Medical risk automatic screening system

Abstract

The invention discloses an automatic medical risk screening system, which relates to the technical field of medical information and comprises a data acquisition module, an intelligent analysis module, a risk early warning module, an intervention management module, a complaint management module and a data visualization module. According to the medical risk automatic screening system provided by the invention, risk data sources in a hospital HIS system are collected in real time and are preprocessed and intelligently analyzed, so that the system can more accurately identify potential risk patients and provide timely early warning prompts for medical staff, comprehensive management of medical risks is realized through each module, and through PDCA circulation, the medical risk screening efficiency is improved. According to the technical scheme, the system automatically generates disposal suggestions, distributes disposal tasks, evaluates the intervention effect and optimizes early warning rules, a continuously improved management process is formed, and a multi-dimensional interactive analysis interface is also provided, so that medical staff can intuitively understand information such as risk distribution, intervention effect tracking and complaint causes in each inpatient area.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and specifically relates to a medical risk automatic screening system. Background Art

[0002] With the deepening of medical informatization construction, medical institutions have gradually established a risk monitoring system based on the hospital information system. The current mainstream medical risk management systems usually rely on multi-source data such as electronic medical records, expense settlement, and nursing records, and use statistical analysis and machine learning algorithms to construct risk assessment models to achieve monitoring and early warning of typical risk indicators such as abnormal length of stay, high expense expenditure, and unplanned re-operation. Such systems integrate data from core business systems such as the hospital HIS, LIS, and PACS, use technical means such as time series analysis and clustering algorithms to identify potential risk patients, and send early warning messages to medical staff through a message push mechanism, playing an important role in improving the quality and safety of medical care and optimizing resource allocation.

[0003] Existing technologies mostly focus on the static monitoring of single-dimensional risk factors, and it is difficult to achieve dynamic correlation analysis of multi-modal medical data. There is still room for improvement in the prediction accuracy and timeliness in complex clinical scenarios. The traditional risk scoring models generally use fixed threshold settings for the monitoring of indicators such as length of stay and frequency of transfer to the intensive care unit, lacking an adaptive adjustment mechanism based on individual patient characteristics and being easily affected by departmental differences and the diversity of treatment plans. In addition, most systems do not form a complete closed-loop management path after risk early warning, lack quantitative evaluation means for the implementation effect of intervention measures, and there is an information island phenomenon between complaint data processing and the risk early warning system. In response to this, we propose a medical risk automatic screening system. Summary of the Invention

[0004] To solve the above technical problems and provide a medical risk automatic screening system, this technical solution solves the problems that the traditional risk scoring models generally use fixed threshold settings for monitoring, most systems do not form a complete closed-loop management path after risk early warning, and there is a lack of quantitative evaluation means for the implementation effect of intervention measures.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A medical risk automatic screening system, including: a data acquisition module, an intelligent analysis module, a risk early warning module, an intervention management module, a complaint management module, and a data visualization module; The data acquisition module is used to obtain risk data sources in real time from the hospital HIS system. The risk data sources include: disease record data, data of patients staying in the hospital for more than 30 days, data of patients being transferred to the intensive care unit more than three times, arrears data, high-expense data of more than or equal to 200,000, and unplanned re-operation data, and preprocess the data; The intelligent analysis module is electrically connected to the data acquisition module, and the intelligent analysis module is used to retrieve the data of the risk data source according to a preset algorithm through a risk scoring model and output it; The risk warning module is electrically connected to the intelligent analysis module, and the risk warning module is used to trigger a warning mechanism for patients who reach the preset risk threshold according to the calculation result of the risk scoring model; The intervention management module is used to establish a risk file for the warned patients and record the intervention process, including the intervention time, measure content, and implementing personnel; The complaint management module is used to handle the entire process of medical complaints; The data visualization module is used to provide a multi-dimensional interactive analysis interface.

[0006] Preferably, the preprocessing of the disease condition filing data is specifically as follows: Perform word segmentation, part-of-speech tagging, and named entity recognition on the medical record text, and extract key disease condition information; Judge whether the medical record record is complete and accurate through semantic similarity calculation; The semantic similarity calculation adopts the cosine similarity algorithm, and the formula is: In the formula, A and B are respectively the vector representations of two medical record texts.

[0007] Preferably, when the intelligent analysis module processes the data of patients staying in the hospital for more than 30 days, an improved ARIMA time series analysis model is adopted, specifically as follows: Collect the vital signs, medication records, and treatment plan data of the patient during hospitalization, and construct a multi-variable time series; determine the ARIMA model parameters through the Bayesian optimization algorithm, and its expression is In the formula, is the observed value of the patient at a certain moment t during hospitalization, B is the backshift operator, is the autoregressive parameter, is the autoregressive order, is the differencing operator, is the differencing order, is the moving average parameter, is the moving average order, is the random error term; When the actual hospitalization days exceed the predicted value by 2 standard deviations, a warning is triggered.

[0008] Preferably, when the intelligent analysis module processes the data of multiple admissions to the intensive care unit, a gradient boosting decision tree risk prediction model is constructed, and its method includes: Extract 15-dimensional features such as the frequency of entering the intensive care unit, the vital sign fluctuation coefficient, and the number of treatment plan adjustments from the intensive care records; Rank the feature importance through SHAP values and dynamically adjust the model input dimension; Generate a high-risk warning when the predicted risk probability exceeds 0.7.

[0009] Preferably, when the intelligent analysis module processes high-cost data, it adopts an adaptive threshold detection algorithm: divide the cost interval based on the improved DBSCAN clustering algorithm, and dynamically calculate the cost threshold of each department: In the formula, is the historical cost average of department k, is the standard deviation; trigger a warning when the single-day cost increase exceeds 20% or the cumulative cost breaks through the threshold.

[0010] Preferably, the complaint management module includes: A complaint classification unit that uses a BERT-TextCNN hybrid model for multi-label classification of complaint texts. The model structure includes a BERT embedding layer, a multi-scale convolutional layer, and an Attention weighted layer; A complaint handling tracking unit that automatically generates a reminder of the handling time limit and associates it with the performance appraisal indicators; a complaint knowledge base that stores typical complaint cases and solutions for retrieval and reference.

[0011] Preferably, the risk warning module adopts a hierarchical push mechanism: set three levels of warnings, red, orange, and yellow, corresponding to three push methods: SMS + pop-up window + phone call, SMS + pop-up window, and system message; the warning information includes patient ID, risk type, recommended disposal measures, and time limit requirements.

[0012] Preferably, the data visualization module includes: A risk heat map that shows the risk distribution of each ward based on the geographic information system; An intervention effect tracking panel that shows the change curve of the risk score after the implementation of the intervention measures; A Sankey diagram of the causes of complaints that visually shows the association path of complaint type - department - handling result; supports dynamic drill-down analysis through D3.js.

[0013] Preferably, adopt a containerized microservice architecture, including: data access service, real-time computing service, model service, message service; Each service communicates through ServiceMesh, supporting automatic scaling and gray release.

[0014] Preferably, the intervention management module implements the PDCA cycle: In the planning stage, automatically generate disposal suggestions including legal basis and clinical guidelines; In the execution stage, associate with the hospital OA system to dispatch disposal tasks; The inspection phase is evaluated by comparing the risk indicators before and after the intervention; In the processing phase, effective disposal solutions are stored in the knowledge base and the early warning rules are optimized.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The automatic medical risk screening system proposed by the present invention can more accurately identify potential risk patients by collecting risk data sources in the hospital HIS system in real time, preprocessing and intelligent analysis, and provide timely early warning prompts for medical staff. Through each module, comprehensive management of medical risks is achieved. Through the PDCA cycle, disposal suggestions are automatically generated, disposal tasks are dispatched, the intervention effect is evaluated, and the early warning rules are optimized, forming a continuously improved management process. A multi-dimensional interactive analysis interface is also provided, enabling medical staff to intuitively understand information such as the risk distribution in each ward, the tracking of intervention effects, and the causes of complaints, providing strong data support for medical quality management. The system adopts a containerized microservice architecture, has good scalability and flexibility, and can adapt to the business needs of different medical institutions. Description of the Drawings

[0016] Figure 1 It is the system module framework diagram of the present invention; Figure 2 It is the system working flow chart of the present invention. Detailed Embodiments

[0017] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0018] Referring to Figure 1 As shown, the automatic medical risk screening system includes: a data collection module, an intelligent analysis module, a risk early warning module, an intervention management module, a complaint management module, and a data visualization module; The data collection module is the cornerstone of the entire system. It is responsible for obtaining risk data sources from the hospital HIS system in real time. These data sources cover multiple aspects such as disease record data, data of patients staying in the hospital for more than 30 days, data of patients entering the intensive care unit more than three times, arrears data, high-cost data of more than 200,000 yuan, and data of unplanned reoperations, ensuring the comprehensiveness and accuracy of risk identification. The collected data will go through a preprocessing stage, including data cleaning, format conversion, etc., laying a foundation for subsequent analysis.

[0019] The intelligent analysis module then uses an advanced risk scoring model to conduct in-depth retrieval and analysis on the preprocessed data. This model can comprehensively calculate the risk score of the patient based on the dynamic weight coefficients of each data source and the corresponding risk scores. In this process, the model will fully consider the relevance and influence between different data sources to ensure the accuracy of risk assessment. Among them, the expression of the risk scoring model is: In the formula, is the dynamic weight coefficient of each data source, is the risk score corresponding to each data source; When the risk score reaches the preset risk threshold, the risk warning module will quickly trigger the warning mechanism. This module can automatically identify and screen out high-risk patients and send warning messages to medical staff through various methods such as text messages, pop-up windows, and phone calls. The warning messages contain key information such as the patient's ID, risk type, recommended disposal measures, and time limit requirements, providing timely and effective risk reminders for medical staff.

[0020] The intervention management module is responsible for the subsequent risk management of the warned patients. This module will establish a detailed risk file for each warned patient and record the key information during the intervention process, such as the intervention time, measure content, and executor. This information not only helps medical staff comprehensively understand the risk status of the patient but also provides valuable data support for subsequent evaluation and improvement.

[0021] The complaint management module focuses on handling the entire process of medical complaints. It can improve the efficiency and quality of complaint handling through automated means and reduce errors and delays caused by human factors. At the same time, the establishment of a complaint knowledge base also provides rich case references and solutions for medical staff, helping to improve their complaint handling ability.

[0022] The data visualization module provides an intuitive and convenient risk management interface for medical staff. Through a multi-dimensional interactive analysis interface, medical staff can easily understand information such as the risk distribution in each ward, the tracking of intervention effects, and the causes of complaints, providing strong data support for scientific decision-making.

[0023] The preprocessing of the disease record data is a crucial part of the medical risk automatic screening system, which is directly related to the accuracy and effectiveness of subsequent risk assessment. For the medical record text, the system will first perform word segmentation to cut the continuous text into independent lexical units, which is the basic step of natural language processing. Subsequently, the system will perform part-of-speech tagging on these words to clarify the grammatical role of each word in the sentence, such as noun, verb, adjective, etc., which helps to understand the text content more deeply.

[0024] Next, the system will also perform named entity recognition, which is specifically used to extract key medical condition information from the text, such as disease names, drug names, surgical names, etc. These information are of extremely high value for risk assessment.

[0025] After completing the above steps, the system will also determine whether the medical record is complete and accurate through semantic similarity calculation. In this process, the system adopts the cosine similarity algorithm, and measures the similarity between two medical record texts by calculating the cosine value of the angle between the vector representations of the two medical record texts. Specifically, it is to transform the medical record text into a vector in a high-dimensional space, and then calculate the cosine value of the angle between these vectors. The smaller the angle, the closer the cosine value is to 1, indicating that the two medical record texts are more similar, thereby indirectly judging the integrity and accuracy of the medical record. This step is of great significance for ensuring the reliability of subsequent risk assessment. The formula is: In the formula, A and B are the vector representations of the two medical record texts respectively.

[0026] In the medical risk automatic screening system, the intelligent analysis module plays a core role. Especially when processing data of patients staying in the hospital for more than 30 days, the improved ARIMA time series analysis model it adopts demonstrates high precision and practicality. This model will first comprehensively collect the vital sign data, medication records and details of treatment plans of patients during their hospitalization. These data together constitute a complex and multi-dimensional time series. These time series not only reflect the dynamic changes in the health status of patients, but also contain key information about the treatment process.

[0027] In order to accurately capture the patterns and trends in these time series, the intelligent analysis module uses the Bayesian optimization algorithm to finely determine the parameters of the ARIMA model. As a classic time series prediction method, each symbol in the expression of the ARIMA model represents a key component of the model: is the observed value of a patient at a certain moment during hospitalization, which intuitively reflects the real-time state of the patient; the backshift operator is used to handle the lag effect of the time series; the autoregressive parameter and autoregressive order capture the historical dependence of the time series itself; the difference operator and its order are used to stationary the non-stationary time series; the moving average parameter and moving average order further smooth the random fluctuations and improve the robustness of the prediction.

[0028] Through the Bayesian optimization algorithm, the system can efficiently find the ARIMA model parameter combination that best suits the current data characteristics, thus greatly improving the prediction accuracy. Once the model is constructed, the system will conduct a risk assessment based on the comparison between the prediction results and the actual length of hospital stay. Specifically, when the actual length of hospital stay exceeds the predicted value by 2 standard deviation ranges, the system will immediately trigger the warning mechanism. This setting not only takes into account the uncertainty of the prediction but also ensures the timeliness and sensitivity of the warning, providing valuable decision-making support for medical staff and helping them identify and intervene in a timely manner those patients who may have medical risks and have an overly long hospital stay. Its expression is In the formula,[[]] is the observed value of the patient at a certain moment t during the hospital stay, B is the backshift operator,[[]] is the autoregressive parameter,[[]] is the autoregressive order,[[]] is the differencing operator, d is the differencing order,[[]] is the moving average parameter,[[]] is the moving average order,[[]] is the random error term;[[]] The warning is triggered when the actual length of hospital stay exceeds the predicted value by 2 standard deviations.[[]]

[0029] When the intelligent analysis module processes the data of multiple admissions to the intensive care unit, it constructs a Gradient Boosting Decision Tree (GBDT) risk prediction model, which specifically includes: extracting 15-dimensional features such as the frequency of admission, the coefficient of vital sign fluctuations, and the number of treatment plan adjustments from the intensive care records; performing feature importance ranking through SHAP values and dynamically adjusting the model input dimension; generating a high-risk warning when the predicted risk probability exceeds 0.7. As the area in the hospital where critically ill patients are intensively treated, the frequent admission and discharge of patients are often accompanied by complex conditions and great treatment difficulties. Therefore, it is necessary to construct a risk prediction model that can accurately capture these complex features. The GBDT model, with its powerful non-linear fitting ability and robustness to outliers, becomes an ideal choice for processing such data. By extracting multi-dimensional features from the intensive care records, such as the frequency of admission and the coefficient of vital sign fluctuations, the GBDT model can more comprehensively depict the patient's condition and provide a solid foundation for risk prediction. Through feature importance ranking using SHAP values, it is possible to clearly understand which features contribute the most to risk prediction, thus helping medical staff better understand the patient's condition and formulate more precise treatment plans. Dynamically adjusting the model input dimension can flexibly optimize the model structure according to the actual situation, improving the accuracy and efficiency of prediction. Generating a high-risk warning when the predicted risk probability exceeds 0.7 provides timely risk alerts for medical staff, enabling them to respond quickly and take necessary intervention measures to effectively reduce medical risks and ensure patient safety.[[]]

[0030] When the intelligent analysis module processes high-cost data, it adopts an adaptive threshold detection algorithm: it divides the cost range based on an improved DBSCAN clustering algorithm and dynamically calculates the cost thresholds for each department: In the formula, is the historical cost average of department k, is the standard deviation; when the daily cost increase exceeds 20% or the cumulative cost breaks through the threshold, a warning is triggered.

[0031] The reason for using the adaptive threshold detection algorithm to process high-cost data is that there are significant differences in medical costs among different departments and patients with different conditions, and fixed thresholds are difficult to adapt to this diversity. By dividing the cost range through the improved DBSCAN clustering algorithm and dynamically calculating the cost thresholds for each department, it is possible to more accurately identify abnormal cost situations, improve the accuracy and pertinence of warnings. The advantage of this approach is that it can flexibly cope with the volatility of medical costs and avoid false alarms or missed alarms caused by inappropriate setting of fixed thresholds. At the same time, it can also help the hospital promptly discover potential medical resource waste or fraud behaviors, providing strong support for cost control and compliance management.

[0032] The complaint management module, as an important part of the medical risk automatic screening system, is committed to improving the efficiency and quality of medical complaint handling. This module integrates three core functions: a complaint classification unit, a complaint handling tracking unit, and a complaint knowledge base. The complaint classification unit uses an advanced BERT-TextCNN hybrid model to perform multi-label classification on complaint texts. This model combines the deep semantic understanding ability of BERT and the advantages of the convolutional neural network of TextCNN. It obtains the deep semantic features of the text through the BERT embedding layer, then extracts local features using multi-scale convolutional layers, and weights the key information through the Attention weighting layer, thus achieving high-precision classification of complaint texts. This not only improves the efficiency of complaint handling but also provides data support for subsequent analysis and improvement.

[0033] The complaint handling tracking unit automatically generates reminders of the handling time limit and associates them with the performance appraisal indicators to ensure that complaints are handled promptly and effectively. At the same time, the complaint knowledge base stores a large number of typical complaint cases and solutions, providing rich reference resources for medical staff to help them better handle various complaint situations.

[0034] The risk warning module adopts a hierarchical push mechanism. According to the level of risk scores, the warning information is divided into three levels: red, orange, and yellow, corresponding to three push methods: SMS + pop-up window + phone call, SMS + pop-up window, and system message. This differentiated push strategy ensures that high-risk events can receive timely and sufficient attention, while low-risk events can be properly handled without interfering with the normal work of medical staff. The warning information contains key information such as patient ID, risk type, recommended handling measures, and time limit requirements in detail, providing comprehensive and specific decision-making support for medical staff.

[0035] The data visualization module provides medical staff with a multi-dimensional risk management perspective through intuitive and vivid charts and graphs. The risk heat map shows the risk distribution of each ward area based on the Geographic Information System (GIS), helping medical staff quickly identify high-risk areas; the intervention effect tracking panel displays the change curve of the risk score after the implementation of the intervention measures, intuitively reflecting the effectiveness of the intervention measures; the Sankey diagram of the causes of complaints visually shows the correlation path between the types of complaints, departments, and handling results, providing valuable decision-making basis for the hospital management. At the same time, this module also supports dynamic drill-down analysis through D3.js, further enhancing the depth and breadth of data exploration.

[0036] In terms of technical architecture, this system adopts a containerized microservices architecture, ensuring high scalability and flexibility of the system. The data access service (Docker + Flask) is responsible for efficiently and stably accessing various data sources; the real-time computing service (Spark Streaming) processes and analyzes massive data in real time; the model service (TensorFlow Serving) provides high-performance model deployment and inference capabilities; the message service (Kafka) ensures low-latency and highly reliable message transmission between modules. Each service communicates through ServiceMesh, realizing functions such as automatic service discovery, load balancing, and fault recovery. In addition, the system also supports advanced features such as automatic scaling and gray release, further enhancing the stability and availability of the system.

[0037] The intervention management module realizes the closed-loop management process of the PDCA cycle. In the planning stage (Plan), the system automatically generates handling suggestions containing legal bases and clinical guidelines according to the risk assessment results; in the execution stage (Do), the system associates with the hospital OA system to dispatch handling tasks to relevant medical staff; in the checking stage (Check), the system evaluates the effectiveness of the handling measures by comparing the risk indicators before and after the intervention; in the acting stage (Act), the system stores the effective handling plan in the knowledge base and optimizes the warning rules to improve the accuracy of future risk assessments. The continuous iteration and optimization of this process ensure the continuous improvement of the medical risk management level.

[0038] Refer to Figure 2 As shown, the usage process of the present invention is as follows: The risk data source is obtained in real time from the hospital HIS system through the data acquisition module and preprocessed.

[0039] The intelligent analysis module uses the risk scoring model to retrieve and analyze the preprocessed data. When the risk score reaches the preset threshold, the risk warning module triggers the warning mechanism and sends a warning message to the medical staff.

[0040] The intervention management module establishes a risk file for the warned patients and records the intervention process. At the same time, the complaint management module processes the entire process of medical complaints.

[0041] The data visualization module provides a multi-dimensional interactive analysis interface to display information such as the risk distribution of each ward, the tracking of intervention effects, and the causes of complaints.

[0042] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic medical risk screening system, characterized in that, Including: A data collection module, an intelligent analysis module, a risk warning module, an intervention management module, a complaint management module, and a data visualization module; The data collection module is used to obtain risk data sources in real time from the hospital HIS system. The risk data sources include: disease record data, data of patients staying in the hospital for more than 30 days, data of patients entering the intensive care unit more than three times, arrears data, high-cost data of more than or equal to 200,000 yuan, and data of unplanned re-operation, and preprocess the data; The intelligent analysis module is electrically connected to the data collection module. The intelligent analysis module is used to retrieve and output the data of the risk data source according to a preset algorithm through a risk scoring model; The risk warning module is electrically connected to the intelligent analysis module. The risk warning module is used to trigger an early warning mechanism for patients who reach the preset risk threshold according to the calculation result of the risk scoring model; The intervention management module is used to establish a risk file for the warned patients and record the intervention process, including the intervention time, measure content, and executor; The complaint management module is used to handle the entire process of medical complaints; The data visualization module is used to provide a multi-dimensional interactive analysis interface.

2. The medical risk automatic screening system according to claim 1, characterized in that, The specific preprocessing of the disease record data is as follows: Perform word segmentation, part-of-speech tagging, and named entity recognition on the medical record text, and extract key disease information; Judge whether the medical record record is complete and accurate through semantic similarity calculation; The semantic similarity calculation uses the cosine similarity algorithm, and the formula is: In the formula, A and B are the vector representations of two medical record texts respectively.

3. The medical risk automatic screening system according to claim 1, wherein, When the intelligent analysis module processes the data of patients staying in the hospital for more than 30 days, it uses an improved ARIMA time series analysis model. Specifically: Collect the vital signs, medication records, and treatment plan data of the patient during the hospitalization period, and construct a multi-variable time series; determine the ARIMA model parameters through the Bayesian optimization algorithm, and its expression is In the formula, is the observed value of the patient at a certain moment t during hospitalization, B is the backshift operator, is the autoregressive parameter, is the autoregressive order, is the difference operator, d is the difference order, is the moving average parameter, is the moving average order, is the random error term; Trigger an early warning when the actual hospitalization days exceed the predicted value by 2 standard deviations.

4. The medical risk automatic screening system according to claim 1, characterized in that When the intelligent analysis module processes the data of entering the intensive care unit multiple times, it constructs a gradient boosting decision tree risk prediction model. The method includes: Extract 15-dimensional features such as the frequency of entering the intensive care unit, the vital sign fluctuation coefficient, and the number of treatment plan adjustments from the intensive care records; Rank the feature importance through SHAP values and dynamically adjust the model input dimension; Generate a high-risk warning when the predicted risk probability exceeds 0.

7.

5. The medical risk automatic screening system according to claim 1, characterized in that, When the intelligent analysis module processes the high-cost data, it uses an adaptive threshold detection algorithm: divide the cost interval based on the improved DBSCAN clustering algorithm, and dynamically calculate the cost threshold of each department: In the formula, is the historical average cost of department k, is the standard deviation; an alarm is triggered when the daily cost increase exceeds 20% or the cumulative cost exceeds the threshold.

6. The medical risk automatic screening system according to claim 1, wherein The complaint management module includes: A complaint classification unit that uses a BERT-TextCNN hybrid model for multi-label classification of complaint texts. The model structure includes a BERT embedding layer, a multi-scale convolutional layer, and an Attention weighted layer; A complaint handling tracking unit that automatically generates a reminder of the handling time limit and associates it with the performance appraisal indicators; a complaint knowledge base that stores typical complaint cases and solutions for retrieval and reference.

7. The medical risk automatic screening system according to claim 1, characterized in that, The risk warning module adopts a hierarchical push mechanism: set three levels of warnings: red, orange, and yellow, corresponding to three push methods: SMS + pop-up window + phone call, SMS + pop-up window, and system message; The warning information includes the patient ID, risk type, recommended disposal measures, and time limit requirements.

8. The medical risk automatic screening system according to claim 1, characterized in that, The data visualization module includes: Risk heat map, showing the risk distribution of each ward area based on the geographic information system; Intervention effect tracking panel, showing the risk score change curve after the implementation of the intervention measures; Sankey diagram of complaint causes, visually showing the association path of complaint type - department - handling result; Support dynamic drill-down analysis through D3.js.

9. The automatic medical risk screening system according to claim 1, wherein Adopt a containerized microservices architecture, including: data access service, real-time computing service, model service, message service; Each service communicates through ServiceMesh, supporting automatic scaling and gray release.

10. The medical risk automatic screening system according to claim 1, characterized in that, The intervention management module implements the PDCA cycle: In the planning stage, disposal suggestions including legal basis and clinical guidelines are automatically generated; In the execution stage, the hospital OA system is associated to dispatch disposal tasks; In the inspection stage, evaluation is carried out by comparing risk indicators before and after the intervention; In the handling stage, effective disposal plans are stored in the knowledge base and warning rules are optimized.

Images