Sanitary and medical safety monitoring and notification early warning method

Abstract

The invention relates to the technical field of data processing, and discloses a health and medical safety monitoring and notification early warning method, and the method comprises the steps: obtaining an initial heterogeneous data set of a target terminal; cleaning the initial heterogeneous data set, and filtering a cleaning result to obtain a standard security data stream; performing two-dimensional verification on the real-time data quality and logic consistency of the standard security data stream to obtain a clean data pool; associating the clean data pool, analyzing an associated result, identifying a compound safety risk, and calculating a risk dynamic weight of the compound safety risk; comprehensively judging the potential influence range and the upgrading probability of the composite security risk to obtain a risk situation report; mapping the risk level and the prediction influence range in the risk situation report to a preset differentiated early warning strategy matrix, and generating and distributing a directional early warning instruction to the target terminal; the efficiency of health and medical safety monitoring and notification early warning can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for monitoring, reporting and early warning of health and medical safety. Background Technology

[0002] Existing technologies have significant shortcomings in the data processing stage of health and medical safety monitoring. They fail to systematically integrate and standardize heterogeneous monitoring data such as vital signs, medical order execution, and environmental disinfection, and simply summarize scattered data, resulting in inconsistent data formats and asynchronous time sequences, making it difficult to form a complete and effective monitoring dataset. Furthermore, they fail to purify data through multi-dimensional filtering and dual-dimensional verification, performing only basic missing value filling, which cannot eliminate clinically unreasonable, logically contradictory, and low-reliability data. This leads to inconsistent monitoring data quality, containing a large amount of invalid or interfering information, making it difficult to accurately support subsequent risk identification and analysis, and failing to meet the stringent requirements for data accuracy in medical safety monitoring.

[0003] Existing technologies have significant shortcomings in the identification and early warning notification of medical safety risks. They fail to conduct cross-modal correlation and collaborative analysis of clean data, relying solely on a single dimension to identify complex safety risks stemming from multiple intertwined factors. Furthermore, they do not calculate dynamic risk weights and combine them with historical risk databases for comprehensive judgment, instead relying on fixed thresholds to determine risk levels. This fails to accurately assess the potential impact and escalation probability of risks, resulting in a lack of scientific rigor and foresight in risk assessment. Finally, they lack a differentiated early warning strategy matrix, employing only a standardized notification process and format. This prevents the targeted distribution of early warning instructions based on risk level and impact scope, leading to insufficient targeting and timeliness, and hindering precise prevention and efficient handling of medical safety risks. Summary of the Invention

[0004] This invention provides a method for monitoring, reporting, and early warning of health and medical safety, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for monitoring, reporting, and early warning of health and medical safety, comprising: S1. Acquire and integrate the heterogeneous monitoring data of the target terminal to obtain the initial heterogeneous dataset of the target terminal; S2. Standardize and clean the initial heterogeneous dataset, and perform multi-dimensional data filtering on the cleaned results to obtain the standard secure data stream of the target terminal; S3. Perform a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain the clean data pool of the target terminal; S4. Perform cross-modal data association on the clean data pool, and conduct collaborative analysis on the association results to identify the composite security risks in the target terminal, and calculate the dynamic risk weight of the composite security risks; S5. Based on the aforementioned dynamic risk weights and combined with the historical risk database of the target terminal, a comprehensive assessment is made of the potential impact range and escalation probability of the composite security risks to obtain a risk status report for the target terminal. S6. Map the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix to trigger a notification process that matches the risk scenario in the risk situation report, and generate and distribute targeted early warning instructions to the target terminal.

[0006] In a preferred embodiment, the step of acquiring and integrating the heterogeneous monitoring data of the target terminal to obtain the initial heterogeneous dataset of the target terminal includes: Collect time-series data of vital signs from the target terminal within a unit monitoring cycle; Extract medical order execution record data and environmental disinfection log data corresponding to the unit's monitoring cycle from the medical information records associated with the target terminal; The vital signs time-series data, the medical order execution record data, and the environmental disinfection log data are time-series aligned to obtain the aligned result of the target terminal; The aligned results are formatted and standardized to obtain the initial heterogeneous dataset of the target terminal.

[0007] In a preferred embodiment, the standardization and cleaning of the initial heterogeneous dataset, followed by multi-dimensional data filtering of the cleaned results to obtain the standard secure data stream for the target terminal, includes: Identify missing value entries and anomalous format entries in the initial heterogeneous dataset; Based on a pre-defined medical data rule base, the missing value entries and the format-abnormal entries are standardized and corrected to obtain complete data entries and format-normalized data of the initial heterogeneous dataset; The complete data entries and the formatted data are fused together, and the fused data is then merged with the data entries in the initial heterogeneous dataset that do not require processing to obtain the intermediate dataset of the target terminal. The data entries in the intermediate dataset are sorted by time topology to obtain an ordered data sequence of the target terminal; Based on the clinical rationality dimension of the medical data rule base, the ordered data sequence is initially filtered out to obtain a clinically rational subset of data for the target terminal. Based on the data credibility dimension of the medical data rule base, the ordered data sequence is filtered twice to obtain a highly reliable data subset of the target terminal. The intersection of the clinically reasonable subset of data and the highly reliable subset of data is taken as the effective dataset of the target terminal, and the effective dataset is serialized and encapsulated to obtain the standard secure data stream of the target terminal.

[0008] In a preferred embodiment, the step of performing a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain a clean data pool for the target terminal includes: The historical data quality performance records of the target terminal are associated and mapped with the standard security data stream to obtain the quality tolerance threshold table of the target terminal; Based on the quality tolerance threshold table, a real-time quality scan is performed on the data entries in the standard security data stream to obtain a dynamic quality defect list of the target terminal. The defective data entries in the dynamic quality defect list are removed from the standard security data stream to obtain the data stream to be verified for the target terminal. Perform a deep logical consistency check on the data entries in the data stream to be checked to obtain a list of logical contradictions of the target terminal; Based on the logical contradiction list, logical conflict entries in the data stream to be verified are removed to obtain the clean data pool of the target terminal.

[0009] In a preferred embodiment, the cross-modal data association of the clean data pool includes: Based on the medical data classification rules in the medical data rule base, the heterogeneous data streams in the clean data pool are divided by modal features to obtain the first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream of the target terminal. The first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream are aligned in the time dimension to obtain the aligned data stream of the target terminal. Interactively fuse data from different dimensions in the aligned data stream to obtain a multimodal data sequence for the target terminal; Semantic association parsing is performed on the multimodal data sequence to obtain the semantic association relationship of the target terminal; The semantic association is used as the cross-modal association result of the target terminal.

[0010] In a preferred embodiment, the step of collaboratively analyzing the associated results to identify complex security risks in the target terminal includes: A consistency analysis is performed on the cross-modal association results to obtain the consistency verification results of the target terminal; The cross-modal association results are subjected to association mining to obtain the multimodal association patterns of the target terminal; By coupling the consistency verification result and the multimodal association pattern with risk logic, potential security events of the target terminal can be obtained. Risk assessment is performed on the potential security events to obtain the complex security risks in the target terminal.

[0011] In a preferred embodiment, calculating the dynamic risk weight of the composite security risk includes: The composite security risk is normalized to obtain the risk characterization intensity of the composite security risk; The consistency between modes in the composite security risk is quantified to obtain the risk quantity of the composite security risk; Based on the intensity of the risk characterization, determine the time decay coefficient of the composite security risk within the current time window; Based on the risk characterization intensity, the risk quantity, and the time decay coefficient, the dynamic risk weight of the composite safety risk is calculated, wherein the calculation formula for the dynamic risk weight is as follows: ; In the formula, The dynamic weight of the risk, The intensity of the risk characterization, The time decay coefficient is... The duration of the intensity of the risk characterization. The risk level is... The coefficient representing the degree of consistency of the aforementioned risk quantity. The equilibrium intensity coefficient is the ratio of the intensity of the risk characterization. It is a natural exponential function.

[0012] In a preferred embodiment, the step of comprehensively judging the potential impact range and escalation probability of the complex security risk based on the dynamic risk weight and the historical risk database of the target terminal, and obtaining a risk status report for the target terminal, includes: Retrieve historical risk event records from the target terminal's historical risk database that are similar in type and representation to the composite security risk; Based on the aforementioned dynamic risk weights, a preliminary risk assessment is performed on the composite security risks to obtain their basic risk levels. Extract the actual impact range data and risk escalation path data of the historical risk events from the historical risk event records; The current characteristics of the composite security risk are compared with the initial characteristics of the historical risk event records using a multidimensional feature similarity metric to obtain the feature matching degree evaluation result of the target terminal. The potential impact range of the complex security risk is obtained by performing a weighted correlation mapping between the feature matching degree evaluation results and the actual impact range data of the historical risk events. Using the risk escalation path data and the changing trend of the risk dynamic weight as the judgment benchmark, the probability of the composite security risk escalating within a future time window is assessed; The risk status report of the target terminal is obtained by fusing structured information from the basic risk level, the potential impact range, and the escalation probability.

[0013] In a preferred embodiment, mapping the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix includes: The risk situation report is analyzed to extract the risk level identifier and the description of the predicted impact range. The risk level identifier and the predicted impact range description are jointly encoded to obtain the risk scenario identifier of the target terminal; Based on the risk level identifier, the early warning strategy codes in the preset differentiated early warning strategy matrix are selected; Based on the warning strategy code, the risk scenario identifier is logically mapped to obtain the matrix matching output of the target terminal.

[0014] In a preferred embodiment, the step of triggering a notification process that matches the risk scenario in the risk situation report, and generating and distributing a targeted early warning instruction to the target terminal, includes: Based on the matrix matching output, a notification process matching the risk scenario in the risk situation report is triggered to determine the notification channel and notification content template for the target terminal. Fill the core risk information from the risk situation report into the notification content template to obtain the core risk template for the target terminal; The core risk template is mapped into an instruction format to obtain a targeted early warning instruction for the target terminal; The targeted early warning instruction is distributed to the recipient of the target terminal through the notification channel.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention provides high-quality data support for health and medical safety monitoring through precise integration and purification of multi-source data. It comprehensively acquires heterogeneous monitoring data such as vital signs, medical order execution, and environmental disinfection, and constructs an initial dataset through time-series alignment and formatting. Missing and abnormal data are corrected through standardized cleaning, and multi-dimensional filtering based on clinical rationality and data credibility is performed. Finally, real-time data quality and logical consistency are verified to form a clean data pool, ensuring the integrity, accuracy, and reliability of the monitoring data, laying a solid foundation for risk identification.

[0016] 2. This invention significantly improves the targeting and timeliness of medical safety risk prevention and control by leveraging deep correlation analysis and precise early warning distribution. It performs cross-modal correlation and collaborative analysis on a clean data pool to accurately identify complex safety risks and calculate dynamic weights; combines a historical risk database to determine the potential impact range and escalation probability of risks, generating a structured risk situation report; and based on a differentiated early warning strategy matrix, triggers a notification process matching risk scenarios, generates targeted early warning instructions, and distributes them precisely, achieving early detection, early assessment, and early warning of medical safety risks, providing efficient decision support for medical safety assurance. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for monitoring, reporting, and issuing early warnings for health and medical safety, provided in an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a method for monitoring, reporting, and issuing early warnings for healthcare safety. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cluster of cloud servers. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1The diagram shown is a flowchart illustrating a method for monitoring, reporting, and issuing early warnings for healthcare safety according to an embodiment of the present invention. In this embodiment, the method includes: S1. Acquire and integrate the heterogeneous monitoring data of the target terminal to obtain the initial heterogeneous dataset of the target terminal; In this embodiment of the invention, the step of acquiring and integrating the heterogeneous monitoring data of the target terminal to obtain the initial heterogeneous dataset of the target terminal includes: Collect time-series data of vital signs from the target terminal within a unit monitoring cycle; Extract medical order execution record data and environmental disinfection log data corresponding to the unit's monitoring cycle from the medical information records associated with the target terminal; The vital signs time-series data, the medical order execution record data, and the environmental disinfection log data are time-series aligned to obtain the aligned result of the target terminal; The aligned results are formatted and standardized to obtain the initial heterogeneous dataset of the target terminal.

[0021] The target terminal is equipped with vital sign monitoring equipment, which continuously collects relevant data within a set monitoring cycle. This data includes indicators that reflect the body's condition, such as heart rate, respiratory rate, body temperature, and blood pressure. The changes in the values ​​of each indicator are recorded in chronological order of collection time, forming continuous and complete time-series data of vital signs.

[0022] Based on the medical information management system associated with the target terminal, the medical records corresponding to the terminal within the unit monitoring cycle are retrieved, and the doctor's orders and their actual execution status are extracted to form medical order execution record data. At the same time, the disinfection operation information of the target terminal's environment within the cycle is extracted and integrated to form environmental disinfection log data.

[0023] Based on the timeline of the unit monitoring cycle, the time nodes of vital signs time series data, medical order execution record data and environmental disinfection log data are sorted out respectively. The time information of medical order execution record data and environmental disinfection log data is matched with the collection time of vital signs time series data to ensure that the information of the three types of data at the same time node can be correlated with each other, eliminate the deviation in the time dimension, and obtain the aligned result of the target terminal.

[0024] By adopting a unified data format standard, the three types of data in the aligned result are formatted, and the representation and storage structure of different types of data are adjusted to a consistent standard form. This includes correcting the naming of data fields and the encoding of data content according to unified rules, so that the three types of data can be integrated in an orderly manner within the same dataset framework, and finally the initial heterogeneous dataset of the target terminal is obtained.

[0025] The beneficial effects are that by collecting time-series data of vital signs within the monitoring period, and continuously recording core indicators reflecting the body's condition such as heart rate and respiratory rate, the physiological state change trajectory of the target terminal within the monitoring period can be completely captured, providing direct and continuous physiological data support for medical safety monitoring and avoiding misjudgment of the state due to data fragmentation.

[0026] By extracting corresponding periodic medical order execution record data and environmental disinfection log data from the medical information records associated with the target terminal, medical intervention information and environmental safety information are supplemented in addition to physiological data. This enables the monitoring data to cover three core dimensions: physiological state, medical operation, and environmental conditions, solving the problem that single physiological data cannot fully support safety monitoring and enriching the diversity and relevance of the data.

[0027] The three types of data were time-series aligned. Using the time axis of the unit monitoring cycle as the benchmark, the time nodes of medical order execution and environmental disinfection were precisely matched with the collection time of vital sign data. This eliminated the time deviation of data from different sources and ensured that physiological states, medical operations and environmental conditions at the same time node could be correlated and verified. This laid the foundation for time synchronization for subsequent analysis of causal relationships between data.

[0028] The aligned results are formatted and standardized, and the representation, storage structure and field naming of the three types of data are adjusted according to preset standards. This enables the heterogeneous data to be integrated in a standardized manner within the same dataset framework, eliminating integration obstacles caused by differences in data formats. This results in an initial heterogeneous dataset with a unified structure and logical coherence, providing a standardized and efficient input foundation for subsequent data processing steps such as standardized cleaning and multi-dimensional filtering.

[0029] S2. Standardize and clean the initial heterogeneous dataset, and perform multi-dimensional data filtering on the cleaned results to obtain the standard secure data stream of the target terminal; In this embodiment of the invention, the standardization and cleaning of the initial heterogeneous dataset, followed by multi-dimensional data filtering of the cleaned results to obtain the standard secure data stream for the target terminal, includes: Identify missing value entries and anomalous format entries in the initial heterogeneous dataset; Based on a pre-defined medical data rule base, the missing value entries and the format-abnormal entries are standardized and corrected to obtain complete data entries and format-normalized data of the initial heterogeneous dataset; The complete data entries and the formatted data are fused together, and the fused data is then merged with the data entries in the initial heterogeneous dataset that do not require processing to obtain the intermediate dataset of the target terminal. The data entries in the intermediate dataset are sorted by time topology to obtain an ordered data sequence of the target terminal; Based on the clinical rationality dimension of the medical data rule base, the ordered data sequence is initially filtered out to obtain a clinically rational subset of data for the target terminal. Based on the data credibility dimension of the medical data rule base, the ordered data sequence is filtered twice to obtain a highly reliable data subset of the target terminal. The intersection of the clinically reasonable subset of data and the highly reliable subset of data is taken as the effective dataset of the target terminal, and the effective dataset is serialized and encapsulated to obtain the standard secure data stream of the target terminal.

[0030] A comprehensive scan of the initial heterogeneous dataset was conducted, and the completeness and format of each data entry were checked one by one. By comparing with the preset data entry requirements, missing value entries with missing key information or missing data fields were identified. At the same time, by verifying whether the data expression and structure conformed to the unified standard, entries with disordered or non-compliant formats were identified.

[0031] The system retrieves a pre-defined medical data rule base, which includes clinical data completion standards and format specifications. For missing value entries, it supplements the missing key information based on reasonable data completion logic for similar medical scenarios in the rule base, forming complete data entries. For entries with abnormal formats, it adjusts the data's representation and structure according to the format standardization standards in the rule base to make it conform to the unified specifications, resulting in format-standardized data.

[0032] By employing a data fusion method, complete data entries and formatted data are classified and integrated according to data type and time node to ensure that the two types of data are orderly associated within the same framework. At the same time, the fused data is summarized with the normal data entries in the initial heterogeneous dataset that do not require processing, and duplicate data is removed while retaining all valid information to form an intermediate dataset for the target terminal.

[0033] Based on time, the timestamp information of all data entries in the intermediate dataset is sorted out, and the data entries are arranged in chronological order to ensure that the data sequence can clearly show the relevant data changes at different time points, so that the temporal logic of the data is coherent and orderly, and an ordered data sequence of the target terminal is obtained.

[0034] Referring to the judgment criteria for the clinical rationality dimension in the medical data rule base, which covers the normal range of various medical data and the requirements for clinical logical association, each data entry in the ordered data sequence is checked one by one to determine whether the data conforms to common sense of clinical diagnosis and treatment and whether it is within a reasonable value range. Abnormal data that exceeds the reasonable range or does not conform to clinical logic is eliminated to obtain a subset of clinically rational data for the target terminal.

[0035] Based on the evaluation criteria for data credibility dimension in the medical data rule base, which includes requirements such as the reliability of data acquisition equipment and the completeness and consistency of data records, the credibility of data entries in the ordered data sequence is checked. Data with qualified acquisition equipment, complete records and no logical contradictions are selected, while unreliable data caused by abnormal acquisition equipment or disordered records are excluded, thus obtaining a highly reliable subset of data for the target terminal.

[0036] By comparing the clinically reasonable subset of data with the highly reliable subset of data, the data items common to both subsets are extracted. These intersection data are used as the effective dataset of the target terminal. Then, according to the preset serialization encapsulation rules, the effective dataset is structurally organized and format standardized to form a standardized and directly usable data stream, and finally the standard secure data stream of the target terminal is obtained.

[0037] The beneficial effects are that by comprehensively scanning the initial heterogeneous dataset and comparing it with the preset data entry requirements and format specifications, it accurately identifies missing entries with incomplete key information or missing data fields, as well as entries with disordered or non-compliant formatting. This operation identifies data quality issues at the source, prevents defective data from entering subsequent processing stages, provides clear targets for data standardization and correction, and ensures the targeted and effective nature of subsequent data processing.

[0038] Based on a pre-defined medical data rule base, which includes clinical data completion standards and format specifications, the system supplements missing value entries with key information according to reasonable data completion logic in similar medical scenarios, forming complete data entries. For entries with abnormal formats, the expression and structure are adjusted according to a unified format standard, resulting in well-formatted data. Standardization correction eliminates data integrity defects and format differences, ensuring that the data conforms to medical data specifications and laying a unified foundation for subsequent data fusion and filtering.

[0039] The complete data entries and formatted data are categorized and integrated according to data type and time point, achieving an organic fusion of the two types of corrected data. Simultaneously, the fused data is combined with the normal data entries from the initial heterogeneous dataset that do not require processing, eliminating duplicate data and retaining all valid information to form an intermediate dataset. This process integrates the corrected data with the original valid data, avoiding data omissions, ensuring the integrity of the dataset, and providing comprehensive data support for subsequent multi-dimensional filtering.

[0040] Based on timestamps, all data entries in the intermediate dataset are sorted chronologically to clarify the temporal logical relationships of the data and form an ordered data sequence. The sorted dataset clearly presents the changes in medical-related data at different time points, making the temporal correlation of the data clearer and providing an ordered data foundation for subsequent judgments on clinical rationality and data credibility based on time-series characteristics.

[0041] Referring to the criteria for clinical rationality in the medical data rule base, including the normal range of medical data and requirements for clinical logical correlation, each data entry in the ordered data sequence is verified one by one. Abnormal data that exceeds the reasonable value range or does not conform to common sense in clinical diagnosis and treatment is removed, resulting in a subset of clinically rational data. This filtering process ensures that the data conforms to medical clinical logic, eliminates data interference that obviously violates medical common sense, and improves the clinical applicability of the data.

[0042] Based on the data credibility assessment criteria in the medical data rule base, which cover requirements such as the reliability of data acquisition equipment, the completeness and consistency of records, ordered data sequences are verified. Data with qualified acquisition equipment, complete records, and no logical contradictions are selected, while unreliable data caused by abnormal acquisition equipment or disordered records are excluded, resulting in a highly reliable data subset. A second filtering process further purifies the data, ensuring its reliability and providing high-quality data assurance for subsequent security risk identification.

[0043] The process involves extracting common data entries from both clinically reasonable and highly reliable subsets of data to obtain a valid dataset that simultaneously satisfies clinical rationality and high reliability, ensuring that the data conforms to medical logic and has a reliable source. Subsequently, the valid dataset is structurally organized and formatted according to preset rules, and then serialized and packaged into a standardized, secure data stream. This process ultimately results in a data stream with a unified structure, reliable quality, and direct usability for subsequent analysis, providing accurate and efficient data input for health and medical safety monitoring and early warning.

[0044] S3. Perform a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain the clean data pool of the target terminal; In this embodiment of the invention, the step of performing a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain a clean data pool for the target terminal includes: The historical data quality performance records of the target terminal are associated and mapped with the standard security data stream to obtain the quality tolerance threshold table of the target terminal; Based on the quality tolerance threshold table, a real-time quality scan is performed on the data entries in the standard security data stream to obtain a dynamic quality defect list of the target terminal. The defective data entries in the dynamic quality defect list are removed from the standard security data stream to obtain the data stream to be verified for the target terminal. Perform a deep logical consistency check on the data entries in the data stream to be checked to obtain a list of logical contradictions of the target terminal; Based on the logical contradiction list, logical conflict entries in the data stream to be verified are removed to obtain the clean data pool of the target terminal.

[0045] Retrieve historical data quality performance records of the target terminal. These records cover quality-related information such as the completeness, format standardization, and reliability of historical data. Associate and map this historical quality information with the current standard secure data stream. Based on the quality compliance of different types of data in the historical data, determine the acceptable quality fluctuation range of each type of data in the current standard secure data stream, and form a quality tolerance threshold table for the target terminal.

[0046] Using the quality tolerance threshold table as the judgment standard, each data entry in the standard security data stream is scanned in real time. The integrity of the data is checked one by one to see if it meets the threshold requirements, whether the format is standardized, and whether key information is missing. All unqualified data entries that do not meet the quality tolerance threshold are summarized and recorded. The problem type and specific location of each defective data are identified, and a dynamic quality defect list of the target terminal is obtained.

[0047] Based on the defect data entries recorded in the dynamic quality defect list, each corresponding defect data entry is accurately located in the standard security data stream. All defect data entries that do not meet the quality requirements are removed to ensure that the remaining data meets the basic requirements of the quality tolerance threshold table. The retained valid data is then integrated to form the data stream to be verified for the target terminal.

[0048] Deep logical consistency verification is performed on all data entries in the verification data stream. The inherent logical relationships between data are sorted out, including the matching relationship of different types of data at the same time node and the change pattern of data between different time nodes. It is checked whether there are data contradictions or changes that do not conform to normal logic. The logically conflicting data entries are recorded in detail, the conflict type and the data content involved are identified, and a list of logical contradictions of the target terminal is obtained.

[0049] Based on the logical conflict items clearly defined in the logical contradiction list, the corresponding conflicting data is found in the data stream to be verified and all of them are removed to ensure that there are no logical contradictions among the remaining data items and that the internal relationship of the data conforms to normal rules and actual scenarios. After the dual screening of quality and logic, the remaining high-quality, logically conflict-free data is collected to form a clean data pool for the target terminal.

[0050] The beneficial effects include retrieving historical data quality performance records from the target terminal, covering core quality information such as the completeness, format standardization, and reliability of the historical data, and mapping these historical quality characteristics to the current standard secure data stream. Based on the quality compliance patterns of different types of data in the historical data, and combined with the needs of the current monitoring scenario, the acceptable quality fluctuation range for each type of data in the current standard secure data stream is determined, forming a quality tolerance threshold table for the target terminal. This threshold table provides personalized judgment standards tailored to the actual situation of the target terminal for subsequent data quality scanning, avoiding misjudgments or omissions caused by using uniform fixed thresholds, and improving the targeting and accuracy of data quality verification.

[0051] Using a quality tolerance threshold table as a clear basis for judgment, real-time quality scanning is conducted on each data entry in the standard secure data stream. Each entry is checked to verify whether its completeness meets the threshold requirements, whether its format conforms to specifications, and whether key information is missing. All non-compliant data entries that fail to meet the quality tolerance threshold are systematically compiled, with detailed records of the problem type, specific location, and non-compliance for each defective data entry, forming a dynamic quality defect list for the target terminal. This list clearly presents all data quality issues, providing clear guidance for the subsequent accurate removal of defective data and ensuring that no non-compliant data is overlooked during data quality verification.

[0052] Based on the defect data entries recorded in the dynamic quality defect list, each corresponding defective data entry is precisely located in the standard secure data stream, and all defective data entries that do not meet quality requirements are removed. During the removal process, valid data that meets the requirements of the quality tolerance threshold table is strictly retained, ensuring that the remaining data meets the standards in basic quality dimensions such as completeness and format conformity, and is then integrated to form the data stream to be verified for the target terminal. This step effectively purifies the data source, removes interference factors at the data quality level, provides high-quality basic data for subsequent logical consistency verification, and avoids the impact of defective data on the logical verification results.

[0053] A deep logical consistency check is performed on all data entries in the verification data stream. This involves thoroughly analyzing the inherent logical relationships between data, including the matching relationships of different types of data at the same time point, the patterns of data change across different time points, and the degree of alignment between the data and common medical knowledge. A meticulous check is conducted to identify any data contradictions, illogical trends, or discrepancies with clinical practice. Detected logically conflicting data entries are recorded in detail, specifying the type of conflict, the data content involved, and the logical contradiction points, resulting in a list of logical contradictions for the target terminal. This list comprehensively exposes problems at the data logic level, providing a precise basis for subsequent removal of logically conflicting data and ensuring the logical consistency and rationality of the data.

[0054] Based on the clearly defined logical conflict entries in the logical contradiction list, corresponding conflicting data is precisely located in the data stream to be verified and completely removed, ensuring that there are no logical contradictions among the remaining data entries and that the inherent relationships between the data conform to normal patterns and actual medical scenarios. After dual screening for quality and logic, the remaining data is aggregated to form a clean data pool for the target terminal. The data in this pool meets basic quality requirements and possesses logical consistency, ensuring the integrity, accuracy, and reliability of the data. The clean data pool provides high-quality data support for subsequent core processes such as cross-modal data association and identification of complex security risks, ensuring the accuracy of medical safety monitoring and early warning analysis from the source.

[0055] S4. Perform cross-modal data association on the clean data pool, and conduct collaborative analysis on the association results to identify the composite security risks in the target terminal, and calculate the dynamic risk weight of the composite security risks; In this embodiment of the invention, the cross-modal data association of the clean data pool includes: Based on the medical data classification rules in the medical data rule base, the heterogeneous data streams in the clean data pool are divided by modal features to obtain the first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream of the target terminal. The first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream are aligned in the time dimension to obtain the aligned data stream of the target terminal. Interactively fuse data from different dimensions in the aligned data stream to obtain a multimodal data sequence for the target terminal; Semantic association parsing is performed on the multimodal data sequence to obtain the semantic association relationship of the target terminal; The semantic association is used as the cross-modal association result of the target terminal.

[0056] The collaborative analysis of the associated results identifies complex security risks in the target terminal, including: A consistency analysis is performed on the cross-modal association results to obtain the consistency verification results of the target terminal; The cross-modal association results are subjected to association mining to obtain the multimodal association patterns of the target terminal; By coupling the consistency verification result and the multimodal association pattern with risk logic, potential security events of the target terminal can be obtained. Risk assessment is performed on the potential security events to obtain the complex security risks in the target terminal.

[0057] The medical data classification rules in the medical data rule base are retrieved. These rules clarify the classification standards and characteristic identifiers of different types of data, such as physiological parameters, medication orders, and environmental monitoring. Based on these rules, the heterogeneous data streams in the clean data pool are identified one by one. Data reflecting physical condition is classified into the first physiological parameter time series data stream, doctor's guidance and medication-related records are classified into the second medical order and medication record data stream, and environmental monitoring data is classified into the third environmental monitoring data stream.

[0058] Using a unified timeline as a benchmark, the time information of each data entry in the first physiological parameter time series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream is extracted. By comparing timestamps, the data in the same time node or time interval in the three types of data streams are accurately matched to eliminate the time deviation between different data streams and ensure that the three types of data are consistent in the time dimension, so as to obtain the aligned data stream of the target terminal.

[0059] By employing a data interaction and fusion method, data from different dimensions in the aligned data stream are integrated and processed. Physiological parameters, medical prescriptions, and environmental monitoring data at the same time point are bound according to their relationships to form a comprehensive data unit containing multi-dimensional information. These data units are then arranged in chronological order to construct a multimodal data sequence that can comprehensively reflect the relevant situation of the target terminal.

[0060] By deeply analyzing the intrinsic connections between different dimensions of data in multimodal data sequences, and combining medical common sense with data association logic, we can analyze the causal relationship between changes in physiological parameters and the execution of medical orders and medication, as well as the possible impact of environmental monitoring data on physiological parameters. This will clarify the semantic relationships between various types of data and form a semantic relationship that clearly describes the association logic between data.

[0061] The semantic relationships obtained from parsing are directly determined as the cross-modal association results of the target terminal. This result fully presents the inherent logic and mutual influence between different modal data, providing a clear data association basis for subsequent analysis, decision-making and other operations based on multimodal data, and ensuring that the comprehensive value of multi-source heterogeneous data can be fully explored.

[0062] Referring to the consistency judgment criteria in the medical data rule base, and combining clinical diagnosis and treatment logic with data association specifications, a comprehensive analysis of cross-modal association results is conducted. This involves verifying whether the descriptions between different modal data are consistent, whether the data change trends conform to normal patterns, and whether the semantic associations are consistent. By verifying the rationality of each association relationship one by one, it is determined whether there are any conflicts between cross-modal data, and the consistency verification results of the target terminal are obtained.

[0063] By employing association mining methods, we deeply analyze the intrinsic connections between different modal data in cross-modal association results, sort out the synergistic change patterns among physiological parameter time series data, medical order and medication record data, and environmental monitoring data, capture the linkage response patterns of multiple types of data under the same event trigger, summarize the recurring association features and combination forms between data, and extract a multimodal association pattern for the target terminal that can reflect the interaction patterns of multimodal data.

[0064] A risk logic coupling framework is established, taking consistency verification results as the basic constraint and multimodal association patterns as the core analysis basis. The two are deeply integrated. By analyzing the non-conflicting associations in the consistency verification results and the abnormal linkage patterns in the multimodal association patterns, potential causes that may trigger security issues are identified. Combined with the risk occurrence logic in the medical scenario, potential security events of the target terminal that include risk triggering conditions, scope of impact, and manifestation form are integrated.

[0065] Based on the preset risk assessment criteria, which cover the severity classification, probability assessment, and impact consequence definition of medical safety risks, potential safety events are comprehensively assessed to determine whether the event is caused by the combined effect of multiple factors and whether it involves different dimensions of safety hazards reflected by multimodal data. Composite risk events triggered by cross-modal data association are screened out, their risk levels and core influencing factors are clarified, and composite safety risks in the target terminal are obtained.

[0066] The beneficial effects are as follows: relying on the classification rules in the medical data rule base, the characteristic identifiers and classification standards of three types of data—physiological parameters, medication orders, and environmental monitoring—are clearly defined, enabling precise identification and classification of heterogeneous data streams in the clean data pool. Data reflecting physical condition are categorized into the first physiological parameter time-series data stream, medical intervention-related records into the second medication order and record data stream, and environmental condition data into the third environmental monitoring data stream. This achieves the orderly splitting and precise classification of heterogeneous data, avoiding confusion between different types of data and laying a clear data foundation for subsequent cross-modal correlation.

[0067] Using a unified timeline as a benchmark, the time information of each data entry in the three types of data streams is extracted. By comparing timestamps, the three types of data within the same time node or time interval are accurately matched, eliminating time deviations between different modalities. This ensures that data on physiological states, medical procedures, and environmental conditions can be correlated under the same time dimension, enabling subsequent fusion analysis to have temporal consistency, avoiding logical errors in association caused by asynchronous time sequences, and improving the accuracy of cross-modal data association.

[0068] Perform interactive fusion on the three types of time-aligned data streams, bind physiological parameters, doctor's order medication records, and environmental monitoring data at the same time node according to their internal relationships, and form a comprehensive data unit containing multi-dimensional information. Then arrange these data units in order of time sequence to construct a multi-modal data sequence, realizing the organic integration of different modal data, breaking the information limitation of single-modal data, and providing comprehensive information support for mining the deep relationships between data.

[0069] Deeply analyze the internal logic of data in different dimensions in the multi-modal data sequence, combine medical common sense and data association rules, and analyze the causal relationships between changes in physiological parameters, doctor's order execution, and medication conditions, as well as the possible impacts of environmental monitoring data on physiological states. Clarify the semantic associations between various types of data to form clear semantic association relationships, enabling the originally scattered heterogeneous data to have meaningful associations, and providing a key association basis for subsequent composite safety risk identification.

[0070] Directly use the parsed semantic association relationships as cross-modal association results, fully presenting the internal logic and mutual influences between different modal data. This result provides clear data association support for subsequent collaborative analysis, ensuring the full mining of the comprehensive value of multi-source heterogeneous data, solving the problem that traditional single-dimensional data is difficult to identify risks intertwined by multiple factors, and laying a solid foundation for accurately identifying composite safety risks.

[0071] Refer to the consistency determination criteria in the medical data rule library, combine clinical diagnosis and treatment logic and data association specifications, and comprehensively verify the fit of different modal data in the cross-modal association results. By verifying the description consistency, change trend coordination, and semantic association rationality among physiological parameters, doctor's order medication, and environmental monitoring data, clarify whether there are conflicting situations to obtain a consistency verification result. This result eliminates the interference of data contradictions on risk identification, ensuring that subsequent analysis is based on data association relationships without logical conflicts, and laying a foundation for accurately identifying risks.

[0072] Adopt an association mining method to deeply analyze the cross-modal association results, sort out the co-variation rules of different modal data, and capture the联动 reaction mode of multiple types of data triggered by the same event. By summarizing the repeatedly occurring association features and combination forms between data, refine and form a multi-modal association mode, clearly presenting the internal interaction logic between data. This mode breaks the information limitation of single-modal data, can挖掘出 the deep associations hidden in multi-source heterogeneous data, and provides a key basis for identifying risks intertwined by multiple factors.

[0073] A risk logic coupling framework is established, using consistency verification results as the basic constraint and multimodal association patterns as the core analytical basis for deep integration. By analyzing conflict-free associations and abnormal linkage patterns, combined with the risk occurrence logic in medical scenarios, potential triggers that may cause safety issues are identified and integrated into potential safety events that include triggering conditions, scope of impact, and manifestation forms. This process achieves an organic combination of data association and risk logic, transforming abstract data associations into concrete risk events, making risk identification more targeted.

[0074] Based on pre-defined risk assessment criteria, encompassing risk severity grading, probability of occurrence assessment, and definition of impact consequences, a comprehensive evaluation of potential security incidents is conducted. The focus is on determining whether an incident is caused by the combined effects of multiple factors and whether it involves different dimensions of security vulnerabilities reflected in multimodal data. Complex risk events are identified, and their risk levels and core influencing factors are clearly defined, resulting in complex security risks. This assessment process accurately distinguishes between single and complex risks, ensuring a focus on complex security vulnerabilities involving multiple intertwined factors, and providing precise targeting for subsequent risk weight calculations and early warning responses.

[0075] S5. Based on the aforementioned dynamic risk weights and combined with the historical risk database of the target terminal, a comprehensive assessment is made of the potential impact range and escalation probability of the composite security risks to obtain a risk status report for the target terminal. In this embodiment of the invention, calculating the dynamic risk weight of the composite security risk includes: The composite security risk is normalized to obtain the risk characterization intensity of the composite security risk; The consistency between modes in the composite security risk is quantified to obtain the risk quantity of the composite security risk; Based on the intensity of the risk characterization, determine the time decay coefficient of the composite security risk within the current time window; Based on the risk characterization intensity, the risk quantity, and the time decay coefficient, the dynamic risk weight of the composite safety risk is calculated, wherein the calculation formula for the dynamic risk weight is as follows: ; In the formula, The dynamic weight of the risk, The intensity of the risk characterization, The time decay coefficient is... The duration of the intensity of the risk characterization. The risk level is... The coefficient representing the degree of consistency of the aforementioned risk quantity. The equilibrium intensity coefficient is the ratio of the intensity of the risk characterization. It is a natural exponential function.

[0076] Based on the dynamic risk weights and combined with the historical risk database of the target terminal, a comprehensive assessment is made of the potential impact range and escalation probability of the complex security risks, resulting in a risk situation report for the target terminal, including: Retrieve historical risk event records from the target terminal's historical risk database that are similar in type and representation to the composite security risk; Based on the aforementioned dynamic risk weights, a preliminary risk assessment is performed on the composite security risks to obtain their basic risk levels. Extract the actual impact range data and risk escalation path data of the historical risk events from the historical risk event records; The current characteristics of the composite security risk are compared with the initial characteristics of the historical risk event records using a multidimensional feature similarity metric to obtain the feature matching degree evaluation result of the target terminal. The potential impact range of the complex security risk is obtained by performing a weighted correlation mapping between the feature matching degree evaluation results and the actual impact range data of the historical risk events. Using the risk escalation path data and the changing trend of the risk dynamic weight as the judgment benchmark, the probability of the composite security risk escalating within a future time window is assessed; The risk status report of the target terminal is obtained by fusing structured information from the basic risk level, the potential impact range, and the escalation probability.

[0077] By adopting a pre-defined normalization process, the original risk characteristics of complex safety risks are transformed into a unified standard. First, the risk intensity range corresponding to various manifestations of complex safety risks is clarified. Risk characteristics of different dimensions and magnitudes are mapped to a unified assessment range according to a fixed ratio, eliminating the influence of dimensional differences and numerical spans among various risk characteristics. This allows the transformed results to intuitively reflect the strength of the risk, ultimately yielding the risk characterization intensity of complex safety risks.

[0078] Based on the quantitative standards for modal consistency in the medical data rule base, this study analyzes the degree of fit between different modal data in complex safety risks, including the logical consistency and trend coordination of data such as physiological parameter modalities, medication prescription modalities, and environmental monitoring modalities. By comparing the consistency of each modal data in representing the same risk event, the risk value corresponding to the degree of consistency is quantified according to the standard. This value directly reflects the joint support of different modalities for the risk, thus obtaining the risk quantity of complex safety risks.

[0079] Referring to the standard for the correspondence between risk characterization intensity and time decay coefficient, this standard is formulated based on the natural decay law of risk in different time windows. The higher the risk characterization intensity, the more significant the current impact of the risk, the smaller the corresponding time decay coefficient, and the slower the risk decay rate. Conversely, the lower the risk characterization intensity, the larger the time decay coefficient, and the faster the risk decay rate. Based on the risk characterization intensity of the current complex safety risk, the corresponding time decay coefficient is accurately matched to determine the time decay coefficient of the complex safety risk in the current time window.

[0080] Based on the risk characterization intensity, and combined with the reinforcing effect of risk quantity on risk and the weakening effect of time decay coefficient on risk, a comprehensive calculation is performed. The risk characterization intensity is multiplied by the risk quantity to strengthen the core degree of risk reflected by both, and then fused with the time decay coefficient to balance the impact of risk changes over time. Through this logical integration, a value that can dynamically reflect the current actual importance of risk is obtained, which is the dynamic risk weight of composite safety risk.

[0081] The risk characterization intensity comes from the normalization of complex safety risks. First, the original risk characteristics and their respective risk intensity ranges corresponding to various manifestations of complex safety risks are identified. Then, a preset normalization method is used to map the original risk characteristics of different dimensions and magnitudes to a unified assessment range in a fixed proportion, eliminating the influence of differences in dimensions and numerical spans. Finally, the risk characterization intensity that can intuitively reflect the strength of the risk is obtained.

[0082] The time decay coefficient is determined based on the risk characterization intensity and refers to the preset standard of correspondence between risk characterization intensity and time decay coefficient. This standard is formulated based on the natural decay law of risk in different time windows. The higher the risk characterization intensity, the smaller the corresponding time decay coefficient and the slower the risk decay rate. Conversely, the lower the risk characterization intensity, the larger the time decay coefficient. The corresponding time decay coefficient is accurately matched according to the risk characterization intensity of the current complex safety risk.

[0083] The duration of risk characterization intensity refers to the time elapsed from the start of the risk characterization intensity of a complex security risk reaching its current level to the present moment. This duration is calculated directly by recording the start time of the risk characterization intensity reaching the current level and combining it with the current system time.

[0084] The risk quantity is derived from the risk quantification of the consistency between modalities in complex safety risks. Based on the quantification standard of modal consistency in the medical data rule base, the logical consistency and trend coordination between different modal data such as physiological parameter modality, medication order modality, and environmental monitoring modality are analyzed. The degree of consistency of each modal data in representing the same risk event is compared, and the corresponding risk value is obtained by quantification according to the standard, that is, the risk quantity.

[0085] The consistency coefficient of risk quantity is a preset fixed coefficient. Based on a large amount of medical safety risk sample data, combined with the degree of influence of risk quantity on the dynamic weight of risk in different scenarios, the optimal value is determined through statistical analysis. Its role is to adjust the contribution of risk quantity in the calculation of dynamic weight of risk, and to ensure that the influence of risk quantity meets the actual medical safety assessment needs.

[0086] The balance strength coefficient of risk characterization intensity is a preset fixed coefficient. Based on the core importance of risk characterization intensity in the dynamic risk weight and combined with the actual focus of medical safety assessment, an appropriate value is determined through multiple rounds of scenario verification. This value is used to balance the weight ratio of risk characterization intensity and other parameters in the calculation, so that the final dynamic risk weight can accurately reflect the core characteristics of the risk.

[0087] The significance of this formula lies in its ability to accurately determine the dynamic risk weights of complex safety risks through multi-parameter collaborative calculation, comprehensively considering the core intensity of the risk, its time decay effect, and the supporting strength of modal consistency. With risk representation intensity as the core, it dynamically reflects the decay law of risk over time through a natural exponential function combined with a time decay coefficient and duration; the longer the risk duration, the more significant the decay effect. Simultaneously, the risk quantity and its consistency coefficient reflect the joint supporting role of different modal data on the risk, while the balance intensity coefficient ensures that the core position of risk representation intensity is not weakened. The final calculated dynamic risk weights can reflect the current actual importance of complex safety risks in real time, comprehensively, and objectively, providing accurate quantitative basis for subsequent medical safety decisions.

[0088] The historical risk database of the target terminal is retrieved. This database stores complete records of various risk events that have occurred in the past, including information such as risk type, specific characteristics, and handling process. By comparing the type attributes and core characteristics of complex security risks, historical risk event records with the same type and similar characteristics in the historical risk database are selected to ensure that the selected records have reference value.

[0089] Using dynamic risk weights as the core criterion and referring to the preset risk level classification standard, which clarifies the risk level corresponding to different weight ranges, the dynamic risk weights of complex safety risks are compared with the standard. Based on the range of the weights, the corresponding risk level is determined, thus obtaining the basic risk level of complex safety risks, which intuitively reflects the current severity of the risks.

[0090] We conduct in-depth analysis of the selected historical risk event records, extracting data on the actual impact scope of each historical risk event, including the scope of objects affected, the business processes involved, and the areas covered by the consequences. At the same time, we sort out the risk escalation path data, including the key nodes, triggering conditions, and development paths of historical risk events from their initial state to their escalation, to provide historical reference for subsequent analysis.

[0091] The current risk type, characterization features, triggering factors, and involved modalities are extracted from the complex security risks. At the same time, the initial features in the historical risk event records are extracted. The feature similarity quantification method is used to compare the degree of fit between the two types of features in each dimension. The feature matching degree evaluation result of the target terminal is quantified based on the comparison results. This result reflects the similarity between the current risk and the historical risk.

[0092] Based on the feature matching degree evaluation results, a weighting rule is set so that the higher the matching degree, the greater the weight. The feature matching degree evaluation results are weighted and correlated with the actual impact range data of the corresponding historical risk events. By integrating the impact range information of similar historical risks and adjusting it in combination with the specific scenario of the current risk, the objects, links and fields that may be affected by the composite security risk are deduced, and the potential impact range of the composite security risk is obtained.

[0093] Using historical risk escalation path data as a reference, this study analyzes the consistency between the current development status of complex security risks and the initial stage in the escalation path data. It also tracks the changing trend of risk dynamic weights to determine whether the weights are continuously rising, remaining stable, or gradually declining. By combining these two factors, the study assesses the likelihood of complex security risks evolving from the current level to a higher level within a set future time window, thus obtaining the corresponding escalation probability.

[0094] By adopting a structured information fusion approach, the basic risk level, potential impact scope, and escalation probability are integrated according to a pre-set reporting framework. The presentation logic and correlation of each part of the information are clarified, and auxiliary information such as the core characteristics of the risk and key triggering factors are supplemented to form a risk situation report for the target terminal that is clear in structure, comprehensive in content, and accurate in data, providing complete information support for subsequent risk response decisions.

[0095] The beneficial effects are that by adopting a unified normalization method, the original risk characteristics and their respective risk intensity ranges corresponding to various manifestations of complex safety risks are clearly defined, and original risk characteristics of different dimensions and magnitudes are mapped to a unified assessment interval according to a fixed ratio. This operation completely eliminates the influence of dimensional differences and numerical ranges among various risk characteristics, enabling the transformed risk representation intensity to intuitively and accurately reflect the strength of the risk, and providing core basic data with a unified dimension for the subsequent collaborative calculation of various parameters.

[0096] Based on the quantitative standards for modal consistency in the medical data rule base, this study conducts an in-depth analysis of the logical consistency and synergistic trends among different modalities of data, such as physiological parameters, medication orders, and environmental monitoring, in complex safety risks. By comparing the degree of consistency in the representation of the same risk event by each modal data, the corresponding risk value, i.e., the risk quantity, is quantified according to the standards. This value directly reflects the combined support of multimodal data for risk, compensating for the limitations of single-dimensional risk assessment and making risk assessment more comprehensive and persuasive.

[0097] Referring to a correspondence standard based on the natural decay law of risk, a time decay coefficient is precisely matched according to the intensity of risk representation. The higher the intensity of risk representation, the more significant the current impact of the risk, and the smaller the corresponding time decay coefficient, indicating a slower rate of risk decay; conversely, the lower the intensity of risk representation, the larger the time decay coefficient. This coefficient dynamically reflects the changing pattern of risk over time, enabling risk assessment to conform to dynamic changes in the time dimension, avoiding risk misjudgment caused by static assessment, and improving the timeliness and accuracy of risk weight calculation.

[0098] By integrating key parameters such as risk characterization intensity, risk quantity, and time decay coefficient through a pre-set formula, and introducing a balance intensity coefficient and a consistency coefficient to adjust the contribution ratio of each parameter, this method uses risk characterization intensity as the core, combined with risk quantity to strengthen the supporting role of multimodal data. The time decay coefficient balances the time decay effect of risk, resulting in a dynamic risk weight that reflects the actual importance of complex safety risks in real time, comprehensively, and objectively. This calculation method takes into account the core intensity of risk, the strength of multimodal support, and dynamic changes over time, solving the problems of inflexibility and accuracy in traditional fixed threshold assessments, and providing a scientific and quantitative decision-making basis for subsequent risk situation analysis.

[0099] The target terminal's historical risk database stores complete information on various past risk events, including key data such as risk type, specific characteristics, and scope of impact. By comparing the type attributes and core characteristics of complex security risks, it accurately filters out historical risk event records with similarity, providing referable historical case studies for current risk analysis. This avoids isolated judgments divorced from real-world scenarios and makes risk assessments more relevant and practical.

[0100] The dynamic risk weighting comprehensively considers core factors such as the intensity of risk characteristics, the amount of risk, and the time decay coefficient, objectively reflecting the current severity of complex safety risks. Based on this weighting and combined with pre-defined risk level classification standards, the basic risk level is directly determined without relying on subjective judgment or a single indicator, ensuring the accuracy and objectivity of the initial risk level assessment and laying the foundation for subsequent analysis of the scope of impact and the probability of escalation.

[0101] A deep analysis of selected historical risk event records was conducted to extract data on the actual scope of impact, clarifying the targets, business processes, and areas affected by historical risks. Simultaneously, data on risk escalation paths was analyzed to understand the key nodes, triggering conditions, and development patterns of historical risks from their initial state to their escalation. This extracted key data provides direct historical data support for deducing the potential scope of current risks and assessing the probability of escalation, making the judgment process more scientific.

[0102] This study extracts multi-dimensional current features, including risk type, triggering factors, and involved modalities, from complex security risks. Simultaneously, it extracts initial features from historical risk event records. By comparing the degree of fit between these two types of features across each dimension, a feature matching degree assessment result is quantified. This result accurately reflects the similarity between current and historical risks, providing a quantitative basis for subsequent weighted correlation mapping and ensuring the rationality of inferring the current risk situation based on historical data.

[0103] A weighting rule is established, with higher feature matching scores receiving greater weight. The feature matching score evaluation results are then linked and integrated with the corresponding historical data on the actual impact range of risks. Adaptive adjustments are made to the historical impact range data based on the specific scenario of the current risk, deriving the objects, processes, and areas that may be affected by the current complex safety risks. This yields a precise potential impact range, resolving the issue of ambiguous impact ranges in traditional risk assessments and providing a clear target for subsequent early warning and response.

[0104] Using historical risk escalation path data as a reference, this study compares the current development status of complex security risks with the initial stages of historical risks; simultaneously, it tracks the changing trends of risk dynamic weights to determine whether they are continuously rising, remaining stable, or gradually declining. By combining these two factors, a scientific assessment of the likelihood of current risks escalating within a future time window is conducted, making risk assessment forward-looking, avoiding misjudgments of risk development trends, and providing a basis for developing preventative measures in advance.

[0105] A structured information fusion approach is adopted to integrate basic risk levels, potential impact scope, and escalation probabilities according to a pre-set framework, clarifying the logical connections between different parts of the information, while supplementing auxiliary information such as core risk characteristics and key triggering factors. The resulting risk situation report is clearly structured and comprehensive, including both quantitative risk indicators and specific risk descriptions. This provides complete and accurate information support for subsequent mapping of differentiated early warning strategies and generation of targeted early warning instructions, ensuring the pertinence and efficiency of early warning response.

[0106] S6. Map the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix to trigger a notification process that matches the risk scenario in the risk situation report, and generate and distribute targeted early warning instructions to the target terminal.

[0107] In this embodiment of the invention, mapping the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix includes: The risk situation report is analyzed to extract the risk level identifier and the description of the predicted impact range. The risk level identifier and the predicted impact range description are jointly encoded to obtain the risk scenario identifier of the target terminal; Based on the risk level identifier, the early warning strategy codes in the preset differentiated early warning strategy matrix are selected; Based on the warning strategy code, the risk scenario identifier is logically mapped to obtain the matrix matching output of the target terminal.

[0108] The process of triggering a notification matching the risk scenario in the risk situation report, and generating and distributing targeted early warning instructions to the target terminal, includes: Based on the matrix matching output, a notification process matching the risk scenario in the risk situation report is triggered to determine the notification channel and notification content template for the target terminal. Fill the core risk information from the risk situation report into the notification content template to obtain the core risk template for the target terminal; The core risk template is mapped into an instruction format to obtain a targeted early warning instruction for the target terminal; The targeted early warning instruction is distributed to the recipient of the target terminal through the notification channel.

[0109] A comprehensive analysis of the risk situation report is conducted, focusing on the risk level information clearly marked in the report. The risk level indicators that reflect the severity of the risk are accurately extracted. At the same time, the descriptions in the report about the objects, scope, and fields that the risk may affect are carefully reviewed, and the descriptions of the predicted impact range are fully extracted to ensure that there are no omissions or deviations in the two types of core information.

[0110] By adopting a joint coding approach, the extracted risk level identifier and the predicted impact range description are used as unified coding objects. According to the preset coding rules, the core features of the two types of information are transformed into standardized identifiers. Through coding integration, the risk level and the impact range form a unique corresponding identifier, and finally the risk scenario identifier of the target terminal is obtained, so as to achieve an accurate definition of the current risk scenario.

[0111] The system retrieves a pre-defined differentiated early warning strategy matrix, which contains the mapping relationship between different risk levels and corresponding early warning strategy codes. Based on the extracted risk level identifiers, the system matches the corresponding entries one by one in the matrix and filters out the early warning strategy codes that are perfectly matched with the current risk level, ensuring that the filtered codes can accurately correspond to the early warning requirements of the risk level.

[0112] Based on the selected early warning strategy codes, a mapping relationship between the codes and risk scenario identifiers is established. By comparing the early warning scenario requirements corresponding to the early warning strategy codes with the actual risk scenarios represented by the risk scenario identifiers, the compatibility between the two is confirmed. The risk scenario identifiers are then accurately mapped to the specific early warning strategies in the differentiated early warning strategy matrix, resulting in the matrix matching output of the target terminal, which provides a clear basis for the subsequent differentiated early warning execution.

[0113] Based on the early warning strategy corresponding to the matrix matching output, a notification process that is fully adapted to the risk scenario in the risk situation report is directly triggered. This process has preset notification channel categories and notification content template specifications for different risk scenarios. By identifying the core characteristics of the risk scenario, the most suitable notification channel for the current risk communication is selected from the preset channel set. At the same time, a notification content template that can clearly present the key risk information is matched, clarifying the notification channel and notification content template for the target terminal.

[0114] Extract core risk information from the risk situation report, including basic risk level, potential impact range, escalation probability, key triggering factors, and other key content. According to the information filling format and order preset in the notification content template, fill in the extracted core risk information one by one into the designated positions of the template to ensure that the information is filled accurately and completely without omissions or misalignments. This allows the template to comprehensively and clearly present the core situation of the current risk, resulting in the core risk template for the target terminal.

[0115] Referring to the preset instruction mapping rules, the rules clarify the correspondence between various information in the core risk template and the elements of the early warning instructions. The textual description information in the core risk template is transformed into a standardized and executable instruction format, including key instruction elements such as instruction number, risk type identifier, early warning level code, and execution requirements. This ensures that the transformed instructions have clear execution guidance and operation instructions, resulting in targeted early warning instructions for the target terminal.

[0116] Activate the established notification channels and adapt the format of the targeted early warning instructions according to the information transmission specifications of the channels to ensure that the instructions can be transmitted stably in the channels without information distortion. At the same time, confirm the accurate information of the target terminal recipient, including the receiving address, contact person, and receiving terminal. Send the targeted early warning instructions accurately to the recipient through the notification channels to ensure that the recipient can obtain the early warning information in a timely manner and take appropriate countermeasures.

[0117] The beneficial effects include a comprehensive and detailed analysis of risk situation reports, focusing on the risk levels explicitly marked in the reports, accurately extracting risk level indicators that directly reflect the severity of the risks, and systematically reviewing the reports' descriptions of the objects, areas of coverage, and boundaries of impact that may be affected by the risks, thus fully extracting a description of the predicted impact range. This operation ensures that core risk information is obtained without omission or bias, providing accurate raw input for subsequent risk scenario definition and strategy matching, and avoiding errors in early warning strategy adaptation due to incomplete information extraction.

[0118] A standardized joint coding method is adopted, using the extracted risk level identifiers and predicted impact range descriptions as unified coding objects. Following pre-defined coding rules, the core features of these two types of information are transformed into unique and standardized identifiers. Through coding integration, a strong correlation is formed between risk level and impact range, creating a unique identifier—the risk scenario identifier. This achieves accurate definition and quantitative representation of the current risk scenario, transforming the originally abstract risk description into standardized information that can be directly used for strategy matching, thus improving the efficiency and accuracy of subsequent mapping processes.

[0119] The system retrieves a pre-defined matrix of differentiated early warning strategies, which establishes a precise mapping between different risk levels and corresponding early warning strategy codes. Using the extracted risk level identifiers as the primary search criteria, the system matches each strategy against the matrix to identify early warning strategy codes that perfectly match the current risk level. This step quickly pinpoints the core early warning strategy direction that aligns with the severity of the risk, laying the foundation for subsequent precise matching of specific early warning procedures. It avoids blindly searching through a vast number of strategies and improves the targeting of early warning strategy matching.

[0120] Using the selected early warning strategy codes as the core logical link, a correspondence is established between the codes and risk scenario identifiers. The compatibility is verified by comparing the early warning scenario requirements corresponding to the early warning strategy codes with the actual risk scenarios represented by the risk scenario identifiers. The risk scenario identifiers are then precisely mapped to specific early warning strategies in the differentiated early warning strategy matrix, resulting in matrix matching output. This process achieves precise binding between risk scenarios and early warning strategies, ensuring that the output early warning strategies not only meet risk level requirements but also adapt to the characteristics of the impact range. This provides a clear and reliable basis for subsequently triggering targeted notification processes and generating directed early warning instructions.

[0121] The matrix matching output clearly defines the early warning strategy for the corresponding risk scenario. Based on this output, an appropriate notification process is directly triggered. This process pre-defines the notification channel categories and notification content template specifications for different risk scenarios. By identifying the core characteristics of the risk scenario, the most suitable method for conveying the current risk is selected from the pre-defined channel set. Simultaneously, a template that clearly presents key risk information is matched, accurately determining the notification channel and notification content template for the target terminal. This operation avoids the blindness of uniform notifications, ensuring that the notification channel adapts to the risk dissemination needs, and that the template can specifically present core information, laying the foundation for subsequent early warning instruction generation.

[0122] Core risk information, including basic risk level, potential impact scope, escalation probability, and key triggering factors, is precisely extracted from the risk situation report. Following the pre-defined information filling format and order in the notification content template, the extracted information is then filled into the designated locations in the template. This ensures accurate, complete, and comprehensive information filling, presenting a clear and comprehensive picture of the current risk situation and forming a core risk template for the target terminal. This template integrates key risk information, avoiding information fragmentation and providing a well-structured and detailed foundation for subsequent instruction-based conversion.

[0123] Referring to pre-defined instruction mapping rules, these rules clarify the correspondence between various textual descriptions in the core risk template and the elements of the early warning instructions. This transforms the risk descriptions in the template into standardized, executable instruction formats, including key elements such as instruction number, risk type identifier, early warning level code, and execution requirements. The resulting targeted early warning instructions have clear execution guidance and operational instructions, overcoming the ambiguity of purely textual descriptions and ensuring that recipients can quickly and accurately understand risk requirements, providing a clear basis for action in efficiently responding to risks.

[0124] Established notification channels were activated, and the targeted early warning instructions were formatted and adapted according to the channel's information transmission specifications to ensure stable transmission without information distortion. Simultaneously, the accurate information of the target terminal recipient was confirmed, including the receiving address, contact person, and receiving terminal. The targeted early warning instructions were then accurately sent to the recipient through this channel. This process ensured precise delivery of early warning instructions, guaranteeing that recipients received early warning information promptly and could quickly take appropriate countermeasures, improving the timeliness and efficiency of medical safety risk early warning and ensuring timely risk management.

[0125] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0126] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for monitoring, reporting, and early warning of health and medical safety, characterized in that, The method includes: S1. Acquire and integrate the heterogeneous monitoring data of the target terminal to obtain the initial heterogeneous dataset of the target terminal; S2. Standardize and clean the initial heterogeneous dataset, and perform multi-dimensional data filtering on the cleaned results to obtain the standard secure data stream of the target terminal; S3. Perform a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain the clean data pool of the target terminal; S4. Perform cross-modal data association on the clean data pool, and conduct collaborative analysis on the association results to identify the composite security risks in the target terminal, and calculate the dynamic risk weight of the composite security risks; S5. Based on the aforementioned dynamic risk weights and combined with the historical risk database of the target terminal, a comprehensive assessment is made of the potential impact range and escalation probability of the composite security risks to obtain a risk status report for the target terminal. S6. Map the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix to trigger a notification process that matches the risk scenario in the risk situation report, and generate and distribute targeted early warning instructions to the target terminal.

2. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The acquisition and integration of heterogeneous monitoring data from the target terminal to obtain the initial heterogeneous dataset of the target terminal includes: Collect time-series data of vital signs from the target terminal within a unit monitoring cycle; Extract medical order execution record data and environmental disinfection log data corresponding to the unit's monitoring cycle from the medical information records associated with the target terminal; The vital signs time-series data, the medical order execution record data, and the environmental disinfection log data are time-series aligned to obtain the aligned result of the target terminal; The aligned results are formatted and standardized to obtain the initial heterogeneous dataset of the target terminal.

3. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The process of standardizing and cleaning the initial heterogeneous dataset, and then performing multi-dimensional data filtering on the cleaned results to obtain the standard secure data stream for the target terminal includes: Identify missing value entries and anomalous format entries in the initial heterogeneous dataset; Based on a pre-defined medical data rule base, the missing value entries and the format-abnormal entries are standardized and corrected to obtain complete data entries and format-normalized data of the initial heterogeneous dataset; The complete data entries and the formatted data are fused together, and the fused data is then merged with the data entries in the initial heterogeneous dataset that do not require processing to obtain the intermediate dataset of the target terminal. The data entries in the intermediate dataset are sorted by time topology to obtain an ordered data sequence of the target terminal; Based on the clinical rationality dimension of the medical data rule base, the ordered data sequence is initially filtered out to obtain a clinically rational data subset for the target terminal. Based on the data credibility dimension of the medical data rule base, the ordered data sequence is filtered twice to obtain a highly reliable data subset of the target terminal. The intersection of the clinically reasonable subset of data and the highly reliable subset of data is taken as the effective dataset of the target terminal, and the effective dataset is serialized and encapsulated to obtain the standard secure data stream of the target terminal.

4. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The process of performing a two-dimensional verification of the real-time data quality and logical consistency of the standard secure data stream to obtain a clean data pool for the target terminal includes: The historical data quality performance records of the target terminal are associated and mapped with the standard security data stream to obtain the quality tolerance threshold table of the target terminal; Based on the quality tolerance threshold table, a real-time quality scan is performed on the data entries in the standard security data stream to obtain a dynamic quality defect list of the target terminal. The defective data entries in the dynamic quality defect list are removed from the standard security data stream to obtain the data stream to be verified for the target terminal. Perform a deep logical consistency check on the data entries in the data stream to be checked to obtain a list of logical contradictions of the target terminal; Based on the logical contradiction list, logical conflict entries in the data stream to be verified are removed to obtain the clean data pool of the target terminal.

5. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The cross-modal data association of the clean data pool includes: Based on the medical data classification rules in the medical data rule base, the heterogeneous data streams in the clean data pool are divided by modal features to obtain the first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream of the target terminal. The first physiological parameter time-series data stream, the second medical order and medication record data stream, and the third environmental monitoring data stream are aligned in the time dimension to obtain the aligned data stream of the target terminal. Interactively fuse data from different dimensions in the aligned data stream to obtain a multimodal data sequence for the target terminal; Semantic association parsing is performed on the multimodal data sequence to obtain the semantic association relationship of the target terminal; The semantic association is used as the cross-modal association result of the target terminal.

6. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 5, characterized in that, The collaborative analysis of the associated results identifies complex security risks in the target terminal, including: A consistency analysis is performed on the cross-modal association results to obtain the consistency verification results of the target terminal; The cross-modal association results are subjected to association mining to obtain the multimodal association patterns of the target terminal; By coupling the consistency verification result and the multimodal association pattern with risk logic, potential security events of the target terminal can be obtained. Risk assessment is performed on the potential security events to obtain the complex security risks in the target terminal.

7. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The calculation of the dynamic risk weights for the composite security risks includes: The composite security risk is normalized to obtain the risk characterization intensity of the composite security risk; The consistency between modes in the composite security risk is quantified to obtain the risk quantity of the composite security risk; Based on the intensity of the risk characterization, determine the time decay coefficient of the composite security risk within the current time window; Based on the risk characterization intensity, the risk quantity, and the time decay coefficient, the dynamic risk weight of the composite safety risk is calculated, wherein the calculation formula for the dynamic risk weight is as follows: ； In the formula, The dynamic weight of the risk, The intensity of the risk characterization, The time decay coefficient is... The duration of the intensity of the risk characterization. The risk level is... The coefficient representing the degree of consistency of the aforementioned risk quantity. The equilibrium intensity coefficient is the ratio of the intensity of the risk characterization. It is a natural exponential function.

8. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, Based on the dynamic risk weights and combined with the historical risk database of the target terminal, a comprehensive assessment is made of the potential impact range and escalation probability of the complex security risks, resulting in a risk situation report for the target terminal, including: Retrieve historical risk event records from the target terminal's historical risk database that are similar in type and representation to the composite security risk; Based on the aforementioned dynamic risk weights, a preliminary risk assessment is performed on the composite security risks to obtain their basic risk levels. Extract the actual impact range data and risk escalation path data of the historical risk events from the historical risk event records; The current characteristics of the composite security risk are compared with the initial characteristics of the historical risk event records using a multidimensional feature similarity metric to obtain the feature matching degree evaluation result of the target terminal. The potential impact range of the complex security risk is obtained by performing a weighted correlation mapping between the feature matching degree evaluation results and the actual impact range data of the historical risk events. Using the risk escalation path data and the changing trend of the risk dynamic weight as the judgment benchmark, the probability of the composite security risk escalating within a future time window is assessed; The risk status report of the target terminal is obtained by fusing structured information from the basic risk level, the potential impact range, and the escalation probability.

9. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 1, characterized in that, The process of mapping the risk level and predicted impact range in the risk situation report to a preset differentiated early warning strategy matrix includes: The risk situation report is analyzed to extract the risk level identifier and the description of the predicted impact range. The risk level identifier and the predicted impact range description are jointly encoded to obtain the risk scenario identifier of the target terminal; Based on the risk level identifier, the early warning strategy codes in the preset differentiated early warning strategy matrix are selected; Based on the warning strategy code, the risk scenario identifier is logically mapped to obtain the matrix matching output of the target terminal.

10. The method for monitoring, reporting, and early warning of health and medical safety as described in claim 9, characterized in that, The process of triggering a notification matching the risk scenario in the risk situation report, and generating and distributing targeted early warning instructions to the target terminal, includes: Based on the matrix matching output, a notification process matching the risk scenario in the risk situation report is triggered to determine the notification channel and notification content template for the target terminal. Fill the core risk information from the risk situation report into the notification content template to obtain the core risk template for the target terminal; The core risk template is mapped into an instruction format to obtain a targeted early warning instruction for the target terminal; The targeted early warning instruction is distributed to the recipient of the target terminal through the notification channel.

Images