rate fluctuation tolerance of different users and eliminate the core defect of high misjudgment rate of the static threshold method.

[0021] Optionally, after analyzing the physiological abnormality features and the behavior status identifier to determine the trigger condition of the abnormal physiological event, it further includes: When the trigger condition is satisfied, obtain user interaction data; Analyze the user interaction data to determine the user terminal interaction ability; Determine a dynamic return visit strategy according to the user terminal interaction ability.

[0022] Through this solution, when the trigger condition is satisfied, obtain user interaction data, avoid the randomness of manual input, and ensure the complete acquisition of user behavior characteristics when an abnormal event is triggered. Analyze the user interaction data to determine the user terminal interaction ability, which helps to eliminate the problem of insufficient sensitivity of single-dimensional matching. Determine a dynamic return visit strategy according to the user terminal interaction ability, which helps to eliminate the problem of low user response rate caused by a unified questionnaire and ensure the timely replenishment of data.

[0023] In a second aspect, the present application provides a health data management service system, and the system includes: A data acquisition module for acquiring a historical multi-source user data set, where the historical multi-source user data set includes physiological index data and behavior status identifiers; A data analysis module for analyzing the physiological index data to determine physiological abnormality features; An identifier analysis module for analyzing the behavior status identifier to determine the trigger condition of the physiological abnormality feature; A feature determination module for acquiring a real-time multi-source user data set of the user to be managed, analyzing the real-time multi-source user data set, and determining the user individual features of the user to be managed; A solution determination module for determining a personalized intervention solution based on the trigger condition and according to the user individual features.

[0024] Optionally, when the data acquisition module acquires the historical multi-source user data set, it is used for: Obtain data attributes; parse the data attributes to determine sensitive data types; According to the sensitive data types, perform dynamic desensitization processing on the original data to obtain the desensitized original data; Use the desensitized original data as the historical multi-source user dataset.

[0025] Optionally, when the data analysis module analyzes the physiological index data to determine physiological abnormality characteristics, it is used for: Parse the physiological index data to determine historical heart rate data; Parse the historical heart rate data to determine the heart rate period distribution characteristics; Analyze the heart rate period distribution characteristics to determine the dynamic analysis threshold; Based on the dynamic analysis threshold, analyze the physiological index data to determine physiological abnormality characteristics.

[0026] Optionally, when the identification analysis module analyzes the behavior status identifier to determine the trigger conditions for the physiological abnormality characteristics, it is used for: Based on the dynamic analysis threshold, analyze the physiological abnormality characteristics to determine the abnormal deviation degree; According to the abnormal deviation degree, determine the physiological abnormality level; Match the physiological abnormality level with the behavior status identifier to obtain the matching degree; According to the matching degree, determine the trigger conditions for the physiological abnormality characteristics.

[0027] Optionally, the real-time multi-source user dataset includes life records. When the solution determination module determines a personalized intervention solution based on the trigger conditions and according to the user individual characteristics, it is used for: Analyze the life records to determine the user's living habits; Based on the trigger conditions, analyze the user individual characteristics and the physiological abnormality characteristics to determine the feature-abnormality mapping relationship; According to the mapping relationship, infer the stress source; According to the stress source, determine the personalized intervention solution.

[0028] Optionally, when the solution determination module determines the feature-abnormality mapping relationship based on the trigger conditions by analyzing the user individual characteristics and the physiological abnormality characteristics, it is used for: Based on the trigger conditions, analyze the user individual characteristics to determine event-related characteristics; Analyze the event-related characteristics to determine the co-occurrence frequency with the physiological abnormality characteristics; According to the co-occurrence frequency, determine the set of strong association rules; Determine the mapping relationship between features and anomalies according to the set of strong association rules.

[0029] Optionally, the historical multi-source user dataset includes user attribute data; when the data analysis module analyzes the heart rate period distribution characteristics and determines the dynamic analysis threshold, it is used for: Cluster the heart rate period distribution characteristics according to a preset continuous period through a data statistical algorithm to obtain a set of heart rate data for the same period of several consecutive N days; Calculate the standard deviation and mean of the set of heart rate data; Analyze the user attribute data to determine the age group and physical examination data to which the user belongs; Determine the user's disease history according to the physical examination data; Determine the threshold relationship index according to the age group to which the user belongs and the user's disease history; Determine the dynamic analysis threshold according to the standard deviation, the mean, and the threshold relationship index.

[0030] Optionally, when the data analysis module determines the threshold relationship index according to the age group to which the user belongs and the user's disease history, it is used for: Establish a feature vector including age, BMI index, and cardiovascular disease markers; Input the feature vector into a pre-trained risk coefficient prediction model and output the threshold relationship index.

[0031] Optionally, the health data management service system further includes a policy determination module, which is used for: When the trigger condition is satisfied, obtain user interaction data; Analyze the user interaction data to determine the user terminal interaction ability; Determine the dynamic return visit policy according to the user terminal interaction ability. Description of the Drawings

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a schematic diagram of an application scenario provided by an embodiment of the present application; Figure 2 It is a flowchart of a health data management service method provided by an embodiment of the present application; Figure 3Schematic diagram of a health data management service system provided by an embodiment of the present application. Detailed implementation manners

[0034] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0035] In addition, the term "and / or" in this article is only a kind of association relationship describing associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after unless otherwise specified.

[0036] The embodiments of the present application will be further described in detail below with reference to the accompanying drawings of the specification.

[0037] In the field of health monitoring, in the prior art, the determination of abnormal trigger conditions is mostly based on single-dimensional matching. However, due to individual risk coefficient differences, its clinical significance is small, and it is difficult to meet the actual health management needs.

[0038] Based on this, the present application provides a health data management service system and method, which acquires a historical multi-source user data set, avoids the lack of behavioral state data caused by manual input, and eliminates the problem of broken attribute associations caused by global anonymization. Analyze physiological index data to determine physiological abnormal characteristics, reduce the misjudgment rate, and improve the detection sensitivity of concealed abnormalities. Analyze behavioral state identifiers to determine the trigger conditions for physiological abnormal characteristics, which helps to avoid the overfitting problem caused by a single statistic and improve the accuracy of stress source derivation. Acquire the real-time multi-source user data set of the user to be managed, analyze the real-time multi-source user data set, and determine the user individual characteristics of the user to be managed, which helps to achieve individualized threshold adjustment under privacy protection and eliminate the problem that group rules cannot adapt to individual risk coefficients. Based on the trigger conditions, according to the user individual characteristics, determine a personalized intervention plan, reduce the error in the assessment of intervention urgency, and improve the user response rate of the data complement task.

[0039] Figure 1 Schematic diagram of an application scenario provided by the present application. When performing health data management, the method provided by the present application is applied.

[0040] Specifically, the method provided in this application is applied to any server. The server interacts with the device of the user to be managed. First, the internal database is retrieved to obtain a historical multi-source user dataset. The physiological index data therein is analyzed to determine physiological abnormality characteristics, reducing the misjudgment rate. The behavior status identifier is analyzed to determine the triggering conditions of the physiological abnormality characteristics, avoiding the overfitting problem caused by a single statistic. The real-time multi-source user dataset generated by the device of the user to be managed is obtained and analyzed to determine the user individual characteristics of the user to be managed. According to the user individual characteristics, a personalized intervention plan is determined, reducing the intervention urgency assessment error and improving the user response rate of the data complement task. The specific implementation manner can refer to the following embodiments.

[0041] Figure 2 FIG. 4 is a flowchart of a health data management service method provided by an embodiment of this application. The method of this embodiment can be applied to the server in the above scenario. As Figure 2 shown, the method includes: S201. Obtain a historical multi-source user dataset, where the historical multi-source user dataset includes physiological index data and a behavior status identifier; The historical multi-source user dataset can be a time-series dataset that has been desensitized and contains physiological index data and a behavior status identifier.

[0042] The physiological index data can be vital sign parameters.

[0043] The behavior status identifier can be a behavior classification label.

[0044] Specifically, through the medical device API, the wearable device SDK, and the user interaction log, the physiological index data and the behavior status identifier including time stamps are collected and stored in the internal database as the historical multi-source user dataset.

[0045] In the specific implementation manner, if the historical multi-source user dataset is to be retrieved, a dynamic K-anonymous desensitization strategy can be adopted to perform attribute-preserving replacement on the user identity information. Among them, the user attribute data is subjected to attribute generalization processing so that it can be associated with individual characteristics but cannot reverse-identify the user.

[0046] S202. Analyze the physiological index data to determine physiological abnormality characteristics; The physiological abnormality characteristics can be the corresponding characteristics of the deviation of the physiological index.

[0047] Specifically, the multi-source user dataset is divided into subsets such as the exercise period and the resting period according to the behavior status identifier; then, the probability distribution of the physiological index data is calculated for each subset respectively, and the percentile is taken as the normal interval; furthermore, when the continuous sampling points exceed the normal interval, physiological abnormality characteristics including the deviation amplitude and the duration are generated.

[0048] S203. Analyze the behavior status identifier to determine the triggering conditions for physiological abnormality features; The triggering conditions can be a set of strong association rules for behavior status - physiological abnormality.

[0049] Specifically, perform temporal segmentation on the behavior status identifier to extract high-frequency behavior combination patterns; subsequently, attach time decay weights to the high-frequency behavior combination patterns, quickly calculate the time-weighted support degree in a vertical data format, and use an improved APRIORI algorithm to calculate the conditional probability and lift index of different behavior patterns and physiological abnormality features; then, screen the association rules that meet the minimum support and confidence to construct the triggering conditions for physiological abnormality features.

[0050] S204. Obtain the real-time multi-source user dataset of the user to be managed, analyze the real-time multi-source user dataset, and determine the user individual characteristics of the user to be managed; The user to be managed can be the target user currently accessing the health data management service.

[0051] The real-time multi-source user dataset can be a multi-dimensional data set currently generated by the user to be managed.

[0052] The user individual characteristics can be a feature vector for personalized decision-making.

[0053] Specifically, extract the dynamic biomarkers related to chronic disease management from the real-time multi-source user dataset, and then perform sliding window normalization on the dynamic biomarkers; subsequently, perform hierarchical clustering on the historical data according to demographic attributes and health status to generate a population feature template library; furthermore, calculate the weighted difference degree between the normalization result and the population feature template library; finally, perform exponential smoothing on the difference degree sequence to generate the smoothed user individual characteristics.

[0054] S205. Based on the triggering conditions, determine a personalized intervention plan according to the user individual characteristics.

[0055] The personalized intervention plan can be a set of action instructions generated by matching the triggering conditions and the user individual characteristics.

[0056] Specifically, based on the triggering conditions, perform multi-layer matching: First, calculate the abnormality level according to the deviation degree and duration to determine the response priority; second, combine the risk coefficient calculated from the disease history and physiological baseline in the user individual characteristics to adjust the intervention intensity; then, select the data complement method based on the interaction preference characteristics; finally, generate a personalized intervention plan including medical advice, behavior adjustment instructions, and reexamination reminders.

[0057] Through this solution, a historical multi-source user dataset is obtained, avoiding the lack of behavioral status data caused by manual entry and eliminating the problem of broken attribute associations caused by global anonymization. Analyze the physiological index data to determine physiological abnormal characteristics, reduce the misjudgment rate, and improve the detection sensitivity for hidden abnormalities. Analyze the behavioral status identifiers to determine the triggering conditions of physiological abnormal characteristics, which helps to avoid the overfitting problem caused by a single statistic and improve the accuracy of stress source derivation. Obtain the real-time multi-source user dataset of the user to be managed, analyze the real-time multi-source user dataset, and determine the individual characteristics of the user to be managed, which helps to achieve individualized threshold adjustment under privacy protection and eliminate the problem that group rules cannot adapt to individual risk coefficients. Based on the triggering conditions, according to the individual characteristics of the user, determine a personalized intervention plan, reduce the error in the assessment of intervention urgency, and increase the user response rate of the data complement task.

[0058] In some embodiments, data attributes are obtained; the data attributes are parsed to determine sensitive data types; according to the sensitive data types, the original data is dynamically desensitized to obtain desensitized original data; the desensitized original data is used as the historical multi-source user dataset.

[0059] The data attributes can be descriptive features and classification labels carried by the data itself.

[0060] The sensitive data types can be data categories that require special protection.

[0061] The original data can be multi-source user data that has not been desensitized.

[0062] Specifically, data attributes are extracted from the metadata layer of the historical multi-source user dataset. According to a predefined sensitive word library, the data attributes are parsed field by field, and combined with a sensitivity weight calculation model constructed based on information entropy theory and attribute correlation analysis, three sensitive data types, namely direct identifiers, highly correlated indirect identifiers, and low-sensitive time-series data, are output. According to the sensitive data types, the SHA-3 hash algorithm defined by the NIST FIPS 202 standard is applied to the fields of the direct identifiers to generate a fixed-length ciphertext string, and the anti-rainbow table attack ability is enhanced through salt value confusion while retaining the cross-table association ability; then, according to the data distribution characteristics and business requirements, dynamic generalization is performed on the fields of the highly correlated indirect identifiers; subsequently, the order-preserving encryption and statistical feature retention technology are used for the low-sensitive time-series data; finally, the desensitized original data is obtained. The desensitized original data is re-encoded according to a unified spatio-temporal benchmark and aligned with the historical multi-source user dataset in terms of time stamps and space grid mapping.

[0063] Through this solution, data attributes are obtained to provide a complete metadata foundation for sensitive data identification and ensure cross-source data relevance. Parsing data attributes and determining sensitive data types can help eliminate the problem of individual attribute association loss caused by global desensitization and avoid the inability to adapt individual characteristics after complete anonymization. According to the type of sensitive data, the original data is dynamically desensitized to obtain the desensitized original data, which helps to balance data utility and privacy security, maintain the availability of time series features, and avoid desensitization from destroying short-term variation detection such as ventricular premature beats. Using the desensitized original data as a historical multi-source user data set can help eliminate the spatiotemporal offset errors caused by device differences.

[0064] In some embodiments, physiological indicator data are analyzed to determine historical heart rate data; historical heart rate data are analyzed to determine heart rate time period distribution characteristics; heart rate time period distribution characteristics are analyzed to determine dynamic analysis thresholds; based on the dynamic analysis thresholds, physiological indicator data are analyzed to determine physiological abnormality characteristics.

[0065] The historical heart rate data may be heart rate time series data for a period of time in the past, which may be obtained based on experience or may be manually determined.

[0066] The heart rate time period distribution feature may be a differentiated distribution law obtained by statistically classifying historical heart rate data.

[0067] The dynamic analysis threshold may be a scene-adaptive decision threshold generated according to distribution characteristics of different physiological time periods.

[0068] Specifically, the second-level heart rate value is parsed from the physiological indicator data, the device abnormal value is eliminated, and the structured historical heart rate data with timestamp alignment is generated. Subsequently, the historical heart rate data is matched with the user behavior status data of the same period, and the time period type corresponding to each heart rate sampling point is marked; then, the heart rate time period distribution characteristics are grouped according to the marked time period type, and the heart rate time period distribution characteristics are calculated. Then, the same attribute group is constructed according to age, gender, and disease history. Then, the median and interquartile range of the heart rate time period distribution characteristics in the same attribute group are statistically calculated according to the time period type to determine the same attribute group baseline; then, according to the current user's heart rate time period distribution characteristics, the relative offset from the median of the group baseline is calculated; then, the historical offset rate sequence is weighted by exponential weighted moving average, and the smoothed individual offset coefficient is finally output; the dynamic analysis threshold is determined according to the time period type and the individual offset coefficient. Finally, the real-time collected physiological indicator data is compared with the dynamic analysis threshold of the corresponding time period, and the data points exceeding the dynamic analysis threshold are marked; then, the continuous abnormal data points are clustered to determine the physiological abnormal characteristics.

[0069] Through this solution, parsing physiological index data and determining historical heart rate data helps to eliminate the sampling frequency differences between devices. Analyzing the historical heart rate data and determining the distribution characteristics of heart rate time periods helps to eliminate the problem of ignoring the fluctuation rules of time periods and avoid misjudging the resting heart rate of elderly patients caused by the general rules for the group. Analyzing the distribution characteristics of heart rate time periods and determining the dynamic analysis threshold helps to eliminate the problems of over-sensitivity or missed detection caused by a single interval. Based on the dynamic analysis threshold, analyzing the physiological index data and determining the physiological abnormality characteristics helps to eliminate the problem of being insensitive to progressive abnormalities.

[0070] In some embodiments, based on the dynamic analysis threshold, analyze the physiological abnormality characteristics to determine the abnormal deviation degree; according to the abnormal deviation degree, determine the physiological abnormality level; match the physiological abnormality level with the behavior status identifier to obtain the matching degree; according to the matching degree, determine the trigger condition of the physiological abnormality characteristics.

[0071] The abnormal deviation degree can be a standardized deviation amount used to quantify the abnormal degree of physiological characteristics relative to the current time period.

[0072] The physiological abnormality level can be a classification of the severity of abnormal events.

[0073] The matching degree can be the association strength between the current physiological abnormal event and the user's behavior status.

[0074] Specifically, compare the physiological abnormality characteristics with the dynamic analysis threshold of the corresponding time period point by point; then, for the data points that exceed the dynamic analysis threshold, calculate the absolute difference between the actual value and the dynamic analysis threshold, which is recorded as the single-point deviation degree; if the number of continuously over-limit data points is too large, accumulate the single-point deviation degrees to generate the window deviation degree; subsequently, divide the window deviation degree by the dynamic analysis threshold to output the standardized abnormal deviation degree. Based on the heart rate deviation risk stratification standard in the AHA clinical abnormality classification guidelines, determine the physiological abnormality level according to the abnormal deviation degree. Based on the co-occurrence probability of abnormal events and behavior status in historical data, statistically generate a matching weight coefficient table; then, extract the weights of the corresponding levels from the matching weight coefficient table according to the user's behavior status identifier, and perform a multiplication operation with the physiological abnormality level to output the matching degree. According to the matching degree, set multi-level trigger thresholds: First, confirm the level that only generates abnormal event logs as the log record level; second, confirm the level that triggers the health warning of the user terminal as the warning prompt level; then, confirm the level that starts the emergency medical intervention process as the medical alarm level; integrate the log record level, warning prompt level, and medical alarm level to generate the final trigger condition of the physiological abnormality characteristics.

[0075] Through this solution, based on the dynamic analysis threshold, physiological abnormal features are analyzed to determine the abnormal deviation degree, ensuring the comparability of abnormalities across time periods and overcoming the problem of undifferentiated evaluation of the same deviation degree in young and elderly patients. According to the abnormal deviation degree, the physiological abnormal level is determined, which helps to break through the limitation of not constructing a multi-layer model of abnormal level - individual attributes and provides a basis for differentiated intervention. The physiological abnormal level is matched with the behavior state identifier to obtain the matching degree, which reflects the consistency between the abnormal event and the current behavior pattern and eliminates the problem that the speculation of stress sources stays on the superficial level. According to the matching degree, the triggering conditions of physiological abnormal features are determined to achieve a gradient response from low-urgency recording to high-urgency intervention.

[0076] In some embodiments, the life records are analyzed to determine the user's living habits; based on the triggering conditions, the user's individual characteristics and physiological abnormal features are analyzed to determine the mapping relationship between features and abnormalities; according to the mapping relationship, the stress source is speculated; according to the stress source, a personalized intervention plan is determined.

[0077] The life records can be a set of structured multimodal behavior data.

[0078] The user's living habits can be a combination of high-frequency repetitive behaviors extracted from the life records.

[0079] Feature-abnormality can be an association pair formed by coupling the user's individual characteristics and physiological abnormal features.

[0080] The mapping relationship can be a causal relationship established through strong association rule mining.

[0081] The stress source can be a living habit with a high confidence association with physiological abnormal events.

[0082] Specifically, the multi-source user data sets in the life records are aligned by time stamps to generate a multimodal behavior sequence with a unified time axis; then, partial desensitization is performed on the sensitive fields while retaining the relevant fields such as time stamps and durations; subsequently, clustering analysis is performed on the desensitized behavior sequence based on a time window to extract the combination of high-frequency repetitive behaviors and label it as the user's living habits. Then, the triggering conditions are bound to the physiological abnormal features and the user's individual characteristics in the corresponding time period to form a three-dimensional feature matrix; the mapping relationship between features and abnormalities is generated by mining strong association rules through the FP-Growth algorithm. According to the mapping relationship, the stress source is speculated from the user's living habits. Finally, according to the stress source, a preset strategy library constructed based on evidence-based medical guidelines and health behavior theory is called, and the intervention intensity is adjusted in combination with the triggering condition level, thereby generating a personalized intervention plan.

[0083] Through this solution, analyzing life records and determining users' living habits helps eliminate the problem of missing baseline data caused by manual input. Based on the trigger conditions, analyzing users' individual characteristics and physiological abnormality characteristics to determine the mapping relationship between characteristics and abnormalities, replacing the traditional single-index matching mechanism, and improving the accuracy of abnormality attribution. According to the mapping relationship, inferring stress sources helps avoid misjudgment of stress sources caused by shallow statistics. According to the mapping relationship, inferring stress sources; according to stress sources, determining personalized intervention plans helps correct the dislocation of intervention strategies caused by fixed follow-up templates and improve the executability of the plans.

[0084] In some embodiments, based on the trigger conditions, analyze users' individual characteristics to determine event-related characteristics; analyze the event-related characteristics to determine the co-occurrence frequency with physiological abnormality characteristics; according to the co-occurrence frequency, determine the set of strong association rules; according to the set of strong association rules, determine the mapping relationship between characteristics and abnormalities.

[0085] The event-related characteristics can be a set of characteristics screened from users' individual characteristics that have potential associations with physiological abnormality events.

[0086] The co-occurrence frequency can be the statistical number of times that event-related characteristics and physiological abnormality characteristics appear simultaneously in the historical multi-source user dataset.

[0087] The set of strong association rules can be a set of rules representing the high-probability causal relationship between event-related characteristics and physiological abnormality characteristics.

[0088] Specifically, according to the hierarchical division of the trigger conditions, extract users' individual characteristics within the corresponding time period; furthermore, bind the users' individual characteristics with the physiological abnormality characteristics within the same time window to determine the event-related characteristics. Based on the corresponding time window of the trigger conditions, count the co-occurrence frequency between the event-related characteristics and the physiological abnormality characteristics in the historical multi-source user dataset. Then, based on the causal association strength grading of abnormality events and living habits in the clinical guidelines, construct the support threshold and confidence threshold; set the minimum support threshold and confidence threshold; subsequently, use the FP-Growth algorithm to mine frequent item sets from the co-occurrence probability matrix to determine the set of strong association rules. Furthermore, sort the association rules in descending order of confidence, eliminate the rules containing contradictory items, and retain the valid rules to form the mapping relationship between characteristics and abnormalities.

[0089] Through this solution, based on the triggering conditions, analyze the individual characteristics of the user, determine the event-related characteristics, realize the multi-level binding of abnormal events and individual characteristics, and avoid the waste of computing power caused by global feature scanning. Analyze the event-related characteristics, determine the co-occurrence frequency with physiological abnormal characteristics, and solve the problem of lack of causal basis in shallow statistics. According to the co-occurrence frequency, determine the set of strong association rules, which helps to eliminate the problem of insufficient timeliness in the analysis of historical multi-source user data sets. According to the set of strong association rules, determine the mapping relationship between features and abnormalities, which helps to eliminate the problem that group rules cannot be adapted to individuals.

[0090] In some embodiments, through a data statistical algorithm, cluster the heart rate period distribution characteristics according to a preset continuous period to obtain a set of heart rate data for the same period of consecutive N days; calculate the standard deviation and mean of the set of heart rate data; parse the user attribute data to determine the age group to which the user belongs and the physical examination data; determine the user's disease history according to the physical examination data; determine the threshold relationship index according to the age group to which the user belongs and the user's disease history; determine the dynamic analysis threshold according to the standard deviation, mean and threshold relationship index.

[0091] The data statistical algorithm can be a mathematical processing method for clustering and analyzing the user's heart rate data according to a preset continuous period and calculating the standard deviation and mean.

[0092] The preset continuous period can be a set of continuous time periods generated by dividing the daily cycle according to a fixed time length in advance. It is stored in the server in advance and called when used.

[0093] The set of heart rate data can be a time series data set composed of heart rate measurement values collected continuously for N days in the same preset continuous period.

[0094] The standard deviation can be a statistic that measures the degree of fluctuation of heart rate data within the same period.

[0095] The mean can be the arithmetic mean of heart rate data within the same period.

[0096] The user attribute data can be structured metadata containing the user's age group code and the conclusion of the physical examination report.

[0097] The age group to which the user belongs can be the user's age interval code.

[0098] The physical examination data can be the user's health examination results.

[0099] The user's disease history can be a list of chronic disease labels extracted from the physical examination data.

[0100] The threshold relationship index can be a numerical correction coefficient used to adjust the multiple of the standard deviation.

[0101] Specifically, slice the heart rate period distribution characteristics continuously collected by the user according to a preset continuous period; use the dynamic time warping distance as the similarity measure, and perform cross-day analysis on the same user in the historical data through a time series clustering algorithm to extract the heart rate data at the same time period for N consecutive days, forming a heart rate data set. Perform statistical calculations of the standard deviation and mean on the data set of the same time period generated by clustering: First, use the unbiased estimation formula to calculate the standard deviation; then, take the mean of the heart rate data set within the same time period. Analyze the date of birth field and physical examination data in the user attribute data, and divide the age group to which the user belongs according to clinical standards. Extract the ICD-10 disease code according to the diagnosis conclusion in the physical examination data to construct the user's disease history. Load the basic correction coefficient table set by clinical epidemiology research according to the age group; then, load the disease-specific adjustment rules corresponding to different diseases in the disease diagnosis and treatment guidelines according to the user's disease history; subsequently, integrate the basic correction coefficient table and the disease-specific adjustment rules to calculate the threshold relationship index. Based on the theory of statistical confidence intervals, with the mean as the center and the product of the threshold relationship index and the standard deviation as the radius, construct the period benchmark threshold; then, introduce a standard deviation amplification factor verified by experimental data for dynamic adjustment to generate the final dynamic analysis threshold.

[0102] Through this solution, through the data statistical algorithm, cluster the heart rate period distribution characteristics according to the preset continuous period to obtain several data sets of heart rate data at the same time period for N consecutive days, which helps to eliminate the interference of single-day accidental fluctuations, ensure data stability, and avoid statistical biases caused by cross-period data mixing. Calculate the standard deviation and mean of the heart rate data set, replace the fixed heart rate interval, and provide a basis for differential calculation of the dynamic threshold in each period. Analyze the user attribute data to determine the age group to which the user belongs and the physical examination data, avoiding the loss of individual characteristics caused by global anonymization. Determine the user's disease history based on the physical examination data to distinguish the clinical significance of the same deviation events between young healthy people and elderly patients. Determine the threshold relationship index according to the age group to which the user belongs and the user's disease history, which helps to replace the coarse-grained correction mode of a single dimension. Determine the dynamic analysis threshold according to the standard deviation, mean, and threshold relationship index, which helps to replace the fixed percentage deviation determination and achieve the matching of risk grading and individual tolerance.

[0103] In some embodiments, establish a feature vector including age, BMI index, and cardiovascular disease marker; input the feature vector into a pre-trained risk coefficient prediction model to output the threshold relationship index.

[0104] The BMI index can be a continuous physiological parameter used to quantify the physical characteristics of the user.

[0105] The cardiovascular disease marker can be a binary identifier generated based on whether the user's disease history contains diseases with ICD-10 codes 100-199.

[0106] The feature vector can be a vector composed of the age group code to which the user belongs, the BMI index, and the cardiovascular disease marker.

[0107] The risk coefficient prediction model can be a machine learning model used to dynamically adjust the physiological abnormality determination threshold.

[0108] Specifically, map the age group to which the user belongs to a numerical code, that is, age; then, calculate the BMI index according to the height and weight in the physical examination data using a formula; subsequently, screen for diseases with ICD-10 codes I00-I99 from the user's medical history, and if there are cardiovascular-related disease codes, mark them to generate a cardiovascular disease marker; finally, splice the age, BMI index, and cardiovascular disease marker in a fixed order into a feature vector. Based on the clinical cohort data, use the multiple linear regression algorithm to train and generate a risk coefficient prediction model; then, input the feature vector into the pre-trained risk coefficient prediction model to output a threshold relationship index.

[0109] Through this solution, a feature vector including age, BMI index, and cardiovascular disease marker is established, eliminating the problem that individual attributes cannot be associated after data desensitization of the age group, quantifying the potential impact of body fat percentage on heart rate variability, corresponding to the problem of not establishing the association between physiological abnormalities and living habits, providing a basis for quantifying obesity-related risks, and at the same time eliminating the defect of undifferentiated assessment of individual risk coefficients. Inputting the feature vector into the pre-trained risk coefficient prediction model and outputting a threshold relationship index helps to dynamically adjust the heart rate fluctuation tolerance of different users and eliminate the core defect of high misjudgment rate of the static threshold method.

[0110] In some embodiments, when a trigger condition is satisfied, obtain user interaction data; analyze the user interaction data to determine the user terminal interaction ability; and determine a dynamic follow-up strategy according to the user terminal interaction ability.

[0111] The user interaction data can be quantifiable operation records generated by the user during use.

[0112] The user terminal interaction ability can be a quantitative evaluation result generated by analyzing the user interaction data.

[0113] The dynamic follow-up strategy can be a differentiated data complementation scheme generated in real time according to the user terminal interaction ability.

[0114] Specifically, when it is determined that the triggering conditions for physiological abnormal events are met, the user terminal device is automatically called to send a data collection instruction, the mobile operation log is retrieved, and user interaction data such as the mobile usage frequency, voice input records, and questionnaire response timeliness are extracted. Furthermore, based on the user interaction data, a standardized interaction ability feature vector is defined, and the interaction ability feature vector is input into a predefined hierarchical threshold rule library to perform a multi-dimensional joint determination of the user terminal interaction ability. Finally, according to the determination result of the user terminal interaction ability, a standardized process framework referring to the clinical diagnosis and treatment path is matched from the mapping relationship library between the interaction ability features and the follow-up strategy, and a preset template generated in combination with the usability design criteria of the interaction interface; furthermore, based on the matching result, combined with the terminal real-time state data injection strategy parameters, a dynamic follow-up strategy is determined.

[0115] Through this solution, when the triggering conditions are met, user interaction data is obtained, avoiding the randomness of manual entry and ensuring the complete acquisition of user behavior characteristics when abnormal events are triggered. Analyzing the user interaction data and determining the user terminal interaction ability helps to eliminate the problem of insufficient matching sensitivity in a single dimension. Determining the dynamic follow-up strategy based on the user terminal interaction ability helps to eliminate the problem of low user response rate caused by a unified questionnaire and ensure the timely replenishment of data.

[0116] Figure 3 The following is a schematic structural diagram of a health data management service system provided by an embodiment of the present application, as Figure 3 shown, the health data management service system 300 of this embodiment includes: a data acquisition module 301, a data analysis module 302, an identification analysis module 303, a feature determination module 304, and a solution determination module 305.

[0117] The data acquisition module 301 is used to obtain a historical multi-source user data set, and the historical multi-source user data set includes physiological index data and behavior state identifiers; The data analysis module 302 is used to analyze the physiological index data and determine physiological abnormal characteristics; The identification analysis module 303 is used to analyze the behavior state identifiers and determine the triggering conditions for the physiological abnormal characteristics; The feature determination module 304 is used to obtain a real-time multi-source user data set of the user to be managed, analyze the real-time multi-source user data set, and determine the user individual characteristics of the user to be managed; The solution determination module 305 is used to determine a personalized intervention solution based on the triggering conditions and according to the user individual characteristics.

[0118] Optionally, when the data acquisition module 301 obtains the historical multi-source user data set, it is used for: Obtain data attributes; parse the data attributes to determine sensitive data types; Perform dynamic desensitization processing on the original data according to the sensitive data type to obtain the desensitized original data; Use the desensitized original data as the historical multi-source user dataset.

[0119] Optionally, when the data analysis module 302 analyzes the physiological index data to determine physiological abnormality characteristics, it is used for: Parse the physiological index data to determine historical heart rate data; Parse the historical heart rate data to determine the heart rate period distribution characteristics; Analyze the heart rate period distribution characteristics to determine the dynamic analysis threshold; Based on the dynamic analysis threshold, analyze the physiological index data to determine physiological abnormality characteristics.

[0120] Optionally, when the identification analysis module 303 analyzes the behavior status identifier to determine the trigger condition of the physiological abnormality characteristic, it is used for: Based on the dynamic analysis threshold, analyze the physiological abnormality characteristic to determine the abnormal deviation degree; Determine the physiological abnormality level according to the abnormal deviation degree; Match the physiological abnormality level with the behavior status identifier to obtain the matching degree; Determine the trigger condition of the physiological abnormality characteristic according to the matching degree.

[0121] Optionally, the real-time multi-source user dataset includes life records. When the solution determination module 305 determines a personalized intervention solution based on the trigger condition and according to the user individual characteristics, it is used for: Analyze the life records to determine the user's living habits; Based on the trigger condition, analyze the user individual characteristics and the physiological abnormality characteristics to determine the feature-abnormality mapping relationship; Speculate the stress source according to the mapping relationship; Determine a personalized intervention solution according to the stress source.

[0122] Optionally, when the solution determination module 305 determines the feature-abnormality mapping relationship based on the trigger condition and analyzes the user individual characteristics and the physiological abnormality characteristics, it is used for: Based on the trigger condition, analyze the user individual characteristics to determine event-related characteristics; Analyze the event-related characteristics to determine the co-occurrence frequency with the physiological abnormality characteristics; Determine the strong association rule set according to the co-occurrence frequency; Determine the mapping relationship between features and anomalies according to the set of strong association rules.

[0123] Optionally, the historical multi-source user dataset includes user attribute data; when the data analysis module 302 analyzes the heart rate period distribution characteristics and determines the dynamic analysis threshold, it is used for: Cluster the heart rate period distribution characteristics according to a preset continuous period through a data statistical algorithm to obtain a set of heart rate data for the same period of several consecutive N days; Calculate the standard deviation and mean of the set of heart rate data; Analyze the user attribute data to determine the age group and physical examination data to which the user belongs; Determine the user's medical history according to the physical examination data; Determine a threshold relationship index according to the age group to which the user belongs and the user's medical history; Determine the dynamic analysis threshold according to the standard deviation, the mean, and the threshold relationship index.

[0124] Optionally, when the data analysis module 302 determines the threshold relationship index according to the age group to which the user belongs and the user's medical history, it is used for: Establish a feature vector including age, BMI index, and cardiovascular disease markers; Input the feature vector into a pre-trained risk coefficient prediction model to output a threshold relationship index.

[0125] Optionally, the health data management service system further includes a policy determination module 306, which is used for: When the trigger condition is satisfied, obtain user interaction data; Analyze the user interaction data to determine the user terminal interaction ability; Determine a dynamic return visit policy according to the user terminal interaction ability.

[0126] The system of this embodiment can be used to execute the method of any of the above embodiments, and its implementation principle and technical effects are similar, which will not be elaborated here.

Claims

1. A method for health data management service, characterized in that, Including: Obtain a historical multi-source user dataset, where the historical multi-source user dataset includes physiological index data and behavior status identifiers; Analyze the physiological index data to determine physiological abnormality characteristics; Analyze the behavior status identifiers to determine the triggering conditions of the physiological abnormality characteristics; Obtain the real-time multi-source user dataset of the user to be managed, analyze the real-time multi-source user dataset, and determine the user individual characteristics of the user to be managed; Based on the triggering conditions, according to the user individual characteristics, determine a personalized intervention plan.

2. The method according to claim 1, wherein The obtaining of the historical multi-source user dataset includes: Obtain data attributes; parse the data attributes to determine sensitive data types; According to the sensitive data types, perform dynamic desensitization processing on the original data to obtain desensitized original data; Use the desensitized original data as the historical multi-source user dataset.

3. The method according to claim 1, wherein The analyzing of the physiological index data to determine physiological abnormality characteristics includes: Parse the physiological index data to determine historical heart rate data; Parse the historical heart rate data to determine the heart rate period distribution characteristics; Analyze the heart rate period distribution characteristics to determine the dynamic analysis threshold; Based on the dynamic analysis threshold, analyze the physiological index data to determine physiological abnormality characteristics.

4. The method according to claim 3, wherein The analyzing of the behavior status identifiers to determine the triggering conditions of the physiological abnormality characteristics includes: Based on the dynamic analysis threshold, analyze the physiological abnormality characteristics to determine the abnormal deviation degree; According to the abnormal deviation degree, determine the physiological abnormality level; Match the physiological abnormality level with the behavior status identifiers to obtain a matching degree; According to the matching degree, determine the triggering conditions of the physiological abnormality characteristics.

5. The method according to claim 1, wherein The real-time multi-source user dataset includes life records. The determining of the personalized intervention plan based on the triggering conditions and according to the user individual characteristics includes: Analyze the life records to determine the user's living habits; Based on the triggering conditions, analyze the user individual characteristics and the physiological abnormality characteristics to determine the mapping relationship between characteristics and abnormalities; According to the mapping relationship, infer the stress source; According to the stress source, determine a personalized intervention plan.

6. The method according to claim 5, wherein The determining of the mapping relationship between characteristics and abnormalities based on the triggering conditions, analyzing the user individual characteristics and the physiological abnormality characteristics includes: Based on the triggering conditions, analyze the user individual characteristics to determine event-related characteristics; Analyze the event-related characteristics to determine the co-occurrence frequency with the physiological abnormality characteristics; According to the co-occurrence frequency, determine a set of strong association rules; According to the set of strong association rules, determine the mapping relationship between characteristics and abnormalities.

7. The method according to claim 3, wherein The historical multi-source user dataset includes user attribute data. The analyzing of the heart rate period distribution characteristics Determining the dynamic analysis threshold includes: Through a data statistical algorithm, cluster the heart rate period distribution characteristics according to a preset continuous period to obtain several heart rate data sets for the same period of consecutive N days; Calculate the standard deviation and mean of the heart rate data set; Parse the user attribute data to determine the age group and physical examination data to which the user belongs; According to the physical examination data, determine the user's medical history; Determine a threshold relationship index according to the age group to which the user belongs and the user's medical history; Determine a dynamic analysis threshold according to the standard deviation, the mean value, and the threshold relationship index.

8. The method according to claim 7, wherein The determining the threshold relationship index according to the age group to which the user belongs and the user's medical history includes: Establish a feature vector including age, BMI index, and cardiovascular disease markers; Input the feature vector into a pre-trained risk coefficient prediction model to output a threshold relationship index.

9. The method according to claim 1, wherein After analyzing the physiological abnormality features and the behavior status identifier and determining the triggering conditions for abnormal physiological events, it further includes: When the triggering conditions are met, obtain user interaction data; Analyze the user interaction data to determine the user terminal interaction ability; Determine a dynamic return visit strategy according to the user terminal interaction ability.

10. A health data management service system, characterized in that, Applied to the method according to any one of claims 1-9, it is characterized by including: A data acquisition module for acquiring a historical multi-source user data set, where the historical multi-source user data set includes physiological index data and behavior status identifiers; A data analysis module for analyzing the physiological index data to determine physiological abnormality features; An identifier analysis module for analyzing the behavior status identifier to determine the triggering conditions for the physiological abnormality features; A feature determination module for acquiring a real-time multi-source user data set of a user to be managed, analyzing the real-time multi-source user data set, and determining the user individual features of the user to be managed; A solution determination module for determining a personalized intervention solution based on the triggering conditions and according to the user individual features.

Images