Digital intelligent medical case whole-process management system based on data driving

Abstract

The invention relates to the technical field of data processing, in particular to a digital intelligent medical case whole-course management system based on data driving, which comprises a dynamic threshold classification module for updating disease attack rates of different seasons according to different seasons, inputting medical files of different patients in corresponding seasons, comparing data in a normal range, and classifying the medical files according to the dynamic threshold classification module. And classifying data which do not exceed the normal range and data which exceed the normal range. According to the method, an age-complication-physiological index association model is established, a random forest algorithm is utilized, the influence of individual features of patients on different ranges of data is determined, feature importance is subjected to dynamic sorting analysis, and data access permission of medical personnel is controlled, so that the accuracy of the data access permission of the medical personnel is improved on the basis of ensuring patient privacy compliance. Therefore, medical staff can pay more attention to the critical patient more quickly, visual early warning of critical risk characteristics is achieved, and meanwhile the risk of data leakage can be effectively prevented.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data-driven digital medical case management system. Background Art

[0002] Medical management refers to a series of activities that plan, organize, coordinate, control and evaluate medical institutions, medical resources, and medical service processes. By conducting refined distributed management of medical data, it can reduce operating costs, improve the profitability and competitiveness of medical institutions, ensure the efficiency, safety, quality and sustainability of medical services, and optimize the allocation and use of medical resources.

[0003] The data-driven digital medical case management system is mainly used to improve medical record management efficiency, optimize resource allocation, and reduce operating costs. It also supports the construction of maternal and child health information platforms, realizes health management throughout the life cycle, breaks down information silos, integrates and shares data, and manages data security. It also analyzes medical record data based on machine learning algorithms to provide doctors with personalized treatment plan recommendations.

[0004] The medical data stored in the existing medical case management system is usually stored on hard drives or data centers. It is impossible to effectively and securely manage the collected medical data, and it is impossible to effectively analyze and sort the disease characteristics based on data such as the different patients' own conditions. Medical personnel fail to discover and pay attention to critically ill patients in a timely manner, which leads to the problem that some critically ill patients cannot receive timely assistance. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a data-driven digital medical case management system.

[0006] To achieve the above objectives, the present invention adopts the following technical solutions: A data-driven digital medical case management system includes:

[0007] Dynamic threshold classification module: Updates seasonal symptom rates based on the seasons, records the medical records of different patients in the corresponding seasons, compares the data within the normal range, and classifies the data within and outside the normal range to form multiple groups of data;

[0008] Data storage encryption module: Based on the multiple sets of data, a multi-level authority hierarchy is constructed, different identities are established for medical personnel and patients to complete identity classification, and data sensitivity classification is performed accordingly based on the identity classification. Data is encrypted and stored according to the identity classification and data sensitivity classification. Medical personnel and patients with different identity classifications can view the data after the corresponding data sensitivity classification;

[0009] Multi-feature information analysis module: Based on multiple data sets, an age-complication-physiological indicator correlation model is established. The random forest algorithm is used to combine the impact of individual patient characteristics on the generation of data in different ranges to perform feature importance ranking analysis and obtain ranking analysis results.

[0010] Intelligent hierarchical warning module: Based on the sorting analysis results, it conducts risk assessment on the treatment plans of different case information, issues warnings of different degrees, and draws allocation results;

[0011] Dynamic resource optimization module: Based on the allocation results, BIM technology is used to build a three-dimensional resource topology map of the hospital. The location of mobile devices is tracked in real time through locators to determine whether to use mobile devices or other backup devices, or to book usage time based on the patient's needs.

[0012] As a further solution of the present invention: the dynamic threshold classification module includes:

[0013] Environmental coupling data acquisition submodule: Based on the different air quality in different seasons, temperature and humidity sensors and air quality sensors are used to collect environmental parameters such as temperature, humidity, and air quality in the ward in real time to obtain information about the current season;

[0014] Adaptive threshold generation submodule: Based on the current seasonal information, it automatically updates the symptom incidence rate and compares it with the data in the medical files of different patients to obtain multiple sets of data on whether the patient's data exceeds the normal range, and classifies the data in the medical files of different patients separately.

[0015] As a further solution of the present invention: the data storage encryption module includes:

[0016] Permission grading module: Based on the multiple sets of data, different identities are established for medical personnel and patients to complete identity grading, specific operation permission thresholds are set for each identity grade, and a unique authentication identifier is assigned to each identity grade;

[0017] Data sensitivity classification module: Based on the identity classification combined with the data generation time, data source department and the disease type involved in the data, the sensitivity of the data is classified accordingly;

[0018] Data encryption storage module: adopts a layered encryption strategy, and different identity levels can access data of different data sensitivity levels, which are stored in different encrypted storage areas respectively, and independent encryption algorithms and keys are used for the encrypted storage areas.

[0019] As a further solution of the present invention: the multi-feature information analysis module includes:

[0020] Data fusion analysis submodule: Based on multiple sets of data, NLP is used to process text to uniformly identify and encode machine-tested indicators and doctors' handwritten medical records to obtain coding results;

[0021] Model generation submodule: Based on the coding results, the age-complication-physiological indicator correlation model is established and generated according to the different ages, other complications and physiological conditions of different patients;

[0022] Progressive random forest submodule: Based on the association model, different disease characteristics are assigned every 24 hours according to clinical guidelines, and arranged in order from severe to mild to form a dynamic random forest.

[0023] As a further embodiment of the present invention: the progressive random forest submodule includes:

[0024] Clinical knowledge injection unit: Based on clinical guidelines, it analyzes clinical guidelines to build a knowledge graph of decision rules and assigns initial feature weights according to the decision rules;

[0025] Dynamic feature sampling unit: Based on the initial feature weights, feature subsets are selected to assign different pathological features, and important features are retained for progressive optimization.

[0026] As a further solution of the present invention: the dynamic feature sampling unit includes:

[0027] Feature importance assessment subunit: Based on the initial feature weights, retrieve the feature trigger weight ratios of cases in recent years and generate a feature importance trend graph;

[0028] Feature selection subunit: Based on the trend graph, the feature importance is divided into heavy-medium-light, and important features are retained for progressive optimization.

[0029] As a further solution of the present invention: the intelligent hierarchical warning module includes:

[0030] Multi-angle analysis submodule: Based on the ranking analysis results, combined with the treatment results and causes of different patients, the recovery and recurrence rates are determined;

[0031] Situation perception and early warning submodule: Based on the recovery and relapse ratio, the relapse probability of patients with relapse is ranked, and the ranking is updated every six hours to obtain the ranking results;

[0032] Intelligent triage and scheduling submodule: Based on the sorting results, the locator is used to determine the specific locations of different patients and develop the most time-saving treatment route.

[0033] As a further solution of the present invention: the resource dynamic optimization module includes:

[0034] Medical digital twin submodule: This module uses Revit software to build a hospital BIM model, monitor the usage status of key resources such as ICU beds and ventilators in real time, and automatically maps each resource to a specific department, generating a Sankey diagram showing utilization rates by department, equipment, and time period.

[0035] Backup trigger module: Based on the Sankey diagram, if all key resources are in use and there is an emergency, the backup resources will be triggered and other backup equipment will be used in a timely manner.

[0036] As a further solution of the present invention: the medical digital twin submodule includes:

[0037] State mapping unit: Encodes the usage status of key resources such as ICU beds and ventilators into model material attributes to generate a state mapping map;

[0038] Predictive Status Monitoring Unit: This unit retrieves and analyzes the spatiotemporal patterns of equipment usage over the past three years, constructs a department-equipment-time period utilization rate map, uses the Attention-GRU network to predict utilization rates for the next two hours, and generates a heat map of equipment usage, which is updated every fifteen minutes.

[0039] As a further solution of the present invention: the backup trigger module includes:

[0040] Emergency Grading Assessment Unit: Based on the Sankey diagram, SOFA, APACHE II and other scores are obtained from the patient's electronic medical record in real time to calculate the 24-hour survival probability;

[0041] Cross-domain resource discovery unit: Based on the survival probability, it filters out data with a probability less than 89%, extracts constraints such as the required equipment type and usage duration, calculates the equipment distance × available time window × disinfection preparation time, and issues an automatic equipment reservation request.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are:

[0043] 1. The present invention uses a random forest algorithm to establish an age-complication-physiological indicator association model based on multiple groups of data formed by classifying data within and outside the normal range. At the same time, it combines with the patient's individual characteristics to determine the impact of the patient's individual characteristics on data in different ranges, thereby realizing dynamic feature importance ranking analysis. By controlling the data access rights of medical personnel, on the basis of ensuring patient privacy compliance, medical personnel can pay attention to critically ill patients more quickly, realize visual early warning of critical risk characteristics, and effectively prevent data leakage risks.

[0044] 2. The present invention dynamically optimizes resources and uses BIM technology and locators to determine the specific location and usage of key resources such as ICU beds and ventilators in real time, so as to promptly allocate corresponding key resources or spare resources to critically ill patients, thereby further shortening the time for rescuing critically ill patients and improving the rescue effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the overall system flow of the data-driven digital medical case management system of the present invention;

[0046] Figure 2 Schematic diagram of the system framework of the data-driven digital medical case management system of the present invention;

[0047] Figure 3 This is a schematic diagram of the architecture of the multi-feature information analysis module in the data-driven digital medical case management system of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0050] Example 1

[0051] See also Figure 1 、 Figure 2 and Figure 3 The present invention provides a technical solution: a data-driven digital medical case management system including:

[0052] Dynamic threshold classification module: Updates seasonal symptom rates based on the seasons, records the medical records of different patients in the corresponding seasons, compares the data within the normal range, and classifies the data within and outside the normal range to form multiple groups of data;

[0053] Data storage encryption module: Based on the multiple sets of data, a multi-level authority hierarchy is constructed, different identities are established for medical personnel and patients to complete identity classification, and data sensitivity classification is performed accordingly based on the identity classification. Data is encrypted and stored according to the identity classification and data sensitivity classification. Medical personnel and patients with different identity classifications can view the data after the corresponding data sensitivity classification;

[0054] Multi-feature information analysis module: Based on multiple data sets, an age-complication-physiological indicator correlation model is established. The random forest algorithm is used to combine the impact of individual patient characteristics on the generation of data in different ranges to perform feature importance ranking analysis and obtain ranking analysis results.

[0055] Intelligent hierarchical warning module: Based on the sorting analysis results, it conducts risk assessment on the treatment plans of different case information, issues warnings of different degrees, and draws allocation results;

[0056] Dynamic resource optimization module: Based on the allocation results, BIM technology is used to build a three-dimensional resource topology map of the hospital. The location of mobile devices is tracked in real time through locators to determine whether to use mobile devices or other backup devices, or to book usage time based on the patient's needs.

[0057] The dynamic threshold classification module includes:

[0058] Environmental coupling data acquisition submodule: Based on the different air quality in different seasons, temperature and humidity sensors and air quality sensors are used to collect environmental parameters such as temperature, humidity, and air quality in the ward in real time to obtain information about the current season;

[0059] Adaptive Threshold Generation Submodule: Based on the current season information, it automatically updates the symptom onset rate and compares it with the data in the medical files of different patients to obtain multiple sets of data to determine whether the patient's data exceeds the normal range, and classifies the data in the medical files of different patients separately;

[0060] By using temperature and humidity sensors and air quality sensors to collect environmental parameters such as temperature, humidity, and air quality in the ward in real time, information about the current season can be obtained. The onset of symptoms will change according to environmental parameters such as temperature, humidity, and air quality in the ward. Therefore, the onset rate of symptoms can be automatically updated, and the onset rate of symptoms can be compared with the data in the medical files of different patients. Then, the results can be obtained to determine whether the data of different patients exceeds the normal range. The data in the medical files of different patients are classified and processed separately to obtain multiple groups of different data that exceed and do not exceed the normal range.

[0061] The data storage encryption module includes:

[0062] Permission grading module: Based on the multiple sets of data, different identities are established for medical personnel and patients to complete identity grading, specific operation permission thresholds are set for each identity grade, and a unique authentication identifier is assigned to each identity grade;

[0063] Different identities are established for medical personnel and patients to complete identity classification, and specific operation permission thresholds are set for each identity classification. For example, attending physicians can access the complete medical records, examination reports and other core data of the patients they are responsible for, other doctors and nurses can view some basic information and desensitized data, and patients can view all their personal data; when medical personnel or patients initiate data operation requests, the system automatically determines whether the operation corresponding to the request exceeds the operation permission threshold corresponding to their identity classification. If it exceeds, the operation request is rejected and the illegal operation behavior and the initiator's identity information are recorded. During the data access process, medical personnel and patients need to authenticate their identities through the identity authentication identifier. Only after the verification is passed can they view the data of the corresponding data sensitivity classification according to their identity classification;

[0064] Data sensitivity classification module: Based on the identity classification combined with the data generation time, data source department and the disease type involved in the data, the sensitivity of the data is classified accordingly;

[0065] Data encryption storage module: adopts a layered encryption strategy, and different identity levels can access data of different data sensitivity levels, which are stored in different encrypted storage areas. Independent encryption algorithms and keys are used for the encrypted storage areas. Data of different sensitivity levels are stored in different areas and independent encryption algorithms and keys are used, which enhances the security of data storage. All encryption keys are uniformly managed and securely stored, and keys are regularly updated and backed up to ensure the security and availability of keys. Regular updates and backups of keys further improve data security.

[0066] The multi-feature information analysis module includes:

[0067] Data fusion analysis submodule: Based on multiple sets of data, NLP is used to process text to uniformly identify and encode machine-tested indicators and doctors' handwritten medical records to obtain coding results;

[0068] By analyzing these multiple groups, which may contain machine-verified results and medical staff's manual records, NLP is used to process the text to uniformly identify and encode the machine-verified indicators and the doctor's handwritten medical records to obtain the coding results;

[0069] Model generation submodule: Based on the coding results, the age-complication-physiological indicator correlation model is established and generated according to the different ages, other complications and physiological conditions of different patients;

[0070] By coding the results, we calculated the Pearson correlation coefficient matrix between age and complications, and then used a heat map to visualize the data results more intuitively. We then marked the values ​​with absolute values ​​close to 1 and marked them in red. Then, by accessing the patient files, we obtained the ages, other complications, and physiological conditions of different patients. Based on the differences in age, other complications, and physiological conditions of different patients, we established and generated an age-complication-physiological indicator correlation model by consulting clinical guidelines.

[0071] Progressive random forest submodule: Based on the association model, different disease characteristics are assigned every 24 hours according to clinical guidelines, and arranged in order from severe to mild to form a dynamic random forest;

[0072] By analyzing the decision rules constructed in the clinical guidelines, a knowledge graph is obtained. The initial feature weights are assigned according to the decision rules. Based on the initial feature weights, the trigger weight ratios of case features in recent years are retrieved, and feature subsets are selected to generate a feature importance trend graph. The feature importance is divided into severe-moderate-mild to assign different pathological characteristics. Each feature is screened, and important features are retained. These important features are then progressively optimized. Then, the pathological characteristics are obtained based on the patient's file status. However, because the patient's file status may change in real time, it is necessary to adjust every 24 hours according to the clinical guidelines to assign different pathological characteristics. These characteristics are arranged in descending order from severe to moderate and mild to form a dynamic random forest.

[0073] The model generation submodule includes:

[0074] Dynamic feature interaction unit: Based on the coding results, the Pearson correlation coefficient matrix between age and complications is calculated and visualized using a heat map, with values ​​close to 1 being marked.

[0075] The Progressive Random Forest submodules include:

[0076] Clinical knowledge injection unit: Based on clinical guidelines, it analyzes clinical guidelines to build a knowledge graph of decision rules and assigns initial feature weights according to the decision rules;

[0077] Dynamic feature sampling unit: Based on the initial feature weights, feature subsets are selected to assign different pathological features, and important features are retained for progressive optimization.

[0078] The dynamic feature sampling unit includes:

[0079] Feature importance assessment subunit: Based on the initial feature weights, retrieve the feature trigger weight ratios of cases in recent years and generate a feature importance trend graph;

[0080] Feature selection subunit: Based on the trend graph, the feature importance is divided into heavy-medium-light, and important features are retained for progressive optimization.

[0081] The intelligent graded warning module includes:

[0082] Multi-angle analysis submodule: Based on the ranking analysis results, combined with the treatment results and causes of different patients, the recovery and recurrence rates are determined;

[0083] By sorting and analyzing the results and reviewing the patient's medical records, combined with the treatment results and causes of different patients, we can determine the recovery and recurrence rates of different patients and screen out patients who are likely to relapse.

[0084] Situation perception and early warning submodule: Based on the recovery and relapse ratio, the relapse probability of patients with relapse is ranked, and the ranking is updated every six hours to obtain the ranking results;

[0085] Intelligent triage and scheduling submodule: Based on the sorting results, the locator is used to determine the specific locations of different patients and develop the most time-saving treatment route.

[0086] The likelihood of recurrence is ranked based on the recovery and relapse rates of patients with recurrence. However, patients' physical conditions change daily, so the ranking is updated every six hours to determine the ranking results. A locator is then attached to each patient, and the locator is used to determine the specific location of each patient. The distance between the patient and different available clinics, as well as the elevator waiting time, are then calculated to develop the most time-saving treatment route.

[0087] The resource dynamic optimization module includes:

[0088] Medical digital twin submodule: This module uses Revit software to build a hospital BIM model, monitor the usage status of key resources such as ICU beds and ventilators in real time, and automatically maps each resource to a specific department, generating a Sankey diagram showing utilization rates by department, equipment, and time period.

[0089] Backup trigger module: Based on the Sankey diagram, if all key resources are in use and there is an emergency, the backup resources will be triggered and other backup equipment will be used in a timely manner.

[0090] The medical digital twin sub-module includes:

[0091] State mapping unit: Encodes the usage status of key resources such as ICU beds and ventilators into model material attributes to generate a state mapping map;

[0092] By using Revit software to build a hospital BIM model, the usage status of key resources such as ICU beds and ventilators is monitored in real time. Each resource is automatically mapped to a specific department, generating a Sankey diagram of department-equipment-time period utilization. By encoding the usage status of key resources such as ICU beds and ventilators into model material properties (e.g., red = in use, yellow = about to be used, green = unused), a status mapping diagram can be generated.

[0093] Predictive Status Monitoring Unit: This unit retrieves and analyzes the spatiotemporal patterns of equipment usage over the past three years, constructs a department-equipment-time period utilization rate map, uses the Attention-GRU network to predict utilization rates for the next two hours, and generates a heat map of equipment usage, which is updated every fifteen minutes.

[0094] The backup trigger module includes:

[0095] Emergency Grading Assessment Unit: Based on the Sankey diagram, SOFA, APACHE II and other scores are obtained from the patient's electronic medical record in real time to calculate the 24-hour survival probability;

[0096] Cross-domain resource discovery unit: Based on the survival probability, it filters out data with a probability less than 89%, extracts constraints such as the required equipment type and usage duration, calculates the equipment distance × available time window × disinfection preparation time, and issues an automatic equipment reservation request;

[0097] By retrieving and analyzing the spatiotemporal patterns of equipment usage over the past three years, a department-equipment-time period utilization map was constructed. The Attention-GRU network was used to predict utilization rates for the next two hours, generating a heat map of equipment usage that was updated every fifteen minutes. Based on the results displayed in the Sankey diagram, the patient's electronic medical record was retrieved, and SOFA, APACHE II, and other scores were obtained in real time. The patient's 24-hour survival probability was calculated, and then data with a probability less than 89% was screened out. Constraints such as the required equipment type and usage duration were extracted, and the equipment distance × available time window × disinfection preparation time was calculated. An automatic equipment reservation request was issued. If all key resources are in use and there is an emergency, backup resources are immediately triggered and other backup equipment is used to ensure that critically ill patients can receive timely assistance.

[0098] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A data-driven, digital, and intelligent medical case management system, featuring: The system comprises: Dynamic threshold classification module: Updates seasonal symptom rates based on the seasons, records the medical records of different patients in the corresponding seasons, compares the data within the normal range, and classifies the data within and outside the normal range to form multiple groups of data; Data storage encryption module: Based on the multiple sets of data, a multi-level authority hierarchy is constructed, different identities are established for medical personnel and patients to complete identity classification, and data sensitivity classification is performed accordingly based on the identity classification. Data is encrypted and stored according to the identity classification and data sensitivity classification. Medical personnel and patients with different identity classifications can view the data after the corresponding data sensitivity classification; Multi-feature information analysis module: Based on the multiple sets of data, an age-complication-physiological indicator association model is established. A random forest algorithm is used to analyze the impact of individual patient characteristics on the generation of data in different ranges, and feature importance ranking is performed to obtain ranking analysis results. Intelligent graded warning module: Based on the ranking analysis results, it conducts risk assessment on treatment plans of different case information, issues warnings of different degrees, and obtains allocation results; Resource dynamic optimization module: Based on the allocation results, use BIM technology to build a three-dimensional resource topology map of the hospital, track the location of mobile devices in real time through locators, determine whether to use the mobile devices or other backup devices, or book usage time according to the patient's needs.

2. The data-driven, digital, and intelligent medical case management system according to claim 1, characterized in that: The dynamic threshold classification module includes: Environmental coupling data acquisition submodule: Based on the different air quality in different seasons, temperature and humidity sensors and air quality sensors are used to collect environmental parameters such as temperature, humidity, and air quality in the ward in real time to obtain information about the current season; Adaptive threshold generation submodule: Based on the current seasonal information, the symptom incidence rate is automatically updated, and compared with the data in the medical files of different patients to obtain multiple sets of data on whether the patient's data exceeds the normal range, and the data in the medical files of the different patients are classified respectively.

3. The data-driven, digital, and intelligent medical case management system according to claim 1, characterized in that: The data storage encryption module includes: Permission grading module: Based on the multiple sets of data, different identities are established for medical personnel and patients to complete identity grading, specific operation permission thresholds are set for each identity grade, and a unique authentication identifier is assigned to each identity grade; Data sensitivity classification module: Based on the identity classification combined with the data generation time, data source department and the disease type involved in the data, the sensitivity of the data is classified accordingly; Data encryption storage module: adopts a layered encryption strategy, and different identity levels can access data of different data sensitivity levels, which are stored in different encrypted storage areas respectively, and independent encryption algorithms and keys are used for the encrypted storage areas.

4. The data-driven, digital, and intelligent medical case management system according to claim 1, characterized in that: The multi-feature information analysis module includes: Data fusion analysis submodule: Based on the multiple sets of data, NLP is used to process text to uniformly identify and encode machine inspection indicators and doctor's handwritten medical records to obtain coding results; Model generation submodule: Based on the coding results, according to the different ages of different patients, the different complications they suffer from, and the different physiological conditions, establish and generate an age-complication-physiological indicator correlation model; Progressive random forest submodule: Based on the association model, different disease characteristics are assigned every 24 hours according to clinical guidelines, and arranged in order from severe to mild to form a dynamic random forest.

5. The data-driven digital medical case management system according to claim 4 is characterized by: The progressive random forest submodule includes: Clinical knowledge injection unit: Based on the clinical guidelines, the clinical guidelines are parsed to construct a knowledge graph of decision rules, and initial feature weights are assigned according to the decision rules; Dynamic feature sampling unit: Based on the initial feature weights, feature subsets are selected to assign different pathological features, and important features are retained for progressive optimization.

6. The data-driven, digital, and intelligent medical case management system according to claim 5, characterized in that: The dynamic feature sampling unit includes: Feature importance assessment subunit: Based on the initial feature weights, retrieve the case feature trigger weight ratios in recent years and generate a feature importance trend graph; Feature selection subunit: Based on the trend graph, the feature importance is divided into heavy-medium-light, and important features are retained for progressive optimization.

7. The data-driven, digital, and intelligent medical case management system according to claim 1, characterized in that: The intelligent hierarchical warning module includes: Multi-angle analysis submodule: Based on the ranking analysis results, combined with the treatment results and causes of different patients, the recovery and recurrence rates are determined; Situation perception and early warning submodule: Based on the recovery and relapse ratios, the relapse probability of patients with relapse is ranked, and the ranking is updated every six hours to obtain the ranking results; Intelligent triage scheduling submodule: Based on the sorting results, the locator is used to determine the specific locations of different patients and formulate the most time-saving treatment route.

8. The data-driven digital medical case management system according to claim 1 is characterized by: The resource dynamic optimization module includes: Medical digital twin submodule: This module uses Revit software to build a hospital BIM model, monitor the usage status of key resources such as ICU beds and ventilators in real time, and automatically maps each resource to a specific department, generating a Sankey diagram showing utilization rates by department, equipment, and time period. Backup trigger module: Based on the Sankey diagram, if all key resources are in use and there is an emergency, the backup resources are triggered and other backup devices are used in a timely manner.

9. The data-driven digital medical case management system according to claim 8, characterized in that: The medical digital twin submodule includes: State mapping unit: Encodes the usage status of key resources such as ICU beds and ventilators into model material attributes to generate a state mapping map; Predictive Status Monitoring Unit: This unit retrieves and analyzes the spatiotemporal patterns of equipment usage over the past three years, constructs a department-equipment-time period utilization rate map, uses the Attention-GRU network to predict utilization rates for the next two hours, and generates a heat map of equipment usage, which is updated every fifteen minutes.

10. The data-driven digital medical case management system according to claim 8, characterized in that: The standby trigger module includes: Emergency grading assessment unit: Based on the Sankey diagram, SOFA, APACHE II and other scores are obtained from the patient's electronic medical record in real time to calculate the 24-hour survival probability; Cross-domain resource discovery unit: Based on the survival probability, it filters out data with a probability less than 89%, extracts constraints such as the required equipment type and usage duration, calculates the equipment distance × available time window × disinfection preparation time, and issues an automatic equipment reservation request.

Images