A digital management system for the whole life cycle of a drug instruction manual

Abstract

The application discloses a kind of whole life cycle digital management systems of drug instruction manual, it is related to medical digital and medical information management technical field, including the standardized data acquisition layer, intelligent data management layer, hierarchical scene service layer and compliance data opening layer of sequential data communication;The acquisition layer realizes the compliance filling of instruction manual template and version full-link traceability, the built-in semantic self-adapting hierarchical matching module in management layer, the intensity of data management and opening grade are matched simultaneously by calculating field semantic risk entropy, output structured standard data set with risk label, the service layer provides differentiated services to multiple subjects, the opening layer realizes hierarchical dynamic data opening and calls full-link audit, the application covers whole life cycle closed-loop management of drug instruction manual, consider data management quality and efficiency, balance the compliance and industry value of data opening.

Description

Technical Field

[0001] This invention relates to the field of pharmaceutical digitization and medical information management technology, and more specifically, to a digital management system for the entire lifecycle of drug instructions. Background Technology

[0002] With the digital transformation of the pharmaceutical industry and the increasing standardization of drug regulation, drug instructions, as the legally mandated carrier of core information on drug safety and efficacy, are receiving increasing attention from the industry regarding their management efficiency and data value. Currently, most drug marketing authorization holders still use a decentralized document management model, with drug instructions primarily in unstructured formats such as PDFs and paper documents. This makes version iteration and traceability difficult, and data quality cannot be guaranteed. At the same time, the public's access to drug information is fragmented and lacks professionalism, and there is a lack of standardized channels for data sharing within the industry, making it difficult to support scenarios such as pharmaceutical company R&D and clinical drug use decisions.

[0003] While some existing drug information management systems have implemented electronic storage and basic retrieval functions for drug instructions, significant shortcomings remain: First, the structured management of unstructured drug instructions largely relies on manual input and organization, resulting in low conversion rates, inconsistent data quality, and the widespread use of uniform threshold quality verification models. This makes it impossible to match governance resources based on content risk levels, hindering a balance between efficiency and quality. Second, data sharing often employs static tagging and hierarchical models, with fixed, manually preset permission granularity. This fails to adapt to dynamic risk changes brought about by version iterations of drug instructions, making it difficult to balance data compliance and industry value. Third, the data governance and data sharing stages are independent of each other. Governance results cannot directly support permission control in the sharing stage, and the needs of the sharing stage cannot be used to optimize the governance system, hindering the full realization of data asset value. Summary of the Invention

[0004] The purpose of this invention is to provide a digital management system for the entire lifecycle of drug instructions to solve the above-mentioned problems.

[0005] To achieve the above objectives, an embodiment of the present invention provides the following technical solution: A digital management system for the entire lifecycle of drug instructions includes a standardized data acquisition layer, an intelligent data governance layer, a hierarchical scenario service layer, and a compliant data opening layer, which are connected sequentially. The intelligent data governance layer has a built-in semantic adaptive hierarchical matching module, which is connected to the data quality closed-loop governance module and the hierarchical tagging data interface unit of the compliant data opening layer. The standardized data acquisition layer is configured with a templated data entry engine and a version full-link traceability module for pharmaceutical companies' terminals. The templated data entry engine has a built-in set of drug regulatory compliance verification rules for templated input of instruction manual information, real-time compliance verification, and error prompts. The version full-link traceability module is used to record the changes, operating entities, and time nodes of each version of the instruction manual, and generate a field-level full-link version traceability chain. The intelligent data governance layer is configured with a multimodal structured parsing engine, a data quality closed-loop governance module, and a semantic adaptive hierarchical matching module. It is used to automatically extract core fields from the unstructured instruction manual files output by the acquisition layer, and complete data cleaning and quality scoring. The semantic adaptive hierarchical matching module is used to calculate the semantic risk entropy value of each field of the instruction manual, and synchronously match the corresponding data governance strength and data openness level according to the entropy value to generate a multi-label classification system and output a structured, risk-labeled standard dataset of instruction manuals. The tiered scenario service layer connects to three types of terminals: public terminals, pharmaceutical company terminals, and medical institution terminals. It is used to provide differentiated instruction manual retrieval, management, and decision support services to different entities based on the standard dataset. The compliant data opening layer is configured with a hierarchical and tagged data interface unit, a permission approval unit, and a call end-to-end audit unit. The hierarchical and tagged data interface unit is connected to the risk tag data of the semantic adaptive hierarchical matching module and is used to adjust the opening permissions according to the semantic risk level. It outputs the corresponding hierarchical instruction manual data to external entities through a standardized API. The call end-to-end audit unit is used to log and audit all data call behaviors.

[0006] As a further improvement of the present invention, the template-based data entry engine incorporates regulatory standard verification rules into its set of drug instruction writing standards issued by drug regulatory authorities. The verification dimensions include the completeness of required fields, terminology standardization, format consistency, and content compliance. The template-based data entry engine triggers verification in real time during user input, marks content that does not conform to the rules, and simultaneously outputs compliance correction guidelines.

[0007] As a further improvement of the present invention, the version full-link traceability module generates a unique version identifier and change snapshot for each version iteration, automatically compares the field-level differences between adjacent versions, and generates a change comparison file; the traceability chain contains full node information from the initial draft, review, effective to revision, supports backtracking retrieval by time, operator, and changed field dimensions, and can export traceability reports that meet compliance audit requirements.

[0008] As a further improvement of the present invention, the multimodal structured parsing engine integrates an optical character recognition unit and a semantic parsing unit in the pharmaceutical field, supporting the input of instruction manual documents in PDF, scanned images, and Word formats; the semantic parsing unit is built based on a pre-trained language model in the pharmaceutical field, and can automatically identify and extract seven core fields, including drug name, indications, contraindications, usage and dosage, adverse reactions, precautions, and drug interactions, while parsing the table content in the instruction manual and mapping it into structured fields.

[0009] As a further improvement of the present invention, the formula for calculating the semantic risk entropy (SRE) of the semantic adaptive hierarchical matching module is as follows: Where pi is the proportion of the i-th type of risk semantic in the current field text, n is the total number of risk semantic categories, and Wr is the inherent risk weight of the field; The semantic adaptive hierarchical matching module matches multi-level collaborative schemes according to the range of semantic risk entropy (SRE). Each scheme is simultaneously bound to the corresponding data governance strength and data openness level, realizing the linkage and adaptation of governance and openness.

[0010] As a further improvement of the present invention, the data quality closed-loop governance module dynamically adjusts the quality verification dimension, review process and quality qualification threshold according to the governance intensity level output by the semantic adaptive hierarchical matching module; and automatically returns the instruction manual data below the corresponding threshold to the collection layer and marks the problematic fields.

[0011] As a further improvement of the present invention, the multi-label classification system includes three levels of labels: the first-level label is a semantic risk level label, which is divided into three categories: public level, industry level, and internal level, and is updated synchronously with the semantic risk entropy calculation results; the second-level label is a business scenario label, which is divided into three categories: regulatory compliance, clinical drug use, and public science popularization; the third-level label is a drug attribute label, which includes drug category, indication classification, and dosage form classification; the first-level labels are mapped one-to-one with the hierarchical interface permissions of the compliance data open layer.

[0012] As a further improvement of the present invention, the public terminal of the tiered scenario service layer is configured with a multi-condition combination search unit and a personalized medication reminder unit, which supports searching by drug name, ingredients, and indications, and outputs medication safety tips based on the drug combinations queried by the user; the pharmaceutical company terminal of the tiered scenario service layer is configured with a version statistical analysis unit, which is used to output statistical data related to instruction manual iteration and compliance; the medical institution terminal of the tiered scenario service layer is configured with a clinical rapid query unit, which supports matching the corresponding instruction manual by drug code and doctor's order name.

[0013] As a further improvement of the present invention, the hierarchical tagging data interface unit divides the interface permissions into three levels according to the first-level risk tag: The public-level interface is open to registered entities, providing basic scientific information about drugs; After qualification approval, industry-level interfaces will be opened to medical institutions and pharmaceutical R&D entities to provide fields related to clinical drug use; The internal-level interface is only open to the drug marketing authorization holder and provides full version traceability and internal management data. When a change in the specification version causes a change in the semantic risk entropy of a field, the system automatically adjusts the interface openness level of the corresponding data.

[0014] As a further improvement of the present invention, the permission approval unit is configured with an approval process that combines automatic qualification verification and manual review to conduct qualification review and permission classification for interface call applications from external entities; the call end-to-end audit unit records the entity information, time, interface, data range and frequency of each call, generates a compliance audit log, triggers permission freeze warning for abnormal call behavior, and simultaneously inputs the compliance feedback data back into the semantic adaptive hierarchical matching module to optimize risk calculation parameters.

[0015] Compared with the prior art, the advantages of this invention are: (1) This invention quantifies the security risk level of each field in the specification by semantic risk entropy, dynamically matches the corresponding governance intensity and verification standard, enables strict verification and manual review for high-risk content to ensure data accuracy, and enables lightweight verification for low-risk content to improve processing efficiency. Under the premise of ensuring data quality, it greatly improves the overall governance efficiency and reduces governance costs.

[0016] (2) The present invention updates the data openness level in real time based on semantic risk entropy. When the risk increases due to changes in the specification version, the openness permission is automatically converged, and when the risk decreases, the openness scope is automatically expanded. This not only avoids the compliance risk of sensitive data leakage, but also maximizes the release of the industrial value of publicly available data, and achieves a balance between data security and openness value.

[0017] (3) The present invention uses the semantic risk calculation results as a unified basis for governance classification and open classification, avoiding the duplication of calibration work in the two stages and improving the system operation efficiency; the call and compliance feedback data accumulated in the open stage can be used to iteratively optimize the risk calculation model, further improve the accuracy of governance and the rationality of openness, form a positive cycle, and improve the management efficiency of the system throughout its entire life cycle. Attached Figure Description

[0018] Figure 1 This is a diagram of the overall system architecture of the present invention; Figure 2 This is an internal flowchart of the semantic adaptive hierarchical matching module of the present invention; Figure 3This is a flowchart illustrating the data quality closed-loop governance and hierarchical openness linkage of the present invention. Figure 4 This is a flowchart illustrating the overall system workflow of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, the system in this embodiment is developed based on a B / S microservice architecture. The backend uses the Spring Boot framework, the database uses MySQL to store structured business data, Elasticsearch to implement full-text search, and MinIO object storage service to store original instruction manual files. The frontend uses the Vue.js framework to develop the interactive interfaces for each terminal, and the whole system complies with relevant industry standards for drug regulation and medical data security.

[0021] I. Standardized Data Acquisition Layer This layer serves as the system's data entry point, providing drug marketing authorization holders with instruction manual entry and version management services, and is deployed within the system's business access cluster.

[0022] Template-based data entry engine Template Configuration: In accordance with the "Regulations on the Management of Drug Instructions and Labels", pre-set instruction filling templates for three major categories and 12 subcategories of chemical drugs, traditional Chinese medicines and biological products are provided. Each template corresponds to a legal field structure, including 32 fixed fields such as drug name, ingredients, properties, indications, specifications, usage and dosage, adverse reactions, contraindications and precautions, of which 18 are mandatory fields.

[0023] Verification rule set: Built-in compliance verification rules, divided into four categories: ① Completeness verification: Instantly triggers prompts when required fields are empty; ② Terminology standardization verification: Connects to the National Pharmacopoeia Commission's pharmaceutical terminology standard library to verify the standardization of terminology expressions in fields such as indications and adverse reactions; ③ Format consistency verification: Verifies the uniformity of dosage format, numerical units, and punctuation marks in usage and dosage; ④ Content compliance verification: Identifies expressions that do not meet regulatory requirements.

[0024] Interaction logic: During the user's data entry process, the system triggers real-time validation, marks fields that do not meet the rules with prompts, displays the specific reasons for the errors and correction instructions, and the data can only be submitted for review after all validation items pass.

[0025] Version end-to-end traceability module Version identification mechanism: Each submitted version of the manual generates a unique version number, and at the same time, a hash value is generated for the entire content of the version for storage, ensuring that the version content cannot be tampered with.

[0026] Change Snapshot and Difference Comparison: The system automatically compares all field contents of the current version with the previous effective version, generates a field-level change comparison table, marks the added, modified, and deleted field contents and specific change values, supports filtering and viewing differences by field, and synchronously inputs the difference text into the semantic adaptive hierarchical matching module for risk calculation.

[0027] Traceability Chain and Audit Export: The traceability chain covers the entire process of the instruction manual from initial draft creation, internal multi-level review, drug regulatory filing, official effectiveness to subsequent revisions. Each node records the operator, operation time, operation type, and review comments; it supports multi-dimensional retrieval and can export traceability reports that meet compliance audit requirements.

[0028] II. Intelligent Data Governance Layer This layer is the core processing layer of the system, responsible for converting unstructured instructions into high-quality structured standard data and deploying it on the system's computing power processing cluster.

[0029] Multimodal structured parsing engine Input Support: Supports uploading instruction manual files in four formats: copyable PDF, scanned PDF, JPG / PNG images, and Word documents.

[0030] Optical Character Recognition Unit: Employs an OCR model finely tuned from a pharmaceutical industry corpus to recognize text in scanned documents and image-format instruction manuals, and supports the reconstruction of the row and column structures of complex tables.

[0031] Pharmaceutical Semantic Parsing Unit: Based on a pre-trained language model, fine-tuning and training are performed on publicly available drug instruction manual corpora to achieve entity extraction and semantic classification of core fields; semantic segmentation is performed on hierarchical headings and paragraph content of the instruction manual, automatically mapping them to the corresponding structured fields, while extracting information such as dosage and contraindications from the tables to fill the corresponding fields.

[0032] Confidence verification: After parsing, the system outputs structured data in standard JSON format. Each field is accompanied by an identification confidence level. Fields with confidence levels below a preset threshold are automatically marked as requiring manual review.

[0033] Semantic adaptive hierarchical matching module like Figure 2 As shown, this module is the core of the system linkage. By calculating semantic risk entropy, it synchronously matches governance intensity and openness level to achieve unified linkage between the two links.

[0034] (1) Semantic risk entropy calculation The module calculates the semantic risk entropy for each structured field individually, using the following formula: p i : The proportion of the i-th type of risk semantics in the current field text. Risk semantics are divided into three categories: severe risk, medium risk, and low risk (n=3). The identification and statistics are obtained through a pre-trained medical risk semantic classification model. For example, "fatal" and "shock" belong to severe risk semantics, "dizziness" and "nausea" belong to medium risk semantics, and "light-proof preservation" belongs to low risk semantics.

[0035] W r : Inherent risk weights for fields, preset according to the legal safety level of the field, including contraindications and adverse reactions fields W. r =1.0, Precautions, Drug Interaction Field W r =0.7, Usage and Quantity Field W r =0.6, Drug Name, Packaging Specifications Field W r =0.2, the remaining fields W r =0.1.

[0036] The higher the SRE value of the calculated result, the higher the semantic security risk of the field.

[0037] (2) Mapping of hierarchical collaborative schemes The module has 5 preset collaborative schemes, corresponding to the SRE value range. Each scheme is simultaneously bound to governance strength and openness level, enabling the two to be linked in a mandatory manner. Level 1 (Very Low Risk): Governance intensity is only basic integrity verification, which is automatically passed; Openness level is public, accessible to all entities.

[0038] Level 2 (Low Risk): Governance intensity is based on rule verification and automatic algorithm review; the openness level is public, accessible to all entities.

[0039] Level 3 (Medium Risk): Governance intensity includes full-dimensional rule verification and automatic algorithm review; Openness level is industry-level, accessible to qualified entities.

[0040] Level 4 (High Risk): Governance intensity includes full-dimensional rule verification and basic manual review; Openness level is industry-level, accessible to entities with advanced qualifications.

[0041] Level 5 (Extremely High Risk): Governance intensity includes full-dimensional rule verification and dual expert review; access level is internal, accessible only to the holder.

[0042] When the manual version is updated or the field content changes, the module automatically recalculates the SRE value and adjusts the governance intensity and openness level accordingly.

[0043] Data quality closed-loop governance module like Figure 3 As shown, this module dynamically adjusts the quality verification rules and qualification thresholds based on the governance intensity level output by the semantic adaptive hierarchical matching module.

[0044] The quality scoring uses a multi-dimensional weighted scoring model, and the comprehensive quality score is calculated using the following formula: Q = W1×S1 + W2×S2 + W3×S3 + W4×S4 S1 is the field completeness score, S2 is the content accuracy score, S3 is the format standardization score, and S4 is the tag matching score. W1, W2, W3, and W4 are the weight coefficients of the corresponding dimensions. The weights are dynamically adjusted according to the governance intensity level. The weight of content accuracy increases accordingly under the high risk level, and the weight of completeness increases accordingly under the low risk level.

[0045] Different treatment intensities correspond to different acceptable thresholds: Level 1-2 governance intensity: The passing threshold is 75 points. After passing the verification, it will be automatically entered into the standard dataset. Level 3 governance intensity: The passing threshold is 85 points. After passing the verification, it will automatically be included in the standard dataset. Level 4-5 governance intensity: The qualified threshold is 95 points. After the algorithm verification is passed, it will be transferred to the manual review station. Only after the review is passed can it be entered into the standard dataset.

[0046] Data in the instruction manual that scores below the corresponding threshold is automatically returned to the data collection layer, along with a list of problematic fields and correction suggestions. After correction, it re-enters the governance process, forming a dynamic quality closed loop.

[0047] Multi-label classification system The system automatically generates a three-level tagging system for each qualified instruction manual data entry, and the tags are stored in conjunction with the structured data. Level 1 tags (semantic risk level tags): output by the semantic adaptive hierarchical matching module, divided into public level, industry level, and internal level, and dynamically updated with version changes; Secondary tags (business scenario tags): divided into three categories: regulatory compliance, clinical drug use, and public science education, to match the data needs of different business scenarios; Level 3 tags (drug attribute tags): including drug category, indication classification, and dosage form classification, used for data classification retrieval and precise push.

[0048] III. Tiered Scenario Service Layer This layer provides differentiated services to three types of entities and is deployed in the system application service cluster.

[0049] Public service Multi-condition combined search: Supports multi-dimensional fuzzy search by generic name, brand name, ingredients, and indications of drugs. Search results are sorted by OTC drugs and prescription drugs. By default, three types of safety information are displayed first: contraindications, adverse reactions, and precautions. Non-core content is collapsed, and users can click to expand and view the full instruction manual.

[0050] Personalized medication reminders: When users search for multiple medications at the same time, the system automatically matches them based on the drug interaction field and alerts users to the risks of combined use; for special populations, it automatically highlights the contraindications of the medications and generates safety reminders.

[0051] Access channels: Supports two public access points: mini-programs and H5 web pages. Basic search functions can be used without registration.

[0052] Pharmaceutical company services Full version management: Enterprise administrator accounts can manage all drug instruction manual versions for the enterprise, view the complete version traceability chain and changes in risk levels, initiate version revision requests, and configure internal review processes.

[0053] Statistical Analysis Dashboard: Provides multi-dimensional data statistics, including the frequency of manual version iterations, statistics of frequently changed fields, compliance verification pass rate, risk level distribution, data quality score trends, etc., to support enterprise compliance management and R&D optimization.

[0054] Internal access control: Enterprises can configure three types of roles: data entry clerk, reviewer, and administrator, and distinguish the operation and viewing permissions of different positions.

[0055] Medical institution services Quick Clinical Search: Supports quick matching of corresponding instructions by national drug code, prescription name, and generic drug name.

[0056] Clinical support functions: It provides functions for querying drug dosages for special populations and batch querying drug interactions. It can be connected to the hospital HIS system and retrieve the corresponding instruction manual content with one click on the prescription interface to assist clinical medication decision-making.

[0057] Access method: After the medical institution submits its practice qualification for review and approval, it can access the system through its web page account or the hospital's internal system interface.

[0058] IV. Compliant Data Open Layer This layer is responsible for providing standardized data services to external parties. It is deployed on the system's open gateway cluster and complies with relevant network security level protection requirements.

[0059] Hierarchical and Tag-Based Data Interface System The interface adopts the RESTful standard design, and is divided into three levels based on the first-level semantic risk tags. Permissions are mapped one-to-one with tags, and are dynamically adjusted according to the tags. Public-level interface: The data fields corresponding to the public-level tags are open and can be accessed by all registered entities. It is suitable for science popularization applications and public query scenarios.

[0060] Industry-level interface: Data fields with open and public levels plus industry-level tags are only open to qualified entities such as medical institutions, pharmaceutical R&D companies, and research institutions.

[0061] Internal-level interface: Provides access to all fields and version traceability data, but is only accessible to the marketing authorization holder of the corresponding drug.

[0062] When the version of the instruction manual is updated, resulting in changes to semantic risks or adjustments to tag levels, the system automatically adjusts the corresponding entity's interface call permissions and simultaneously pushes a notification of the permission change.

[0063] Authorization Approval Unit The approval process employs a two-tiered mechanism: automatic verification and manual review. ①Automatic qualification verification: The system connects to the official qualification query platform to automatically verify the validity of the applicant's qualification documents and match the initial interface level; ② Manual review: The platform administrator reviews the applicant's usage scenarios and data requirements. After confirming compliance, the corresponding permissions are granted, and a unique authentication key is assigned to the applicant. The interface call uses a signature authentication mechanism to ensure security.

[0064] Call the end-to-end audit unit Full logging: All API calls are recorded with complete audit logs, including the caller ID, call time, IP address, API called, request parameters, return data range, and call status. Logs are stored periodically according to regulatory requirements.

[0065] Abnormal behavior auditing: The system automatically audits call logs to identify behaviors such as excessive calls, unauthorized data requests, and abnormal batch calls. After triggering an alert, it automatically freezes the call permissions of the corresponding account and notifies the administrator for review.

[0066] Reverse optimization feedback: The system regularly collects statistics on the call scenarios, compliance events, and data request feedback of each interface, and inputs the statistical results into the semantic adaptive hierarchical matching module to iteratively optimize the risk semantic classification weights and hierarchical threshold parameters, thereby continuously improving the accuracy of risk identification and the rationality of risk classification.

[0067] V. Overall System Workflow like Figure 4 As shown, during the data collection and input phase: pharmaceutical company users enter instruction manual information through a template-based data entry engine or upload the original instruction manual file. The system performs compliance verification in real time. After the verification is passed, the data is submitted, and the version full-chain traceability module synchronously records the version nodes and field changes.

[0068] Analysis and Risk Quantification Stage: The multimodal structured analysis engine extracts the structure of the submitted specification document, and the semantic adaptive hierarchical matching module calculates the semantic risk entropy of each field, matches the corresponding governance strength and openness level, and generates a tag system.

[0069] Tiered governance phase: The data quality closed-loop governance module executes the corresponding dimension of quality verification and review process according to the matched governance intensity level; qualified data enters the standard dataset, and unqualified data is returned to the collection layer for correction.

[0070] Scenario Service Phase: The tiered scenario service layer provides differentiated query, management, and decision support services to three types of terminals based on labeled standard datasets.

[0071] Compliance, Openness and Feedback Phase: External entities submit interface applications, obtain corresponding initial permissions after approval, and call the specification data through standardized APIs. All call behaviors are fully audited and recorded. Call feedback data is used to optimize the semantic risk model, forming a closed loop for full lifecycle management.

[0072] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0073] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A digital management system for the entire lifecycle of drug instructions, characterized in that, It includes a standardized data acquisition layer, an intelligent data governance layer, a hierarchical scenario service layer, and a compliant data opening layer, which are connected sequentially. The intelligent data governance layer has a built-in semantic adaptive hierarchical matching module, which is connected to the data quality closed-loop governance module and the hierarchical labeled data interface unit of the compliant data opening layer. The standardized data acquisition layer is configured with a templated data entry engine and a version full-link traceability module for pharmaceutical companies' terminals. The templated data entry engine has a built-in set of drug regulatory compliance verification rules for templated input of instruction manual information, real-time compliance verification, and error prompts. The version full-link traceability module is used to record the changes, operating entities, and time nodes of each version of the instruction manual, and generate a field-level full-link version traceability chain. The intelligent data governance layer is configured with a multimodal structured parsing engine, a data quality closed-loop governance module, and a semantic adaptive hierarchical matching module. It is used to automatically extract core fields from the unstructured instruction manual files output by the acquisition layer, and complete data cleaning and quality scoring. The semantic adaptive hierarchical matching module is used to calculate the semantic risk entropy value of each field of the instruction manual, and synchronously match the corresponding data governance strength and data openness level according to the entropy value to generate a multi-label classification system and output a structured, risk-labeled standard dataset of instruction manuals. The tiered scenario service layer connects to three types of terminals: public terminals, pharmaceutical company terminals, and medical institution terminals. It is used to provide differentiated instruction manual retrieval, management, and decision support services to different entities based on the standard dataset. The compliant data opening layer is configured with a hierarchical and tagged data interface unit, a permission approval unit, and a call end-to-end audit unit. The hierarchical and tagged data interface unit is connected to the risk tag data of the semantic adaptive hierarchical matching module and is used to adjust the opening permissions according to the semantic risk level. It outputs the corresponding hierarchical instruction manual data to external entities through a standardized API. The call end-to-end audit unit is used to log and audit all data call behaviors.

2. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The template-based data entry engine has a built-in set of regulatory compliance verification rules, which includes entries for drug instruction manual writing standards issued by drug regulatory authorities. The verification dimensions include the completeness of required fields, the standardization of terminology, the consistency of format, and the compliance of content. The template-based data entry engine triggers real-time verification during user input, marks content that does not conform to the rules, and simultaneously outputs compliance correction guidelines.

3. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The version full-link traceability module generates a unique version identifier and change snapshot for each version iteration, automatically compares the field-level differences between adjacent versions, and generates a change comparison file. The traceability chain contains full node information from the initial draft, review, effective date to revision, supports backtracking retrieval by time, operator, and changed field dimensions, and can export traceability reports that meet compliance audit requirements.

4. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The multimodal structured parsing engine integrates an optical character recognition unit and a semantic parsing unit in the pharmaceutical field, supporting input of instruction manual documents in PDF, scanned images, and Word formats. The semantic parsing unit is built based on a pre-trained language model in the pharmaceutical field, which can automatically identify and extract seven core fields: drug name, indications, contraindications, usage and dosage, adverse reactions, precautions, and drug interactions. At the same time, it parses the table content in the instruction manual and maps it into structured fields.

5. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The formula for calculating the semantic risk entropy (SRE) of the semantic adaptive hierarchical matching module is as follows: Where, p i W represents the proportion of the i-th type of risk semantic in the current field text, where n is the total number of risk semantic categories. r This represents the inherent risk weight of the field. The semantic adaptive hierarchical matching module matches multi-level collaborative schemes according to the range of semantic risk entropy (SRE). Each scheme is simultaneously bound to the corresponding data governance strength and data openness level, realizing the linkage and adaptation of governance and openness.

6. The digital management system for the entire lifecycle of drug instructions according to claim 5, characterized in that, The data quality closed-loop governance module dynamically adjusts the quality verification dimensions, review process, and quality qualification threshold based on the governance intensity level output by the semantic adaptive hierarchical matching module; and automatically returns instruction manual data below the corresponding threshold to the collection layer, marking problematic fields.

7. The digital management system for the entire lifecycle of drug instructions according to claim 5, characterized in that, The multi-label classification system includes three levels of labels: Level 1 labels are semantic risk level labels, which are divided into three categories: public, industry, and internal, and are updated synchronously with the semantic risk entropy calculation results; Level 2 labels are business scenario labels, which are divided into three categories: regulatory compliance, clinical drug use, and public science popularization; Level 3 labels are drug attribute labels, which include drug category, indication classification, and dosage form classification; The Level 1 labels are mapped one-to-one with the hierarchical interface permissions of the compliance data open layer.

8. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The public client of the tiered scenario service layer is configured with a multi-condition combination search unit and a personalized medication reminder unit, supporting searches by drug name, ingredients, and indications, and outputting medication safety tips based on the drug combinations queried by the user; the pharmaceutical company client of the tiered scenario service layer is configured with a version statistical analysis unit, used to output statistical data related to instruction manual iteration and compliance; the medical institution client of the tiered scenario service layer is configured with a clinical rapid query unit, supporting matching the corresponding instruction manual by drug code and doctor's order name.

9. The digital management system for the entire lifecycle of drug instructions according to claim 7, characterized in that, The hierarchical and tagged data interface unit divides interface permissions into three levels according to the first-level risk tag: The public-level interface is open to registered entities, providing basic scientific information about drugs; After qualification approval, industry-level interfaces will be opened to medical institutions and pharmaceutical R&D entities to provide fields related to clinical drug use; The internal-level interface is only open to the drug marketing authorization holder and provides full version traceability and internal management data. When a change in the specification version causes a change in the semantic risk entropy of a field, the system automatically adjusts the interface openness level of the corresponding data.

10. The digital management system for the entire lifecycle of drug instructions according to claim 1, characterized in that, The permission approval unit is configured with an approval process that combines automatic qualification verification and manual review to conduct qualification review and permission classification for interface call applications from external entities; the call end-to-end audit unit records the entity information, time, interface, data range and frequency of each call, generates a compliance audit log, triggers permission freeze warnings for abnormal call behavior, and simultaneously inputs the compliance feedback data back into the semantic adaptive hierarchical matching module to optimize risk calculation parameters.

Images