An artificial intelligence-based drug safety information system

Abstract

The present application relates to the technical field of drug information analysis, and more particularly to a drug alert information system based on artificial intelligence. The system comprises a medicine variety acquisition module, a drug safety data acquisition module, an existing adverse reaction statistical module and a safety signal evaluation module, can obtain drug alert service demand and target drug variety information; interfaces with a medical big data source, collects the whole life cycle safety data corresponding to the target drug variety and carries out existing adverse reaction statistical analysis, generates the existing adverse reaction statistical analysis result of the variety; based on the existing adverse reaction statistical analysis result of the variety, the safety signal evaluation is carried out, the safety signal evaluation report is generated, and the existing adverse reaction statistical analysis result of the variety is integrated, combined with literature retrieval evaluation data and drug alert system files, product safety related data review is generated, and drug alert commission work and system construction are supported simultaneously. The present application can significantly improve the drug alert system and compliance management level.

Description

Technical Field

[0001] This invention relates to the field of drug information analysis technology, and in particular to a drug vigilance information system based on artificial intelligence. Background Technology

[0002] Currently, most pharmacovigilance services employ a manual retrieval, decentralized statistical, and experience-based evaluation model. This involves manually collecting drug safety data from limited data sources, manually compiling existing adverse reactions and writing evaluation reports, and then integrating these data to form a product safety overview. However, existing methods are hampered by a lack of artificial intelligence technology support and a limited data collection scope, making it impossible to complete the statistical analysis of existing adverse reactions and the evaluation of safety signals. Furthermore, the integration of data lacks systematicity and standardization, failing to simultaneously support pharmacovigilance outsourcing and system construction, thus reducing the efficiency of pharmacovigilance services. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide an artificial intelligence-based pharmacovigilance information system to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an artificial intelligence-based pharmacovigilance information system includes the following modules: The pharmaceutical product acquisition module is used to obtain the pharmacovigilance service requirements and target drug product information of pharmaceutical companies. The drug safety data acquisition module is used to connect with medical big data sources based on artificial intelligence technology to collect full life-cycle safety data for target drug varieties. The existing adverse reaction statistics module is used to perform statistical analysis of existing adverse reactions on the full life cycle safety data of the target drug and generate statistical analysis results of existing adverse reactions of the drug. The safety signal evaluation module is used to evaluate safety signals based on the statistical analysis results of existing adverse reactions of a product and generate a safety signal evaluation report. It integrates the statistical analysis results of existing adverse reactions of a product with the safety signal evaluation report, combines literature search evaluation data and pharmacovigilance system documents, and generates a summary of relevant information on product safety, while simultaneously supporting pharmacovigilance outsourcing and system construction.

[0005] The beneficial effects of this invention are: The artificial intelligence-based pharmacovigilance information system proposed in this invention consists of a pharmaceutical product data collection module, a drug safety data collection module, an existing adverse reaction statistics module, and a safety signal evaluation module. Compared with existing technologies, the advantages of this application lie in the fact that by clearly defining the collection targets, analysis scope, and service objectives at the initial stage of service initiation, subsequent data collection, statistical analysis, signal evaluation, and other related work can be carried out around the identified pharmaceutical products, avoiding invalid data collection, duplication of work, and waste of resources due to unclear objectives, while making the overall service process more targeted. By locking in service needs and product scope in advance, a clear direction can be provided for subsequent data source docking, analysis dimension design, and evaluation logic construction, ensuring that all work revolves around the core demands of pharmaceutical companies for authentic regulatory reporting, risk control, and system improvement, effectively improving the alignment between service content and enterprise needs. Secondly, by relying on artificial intelligence technology to achieve rapid docking of multi-source medical big data, it can simultaneously cover multiple important data sources such as clinical research, post-marketing surveillance, medical institution reporting, adverse reaction databases, and literature, breaking through the limitations of traditional manual methods that can only dock with a limited number of data sources and have narrow information coverage, and realizing the complete collection of data from the entire process of target drug development, clinical trials, market use, to long-term monitoring. AI-driven data acquisition methods possess automatic identification, batch acquisition, and continuous updating capabilities, enabling the rapid aggregation of massive amounts of multi-source data. This significantly reduces the workload of manual database-by-database retrieval and record-by-record, while also minimizing human errors such as missed or incorrect data collection, effectively enhancing the data support capabilities and overall operational efficiency of pharmacovigilance services. Then, through structured, multi-dimensional statistical analysis of the target drug's full lifecycle safety data, key aspects such as adverse reaction types, affected systems, incidence rates, population distribution, severity, dose-relatedness, and time distribution can be systematically identified. This comprehensively presents the overall characteristics of known adverse reactions, clearly outlining the existing risks associated with the drug, avoiding the limitations of traditional manual statistics which only achieve simple counting and cannot perform multi-dimensional cross-analysis. Automated and systematic statistical analysis methods can rapidly process massive amounts of data, completing complex classification, summarization, and comparison in a short time, significantly shortening the statistical cycle, reducing the intensity of manual calculations and processing, and improving the objectivity and consistency of the analysis results. Finally, by conducting scientific safety signal evaluation based on existing adverse reaction statistics, potential risks can be identified, the importance and correlation of signals can be determined, and professional and standardized evaluation reports can be generated, avoiding judgment biases caused by traditional experience-based evaluations. By integrating statistical results, signal evaluation, literature review and evaluation, and system documents to form a comprehensive overview of product safety-related data, the content structure becomes more complete, the logic clearer, and the format more in line with regulatory and industry standards, thus overcoming the predicament of traditional data being scattered, disorganized, and lacking in compliance.This review can be directly used for external tasks such as pharmacovigilance commissioned projects and regulatory reporting, and can also serve as an internal basis for enterprises to operate their pharmacovigilance systems, manage risks, and optimize processes, significantly improving service efficiency and comprehensive support capabilities, and helping pharmaceutical companies improve their pharmacovigilance systems and compliance management. Attached Figure Description

[0006] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the modules of the artificial intelligence-based pharmacovigilance information system of the present invention; Figure 2 for Figure 1 A functional flowchart of the Chinese medicine safety data acquisition module. Detailed Implementation

[0007] The technical system of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0008] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0009] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0010] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides an artificial intelligence-based pharmacovigilance information system, which includes the following modules: The pharmaceutical product acquisition module is used to obtain the pharmacovigilance service requirements and target drug product information of pharmaceutical companies. In this embodiment of the invention, relying on the system's built-in demand collection unit, a fixed collection method is used to connect with the relevant interfaces of pharmaceutical companies' pharmacovigilance services. This allows for the collection of the companies' corresponding pharmacovigilance service requirements, including statistical analysis of existing adverse reactions, safety signal evaluation, writing of product safety-related data summaries, construction of pharmacovigilance systems, and support for outsourced work. These requirements are categorized and recorded according to type. Simultaneously, target drug information is collected using a fixed extraction method to extract core information such as drug name, dosage form, specifications, indications, market launch date, manufacturing process, and previous pharmacovigilance records. This information is then organized and archived by drug type, eliminating missing or logically contradictory content to ensure the completeness and traceability of the collected demand and drug information. Service demands are linked and bound to target drug information, with each demand corresponding to a specific target drug. This provides a foundation for subsequent data collection, statistical analysis, and service implementation, supporting the orderly conduct of outsourced pharmacovigilance work and ensuring the system aligns with the actual service needs of enterprises.

[0011] The drug safety data acquisition module is used to connect with medical big data sources based on artificial intelligence technology to collect full life-cycle safety data for target drug varieties. In this embodiment of the invention, relying on an AI-based pharmacovigilance information system, the system's built-in AI data docking engine is invoked. This engine, after pharmacovigilance data docking training and calibration, can achieve stable docking with various medical big data sources. The docking process employs encrypted transmission to ensure data transmission security and comply with pharmacovigilance data confidentiality requirements. The medical big data sources include clinical trial databases, post-marketing surveillance databases, medical institution reporting systems, and pharmaceutical literature databases. Based on the collected target drug information, targeted collection requests are sent to various data sources, including core information about the target drug and the scope of data collection. Data related to the target drug returned from various data sources is received through encrypted communication links. This data covers adverse reaction records from clinical trials, post-marketing individual adverse reaction reports, clinical drug use monitoring data, and related literature data. The data is categorized and organized by data type, invalid and abnormal data is removed, and full lifecycle safety data corresponding to the target drug is generated, providing complete data support for subsequent statistical analysis of existing adverse reactions.

[0012] The existing adverse reaction statistics module is used to perform statistical analysis of existing adverse reactions on the full life cycle safety data of the target drug and generate statistical analysis results of existing adverse reactions of the drug. In this embodiment of the invention, the full life-cycle safety data of the target drug is imported into an AI-based adverse reaction statistical model. The model has a built-in feature extraction and statistical analysis unit, employing a fixed extraction method to extract relevant features of adverse reactions, including adverse reaction type, frequency, severity, affected population, and related medication information. Adverse reaction data is categorized and statistically analyzed by type, calculating the proportion of each type, and exploring the correlation between adverse reaction occurrence and the user population, dosage, and duration of medication. Potential patterns in adverse reaction occurrence are analyzed, and a pattern analysis report is generated. The adverse reaction classification statistics, correlation analysis data, and pattern analysis report are integrated, organized in a fixed format, and supplemented with data validity explanations, statistical method explanations, and analytical conclusions. This clarifies the occurrence characteristics, related factors, and potential risks of existing adverse reactions of the target drug, generating statistical analysis results of existing adverse reactions for the drug. This provides core statistical basis for subsequent safety signal evaluation and the writing of product safety-related data summaries.

[0013] The safety signal evaluation module is used to evaluate safety signals based on the statistical analysis results of existing adverse reactions of a product and generate a safety signal evaluation report. It integrates the statistical analysis results of existing adverse reactions of a product with the safety signal evaluation report, combines literature search evaluation data and pharmacovigilance system documents, and generates a summary of relevant information on product safety, while simultaneously supporting pharmacovigilance outsourcing and system construction.

[0014] In this embodiment of the invention, based on the statistical analysis results of existing adverse reactions of a product, abnormal data is extracted as initial clues for safety signal monitoring. The proportion of abnormal data and the credibility of the clues are calculated, potential safety signals are identified and classified, and the built-in GVP safety evaluation standard is invoked to evaluate the risk level of potential safety signals, calculate risk scores, and integrate them to form a safety signal evaluation report. Pharmaceutical literature databases are searched to obtain literature related to the safety signals of the target drug. The literature is structured, quality-evaluated, and analyzed for its supportability to generate literature retrieval evaluation data. The system's built-in pharmacovigilance system file is invoked to extract relevant content on system requirements, calculate system fit, and generate pharmacovigilance system evaluation data. The statistical analysis results of existing adverse reactions of the product, the safety signal evaluation report, the literature retrieval evaluation data, and the pharmacovigilance system evaluation data are integrated, organized according to the fixed format of a product safety-related data summary, supplemented with risk prevention and control suggestions and system optimization directions, to generate a product safety-related data summary. This summary simultaneously supports the pharmacovigilance outsourcing work of pharmaceutical companies, provides data support and directional guidance for the construction of the company's pharmacovigilance system, ensures that the information system meets the company's daily pharmacovigilance work needs, and achieves compliant and efficient operation.

[0015] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1A functional flowchart of the drug safety data acquisition module is shown in this embodiment. The drug safety data acquisition module includes the following functions: S201: Invoke the artificial intelligence data docking engine to establish an encrypted communication link with the medical big data platform; In this embodiment of the invention, by employing artificial intelligence data docking engine construction technology and encrypted communication link establishment method, the data docking process of the AI-based pharmacovigilance information system completes the encrypted communication connection with the medical big data platform. The AI ​​data docking engine is activated; the engine has a built-in identity authentication unit and encryption verification unit. Through identity authentication, two-way identity verification with the medical big data platform is completed. After successful authentication, an end-to-end encrypted communication link is constructed using an encryption protocol. During the link transmission process, the transmitted content is encapsulated using encryption rules to ensure that the data interaction process complies with pharmacovigilance data security requirements, providing a stable and secure transmission foundation for subsequent drug safety data collection, case report processing, and compliant system operation.

[0016] S202: Send a data collection request to the medical big data platform through an encrypted communication link. The data collection request includes information on the target drug variety and the data collection scope. In this embodiment of the invention, a data collection request is generated by the request construction unit built into the artificial intelligence data docking engine. The request content includes target drug information and data collection scope. The drug information includes drug name, dosage form and specifications. The data collection scope covers data related to clinical research, post-marketing monitoring and individual adverse reaction reports. After the request is constructed, it is sent to the medical big data platform through the established encrypted communication link. The request transmission process retains a transmission log for subsequent data traceability and analysis of the holder's pharmacovigilance status.

[0017] S203: Receive initial safety data returned by the medical big data platform, the initial safety data including clinical research data, post-marketing surveillance data, and individual adverse reaction report data; In this embodiment of the invention, initial security data fed back from a medical big data platform is received through an encrypted communication link. The data includes clinical research data, post-marketing monitoring data, and individual adverse reaction report data. The data integrity is verified during the receiving process. After the verification is passed, the data is initially classified and stored according to the data type. The clinical research data is used for drug safety signal evaluation and analysis, while the post-marketing monitoring data and individual adverse reaction report data are used for statistical analysis of existing adverse reaction data of the product, providing original data support for subsequent data screening and writing of product safety-related information reviews.

[0018] S204: The initial security data is screened through an artificial intelligence data docking engine, the percentage of valid data is calculated, invalid and abnormal data are removed, and valid security data is obtained. In this embodiment of the invention, the data filtering unit built into the artificial intelligence data docking engine performs item-by-item verification on the initial safety data. The verification includes field completeness, logical rationality, and content standardization. The percentage of valid data is calculated by dividing the number of valid data entries by the total number of data entries. Data with missing data, logical conflicts, or duplicate content is judged as invalid data and abnormal data and is removed. Valid safety data that meets the requirements for pharmacovigilance data is retained, providing a high-quality data foundation for subsequent structured transformation and standardization processing.

[0019] S205: Perform structured transformation on the effective safety data, combine the data validity ratio, and perform standardization to generate full life cycle safety data for the target drug product.

[0020] In this embodiment of the invention, by performing a structured transformation on valid safety data, unstructured text data is converted into structured data in a standardized format. Field naming, numerical units, and time formats are unified. The converted data is weighted according to the data validity ratio. Then, standardized processing is performed according to the pharmacovigilance data specification to finally generate full life cycle safety data corresponding to the target drug product. This data can be directly used for statistical analysis of existing adverse reaction data of the product, safety signal evaluation analysis, literature retrieval evaluation analysis, and writing of product safety-related information reviews, supporting the improvement of the pharmacovigilance system, professional personnel training, and the compliant and efficient operation of the information system.

[0021] Furthermore, the existing adverse reaction statistics module includes the following functions: The entire lifecycle safety data is input into an AI-based adverse reaction statistical model to extract adverse reaction-related features; In this embodiment of the invention, an AI-based adverse reaction statistical model is activated. This model incorporates a feature recognition unit and a data parsing unit, and after training and calibration with historical pharmacovigilance data, it can accurately capture core information related to adverse reactions. By inputting full lifecycle safety data into the model, it extracts adverse reaction-related features through multi-layered analytical operations. These features include the adverse reaction name, clinical manifestations, occurrence time, severity, outcome, and associated medication information. The extraction process strictly matches the adverse reaction classification standards in the pharmacovigilance system documents, ensuring that the feature extraction aligns with the needs of daily pharmacovigilance work. This provides core feature support for subsequent adverse reaction classification statistics, statistical analysis of existing adverse reaction data for various products, and the writing of product safety-related literature reviews, while also aiding in safety signal evaluation and analysis.

[0022] Furthermore, based on the characteristics of adverse reactions, the occurrence of different types of adverse reactions is classified and statistically analyzed, the proportion of each type of adverse reaction is calculated, and statistical data on adverse reaction classification is generated. In this embodiment of the invention, adverse reactions are categorized and organized according to their extracted characteristics, strictly following the operating procedures outlined in the pharmacovigilance system documents. Adverse reactions are classified into categories such as digestive system reactions, nervous system reactions, and skin and mucous membrane reactions. The frequency and number of cases involved in each category are statistically analyzed. The percentage of each adverse reaction is calculated by dividing the frequency of a single adverse reaction by the total number of adverse reactions, retaining a decimal place. The categorized statistical results and percentages are then organized and archived according to adverse reaction type to generate adverse reaction classification statistics. This data is directly used for statistical analysis of existing adverse reaction data for the product, providing foundational data for safety signal evaluation analysis and the writing of product safety-related literature reviews, supporting the compliant and efficient operation of the information system.

[0023] Furthermore, we explored the correlation between adverse reactions and the user population, dosage, and duration of medication, and generated correlation analysis data by combining the proportion of adverse reactions. In this embodiment of the invention, by relying on adverse reaction-related characteristics and statistical data on adverse reaction classification, an association mining algorithm is used to explore the correlation between adverse reaction occurrence and the user population, dosage, and duration of medication. The user population is divided into groups based on age, gender, and underlying diseases; the dosage is divided into intervals based on standard dosage and doubled dosage; and the duration of medication is divided into stages based on short-term, medium-term, and long-term. The occurrence of adverse reactions is matched one by one for each group, each dosage interval, and each duration stage. Combined with the proportion of each type of adverse reaction, the characteristics and proportion differences of adverse reaction occurrence under different populations, dosages, and durations are recorded. All correlation information is integrated to generate association analysis data, providing data support for subsequent identification of potential patterns, analysis of professional issues, and comprehensive analysis of pharmacovigilance status of pharmacists.

[0024] Furthermore, based on correlation analysis data, potential patterns in adverse reactions are identified, and a pattern analysis report is generated; In this embodiment of the invention, by using correlation analysis data and pattern recognition methods, potential patterns of adverse reactions are identified. The analysis focuses on the high-incidence types and trends of adverse reactions under specific patient populations, dosages, and durations of use. It clarifies the influence of different factors on the occurrence of adverse reactions, eliminates unrelated incidental phenomena, and extracts universal and traceable potential patterns. Following a reporting format, the identified potential patterns, analysis process, data support, and risk warnings are compiled into a pattern analysis report. The report content aligns with the daily needs of pharmacovigilance work, covering high-incidence characteristics of adverse reactions, influencing factors, and preliminary prevention and control directions. This provides a basis for statistical analysis of existing adverse reaction data, safety signal evaluation analysis, and customization of dedicated pharmacovigilance solutions.

[0025] Furthermore, by integrating adverse reaction classification statistics, correlation analysis data, and pattern analysis reports, statistical analysis results of existing adverse reactions for each product are generated.

[0026] In this embodiment of the invention, adverse reaction classification statistics, correlation analysis data, and pattern analysis reports are comprehensively integrated. Following the standardized requirements for statistical analysis of existing adverse reaction data for each product, the data logic is streamlined, the presentation format is unified, and supplementary explanations of data validity and analytical conclusions are added to clarify the characteristics, related factors, and potential patterns of adverse reactions. The integrated results include complete classification statistics, correlation analysis, and pattern summaries, which can be directly used for writing product safety-related data reviews, evaluating and analyzing safety signals, and analyzing professional issues. Simultaneously, it provides complete statistical data support for comprehensive analysis of pharmacovigilance status by pharmacovigilance holders, improvement of the pharmacovigilance system, and training of professional personnel, ensuring that the information system meets the needs of daily pharmacovigilance work.

[0027] Furthermore, the process of mining the correlation between adverse reaction occurrence and the user population, dosage, and duration of medication, and generating correlation analysis data based on the proportion of adverse reaction occurrence, includes: Extract user characteristics, dosage data, and duration of medication from the full life-cycle safety data; In this embodiment of the invention, relying on full life-cycle safety data, an extraction method is used to screen and extract user characteristics, dosage data, and duration data. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the extracted data meets the needs of daily pharmacovigilance work. User characteristics include age groups, gender, underlying disease types, and allergy history, and are grouped according to standards. Age groups are divided into under 18 years old, 18-45 years old, 46-60 years old, and over 60 years old. Underlying diseases are categorized into cardiovascular diseases, digestive system diseases, etc. Dosage data includes single dose, daily dose frequency, and cumulative dose, all converted to a unified unit. Duration data includes single dose duration and total treatment course, with specific durations recorded according to standards. All extracted data is organized by case number, providing basic data support for subsequent correlation mining, statistical analysis of existing adverse reaction data, and professional problem analysis.

[0028] Furthermore, by using artificial intelligence correlation mining algorithms, a model of the correspondence between adverse reaction occurrences and the characteristics of the drug-using population can be established; In this embodiment of the invention, an artificial intelligence-based correlation mining algorithm is activated. This algorithm, trained and calibrated using historical pharmacovigilance data, can accurately capture the intrinsic correlation between adverse reactions and the user population. Extracted user population characteristics are used as input variables, and adverse reaction occurrences are used as output variables. These are fed into the AI-based correlation mining algorithm, which uses multi-layered computation to mine the intrinsic correlation between the two, constructing a model of the correspondence between adverse reaction occurrences and user population characteristics. The model clearly defines the correspondence logic between different population characteristics and adverse reaction occurrences, labels the correlation tendencies between population characteristics and various adverse reactions, and calibrates the model during training using existing historical adverse reaction data for each product. This ensures the model closely matches actual medication safety scenarios, providing model support for subsequent population correlation analysis, safety signal evaluation analysis, and comprehensive analysis of pharmacovigilance status.

[0029] Furthermore, based on the correspondence model, the differences in adverse reactions corresponding to the characteristics of different drug users are analyzed, the correlation degree of population characteristics is calculated, and population correlation data is generated. In this embodiment of the invention, based on an established correspondence model, analytical methods are used to analyze the differences in adverse reaction occurrences corresponding to different drug-using population characteristics one by one. The frequency, type, and severity of adverse reactions are compared among different age groups, genders, and underlying disease groups to clarify the high incidence characteristics of adverse reactions in various population groups. The correlation degree of population characteristics is calculated as the ratio of the number of adverse reactions occurring in a specific population to the total number of drug users in that population, and then corrected by combining the proportion of adverse reactions occurring in that population. The higher the correlation degree, the stronger the association between the population characteristics and the occurrence of adverse reactions. Different population characteristics, corresponding adverse reaction differences, and population characteristic correlation degrees are organized and archived by population group to generate population correlation data, providing data support for subsequent correlation analysis data reconstruction, professional problem analysis, and customized pharmacovigilance solutions.

[0030] Furthermore, by combining medication dosage data and medication duration data, the corresponding relationship model is adjusted to analyze the impact of medication dosage and medication duration on the occurrence of adverse reactions, calculate the dose-duration influence coefficient, and generate dose-duration correlation data; In this embodiment of the invention, extracted medication dosage data and medication duration data are supplemented into the corresponding relationship model. An adjustment method is used to optimize model parameters, clarifying the correlation logic between medication dosage, medication duration, and adverse reaction occurrence. The adjustment process incorporates adverse reaction classification statistics to ensure the model accurately reflects the impact of dosage and duration on adverse reactions. A calculation method is used to analyze the impact of medication dosage and duration on adverse reactions, calculating the dose-duration influence coefficient. The dose influence coefficient is calculated separately for conventional and doubled doses, and the duration influence coefficient is calculated separately for short-term, medium-term, and long-term effects. Higher coefficient values ​​indicate a greater impact of the dosage or duration on adverse reactions. The influence coefficients corresponding to medication dosage and duration, along with differences in adverse reaction occurrence, are compiled and archived to generate dose-duration correlation data. This data supports subsequent correlation analysis, data reconstruction, and safety signal evaluation analysis, contributing to the compliant and efficient operation of the information system.

[0031] Furthermore, correlation analysis data is generated by reconstructing the correlation analysis data based on the proportion of adverse reactions and combining population correlation data and dose-duration correlation data.

[0032] In this embodiment of the invention, by relying on the incidence rate of adverse reactions, combined with population-related data and dose-duration-related data, the inherent logic among the three is analyzed. The population correlation, dose-duration influence coefficient, and incidence rate of adverse reactions are weighted and fused, with the fusion weight determined based on the data's contribution to the adverse reaction analysis. During the reconstruction process, supplementary data validity explanations are provided, logically contradictory data is eliminated, and the data presentation format is standardized. The correlation logic, specific data, and analytical conclusions between population correlation, dose-duration-related data, and incidence rate of adverse reactions are integrated to generate correlation analysis data. This data fully reflects the correlation characteristics between adverse reactions and the user population, dosage, and duration. It can be directly used for statistical analysis of existing adverse reaction data for specific products, writing reviews of product safety-related materials, and analyzing professional issues. Simultaneously, it provides data support for improving pharmacovigilance systems and training professional personnel.

[0033] Furthermore, the generated dose-duration correlation data includes: Extract dose gradient information from medication dosage data and duration segmentation information from medication duration data; In this embodiment of the invention, based on the extracted medication dosage data and medication duration data, dosage gradient information and duration segment information are screened and extracted one by one. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the extracted information meets the needs of daily pharmacovigilance work. Medication dosage data is divided into gradients by interval, with the gradient division criteria being conventional dose, 1.5 times the conventional dose, and 2 times the conventional dose, clearly defining the dosage range, unit, and number of cases covered for each gradient. Medication duration data is divided into segments by period, with the segment criteria being short-term (1-7 days), medium-term (8-30 days), and long-term (31 days and above), clearly defining the duration range and number of cases covered for each segment. The extracted dosage gradient information and duration segment information are organized according to gradient and segment categories, and the corresponding data range and number of cases are labeled. This provides basic information for subsequent model adjustments, adverse reaction probability calculations, and statistical analysis of existing adverse reaction data for various products, supporting the compliant and efficient operation of the information system.

[0034] Furthermore, the dose gradient information and time segmentation information are substituted into the correspondence model to dynamically adjust the model parameters, calculate the parameter adjustment range, and adjust the correspondence model based on the parameter adjustment range to generate the adjusted correspondence model. In this embodiment of the invention, the extracted dose gradient information and duration segmentation information are successively substituted into the established correspondence model. A dynamic adjustment method is used to optimize and adjust the built-in parameters of the model. The adjustment process combines adverse reaction classification statistics and existing adverse reaction historical data of the drug to ensure that the adjustment direction aligns with actual medication safety scenarios. The parameter adjustment range is calculated by dividing the difference between the adjusted parameter value and the original parameter value by the original parameter value. The range reflects the degree of parameter adjustment. Based on the calculated parameter adjustment range, the correspondence model is adjusted as a whole, and the parameters related to dose gradient and duration segmentation in the model are corrected one by one. This ensures that the model can accurately capture the correlation logic between dose, duration, and adverse reaction occurrence, generating an adjusted correspondence model. This provides accurate model support for subsequent adverse reaction probability calculation and safety signal evaluation analysis.

[0035] Furthermore, based on the adjusted correspondence model, the probability of adverse reactions corresponding to different dose gradients and time segments is calculated; In this embodiment of the invention, by activating the adjusted correspondence model, the model has completed parameter optimization and can accurately reflect the correlation between dose gradient, duration segment, and adverse reaction occurrence. Each dose gradient and duration segment is input into the adjusted correspondence model. The model, through multi-layered calculations, combines historical adverse reaction data and current medication data to calculate the probability of adverse reaction occurrence for each dose gradient and duration segment. The calculation process strictly follows the calculation rules: the number of adverse reaction cases within that gradient / segment is divided by the total number of medication cases within that gradient / segment, and then corrected based on the model parameter adjustment range to obtain the final probability of adverse reaction occurrence. Probability values ​​are recorded according to dose gradient and duration segment, generating probability statistics to provide data support for subsequent determination of correlation strength and analysis of professional issues.

[0036] Furthermore, based on the probability of adverse reactions, the correlation strength between drug dosage and duration of medication and the occurrence of adverse reactions is determined, and correlation strength data is generated; In this embodiment of the invention, based on the calculated probabilities of adverse reactions corresponding to different dose gradients and duration segments, a determination method is used to determine the correlation strength between drug dosage, duration of medication, and the occurrence of adverse reactions. The correlation strength is graded according to standards: a probability of occurrence of 0-20% indicates a weak correlation, 21%-50% indicates a moderate correlation, and 51%-100% indicates a strong correlation, clearly defining the correlation strength levels corresponding to different dose gradients and duration segments. The correlation strength value is calculated, equal to the probability of adverse reaction occurrence multiplied by a weight. The weight is determined based on the degree of influence of dosage and duration on the occurrence of adverse reactions; a higher correlation strength value indicates a stronger correlation between dosage or duration and the occurrence of adverse reactions. The dose gradients, duration segments, corresponding probabilities, and correlation strength values ​​are compiled and archived to generate correlation strength data, providing data support for subsequent comprehensive analysis of dose-duration correlation coupling and pharmacovigilance status.

[0037] Furthermore, dose-duration correlation data is generated by coupling dose-duration correlation based on correlation intensity data and combining dose gradient information with duration segmentation information.

[0038] In this embodiment of the invention, relying on correlation strength data and combining extracted dose gradient and duration segment information, a correlation coupling method is used to combine dose gradients and duration segments pairwise, analyzing the correlation strength after coupling different dose gradients with different duration segments. During the coupling process, the comprehensive correlation strength corresponding to each combination is calculated. The calculation method is the weighted sum of the dose gradient correlation strength value and the duration segment correlation strength value in that combination. The weighting coefficient is determined according to the degree of influence of both on the occurrence of adverse reactions. All dose-duration combinations, corresponding comprehensive correlation strengths, adverse reaction probabilities, and correlation levels are compiled and archived to generate dose-duration correlation data. This data fully reflects the impact of dose-duration coupling on the occurrence of adverse reactions and can be directly used for correlation analysis data reconstruction, statistical analysis of existing adverse reaction data for products, and writing reviews of product safety-related materials. At the same time, it supports the improvement of pharmacovigilance systems, professional personnel training, and the compliant and efficient operation of information systems.

[0039] Furthermore, the dose-time correlation coupling based on correlation intensity data and combined with dose gradient information and duration segmentation information includes: Dose influence features and duration influence features are extracted from the correlation strength data, and correlation coupling basic data are generated based on the dose influence features and duration influence features; In this embodiment of the invention, dose-influence features and duration-influence features are extracted one by one by relying on correlation strength data. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the feature extraction meets the needs of daily pharmacovigilance work. Dose-influence features include the correlation strength values, influence ranges, and correspondences with adverse reaction types for each dose gradient; duration-influence features include the correlation strength values, influence cycles, and correspondences with the severity of adverse reactions for each duration segment. The extracted dose-influence features and duration-influence features are organized by gradient and segment categories, and the corresponding correlation strength values ​​and influence details are labeled. Data with ambiguous features or no actual correlation are eliminated, and the data are integrated to form basic correlation coupling data. This provides basic support for subsequent gradient segment mapping, coupling transmission relationship construction, and statistical analysis of existing adverse reaction data for various products, and helps the information system operate in a compliant and efficient manner.

[0040] Furthermore, based on the associated coupling basic data, the correspondence between dose gradient information and duration segment information is determined, and gradient segment mapping data is generated; based on the gradient segment mapping data, the coupling transmission relationship between drug dosage and drug duration is constructed, and coupling transmission feature data is generated. In this embodiment of the invention, the correspondence between dose gradient information and duration segment information is analyzed one by one based on the associated coupling basic data. The adaptability of different dose gradients to different duration segments is clarified, and the association strength characteristics under each correspondence are labeled. The correspondences are then organized by gradient and segment groups to generate gradient segment mapping data. Based on the gradient segment mapping data, a fixed construction method is used to establish the coupling transmission relationship between medication dosage and medication duration. The influence of dose gradient changes on duration segment association strength and the influence of duration segment changes on dose gradient association strength are clarified. The transmission direction, transmission amplitude, and transmission rules are labeled to construct a complete coupling transmission structure. The coupling transmission relationship, transmission rules, and influencing parameters are organized and archived to generate coupling transmission feature data, providing model and data support for subsequent association strength calibration, professional problem analysis, and comprehensive analysis of pharmacovigilance status of pharmacists.

[0041] Furthermore, the correlation strength data is coupled and calibrated based on the coupling transfer characteristic data to generate calibrated correlation coupling data; In this embodiment of the invention, the correlation strength data is calibrated group by group by substituting the coupling transmission characteristic data into the correlation strength data. The calibration process combines adverse reaction classification statistics and existing historical adverse reaction data of the drug to ensure that the calibration results are consistent with actual medication safety scenarios. A fixed calculation method is used to calculate the calibration coefficient, which is the ratio of the transmission amplitude value to the correlation strength value in the coupling transmission characteristic data. The calibration coefficient is used to correct the deviation in the correlation strength data and eliminate the error caused by calculating the dose and duration separately. Based on the calibration coefficient, the correlation strength data is corrected, and the correlation strength value of each dose-duration combination is adjusted one by one. The corrected data retains a fixed number of decimal places and is organized to form calibrated correlation coupling data. This data can more accurately reflect the impact of dose-duration coupling on the occurrence of adverse reactions, and provides accurate data support for subsequent dose-duration correlation data integration and safety signal evaluation analysis.

[0042] Furthermore, dose-duration correlation data is formed by integrating the calibrated correlation coupling data.

[0043] In this embodiment of the invention, based on calibrated correlation coupling data, the calibrated correlation strength values, coupling transmission rules, dose influence characteristics, and duration influence characteristics corresponding to each dose-duration combination are systematically analyzed. The integration process strictly follows the operating procedures in the pharmacovigilance system documents to ensure data logical coherence and completeness. A unified data presentation format is used, organizing data according to the combination order of dose gradient and duration segments. The probability of adverse reactions, correlation strength level, calibration coefficient, and coupling transmission details corresponding to each combination are labeled, supplementing data validity explanations and eliminating logical contradictions and missing data entries. All integrated information is archived to generate dose-duration correlation data. This data fully presents the comprehensive impact of dose-duration coupling on the occurrence of adverse reactions and is directly used for correlation analysis data reconstruction, statistical analysis of existing adverse reaction data for various products, and writing reviews of product safety-related materials. Simultaneously, it supports the improvement of the pharmacovigilance system, professional personnel training, and the compliant and efficient operation of information systems.

[0044] Furthermore, the process of reconstructing and generating correlation analysis data based on the proportion of adverse reactions and combining population correlation data with dose-duration correlation data includes: Based on the incidence rate of adverse reactions, we extract the distribution data of correlation features and then filter the effective correlation factor data based on the correlation feature data. In this embodiment of the invention, by relying on the incidence rate of adverse reactions, correlation feature distribution data is extracted one by one. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the data extraction meets the needs of daily pharmacovigilance work. The correlation feature distribution data includes the incidence rate distribution of each type of adverse reaction, the incidence rate distribution of adverse reactions in different populations, and the incidence rate distribution of adverse reactions in different dose gradients and duration segments, which are classified and organized according to adverse reaction type, population group, dose gradient, and duration segment. Based on the correlation feature distribution data, a fixed screening method is used to remove correlation factors with too low a proportion or no practical correlation significance, and retain correlation factors that are closely related to the occurrence of adverse reactions and have a stable proportion. The proportion value, scope of influence, and corresponding adverse reaction type of each correlation factor are labeled, and the data are integrated to form effective correlation factor data, providing basic support for subsequent feature matching, statistical analysis of existing adverse reaction data of the product, and safety signal evaluation analysis.

[0045] Furthermore, feature matching is performed based on effective correlation factor data and population correlation data to generate population correlation matching data; In this embodiment of the invention, by relying on effective correlation factor data and population correlation data, each effective correlation factor and population correlation feature is compared and matched one by one. The matching process strictly follows the operating procedures in the pharmacovigilance system document to ensure that the matching logic conforms to the actual medication safety scenario. The population-related features in the effective correlation factors are matched one-to-one with the population group, correlation strength value, and adverse reaction differences in the population correlation data. The population characteristics, correlation strength, and adverse reaction occurrence corresponding to each effective correlation factor are clarified, and matching deviations and correction explanations are marked. Data with failed matches or logical contradictions are removed. All successfully matched information is organized and archived according to population group and correlation factor type to generate population correlation matching data. This data clarifies the correspondence between effective correlation factors and the medication user population, providing data support for subsequent spatiotemporal feature fusion, professional issue analysis, and comprehensive analysis of pharmacovigilance status of pharmacists.

[0046] Furthermore, spatiotemporal features are fused based on population association matching data and dose duration association data to generate fused association feature data; In this embodiment of the invention, spatiotemporal feature fusion is performed by relying on population association matching data and dose-duration association data. The fusion process combines temporal features such as adverse reaction occurrence time and medication cycle with spatial features such as population distribution and dose gradient distribution to ensure that the fusion result can comprehensively reflect the comprehensive influencing factors of adverse reaction occurrence. The population association strength and adverse reaction differences in the population association matching data are weighted and fused with the dose-duration coupling features and association strength values ​​in the dose-duration association data. The fusion weight is determined according to the degree of influence of each feature on the occurrence of adverse reactions, clarifying the superposition effect of association features under different populations and different dose-duration combinations. The fused feature data is organized into segments according to population group, dose gradient, and duration, and the fused association strength, adverse reaction probability, and spatiotemporal feature details are labeled to generate fused association feature data, providing accurate data support for subsequent association reconstruction and safety signal evaluation analysis.

[0047] Furthermore, based on the fusion and correlation feature data, the relationship between adverse reactions and the medication population, medication dosage, and medication duration is reconstructed to generate reconstructed correlation mapping data; the reconstructed correlation mapping data is then integrated to form correlation analysis data.

[0048] In this embodiment of the invention, by relying on fused correlation feature data, the relationship between adverse reactions and the user population, dosage, and duration of medication is comprehensively reconstructed. The reconstruction process strictly follows the operating procedures in the pharmacovigilance system documents and combines existing historical adverse reaction data of the product to ensure that the correlation relationship fits the actual medication scenario. The inherent logic in the fused correlation feature data is analyzed to clarify the mutual influence among the user population, dosage, and duration of medication and their comprehensive effect on the occurrence of adverse reactions. A complete correlation mapping from population characteristics, dosage gradient, duration segmentation to adverse reaction occurrence is constructed, generating reconstructed correlation mapping data. Based on the reconstructed correlation mapping data, all correlation information is integrated, the data presentation format is unified, data validity explanations, correlation logic analysis, and risk warnings are added, redundant and contradictory data are eliminated, and correlation analysis data is compiled. This data fully reflects the comprehensive correlation characteristics between adverse reactions and the user population, dosage, and duration of medication. It can be directly used for generating pattern analysis reports, statistical analysis of existing adverse reaction data of the product, and writing summaries of product safety-related materials. Simultaneously, it supports the improvement of the pharmacovigilance system, professional personnel training, and the compliant and efficient operation of information systems.

[0049] Furthermore, the safety signal evaluation based on the statistical analysis results of existing adverse reactions of the product in the safety signal evaluation module includes: Abnormal data were extracted from the statistical analysis results of existing adverse reactions of the product varieties as initial clues for safety signal monitoring, and the proportion of abnormal data was calculated. In this embodiment of the invention, abnormal data is extracted one by one based on the statistical analysis results of existing adverse reactions of the product. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the extracted abnormal data meets the needs of safety signal monitoring. Abnormal data includes adverse reaction data that exceeds the normal occurrence rate, adverse reaction data that occurs in a concentrated manner in a specific population, and adverse reaction data that is abnormally high under a specific dose-duration combination. These are classified and organized according to the abnormality type. The percentage of abnormal data is calculated using a fixed calculation method, which is the number of abnormal data entries divided by the total number of data entries in the statistical analysis results of existing adverse reactions of the product, retaining a fixed number of decimal places. The specific type of abnormal data, the number of cases involved, and related medication information are marked. Abnormal data serves as the initial clue for safety signal monitoring, providing basic clue support for subsequent clue screening, safety signal evaluation and analysis, and the writing of product safety-related data summaries, thus contributing to the compliant and efficient operation of the information system.

[0050] Furthermore, the initial clues are screened and verified based on the proportion of abnormal data, the credibility of the clues is calculated, and potential safety signals are identified from the statistical analysis results of existing adverse reactions of the product based on the credibility of the clues. In this embodiment of the invention, initial clues for safety signal monitoring are screened one by one based on the calculated proportion of abnormal data. Clues with too low a proportion or no actual risk indication are eliminated, while clues with an abnormal data proportion that meets the monitoring standards are retained. A fixed calculation method is used to calculate the credibility of each clue: the proportion of abnormal data multiplied by a data validity coefficient. The data validity coefficient is determined based on the completeness and logical rationality of the abnormal data; a higher credibility value indicates stronger reliability of the clue. Based on the credibility of the clues, potential safety signals are further identified from the existing adverse reaction statistical analysis results of the product. The abnormal data details, associated populations, dosage, and duration of each potential safety signal are clarified, and the source and abnormal characteristics of the signal are marked, generating a list of potential safety signals. This provides core signal support for subsequent signal classification, safety signal evaluation and analysis, and professional problem analysis.

[0051] Furthermore, potential safety signals are classified to clarify the signal type, scope of occurrence, and affected population, generating signal classification data; In this embodiment of the invention, identified potential safety signals are categorized and organized according to signal type. The classification strictly follows the operating procedures outlined in the pharmacovigilance system documents, aligning with the needs of safety signal evaluation and analysis. Potential safety signals are divided into fixed categories such as digestive system abnormalities, nervous system abnormalities, and skin / mucous membrane abnormalities. The occurrence range of each category of potential safety signals is analyzed, clarifying the geographical location, distribution of medical institutions, and medication scenarios where the signals occur. The affected population is identified, categorized by age, gender, and underlying disease type, and the number of affected cases and the degree of impact are noted for each group. The signal types, occurrence ranges, affected populations, and associated abnormal data are compiled and archived to generate signal classification data. This provides classification data support for subsequent risk level assessments, comprehensive analysis of pharmacovigilance status of pharmacists, and customization of customized pharmacovigilance solutions.

[0052] Furthermore, the built-in GVP security evaluation standard is invoked to evaluate the risk level of various potential security signals in the signal classification data, calculate the risk score corresponding to each type of security signal, and generate signal risk evaluation data. In this embodiment of the invention, an AI-based pharmacovigilance information system is activated, invoking the system's built-in Good Pharmacy Values ​​(GVP) safety evaluation standard. This standard covers signal risk level classification, scoring criteria, and evaluation procedures, aligning with the compliance requirements of routine pharmacovigilance work. Various potential safety signals from the signal classification data are individually substituted into the GVP safety evaluation standard. A fixed calculation method is used to calculate the risk score corresponding to each safety signal. The scoring indicators include signal occurrence frequency, affected population size, severity of adverse reactions, and associated medication range. Each indicator is assigned a corresponding score according to a fixed standard, and the risk score is obtained by summing all indicator scores. Risk levels are classified according to the risk score: 0-30 is low risk, 31-60 is medium risk, and 61-100 is high risk. The risk levels of various potential safety signals are clearly defined. The signal type, risk score, risk level, and scoring basis are organized and archived to generate signal risk evaluation data, providing risk data support for subsequent safety signal evaluation report generation and safety signal evaluation analysis.

[0053] Furthermore, signal classification data and signal risk assessment data are integrated to generate a safety signal assessment report.

[0054] In this embodiment of the invention, by relying on signal classification data and signal risk assessment data, detailed information on various potential safety signals is comprehensively integrated. The integration process strictly follows the operating procedures in the pharmacovigilance system documents to ensure data logical coherence and content completeness. The integrated content includes signal classification details, occurrence scope, affected population, risk score, risk level, and evaluation basis, supplements signal risk prevention and control recommendations, and improves the report content by combining literature search evaluation analysis results and existing adverse reaction data of the product. The integrated information is organized into a safety signal evaluation report according to a fixed report format. The report content is tailored to the needs of safety signal evaluation and analysis and can be directly used for writing product safety-related data reviews, analyzing professional issues, and comprehensively analyzing the pharmacovigilance status of holders. At the same time, it supports the improvement of the pharmacovigilance system, the training of professional personnel, and the compliant and efficient operation of the information system, providing a risk basis for developing a customized overall pharmacovigilance solution.

[0055] Furthermore, the integrated product data includes existing adverse reaction statistical analysis results and safety signal evaluation reports, combined with literature search evaluation data and pharmacovigilance system documents, to generate a product safety-related data summary, including: Extract signal information and risk assessment conclusions from the safety signal evaluation report, determine the core direction of literature retrieval, and calculate the signal correlation degree; In this embodiment of the invention, by relying on safety signal evaluation reports, signal information and risk assessment conclusions in the reports are screened and extracted one by one. The extraction process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the extracted information meets the needs of literature retrieval, evaluation, and analysis. The signal information includes the type, scope of occurrence, affected population, and risk level of various safety signals; the risk assessment conclusions include signal risk prevention and control recommendations and an analysis of the degree of risk impact. Based on the extracted information, the core direction of literature retrieval is clarified, focusing on the occurrence mechanism, prevention and control measures, and conclusions of similar studies of various safety signals, and the retrieval direction is divided according to signal type. A fixed calculation method is used to calculate the signal correlation degree, which is calculated by multiplying the signal risk score by the signal occurrence frequency. The higher the correlation degree value, the closer the connection between the signal and the literature retrieval, providing a core basis for subsequent literature retrieval ranking, literature retrieval evaluation and analysis, and the writing of product safety-related data summaries.

[0056] Furthermore, by using a literature search engine and connecting to medical literature databases in conjunction with the core direction of literature retrieval, literature searches are conducted, and literature searches are ranked based on signal relevance to generate literature search results; In this embodiment of the invention, an AI-based pharmacovigilance information system is activated, invoking the system's built-in literature retrieval engine. This engine can stably connect with various medical literature databases, meeting the daily literature retrieval needs of the information system. The determined core directions for literature retrieval are entered into the literature retrieval engine one by one. The engine establishes an encrypted connection with the medical literature database to conduct targeted literature searches. Search logs are maintained during the search process for subsequent data tracing and comprehensive analysis of the pharmacovigilance status of the holder. After the search is completed, the retrieved literature is sorted based on the signal correlation calculated in step S411. Literature with higher correlation values ​​is ranked higher. Literature irrelevant to the core search direction or of substandard quality is removed. The title, author, publication date, core research content, and correlation points with safety signals are labeled for each article. The literature retrieval results are then compiled to provide literature support for subsequent literature evaluation and analysis, and professional problem analysis.

[0057] Furthermore, the literature search results are evaluated and analyzed to extract research conclusions related to safety signals from the literature and generate literature search evaluation data. In this embodiment of the invention, each document in the generated literature search results is evaluated and analyzed individually. The evaluation includes the scientific validity of the research, the completeness of the data, and the reliability of the conclusions. The evaluation strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the evaluation results meet the needs of literature search evaluation and analysis. Research conclusions related to safety signals are extracted from each document, with a focus on extracting relevant content regarding the mechanism, patterns, prevention and control measures, and studies of similar adverse reactions of safety signals. These are compared with the conclusions of the safety signal evaluation report, marking consistent and differing conclusions and explaining the reasons for the differences. The evaluation results, extracted research conclusions, details of the correlation with safety signals, and comparison explanations for each document are compiled and archived to generate literature search evaluation data. This data is directly used for writing reviews of product safety-related materials, evaluating and analyzing safety signals, and analyzing professional issues, while also supporting the improvement of the pharmacovigilance system.

[0058] Furthermore, the pharmacovigilance system documents, including SMP, SOP and pharmacovigilance master documents, are called up to extract the relevant content of system requirements, and the system fit is calculated to generate pharmacovigilance system evaluation data. In this embodiment of the invention, an AI-based pharmacovigilance information system is activated, calling the system's built-in pharmacovigilance system documents, including SMP, SOP, and the main pharmacovigilance document. These documents cover company-level bylaws, management procedures, and operating procedures, aligning with the requirements for a complete pharmacovigilance system and compliant operation. The system requirements related to product safety, safety signal management, and literature retrieval evaluation are extracted from each document, clarifying the specific requirements for adverse reaction statistics, safety signal monitoring, and literature retrieval. A fixed calculation method is used to calculate the system fit: the number of items that actually meet the system requirements is divided by the total number of system requirements, retaining a fixed number of decimal places. Items that meet the requirements and those that do not are marked, with explanations of the reasons for non-compliance and directions for rectification. This data forms the pharmacovigilance system evaluation data, providing system support for subsequent product safety overview integration, comprehensive analysis of pharmacovigilance status by pharmacists, and customized solutions.

[0059] Furthermore, by integrating existing adverse reaction statistical analysis results, safety signal evaluation reports, literature retrieval evaluation data, and pharmacovigilance system evaluation data, a comprehensive summary of product safety-related information is generated.

[0060] In this embodiment of the invention, the core content of various data is comprehensively integrated by relying on the statistical analysis results of existing adverse reactions of the product, safety signal evaluation reports, literature retrieval evaluation data, and pharmacovigilance system evaluation data. The integration process strictly follows the operating procedures in the pharmacovigilance system documents to ensure the logical coherence and completeness of the review content. The integrated content includes detailed statistical information on existing adverse reactions of the product, safety signal evaluation conclusions, literature research support, system compatibility, supplementary professional problem analysis results, risk prevention and control suggestions, and pharmacovigilance status analysis of the holder. Combined with the customized needs of the specific pharmacovigilance solution, the review content is improved. Following the fixed format of product safety-related data reviews, the data presentation method is unified, redundant and contradictory data is eliminated, and supplementary data validity explanations and conclusion summaries are added to generate a product safety-related data review. This review can be directly used for the conduct of pharmacovigilance work by the holder and for the training of professionals. At the same time, it supports the improvement of the pharmacovigilance system and the compliant and efficient operation of the system, providing a complete basis for formulating the most economical overall pharmacovigilance solution under compliance.

[0061] Furthermore, the evaluation and analysis of the literature search results, and the extraction of research conclusions related to security signals from the literature, include: Structured analysis is performed based on the literature retrieval results to extract the research methods, research subjects, and research conclusions of the literature. In this embodiment of the invention, each document is analyzed piecemeal based on the literature search results. The analysis process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the analysis results meet the needs of literature search evaluation and analysis. The analysis focuses on the core content of the documents, using fixed extraction methods to extract the research methods, research subjects, and research conclusions of each document. The research methods include literature research, clinical observation, and data analysis, with specific operating procedures for each method clearly defined. The research subjects include the age, gender, underlying diseases, dosage, and duration of medication use of the drug-using population, and relevant information is organized according to fixed standards. The research conclusions focus on extracting adverse reaction research findings, risk analysis, and prevention and control recommendations related to safety signals. The three core pieces of extracted information are organized by document title, with the extracted content corresponding to each document marked. Entries with missing information or logical contradictions are removed to ensure that the extracted information is complete and traceable. This provides basic information for subsequent calculation of literature support, evaluation of literature quality, and writing of product safety-related literature reviews, supporting the compliant and efficient operation of the information system.

[0062] Furthermore, by comparing the research methods, research subjects, and research conclusions of the literature with the core signal information in the safety signal evaluation report, the support level of the literature is calculated to determine the degree to which the literature supports the safety signal. In this embodiment of the invention, the core signal information in the safety signal evaluation report is compared one by one with the extracted literature research methods, research subjects, and research conclusions. The comparison process strictly follows the operating procedures in the pharmacovigilance system documents to ensure that the comparison logic aligns with the needs of safety signal evaluation and analysis. The comparison focuses on the consistency between the literature research subjects and the population affected by the safety signal, the degree of fit between the literature research conclusions and the safety signal risk assessment conclusions, and the adaptability of the literature research methods to the verification of the safety signal. The literature support level is calculated by dividing the number of matching items by the total number of comparison items, then multiplying by a fixed weight. The higher the support level value, the stronger the support for the safety signal. The degree of support is determined based on the support level value: 0-30% is weak support, 31-60% is moderate support, and 61-100% is strong support. The support level value and degree of support for each literature are marked, providing supporting evidence for subsequent literature quality evaluation, literature annotation, and professional issue analysis.

[0063] Furthermore, the process involves obtaining publication information and extracting journal details and publication dates to determine the timeliness and academic impact of the literature. A timeliness score is calculated using the formula: Timeliness Score = Baseline Score - Deductions for Time Difference. The research design is analyzed, extracting sample size, research period, and control settings to determine the scientific validity of the research design. A scientific validity score is calculated using the formula: Scientific Validity Score = Sample Size Score + Research Period Score + Control Settings Score. The research methods of the literature are compared with the GVP (General Philosophy and Evaluation) literature evaluation standards to generate methodological compliance data. A compliance score is calculated using the formula: Compliance Score = Number of Compliant Methods ÷ Total Number of Research Methods in the Literature. Based on the timeliness score, scientific validity score, and compliance score, a literature rigor score is generated. Finally, the literature rigor score and various evaluation data are integrated to generate literature quality evaluation data. In this embodiment of the invention, publication information for each article is obtained, including the journal information and publication date. A time difference is calculated based on the publication date, representing the difference in years between the current time and the article's publication date. A timeliness score is calculated using a fixed deduction standard: Timeliness Score = Baseline Score - Deduction for Time Difference. The base score is set at 100 points, and 10 points are deducted for each additional year of time difference, with a minimum deduction of 60 points. This determines the timeliness of the article. Academic influence is determined by combining the journal's level with the journal type and influence level. The research design of each article is analyzed, extracting the sample size, research period, and control settings. Sample size, research period, and control settings scores are calculated using a fixed standard, and these three scores are added together to obtain a scientific merit score: Scientific Merit Score = Sample Size Score + Research Period Score + Control Settings Score. The research methods of the articles are compared with the system's built-in GVP (Global Value Parts) literature evaluation standards. Methods that conform to the standards and those that do not are marked, generating method compliance data and calculating a compliance score: Compliance Score = Number of Compliant Methods ÷ Total Number of Research Methods in the Articles. The timeliness score, scientific validity score, and compliance score are weighted and summed according to fixed weights to obtain the literature rigor score. The literature rigor score, the details of each score, the data on methodological compliance, timeliness, and academic influence are integrated to generate literature quality evaluation data, which provides a quality basis for subsequent literature annotation and comprehensive analysis of pharmacovigilance status of holders.

[0064] Furthermore, based on the literature support level and literature quality evaluation data, each literature is classified and labeled to generate literature labeling data; the literature research conclusions, literature support level, literature quality evaluation data and literature labeling data are integrated to generate literature retrieval evaluation data.

[0065] In this embodiment of the invention, each document is classified and labeled individually based on its support level and quality evaluation data. The labeling criteria combine support level and rigor score, categorizing documents into three types: core support documents, general support documents, and unsupported documents. Core support documents must meet the criteria of strong support and a rigor score of 80 or above; general support documents must meet the criteria of moderate support and a rigor score of 60-79; and unsupported documents are those with weak support or a rigor score below 60. The classification results and labeling criteria for each document are then labeled. The extracted research conclusions, support level, quality evaluation data, and the generated labeling data are comprehensively integrated. The integration process strictly follows the operating procedures in the pharmacovigilance system documents, unifying the data presentation format, organizing by document title, labeling complete evaluation information for each document, removing redundant and contradictory data, and supplementing data validity explanations. After integration, literature retrieval evaluation data is generated. This data is directly used for writing reviews of product safety-related materials, evaluating and analyzing safety signals, and analyzing professional issues. It also supports the improvement of the pharmacovigilance system, the training of professional personnel, and the compliant and efficient operation of the information system, providing literature support for a comprehensive analysis of pharmacovigilance status for pharmacists.

[0066] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An artificial intelligence-based pharmacovigilance information system, characterized in that, Includes the following modules: The pharmaceutical product acquisition module is used to obtain the pharmacovigilance service requirements and target drug product information of pharmaceutical companies. The drug safety data acquisition module is used to connect with medical big data sources based on artificial intelligence technology to collect full life-cycle safety data for target drug varieties. The existing adverse reaction statistics module is used to perform statistical analysis of existing adverse reactions on the full life cycle safety data of the target drug and generate statistical analysis results of existing adverse reactions of the drug. The safety signal evaluation module is used to evaluate safety signals based on the statistical analysis results of existing adverse reactions of the product and generate a safety signal evaluation report. By integrating existing adverse reaction statistical analysis results and safety signal evaluation reports for various products, and combining literature search evaluation data with pharmacovigilance system documents, a comprehensive summary of product safety-related information is generated, simultaneously supporting pharmacovigilance outsourcing and system construction.

2. The artificial intelligence-based pharmacovigilance information system according to claim 1, characterized in that, The drug safety data acquisition module includes the following functions: The AI ​​data docking engine is invoked to establish an encrypted communication link with the medical big data platform; A data collection request is sent to the medical big data platform via an encrypted communication link. The data collection request includes information on the target drug and the scope of data collection. Receive initial safety data returned by the medical big data platform, including clinical research data, post-marketing surveillance data, and individual adverse reaction report data; The initial security data is screened by an artificial intelligence data docking engine, the percentage of valid data is calculated, invalid and abnormal data are removed, and valid security data is obtained. The effective safety data is structured and transformed, and after combining the data validity ratio and standardization, the full life cycle safety data corresponding to the target drug is generated.

3. The artificial intelligence-based pharmacovigilance information system according to claim 1, characterized in that, The existing adverse reaction statistics module includes the following functions: The entire lifecycle safety data is input into an AI-based adverse reaction statistical model to extract adverse reaction-related features; Based on the characteristics of adverse reactions, the occurrence of different types of adverse reactions is classified and statistically analyzed, the proportion of each type of adverse reaction is calculated, and statistical data on adverse reaction classification is generated. Explore the correlation between adverse reactions and the user population, dosage, and duration of medication, and generate correlation analysis data by combining the proportion of adverse reactions. Based on correlation analysis data, identify potential patterns in the occurrence of adverse reactions and generate a pattern analysis report; By integrating adverse reaction classification statistics, correlation analysis data, and pattern analysis reports, statistical analysis results of existing adverse reactions for each product are generated.

4. The artificial intelligence-based pharmacovigilance information system according to claim 3, characterized in that, The process of identifying the correlation between adverse reactions and the patient population, dosage, and duration of medication, and generating correlation analysis data based on the proportion of adverse reactions, includes: Extract user characteristics, dosage data, and duration of medication from the full life-cycle safety data; By using artificial intelligence correlation mining algorithms, a model is established to establish the correspondence between the occurrence of adverse reactions and the characteristics of the drug-using population; Based on the correspondence model, we analyze the differences in adverse reactions corresponding to the characteristics of different drug users, calculate the correlation degree of population characteristics, and generate population association data. By combining medication dosage data and medication duration data, the corresponding relationship model is adjusted to analyze the impact of medication dosage and duration on the occurrence of adverse reactions, calculate the dose-duration influence coefficient, and generate dose-duration correlation data. Correlation analysis data is generated by reconstructing the correlation analysis data based on the proportion of adverse reactions and combining population correlation data and dose duration correlation data.

5. The artificial intelligence-based pharmacovigilance information system according to claim 4, characterized in that, The generated dose-duration correlation data includes: Extract dose gradient information from medication dosage data and duration segmentation information from medication duration data; The dose gradient information and time segment information are substituted into the correspondence model, the model parameters are dynamically adjusted, the parameter adjustment range is calculated, and the correspondence model is adjusted based on the parameter adjustment range to generate the adjusted correspondence model. Based on the adjusted correspondence model, the probability of adverse reactions corresponding to different dose gradients and time segments is calculated. Based on the probability of adverse reactions, determine the correlation strength between drug dosage and duration of medication and the occurrence of adverse reactions, and generate correlation strength data; Dose-time correlation data is generated by coupling dose-time correlation based on correlation intensity data and combining dose gradient information with duration segmentation information.

6. The artificial intelligence-based pharmacovigilance information system according to claim 5, characterized in that, The dose-time correlation coupling based on correlation intensity data combined with dose gradient information and duration segmentation information includes: Dose influence features and duration influence features are extracted from the correlation strength data, and correlation coupling basic data are generated based on the dose influence features and duration influence features; Based on the correlation and coupling basic data, the correspondence between dose gradient information and duration segment information is determined, and gradient segment mapping data is generated; based on the gradient segment mapping data, the coupling and transmission relationship between drug dosage and drug duration is constructed, and coupling and transmission feature data is generated. The correlation strength data is coupled and calibrated based on the coupling transfer characteristic data to generate calibrated correlation coupling data. Dose-duration correlation data is generated by integrating the calibrated correlation coupling data.

7. The artificial intelligence-based pharmacovigilance information system according to claim 4, characterized in that, The correlation analysis data reconstructed based on the proportion of adverse reactions and combined with population correlation data and dose-duration correlation data includes: Based on the incidence rate of adverse reactions, we extract the distribution data of correlation features and then filter the effective correlation factor data based on the correlation feature data. Based on feature matching between effective correlation factor data and population correlation data, population correlation matching data is generated. Spatiotemporal features are fused based on population association matching data and dose duration association data to generate fused association feature data; Based on the fusion and correlation feature data, the relationship between adverse reactions and the medication population, medication dosage, and medication duration is reconstructed to generate reconstructed correlation mapping data; the reconstructed correlation mapping data is then integrated to form correlation analysis data.

8. The artificial intelligence-based pharmacovigilance information system according to claim 1, characterized in that, The safety signal evaluation module includes the following: Safety signal evaluation based on statistical analysis results of existing adverse reactions of the product. Abnormal data were extracted from the statistical analysis results of existing adverse reactions of the product varieties as initial clues for safety signal monitoring, and the proportion of abnormal data was calculated. Initial clues are screened and verified based on the proportion of abnormal data, the credibility of the clues is calculated, and potential safety signals are identified from the statistical analysis results of existing adverse reactions of the product based on the credibility of the clues. Classify potential safety signals, identify signal types, their range of occurrence, and the affected populations, and generate signal classification data; The built-in GVP security evaluation standard is invoked to evaluate the risk level of various potential security signals in the signal classification data, calculate the risk score corresponding to each type of security signal, and generate signal risk evaluation data. Integrate signal classification data and signal risk assessment data to generate a safety signal assessment report.

9. The artificial intelligence-based pharmacovigilance information system according to claim 8, characterized in that, The integrated product profile includes both adverse reaction statistical analysis results and safety signal evaluation reports. Combined with literature search evaluation data and pharmacovigilance system documents, a comprehensive summary of product safety-related information is generated, including: Extract signal information and risk assessment conclusions from the safety signal evaluation report, determine the core direction of literature retrieval, and calculate the signal correlation degree; By using a literature search engine and connecting to medical literature databases in conjunction with the core direction of literature search, literature searches are conducted, and literature searches are ranked based on signal relevance to generate literature search results. The literature search results are evaluated and analyzed to extract research conclusions related to safety signals from the literature and generate literature search evaluation data. The pharmacovigilance system documents, including SMP, SOP and pharmacovigilance master file, are called up to extract the relevant content of system requirements, calculate the system fit, and generate pharmacovigilance system evaluation data. By integrating existing adverse reaction statistical analysis results, safety signal evaluation reports, literature retrieval evaluation data, and pharmacovigilance system evaluation data, a comprehensive summary of product safety-related information is generated.

10. The artificial intelligence-based pharmacovigilance information system according to claim 9, characterized in that, The evaluation and analysis of the literature search results, and the extraction of research conclusions related to security signals from the literature, include: Structured analysis is performed based on the literature retrieval results to extract the research methods, research subjects, and research conclusions of the literature. By comparing the research methods, research subjects, and research conclusions of the literature with the core signal information in the safety signal evaluation report, the support level of the literature is calculated to determine the degree to which the literature supports the safety signal. This process involves: acquiring literature publication information and extracting journal details and publication dates to determine the timeliness and academic impact of the literature; calculating a timeliness score using the formula: Timeliness Score = Baseline Score - Deductions for Time Difference; analyzing the research design to extract sample size, research period, and control settings to determine the scientific validity of the research design; calculating a scientific validity score using the formula: Scientific Validity Score = Sample Size Score + Research Period Score + Control Settings Score; comparing the literature's research methods with the GVP (General Philosophy and Evaluation) literature evaluation guidelines to generate methodological compliance data; calculating a compliance score using the formula: Compliance Score = Number of Methods Compliant ÷ Total Number of Research Methods in the Literature; generating a literature rigor score based on the timeliness, scientific validity, and compliance scores; and integrating the literature rigor score and other evaluation data to generate literature quality evaluation data. Based on literature support and quality evaluation data, each document is classified and labeled to generate literature labeling data; literature research conclusions, literature support, literature quality evaluation data and literature labeling data are integrated to generate literature retrieval evaluation data.

Images