Vaccination abnormal reaction investigation and diagnosis system
By designing an adverse reaction investigation and diagnosis system for immunization, we have achieved full-process electronic management and real-time data sharing, which solves the problems of low efficiency in the diagnostic process, inconsistent data and high information security risks in the existing technology, improves the consistency of diagnostic conclusions and information security, and forms a dynamically updated AEFI knowledge graph.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-03-24
AI Technical Summary
In existing technologies, the investigation and diagnosis process for adverse reactions to vaccination is inefficient, relies on offline processes leading to delayed diagnostic conclusions, suffers from insufficient data collaboration, depends on expert experience and lacks standardized auxiliary tools, poses high information security risks, is prone to loss of paper documents, and lacks a structured knowledge base for historical cases, resulting in poor consistency of diagnostic conclusions and inconsistent information sharing.
An adverse reaction investigation and diagnosis system for immunization was designed, including a data acquisition module, an AI-assisted diagnosis module, a three-level collaboration module, an expert review module, a document generation module, and a knowledge base module. It realizes full-process electronic management, and through AI-assisted diagnosis and expert collaboration, combined with encrypted transmission and access control, a structured knowledge base is formed, supporting real-time data sharing and secure storage.
The diagnostic process time was reduced by 80%, data sharing consistency reached 100%, diagnostic conclusion consistency increased to 90%, information security risks were reduced by 99%, knowledge reuse rate was increased by 70%, and overall diagnostic efficiency and accuracy were greatly improved.
Smart Images
Figure CN121726033A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of investigation and diagnosis of abnormal reactions to vaccination, in particular to an investigation and diagnosis system for abnormal reactions to vaccination. BACKGROUND
[0002] Vaccination is the most economical and effective public health measure to control infectious diseases. With the improvement of vaccination coverage, the monitoring, investigation and disposal of suspected abnormal reactions to vaccination (AEFI) have become a key link to ensure vaccination safety. The investigation and diagnosis of AEFI requires the integration of multiple dimensions of information such as the health status of the vaccinated person, the characteristics of the vaccine, and the vaccination process, involving the cooperation of multiple parties such as health departments, disease control centers, vaccination units, medical institutions and expert teams. The accuracy and timeliness of the investigation and diagnosis directly affect the public's trust in vaccines.
[0003] In the prior art, the investigation and diagnosis of AEFI relies on offline processes: the vaccination unit manually fills out the report form, which is submitted to the disease control center by mail or fax, and the expert review requires on-site meetings, with the diagnosis conclusion being passed through paper documents for approval. The average time for offline data transmission is 3-5 days, and the organization cycle of the expert review meeting is as long as 2 weeks, resulting in a lag in diagnosis conclusions. The data of health departments, disease control centers and vaccination units are stored in independent systems, and information sharing requires manual export and import, which may lead to inconsistent data. The diagnosis relies on individual experience of experts, lacks standardized AI auxiliary tools, and the consistency of conclusions for similar cases is insufficient. Paper documents are easily lost, there is no electronic information system support, the protection of the privacy data of the vaccinated person is weak, historical cases have not formed a structured knowledge base, and new cases require repeated literature search for diagnosis, lacking an intelligent reference mechanism. SUMMARY
[0004] In view of the deficiencies of the prior art, the present application provides an investigation and diagnosis system for abnormal reactions to vaccination, which solves the problems of low process efficiency, long offline data transmission time, long organization cycle of expert review, insufficient data collaboration, lack of investigation and diagnosis system support, manual recording and statistics of information at all levels, inconsistent data caused by manual operation for information sharing, poor diagnosis standardization, reliance on expert experience, lack of standardized auxiliary tools, low consistency of conclusions for similar cases, high information security risk, lack of electronic system support, and low knowledge reuse rate, and the like.
[0005] To achieve the above purpose, the present application is implemented by the following technical solution: an investigation and diagnosis system for abnormal reactions to vaccination, comprising: a data acquisition module for acquiring basic information of the vaccinated person, vaccination information and medical records; an AI assisted diagnosis module for automatically extracting key information from medical records and matching AEFI classification standards; A three-level collaborative module is used to realize data sharing and permission control of the health commission-disease control center-vaccination unit. An expert review module is used for expert library management and online consultation. A document generation module is used to automatically generate investigation and diagnosis documents. A knowledge base module is used to store desensitized historical cases and AEFI-related literature. The knowledge base module is connected to the diagnosis model unit through a feedback learning interface, and the knowledge base model training sample is automatically updated every month. A time limit management module is used to control the validity period of the short link and the temporary permission time length. The validity period of the vaccine manufacturer data upload link is set by the user, and the temporary permission time length for expert review is set to 24 hours after the meeting ends.
[0006] Preferably, the data collection module comprises: A picture recognition unit is used to upload pictures and perform recognition. A manual input unit is used to connect with the vaccination unit terminal through wireless communication. A vaccine information association unit is used to automatically match the production manufacturer information and batch data of the vaccinated vaccine. The vaccine information association unit is connected to the vaccine circulation database through an incremental update interface.
[0007] Preferably, the picture recognition unit supports: OCR recognition of identity cards, vaccination cards, medical records, and other documents; Extraction of key information from medical image reports.
[0008] Preferably, the AI-assisted diagnosis module comprises: A natural language processing unit, the output end of which is connected to the input end of the diagnosis model unit, realizing computer understanding and generation of human self-recognition language; The diagnosis model unit is trained by a deep learning algorithm, and the algorithm formula is:
[0009] Wherein, is the input feature vector, , and is the weight matrix, is the bias term, is the sigmoid activation function; A causal association strength calculation subunit is used to output the association probability value of the disease and vaccination. The calculation method is:
[0010] Wherein is the knowledge base model training sample parameter, and the value range is 0.1-0.3.
[0011] Preferably, the third-level coordination module comprises a data transmission unit and an authority management unit, the data transmission unit adopts an SM4 encryption algorithm for data encryption, and the authority management unit is bidirectionally connected with a user role database.
[0012] Preferably, the authority management unit is provided with: District administrator authority: can view individual case data in the jurisdiction and initiate district-level expert review; City administrator authority: can view individual case data in the city, review and approve district-level diagnosis reports, and initiate city-level expert review; Vaccination unit authority: can only view individual case data and supplementary data reported by the unit.
[0013] Preferably, the expert review module comprises an expert extraction unit and a voting unit, the expert extraction unit is connected with a preset expert database through a random number generator, and the output end of the voting unit is connected with a diagnosis conclusion database.
[0014] Preferably, the expert extraction unit is configured with: An expert field tag library, including at least 8 subfields such as immunology, pediatrics and neurology.
[0015] Preferably, the document generation module comprises a template calling unit and an AI writing unit, the template calling unit is connected with a multi-level template library through a version control unit.
[0016] Preferably, the AI writing unit comprises: A demand analysis and logical reasoning subunit that receives user input, analyzes the semantics and context of the input content, supports multi-modal instructions, combines NLP technology for deep semantic analysis, automatically loads knowledge in the background, associates with local knowledge base data, supplements user historical preference data, forms a deep thinking logical framework, the core structure is abstract, method and conclusion, and the extension element is an embedded knowledge base data reference tree diagram; A content planning and content construction subunit that selects an adverse reaction diagnosis template according to the report type, divides chapters and paragraph levels, extracts relevant background information from the individual case information associated with the diagnosis report, integrates multi-modal materials, establishes an association network between contents, generates an initial document based on a large language model, and optimizes the output diagnosis report content according to user feedback during the process; A grammar correction subunit that realizes wrong character recognition and sentence fluency optimization based on NLP technology, Generates text that meets grammar rules and semantic logic according to the required context and situation; A version management subunit that automatically saves more than 10 modified versions of the diagnosis, each version is marked with the modifier and the modification timestamp.
[0017] This system is based on a three-tier collaborative architecture and achieves full-process investigation and diagnosis of adverse reactions to vaccination through a closed-loop process of "data collection - intelligent analysis - expert collaboration - conclusion generation". The core working principle is as follows: Data Acquisition and Initialization Vaccination units can use short, time-sensitive links or the system's reporting module to enter basic information about vaccine recipients, such as age, gender, ID number, medical history, and vaccination information, and upload materials such as vaccination certificates and medical records. The system automatically connects to the Guangdong Provincial Vaccine Circulation Database and the Guangzhou Municipal Vaccination System, synchronizing basic information such as vaccine manufacturers and batches to create basic case files.
[0018] Acceptance and Supplement of Documents The district / municipal CDC receives case applications through the acceptance module and verifies the completeness of basic information. If supplementary materials are required, the system generates a short, time-sensitive link, which is sent to the vaccine manufacturer via SMS. The manufacturer then uploads the vaccine quality inspection report, instructions for use, and other materials via the link, and the materials are automatically archived in the case file database. Simultaneously, the system allows administrators to manually supplement materials from the recipient or vaccination unit to ensure complete diagnostic evidence.
[0019] AI-assisted diagnostic preprocessing After the data is verified, the AI-assisted diagnosis module activates the natural language processing unit to automatically extract key information such as symptoms, onset time, and treatment process from the medical records and generate structured diagnostic evidence. At the same time, the diagnostic model unit matches the extracted information with the AEFI classification standard, calculates the causal association strength between the disease and the vaccination, and forms a preliminary classification suggestion.
[0020] Expert review collaboration The district / municipal CDC randomly selects experts from a pre-set expert database through an expert review module. The system automatically sends short links and accounts with temporary permissions. After logging in, experts can view case files and submit review opinions through the online consultation function. If an offline meeting is held, experts can vote on the diagnostic conclusions within the system, and the results are synchronized to the conclusion form in real time and generate electronic signature records.
[0021] Diagnostic report generation and review The document generation module calls a preset template and automatically generates a draft using the AI writing unit. City / district CDC personnel can then edit and adjust the draft. The adjusted report enters a multi-level review process. Once approved, the system provides feedback to the recipient via SMS or mail, and simultaneously adds the anonymized case to the knowledge base.
[0022] Data security and access control In the whole process, the three-level collaborative module guarantees the data transmission and storage security through the SM4 encryption algorithm, and the permission management unit distributes operation permissions according to roles. The system automatically records user operation logs, ensures data traceability, and meets privacy protection requirements.
[0023] The application provides a vaccination abnormal reaction investigation and diagnosis system. 1、The application shortens the data transmission time from 3-5 days to 2 hours, and compresses the expert review period from 2 weeks to 3 days through whole-process electronic flow management, so that the overall diagnosis timeliness is improved by more than 80%. The average time consumption of AI automatic generation of diagnosis report is only 15 minutes, which is 32 times higher than the efficiency of manual writing, solving the problem of low efficiency in the prior art.
[0024] 2、The application realizes real-time data sharing of the health commission-disease control center-vaccination unit through the encryption transmission and permission control of the three-level collaborative module, and the data consistency reaches 100%. The incremental synchronization mechanism with the vaccine circulation system of Guangdong Province ensures that the vaccine information update delay is less than or equal to 5 minutes.
[0025] 2、The application improves the consistency of similar case diagnosis conclusions to more than 90% through the AI auxiliary diagnosis module, the causality correlation strength algorithm (accuracy is greater than or equal to 92%) and expert collaborative review. The extraction accuracy of the natural language processing unit to medical record information reaches 98%, which is much higher than the extraction accuracy of 75% of manual extraction.
[0026] 2、The application realizes the whole life cycle protection of sensitive data through the storage and transmission mechanism based on the SM4 encryption algorithm and the hierarchical permission management. The audit log is stored for 6 months and supports traceability, and the data leakage risk is reduced by 99% compared with the paper storage method of the prior art.
[0027] 2、The application forms a dynamic updated AEFI knowledge graph through the structured integration of desensitization cases and literature through the knowledge base module. The case retrieval response time is less than or equal to 2 seconds, which provides instant reference for new case diagnosis, and the knowledge reuse rate is improved by 70%, which breaks through the limitation of the knowledge fragmentation of the prior art. BRIEF DESCRIPTION OF DRAWINGS
[0028] Figure 1 The system flowchart of the application. DETAILED DESCRIPTION
[0029] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.
[0030] Embodiments: Please refer to the attached Figure 1 The embodiment of the present application provides a kind of preventive inoculation abnormal reaction investigation diagnosis system, comprising: Data acquisition module for collecting basic information of inoculation, inoculation information and medical record, it includes: Picture recognition unit is used to upload picture and carries out identification, supports the key information extraction of pdf and the like format: OCR identification of identity card, vaccination certificate, medical record book and the like certificate; Key information extraction of medical image report; Manual input unit is used to be connected with the terminal of inoculation unit by wireless communication connection; Vaccine information association unit is used to automatically match the production manufacturer information and batch data of inoculation vaccine, and the vaccine information association unit is connected with vaccine circulation database by incremental update interface.
[0031] AI auxiliary diagnosis module for automatically extracting medical record key information and matching AEFI classification standard, it includes: Natural language processing unit, its output end is connected with the input end of diagnosis model unit, realizes the understanding and generation of computer to human self language; Diagnosis model unit is generated by training deep learning algorithm, and algorithm formula is:
[0032] Among them, is input feature vector, , and is weight matrix, is bias term, sigmoid activation function is; Causal association strength calculation subunit is used to output the correlation probability value of disease and preventive inoculation, and the calculation method is:
[0033] Wherein knowledge base model training sample parameter is 0.1-0.3.
[0034] Three-level collaborative module for realizing data sharing and authority control of health commission-disease control center-inoculation unit, including data transmission unit and authority management unit, the data transmission unit adopts SM4 encryption algorithm to carry out data encryption, and the authority management unit is bidirectionally connected with user role database, and the authority management unit is provided with: District administrator authority: can view individual case data in jurisdiction and initiate district expert review; City-level administrator authority: can view all city case data, approve district-level diagnosis reports, and initiate city-level expert review; Vaccination unit authority: can only view the unit's reported case data and supplementary materials.
[0035] An expert review module for expert library management and online consultation includes an expert extraction unit and a voting unit. The expert extraction unit is connected to a preset expert library through a random number generator. The output end of the voting unit is connected to a diagnosis conclusion database. The expert extraction unit is configured with: An expert field tag library includes at least 8 subfields such as immunology, pediatrics, and neurology.
[0036] A document generation module for automatically generating investigation and diagnosis reports includes a template calling unit and an AI writing unit. The template calling unit is connected to a multi-level template library through a version control unit. The AI writing unit includes: A demand analysis and logical reasoning subunit receives user input, analyzes the semantics and context of the input content, supports multi-modal instructions, combines NLP technology for deep semantic analysis, automatically loads background knowledge, correlates local knowledge base data, supplements user historical preference data, forms a deep thinking logical framework, and the core structure is abstract, method and conclusion, and the extension element is an internal knowledge base data reference tree diagram; A content planning and content construction subunit selects an adverse reaction diagnosis template based on the report type, divides chapters and paragraph levels, extracts relevant background information from the case information associated with the diagnosis report, integrates multi-modal materials, establishes an association network between contents, generates an initial document based on a large language model, and optimizes the output diagnosis report content according to user feedback during the process; A grammar correction subunit based on NLP technology realizes incorrect character recognition and sentence fluency optimization, and Generates text that meets grammar rules and semantic logic according to the required context and situation; A version management subunit automatically saves more than 10 modified versions of the diagnosis report, each version marked with the modifier and the modification timestamp.
[0037] A knowledge base module for storing desensitized historical cases and AEFI-related literature. The knowledge base module is connected to the diagnosis model unit through a feedback learning interface. The knowledge base model training sample is automatically updated every month; A time management module for controlling the validity period of the short link and the duration of temporary authority, wherein the validity period of the vaccine manufacturer's data upload link is set by the user, and the duration of the expert review temporary authority is set to 24 hours after the meeting ends.
[0038] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.
Claims
1. A system for investigating and diagnosing adverse reactions to immunization, characterized in that, include: The data acquisition module is used to collect basic information, vaccination information, and medical records of vaccine recipients; The AI-assisted diagnosis module is used to automatically extract key information from medical records and match them with AEFI classification standards. The three-level collaboration module is used to realize data sharing and access control among the National Health Commission, the Center for Disease Control and Prevention, and vaccination units. The expert review module is used for expert database management and online consultation. The document generation module is used to automatically generate survey and diagnostic reports; The knowledge base module is used to store desensitized historical cases and AEFI-related literature. The knowledge base module is connected to the diagnostic model unit through a feedback learning interface and automatically updates the knowledge base model training samples every month. The validity period management module is used to control the validity period of short links and the duration of temporary permissions. The validity period of vaccine manufacturer data upload links can be set by the user, and the duration of temporary permissions for expert review is set to 24 hours after the end of the meeting.
2. The system for investigating and diagnosing adverse reactions to immunization according to claim 1, characterized in that, The data acquisition module includes: The image recognition unit is used to upload and recognize images; The manual data entry unit is used to connect wirelessly with the terminal of the vaccination unit; The vaccine information association unit is used to automatically match the manufacturer information and batch data of the vaccines to be administered. The vaccine information association unit is connected to the vaccine circulation database through an incremental update interface.
3. The system for investigating and diagnosing adverse reactions to immunization according to claim 2, characterized in that, The image recognition unit supports: OCR recognition of documents such as ID cards, vaccination certificates, and medical records; Extraction of key information from medical imaging reports.
4. The system for investigating and diagnosing adverse reactions to immunization according to claim 1, characterized in that, The AI-assisted diagnostic module includes: The natural language processing unit connects its output to the input of the diagnostic model unit, enabling the computer to understand and generate human-generated language. The diagnostic model units are generated through training using a deep learning algorithm. The algorithm formula is as follows: ； in, For the input feature vector, , and This is the weight matrix. For bias terms, It is the sigmoid activation function; The causal association strength calculation subunit is used to output the association probability value between the disease and vaccination. The calculation method is as follows: ； in The parameters for training samples of the knowledge base model are 0.1-0.
3.
5. The system for investigating and diagnosing adverse reactions to immunization according to claim 1, characterized in that, The three-level collaborative module includes a data transmission unit and a permission management unit. The data transmission unit uses the SM4 encryption algorithm to encrypt data, and the permission management unit is bidirectionally connected to the user role database.
6. The system for investigating and diagnosing adverse reactions to immunization according to claim 5, characterized in that, The access control unit is configured with: District-level administrator privileges: can view case data within the jurisdiction and initiate district-level expert reviews; City-level administrator privileges: can view case data for the entire city and initiate city-level expert reviews; Vaccination unit permissions: Data dimensions are set and bound to accounts according to the organizational structure. The account of this unit can only view the case data and supplementary information reported by this unit.
7. The system for investigating and diagnosing adverse reactions to immunization according to claim 1, characterized in that, The expert review module includes an expert selection unit and a voting unit. The expert selection unit is connected to a preset expert database through a random number generator, and the output of the voting unit is connected to a diagnostic conclusion database.
8. The system for investigating and diagnosing adverse reactions to immunization according to claim 7, characterized in that, The expert extraction unit is configured with: The expert domain tag library includes at least eight sub-fields such as immunology, pediatrics, and neurology.
9. The system for investigating and diagnosing adverse reactions to immunization according to claim 1, characterized in that, The document generation module includes a template calling unit and an AI writing unit. The template calling unit is connected to a multi-level template library through a version control unit.
10. The system for investigating and diagnosing adverse reactions to immunization according to claim 9, characterized in that, The AI writing unit includes: The requirements analysis and logical reasoning subunit receives user input, analyzes the semantics and context of the input content, supports multimodal instructions, combines NLP technology for deep semantic parsing, automatically loads background knowledge, associates local knowledge base data, and supplements user historical preference data to form a deep thinking logical framework. The core structure consists of a summary, methods, and conclusions, and the extended elements are built-in knowledge base data reference tree diagrams. The content planning and content construction sub-unit selects the diagnostic template for adverse reactions based on the report type, divides the report into chapters and paragraphs, extracts relevant background information from the case information associated with the diagnostic report, integrates multimodal materials, establishes a network of connections between content, generates an initial document based on a large language model, and optimizes the output diagnostic report content based on user feedback during the process. The syntax correction subunit, based on NLP technology, identifies typos and optimizes sentence fluency. Generate text that conforms to grammatical rules and semantic logic according to the required context and situation; The version management sub-unit automatically saves more than 10 revised versions of the diagnostic report, with each version marked with the modifier and modification timestamp.