A system and method for analyzing factors affecting adverse events

Through the analysis system for influencing factors of adverse events, the problem that the medical data system cannot effectively manage and analyze adverse event data is solved, and in-depth analysis and prevention support for the causes of adverse events is achieved, and the hospital's prevention capabilities and data utilization efficiency are improved.

CN114238630BActive Publication Date: 2024-12-06RUIJING COMPUTER SCI & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111474365.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-03
Publication Date
2024-12-06
Estimated Expiration
2041-12-03

AI Technical Summary

Technical Problem

The existing medical data system cannot effectively manage and analyze medical order data for adverse events, resulting in the inability to conduct real-time and efficient cause analysis and preventive measures, and insufficient support for resource waste and risk analysis.

Method used

A system for influencing factors of adverse events was designed, including event cause theme library generation module, principal component feature extractor, data quality control and standardization module, influencing factor extraction module, classification model generation module and event factor comprehensive analysis module, through Textured data processing, feature extraction, model training and visual presentation to achieve in-depth analysis and prevention support for the causes of adverse events.

Benefits of technology

It realizes unified management of adverse events and rapid and accurate analysis of causes and influencing factors, provides feasible improvement measures and suggestions, and improves the hospital's prevention capabilities and data utilization efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114238630B_ABST
    Figure CN114238630B_ABST
Patent Text Reader

Abstract

The present invention discloses an adverse event influencing factor analysis system and method, wherein the system obtains the cause data corresponding to clinically diagnosed adverse events, and after quality control, standardization and grouping of clinical data, establishes an adverse event influencing factor analysis model for each clinical theme. When the system receives an analysis request for obtaining adverse event-related influencing factors of clinical data carrying a patient, the adverse event type influencing factor analysis model of the corresponding clinical theme is obtained according to the patient's clinical data matching, and the adverse event classification information, influencing factor specific information and the calculation results of the influencing factors and the ranking results of the influencing factors are obtained, and a comprehensive evaluation is performed according to the analysis information of the causes of each adverse event to obtain an output feasible suggestion. The present invention realizes the unified management of clinical data, and outputs the analysis results based on the influencing factors that lead to the cause data of the adverse event provided by the subject library.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a clinical data processing system and method in the medical industry, and in particular to a system and method for analyzing factors affecting the risk causes of adverse events. Background Art

[0002] In recent years, with the rapid development of computer technology and business expansion in the medical industry, the use of computers to manage clinical data in the medical process has become a development trend in the field of medical systems, especially the problem of medical-related adverse events in the medical industry. The large amount of patient and testamentary diagnostic data generated by its processing is simply stored in the computer and cannot be reasonably applied and processed again. In a sense, it is a waste of resources. The existing data system in the medical industry is only a single data storage management system, and has not provided substantial help and application for the occurrence and prevention of adverse events.

[0003] The main defects are: first, the data related to the registration and processing of adverse events have not been uniformly managed and analyzed, and no separate cause database has been formed, which cannot provide data support for subsequent risk analysis. Secondly, the data are not sorted and disorganized, and there is no corresponding data preprocessing and related data screening. Finally, clinical data is not used for model analysis and feedback, and it is impossible to conduct real-time and efficient cause analysis of the nature of adverse events, and there is no sufficient feedback for the prevention of the next event.

[0004] Therefore, after research, the inventors found that how to create a unified medical system that can organize adverse event medical order data and conduct in-depth analysis of the causes and influencing factors of adverse events based on clinical data has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a system and method for analyzing the causes and influencing factors of adverse events, which can realize unified management of the medical order information generated clinically after the adverse event treatment, and provide the hospital with the analysis results of the causes and influencing factors of adverse events for preventing unknown adverse events based on clinical data.

[0006] The technical solution adopted by the present invention to solve the above technical problems is to provide an adverse event cause influencing factor analysis system, including: an event cause theme library generation module, which is used to identify and extract the input textual data of the cause of each type of clinical diagnosis event; a principal component feature extractor, which is used to analyze the submitted textual cause data and output relevant feature cause results; perform weight probability calculation on the feature results and perform keyword extraction; a data quality control and standardization module, which is used to perform quality control and standardization processing on the clinical data under each theme library after association; an influencing factor extraction module, which is used to parse the clinical data results, extract the relevant influencing factor attributes, and screen them through the word frequency related weight calculation; a classification model generation module, which is used to generate An analysis model of adverse event influencing factors extracted from the subject library of each event cause; a model selector, used to obtain the corresponding clinical theme and the corresponding analysis model of adverse event influencing factors based on the matching of the input newly added clinical data, obtain the weight proportion and ranking of the adverse event influencing factors therein, and send them to the comprehensive analysis module of event influencing factors; the comprehensive analysis module of event factors, used to receive an analysis request of adverse event influencing factors carrying the clinical data of newly added patients, send the patient's clinical data to the model selector, and receive the feature extraction information of the adverse event influencing factors corresponding to the newly added adverse event returned by the model selector, obtain the weight calculation result corresponding to the influencing factor corresponding to the newly added adverse event, and send it in the response of obtaining the analysis of adverse event influencing factors.

[0007] Furthermore, the textual data in the event cause theme library generation module includes: the patient's corresponding adverse event risk category classification and risk value range information, the doctor's medical advice information data for the patient, the medical record clinical pathway, occurrence time and reporting time information.

[0008] Furthermore, the data quality control and standardization processing module includes: a data quality control unit, which is used to screen and clean the clinical data associated in the subject library for abnormal data, and mark the related abnormal data; a data standardization unit, which is used to perform data vectorization processing and standardization on the clinical data set associated in the subject library, and the submodules included are as follows:

[0009] The data binning module is used to adaptively bin the associated clinical data set according to the set adverse event risk type to be predicted; the data vectorization submodule is used to vectorize the read text data of the adverse event type and event cause of the patient, and output the quantitative data type that can be processed by the computer; the abnormal data processing submodule is used to perform relevant data preprocessing operations on the abnormal data marked by the data quality control unit, including screening of abnormal data and elimination of abnormal values, filling in the blanks of abnormal values, performing data attribute statistics on the generated abnormal data, and generating tables.

[0010] Furthermore, the principal component feature extractor includes: an influencing factor active extraction module, which is used to extract the influencing feature factors after screening and analysis in the subject library; an attribute weight calculation unit, which is used to calculate the correlation coefficient of the risk influencing factor text and classify the safety risk of adverse events; a text natural language processing unit, which is used to label the acquired patient data, compare the data set with the training of the historical model to perform factor identification and screening, and realize AI intelligent text classification;

[0011] Furthermore, the classification model generation module includes: a data set training unit, which is used to obtain the safety risk level and type of the adverse event cause data belonging to the subject library; a model generator unit, which is used to generate an adverse event risk influencing factor analysis model corresponding to the model data set training unit. Its main process is to mark the data and then divide the acquired data set into a training set and a test set. According to the specific distribution of the data set, the model is trained and verified by using a logistic regression algorithm, a convolutional neural network algorithm, a decision tree algorithm, etc. The model accuracy is evaluated based on the verification result of the test set, and a decision is made on the selection of the model by comparing the training results of different algorithm models.

[0012] Furthermore, the model generator unit is also used to analyze the system model results according to the adverse event risk influencing factors, using the effectiveness verification dimensions and their importance levels in order of average level accuracy Significance level And the degree of fit The model influence weights w1, w2, and w3 are set to 0.5, 0.3, and 0.2 respectively, and the quantitative standard N for model evaluation selection is:

[0013]

[0014] Through comparative iterative calculations, the adverse event influencing factor analysis algorithm model results that reach the expected value are evaluated.

[0015] Furthermore, the comprehensive adverse event risk influencing factor analysis module includes: a single adverse event influencing factor analysis unit, which is used to receive the newly added specific adverse event type cause data information and adverse event-related influencing factor information sent by the model selector, and perform intelligent analysis on the adverse event cause data, respectively obtain the specific influencing factor weight calculation results and influencing factor ranking information of the newly added detailed adverse events, and perform visual display; a comprehensive adverse event influencing factor analysis unit, which is used to identify the type of the received newly added adverse event type and then analyze and process its cause, and finally obtain the specific identified adverse event type information and the weight information of each influencing factor, and perform visual ranking display on its weight calculation results.

[0016] In order to solve the above technical problems, the present invention also provides a method for analyzing the influencing factors of the causes of adverse events. The above-mentioned system for analyzing the influencing factors of the causes of adverse events is built on the network side of the medical system, including the following steps: first, the cause theme library data is obtained, including data update and data import of the data interface, and the updated data content mainly includes updating the historical unacquired data and real-time updated data; data preprocessing, data control management and data standardization processing are performed on the acquired data; principal component extraction and weight calculation are performed on the clinical medical order data; the processed clinical medical order data set is divided into a training set: a test set = 7:3, and after selecting the type of model, the data is imported, and the model is trained for the training set, and the correlation weight ranking results and event types of the influencing factors are output, and visualization tables and feasible improvement suggestions are given for the results.

[0017] Furthermore, the imported and updated data include data statistics of the hospital, clinical diagnosis data related to its affiliated branches, and internal records of the medical alliance related to it.

[0018] Furthermore, the data preprocessing includes missing value processing, discrete point processing and data normalization of the original clinical data; the inconsistency of the data is mainly solved by filling missing values, smoothing noise data, smoothing or deleting outliers; the data charge management is mainly a data quality assessment process for clinical data, and the measurement dimensions are divided into standard assessment and whether the data indicators meet the quality control standards of the National Health Commission; and the cause data of adverse events managed by data charge management are normalized, which includes reasonable classification operations and risk classification processing of event types; the calculation formula for the weight of clinical medical order data is:

[0019]

[0020] Compared with the prior art, the present invention has the following beneficial effects: the system and method for analyzing the causes and influencing factors of adverse events provided by the present invention can realize unified management of the medical order information generated clinically after the adverse event treatment, and can quickly and accurately provide the hospital with the analysis result information of the influencing factors and feasible improvement measures based on the algorithm model training results. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of the structure of the adverse event cause influencing factor analysis system provided by an embodiment of the present invention;

[0022] Figure 2 A flow chart of a method for analyzing factors affecting adverse causes provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] The embodiment of the present invention builds an adverse event influencing factor analysis system on the medical system platform side. The system obtains data on the causes of adverse events, and after quality control, standardization and grouping of the data, establishes a classification model of the correlation of influencing factors for the causes of different types of adverse events. When the system receives an adverse event risk prediction request for obtaining clinical data of patients, it obtains the adverse event influencing factor analysis model of the corresponding historical training according to the data matching provided by the subject library, obtains the adverse event influencing factor information therein, and obtains the weight of each influencing factor, and performs a comprehensive evaluation based on the correlation numerical ranking of each influencing factor to obtain the analysis result, and provides feasible improvement suggestions. In this way, the embodiment of the present invention provides a unified management of adverse event cause processing data, and provides the analysis result information of the influencing factors and feasible improvement measures to the hospital based on the algorithm model training results.

[0025] Figure 1 The schematic diagram of the structure of the adverse event influencing factor analysis system provided in the embodiment of the present invention includes: an event cause theme library generation module, a data quality control and standardization module, a principal component feature extractor, a classification model generation module, a classification model selection module and an event factor comprehensive analysis module; the main functions and implementations of each module are as follows.

[0026] The event cause subject library generation module is used to identify and extract the input textual data of the causes of various clinical diagnosis events;

[0027] The principal component feature extractor is used to perform principal component sentiment analysis on the submitted reason text data and output relevant feature reason results. It also performs weighted probability calculation on the feature results and extracts keywords;

[0028] Data quality control and standardization module, used to perform quality control and standardization on clinical data under each subject library after association;

[0029] The influencing factor extraction module is used to parse the clinical data results to extract relevant influencing factor attributes and perform word frequency-related weight calculations for secondary screening.

[0030] A classification model generation module, used to generate an analysis model of adverse event influencing factors extracted from the subject library of each event cause;

[0031] A model selector is used to obtain the corresponding clinical theme and the corresponding adverse event influencing factor analysis model according to the matching of the newly input clinical data, obtain the weight ratio and ranking of the adverse event influencing factors, and send them to the event factor comprehensive analysis module;

[0032] The event factor comprehensive analysis module is used to receive an adverse event influencing factor analysis request carrying the clinical data of a newly added patient, send the patient's clinical data to the model selector, and receive the feature extraction information of the adverse event influencing factors corresponding to the newly added adverse event returned by the model selector, obtain the weight calculation results corresponding to the influencing factors corresponding to the newly added adverse event, and send them in the response to obtain the adverse event influencing factor analysis.

[0033] In this system, the cause data subject library module is used to collect the medical clinical data generated after the adverse events are processed. The data content includes the classified specific adverse event clinical data and the unclassified clinical data, and the data interface can be connected to the data subject library under the construction of the associated medical alliance. Based on this subject library, the cause diagnosis data of medical personnel and the adverse event diagnosis results of intelligent judgment can be updated in real time.

[0034] In this system, the subject type refers to the adverse event type, such as adverse drug safety events, medical device adverse event types, and so on.

[0035] In this system, the principal component feature extractor includes an active extraction module for influencing factors, which is used to extract the influencing feature factors after screening and analysis in the subject library in a textual manner; for example, the original influencing factor data of adverse drug events includes influencing factor conditions such as drug production years, drug types, and drug manufacturers. The principal component feature extractor also includes an attribute weight calculation unit, which is used to calculate the correlation coefficient of the risk influencing factor text (such as Pearson correlation coefficient p, spearman correlation coefficient), and classify the safety risks of adverse events.

[0036] In the system, when the clinical data of the subject generation module is updated, the clinical data processing processes of other modules of the system are also updated accordingly.

[0037] In this system, the clinical data associated with the subject generation module are updated in real time. When updating, the updated data content mainly includes updating the historical data that has not been acquired and the real-time updated data. The historical data includes the data statistics of this hospital, the relevant clinical diagnosis data of its associated branches, and the internal records of the medical alliance related to it, including reported and unreported data and reported but not processed data.

[0038] In this system, the feature extraction module includes a sentiment analysis model for the original cause data, extracts the sentiment dictionary of influencing factors related to adverse events, sets the feature influencing factors as sentiment dictionary label types, and uses them to reserve the influencing factor subject library dictionary. When the feature extraction model receives the feature extraction command again, it can extract the factors again by matching all dictionary type labels in the dictionary library. The secondary extraction is performed by calculating the dictionary weights and performing reasonable extraction. The influencing factor dictionary can be discarded if it is below a certain occupancy rate.

[0039] The classification model generation module is used to generate an adverse event influencing factor analysis model corresponding to the model data set unit. The model generator unit is for each adverse event cause type; the main process includes a data set division unit and a model selection unit:

[0040] The dataset training unit, or more precisely, the dataset division level, is a preparatory condition for model training based on the set ratio. The ratio of the training set to the test set is not unique, and its main setting is selected based on the historical model evaluation results.

[0041] The model generator unit generates the adverse event influencing factor analysis model of the current subject library based on different algorithms for the data of the training set. The algorithms used include but are not limited to decision tree algorithm, logistic regression algorithm, neural network algorithm. The algorithm selection interface mainly comes from the algorithm types used in individual literature and the mainstream algorithm library of classification models.

[0042] The model generator unit is also used to analyze the system model results according to the adverse event risk influencing factors, using the effectiveness verification dimensions and their importance levels in order of average level accuracy Significance level And the degree of fit The model influence weights w1, w2, and w3 are set to 0.5, 0.3, and 0.2 respectively, and the quantitative standard N for model evaluation selection is:

[0043]

[0044] The value range of N is between 0 and 1. N is used to evaluate the accuracy of the model in data training and testing. A high N value indicates that the accuracy of the various factor analysis results obtained by the model test is high, indicating that it has a certain practical reference significance. Otherwise, it does not. By comparing iterative calculations, the model results of the adverse event influencing factor analysis algorithm that reach the expected value are evaluated.

[0045] Figure 2 The flowchart of the adverse event influencing factor method provided by the embodiment of the present invention is to build an adverse event influencing factor analysis system on the medical system network side, and the specific steps are as follows:

[0046] (1) First, the data acquisition operation of the cause subject library is performed, which includes data update and data import of the data interface. The updated data content mainly includes the update of historical unobtained data and real-time updated data. The historical data includes the data statistics of the hospital, the clinical diagnosis data related to its associated branches, and the internal records of the medical alliance related to it, including the reported and unreported data and the reported but not processed data.

[0047] (2) Processing the acquired data, including data preprocessing, data control management and data standardization.

[0048] a) Data preprocessing includes missing value processing, discrete point processing and data normalization of the original clinical data. It mainly "cleans" the data by filling missing values, smoothing noisy data, smoothing or deleting outliers, and resolving data inconsistencies.

[0049] b) Data control management mainly involves the data quality assessment process for clinical data. The measurement dimensions are divided into standard assessment and whether the data indicators meet the quality control standards of the National Health Commission. Secondary supervision and judgment of data quality is carried out.

[0050] c) Finally, the data on the causes of adverse events managed through data allegations are normalized, including reasonable classification of event types and risk classification.

[0051] (3) Extract the principal components of clinical medical order data, including sentiment analysis and weight calculation, where the weight calculation formula is:

[0052]

[0053] X1, X2, ..., X n are the influencing factors of specific adverse events, i.e., independent variables, such as date, temperature, place of production, etc.; W1, W2, ..., W n It represents the average proportion of each factor.

[0054] (4) Model training

[0055] If the proportion of the training set is relatively large, the trained model may be closer to the model trained with D. At the same time, T is small and the evaluation result is not accurate enough. If the proportion of T is relatively large, the evaluated model may be significantly different from the previous one, thereby reducing the fidelity of the evaluation. Through repeated experiments, the applicant divided the processed clinical data set into a training set: test set = 7:3. After selecting the type of model, the system accepts the import of data, trains the model for the training set, and outputs the relevance weight ranking results and event types of the influencing factors. The system also provides visualization tables and feasible improvement suggestions for the results.

[0056] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make some modifications and improvements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be based on the definition of the claims.

Claims

1. A system for analyzing factors affecting adverse events, characterized in that: include: The event cause subject library generation module is used to identify and extract the input textual data of the causes of various clinical diagnosis events; The principal component feature extractor is used to analyze the submitted textual cause data and output the relevant feature cause results; Calculate the weight probability of the feature results and extract keywords; Data quality control and standardization module, used to perform quality control and standardization on clinical data under each subject library after association; The impact factor extraction module is used to analyze clinical data results, extract relevant impact factor attributes, and screen them through word frequency-related weight calculation; The classification model generation module is used to generate an analysis model of adverse event influencing factors extracted from the subject library of each event cause; A model selector is used to obtain the corresponding clinical theme and the corresponding adverse event influencing factor analysis model according to the matching of the newly input clinical data, obtain the weight ratio and ranking of the adverse event influencing factors, and send them to the comprehensive analysis module of event influencing factors; The event factor comprehensive analysis module is used to receive an adverse event influencing factor analysis request carrying clinical data of a newly added patient, send the patient's clinical data to the model selector, and receive feature extraction information of adverse event influencing factors corresponding to the newly added adverse event returned by the model selector, obtain the weight calculation result corresponding to the influencing factor corresponding to the newly added adverse event, and send it in the response to obtain the adverse event influencing factor analysis; The principal component feature extractor comprises: The impact factor active extraction module is used to extract the impact characteristic factors after screening and analysis in the subject library in a textual manner; The attribute weight calculation unit is used to calculate the correlation coefficient of the risk influencing factor text and classify the safety risks of adverse events; The text natural language processing unit is used to label the acquired patient data, compare the data set with the training of the historical model to identify factors, and realize AI intelligent text classification; The classification model generation module includes: The data set training unit is used to obtain the safety risk level and type of the adverse event cause data belonging to the subject library; A model generator unit, used to generate an adverse event risk influencing factor analysis model corresponding to the model data set training unit, and also used to perform data labeling processing and then divide the acquired data set into a training set and a test set, and use a logistic regression algorithm, a convolutional neural network algorithm, and a decision tree algorithm to train and verify the model according to the specific distribution of the data set, and evaluate the model accuracy based on the verification result of the test set, and make a decision on the selection of the model by comparing the training results of different algorithm models; The model generator unit is also used to analyze the system model results according to the adverse event risk influencing factors, using the effectiveness verification dimensions and their importance levels in order of average level accuracy , significance level And the degree of fit , model influence weight , , If they are set to 0.5, 0.3, and 0.2 respectively, the quantitative standard N selected for model evaluation is: Through comparative iterative calculations, the adverse event influencing factor analysis algorithm model results that reach the expected value are evaluated.

2. The adverse event risk influencing factor system according to claim 1, characterized in that: The textual data in the event cause theme library generation module includes: the classification of adverse event risk categories and risk value interval information corresponding to the patient, the doctor's medical advice information data for the patient, the clinical pathway of the medical record, the time of occurrence and the reporting time information.

3. The adverse event risk influencing factor analysis system according to claim 1, characterized in that: The data quality control and standardization processing module includes: A data quality control unit, used to screen and clean the clinical data associated with the subject database for abnormal data, and to mark the related abnormal data; The data standardization unit is used to perform data vectorization processing and standardization on the clinical data set associated in the subject library, and includes the following submodules: A data binning module is used to adaptively bin the associated clinical data set according to the set risk type of adverse events to be predicted; The data vectorization submodule is used to vectorize the text data of the adverse event type and event cause of the patient, and output the quantitative data type that can be processed by the computer; The abnormal data processing submodule is used to perform relevant data preprocessing operations on the abnormal data marked by the data quality control unit, including screening of abnormal data and elimination of abnormal values, filling in the blanks of abnormal values, performing data attribute statistics on the generated abnormal data, and generating tables.

4. The adverse event risk influencing factor analysis system according to claim 1, characterized in that: The comprehensive adverse event risk influencing factor analysis module includes: A single adverse event influencing factor analysis unit is used to receive the newly added specific adverse event type occurrence cause data information and adverse event related influencing factor information sent by the model selector, and perform intelligent analysis on the adverse event cause data, respectively obtain the specific influencing factor weight calculation results and influencing factor ranking information of the newly added detailed adverse event, and perform visual display; The comprehensive adverse event influencing factor analysis unit is used to identify the types of newly received adverse events and then analyze and process their causes, ultimately obtaining specific information on the identified adverse event types and the weight information of each influencing factor, and visually ranking and displaying the weight calculation results.

5. A method for analyzing factors affecting the causes of adverse events, characterized in that: On the medical system network side, a system for analyzing the causes and influencing factors of adverse events as described in any one of claims 1 to 4 is established, and the method comprises the following steps: First, obtain the cause theme library data, including data update and data import of the data interface. The updated data content mainly includes the update history data not obtained and the real-time update data; Carry out data preprocessing, data control management and data standardization on the acquired data; Extract principal components and calculate weights of clinical doctor order data; The processed clinical order data set is divided into training set and test set = 7:

3. After selecting the model type, the data is imported, the model is trained on the training set, and the correlation weight ranking results and event types of the influencing factors are output. Visual tables and graphs and feasible improvement suggestions are given for the results.

6. The method for analyzing the causes and influencing factors of adverse events according to claim 5, characterized in that: The imported and updated data include the hospital's data statistics, clinical diagnosis data related to its affiliated branches, and internal records of the medical alliance related to it.

7. The method for analyzing the causes and influencing factors of adverse events according to claim 5, characterized in that: The data preprocessing includes processing missing values, discrete points and normalization of the original clinical data; mainly solving the inconsistency of the data by filling missing values, smoothing noise data, and smoothing or deleting outliers; The data accusation management is mainly a data quality assessment process for clinical data. The measurement dimensions are divided into standard assessment and whether the data indicators meet the quality control standards of the National Health Commission. The data on the causes of adverse events managed by data accusation are normalized, including reasonable classification of event types and risk classification. The calculation formula for the weight of clinical doctor order data is: 。