A data analysis method and system based on drug vigilance system

Abstract

The present application relates to the field of data processing technology, and in particular to a data analysis method and system based on a drug vigilance system. The method includes: obtaining multiple drug vigilance data sets of the same drug produced by multiple manufacturers; determining the first parameter value of the candidate data point in the target drug vigilance data set; determining the second parameter value of the candidate data point; using the inverse of the product of the second parameter value of the candidate data point and the first parameter value as the weight value of the candidate data point, and using the weight value of the candidate data point and the outlier local factor LOF algorithm to determine the local deviation of the candidate data point; using the candidate data point whose local deviation is greater than a preset deviation threshold as the target data point corresponding to the atypical reaction of the drug. The above technical solution can help monitoring personnel more easily discover possible atypical reactions of drugs.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data analysis method and system based on a drug vigilance system. Background Art

[0002] Pharmacovigilance is the process of preventing a drug from posing a threat to patients throughout its life cycle. Pharmacovigilance involves safety monitoring of drugs before, after, and during their use. Due to the limitations of clinical trials in terms of sample size, study duration, and observation time, the adverse reactions and potential risks of some drugs may not be discovered until after they are marketed. Therefore, pharmacovigilance is crucial to ensuring the safety of drug use by the public.

[0003] In order to realize the processing of drug vigilance related data in the drug vigilance process, in the relevant technology, for example, the Chinese patent application document with publication number CN111177516A, a method for a drug vigilance system to process feedback data is provided, including: logging in to the adverse reaction center database according to a pre-configured account and password; identifying updated feedback data reports in the adverse reaction center database, and downloading the updated feedback data reports; importing the updated feedback data reports into the drug vigilance system according to a predefined mapping relationship, and generating a corresponding adverse reaction report; retrieving the MedDRA vocabulary to encode one or more items in the adverse reaction report; judging and entering the knownness of the adverse reaction events in the adverse reaction report according to the pre-configured adverse reaction list; judging and entering the severity of the adverse reaction events in the adverse reaction report according to the pre-configured important medical event list; and submitting the adverse reaction report to the adverse reaction center direct reporting system.

[0004] Some atypical reactions may occur when drugs are used. Atypical reactions refer to adverse reactions that are rare or inconsistent with the known mechanism of action of the drug during the use of the drug. In the relevant technology, known adverse reaction events are obtained from the adverse reaction center database. If the adverse reaction center database does not include atypical reactions of the drug, it is difficult to discover the atypical reactions of the drug. Therefore, it is difficult to determine the possible atypical reactions of the drugs monitored in the drug vigilance system. Summary of the invention

[0005] In order to overcome the problem in the related art that it is difficult to determine possible atypical reactions of drugs monitored in a pharmacovigilance system, the present application provides a data analysis method and system based on a pharmacovigilance system.

[0006] According to a first aspect of an embodiment of the present application, a data analysis method based on a drug vigilance system is provided, comprising: obtaining multiple drug vigilance data sets of the same drug produced by multiple manufacturers, the same drug vigilance data set including multiple data points of a target drug produced by the same manufacturer, and the same data point including multiple sub-data corresponding to multiple adverse reaction monitoring dimensions; determining a first parameter value of the candidate data point for a candidate data point of a target drug vigilance data set in the multiple drug vigilance data sets; the first parameter value is used to characterize the degree of similarity between the candidate data point and data points in other drug vigilance data sets other than the target drug vigilance data set; determining a second parameter value of the candidate data point, the second parameter value is used to characterize the degree of difference between the candidate data point and other data points in the target drug vigilance data set; using the inverse of the product of the second parameter value of the candidate data point and the first parameter value as a weight value of the candidate data point, and using the weight value of the candidate data point and an outlier local factor (LOF) algorithm to determine the local deviation of the candidate data point; the local deviation is used to characterize the degree of isolation of the data point; and using the candidate data point whose local deviation is greater than a preset deviation threshold as a target data point corresponding to the atypical reaction of the drug.

[0007] In this way, multiple drug vigilance data sets for the same drug from multiple manufacturers are obtained, the same drug vigilance data set includes multiple data points of the target drug produced by the same manufacturer, and the same data point includes multiple sub-data corresponding to multiple adverse reaction monitoring dimensions. By determining the local deviation of the candidate data points in the target drug vigilance data set, the candidate data points whose local deviation is greater than the preset deviation threshold have a higher probability of having atypical reactions, which can help monitoring personnel more easily discover possible atypical reactions of the drug.

[0008] At the same time, since the local deviation of the candidate data point is determined based on the weight value of the candidate data point, the weight value of the candidate data point is obtained by comparing the candidate data point with other data points in the target drug vigilance data set where the candidate data point is located, and comparing the candidate data point with other data points in other drug vigilance data sets, these data points are obtained by monitoring the medication performance of the same drug produced by multiple manufacturers, and can more comprehensively evaluate the actual performance of the candidate data points, thereby screening out more accurate atypical reactions of the drug, and multiple drug vigilance data sets correspond to the same drug produced by multiple manufacturers, which can avoid the limitations of the performance of drugs produced by a single manufacturer, thereby avoiding the omission of atypical reactions of the drug.

[0009] Optionally, the plurality of pharmacovigilance data include the same number of data points, and the first parameter value of the candidate data point is determined in the following manner: , F is the first parameter value of the candidate data point, N is the number of pharmacovigilance data sets, norm is the normalization function, is the similarity value between the target pharmacovigilance dataset and the ath other pharmacovigilance dataset, L is the number of adverse reaction monitoring dimensions, M is the number of data points in the pharmacovigilance dataset, is the sub-data of the candidate data point in the bth adverse reaction monitoring dimension, is the sub-data of the i-th data point in the a-th other pharmacovigilance dataset in the b-th adverse reaction monitoring dimension, To take the absolute value.

[0010] In this way, the median of the sub-data of the data points in the target drug vigilance dataset in the same adverse reaction monitoring dimension is compared with the median of the sub-data of the data points in other drug vigilance datasets in the same adverse reaction monitoring dimension. The similarity value obtained can better characterize the similarity between the target drug vigilance dataset and the other drug vigilance datasets.

[0011] Optionally, the similarity value between the target pharmacovigilance dataset and other pharmacovigilance datasets is determined by: , is the similarity value between the target pharmacovigilance dataset and the ath other pharmacovigilance dataset, L is the number of adverse reaction monitoring dimensions, norm is the normalization function, is the median of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the median of the sub-data of the data point in the ath other pharmacovigilance dataset in the bth adverse reaction monitoring dimension.

[0012] In this way, the median of the sub-data of the data points in the target drug vigilance dataset in the same adverse reaction monitoring dimension is compared with the median of the sub-data of the data points in other drug vigilance datasets in the same adverse reaction monitoring dimension. The similarity value obtained can better characterize the similarity between the target drug vigilance dataset and the other drug vigilance datasets.

[0013] Optionally, the second parameter value of the candidate data point is determined by: , where D is the second parameter value of the candidate data point, L is the number of adverse reaction monitoring dimensions, is the normalization function, is the median of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the average value of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the sub-data of the candidate data point in the bth adverse reaction monitoring dimension.

[0014] In this way, the sub-data of the candidate data point in the adverse reaction monitoring dimension are compared with the median of the sub-data of the data point in the target drug vigilance dataset in the adverse reaction monitoring dimension, and the median and average of the sub-data of the data point in the target drug vigilance dataset in the adverse reaction monitoring dimension are compared. The second parameter value can better characterize the degree of difference between the candidate data point and other data points in the target drug vigilance dataset, and can characterize the probability of the existence of extreme data points in the sub-data of the adverse reaction monitoring dimension of the target drug vigilance dataset, so as to better determine the data points corresponding to atypical reactions in the drug vigilance dataset.

[0015] Optionally, the local deviation of the candidate data point is determined using the weight value of the candidate data point and the outlier local factor (LOF) algorithm, including: mapping the data points in multiple drug vigilance data sets to a multidimensional data space; different dimensions of the multidimensional data space correspond to multiple adverse reaction monitoring dimensions; using the outlier local factor (LOF) algorithm to determine the initial local reachable density value of the candidate data point in the multidimensional data space; using the product of the initial local reachable density value of the candidate data point and the weight value as the target local reachable density value of the candidate data point; determining the local deviation of the candidate data point based on the target local reachable density value of the candidate data point and other data points in the neighborhood of the candidate data point in the multidimensional data space.

[0016] Optionally, the initial local reachable density value of the candidate data point in the multidimensional data space is determined by: determining the reachable distance of the data point in the drug vigilance data set to other data points in the K-neighborhood in the multidimensional data space; the K-neighborhood includes K other data points closest to the data point; and taking the reciprocal of the average of the reachable distances of the candidate data point to other data points in the neighborhood in the multidimensional data space as the initial local reachable density value of the candidate data point in the multidimensional data space.

[0017] In this way, by determining the reachable distance of a data point in the drug vigilance data set to other data points in the K-neighborhood in the multidimensional data space, the greater the reachable distance of a data point to other data points in the K-neighborhood in the multidimensional data space, the lower the degree of aggregation of the data point with other data points; the reciprocal of the average value of the reachable distance of the candidate data point to other data points in the neighborhood in the multidimensional data space is used as the initial local reachable density value of the candidate data point in the multidimensional data space, and the initial local reachable density value can be used to better characterize the degree of aggregation of all data points within the K-neighborhood of the data point.

[0018] Optionally, the local deviation of the candidate data point is determined by: , W is the local deviation of the candidate data point, Q is the number of other data points in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the fth other data point in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the candidate data point.

[0019] Optionally, the method also includes: when the number of target data points corresponding to the atypical reactions of the drug is greater than a preset number threshold, sorting all target data points according to specified priority orders in multiple adverse reaction monitoring dimensions; and sending the sorting results and the manufacturer information of the target data points to a display device for display.

[0020] In this way, the sorting results and the producer information of the target data points are sent to the display device for display, which helps relevant personnel to further investigate the causes of atypical drug reactions based on the sorting results and the producer information of the target data points, so as to avoid atypical reactions.

[0021] Optionally, adverse reaction monitoring dimensions include: age, gender, single dosage, frequency of medication, cumulative duration of medication, adverse reaction symptoms, adverse reaction frequency, and cumulative duration of adverse reactions.

[0022] According to a second aspect of an embodiment of the present application, a data analysis system based on a drug vigilance system is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the steps of the data analysis method based on a drug vigilance system provided in the first aspect of the present application are implemented.

[0023] The technical solution provided by the embodiments of the present application may include the following beneficial effects: obtaining multiple drug vigilance data sets for the same drug from multiple manufacturers, the same drug vigilance data set includes multiple data points of the target drug produced by the same manufacturer, the same data point includes multiple sub-data corresponding to multiple adverse reaction monitoring dimensions, and by determining the local deviation of the candidate data points in the target drug vigilance data set, the candidate data points whose local deviation is greater than a preset deviation threshold have a higher probability of having atypical reactions, which can help monitoring personnel more easily discover possible atypical reactions of the drug.

[0024] At the same time, since the local deviation of the candidate data point is determined based on the weight value of the candidate data point, the weight value of the candidate data point is obtained by comparing the candidate data point with other data points in the target drug vigilance data set where the candidate data point is located, and comparing the candidate data point with other data points in other drug vigilance data sets, these data points are obtained by monitoring the medication performance of the same drug produced by multiple manufacturers, and can more comprehensively evaluate the actual performance of the candidate data points, thereby screening out more accurate atypical reactions of the drug, and multiple drug vigilance data sets correspond to the same drug produced by multiple manufacturers, which can avoid the limitations of the performance of drugs produced by a single manufacturer, thereby avoiding the omission of atypical reactions of the drug.

[0025] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0027] Figure 1 is a flow chart of a data analysis method based on a pharmacovigilance system according to an exemplary embodiment;

[0028] Figure 2 It is a structural schematic diagram of a data analysis system based on a drug vigilance system according to an exemplary embodiment. DETAILED DESCRIPTION

[0029] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0030] First, a brief introduction is given to the application scenarios of the embodiments of the present application. In the application scenarios of the present application, the performance of the drug can be monitored throughout its life cycle, such as clinical testing of the drug's performance, or obtaining drug usage reports reported by users. The drug usage reports may include adverse reactions of users when using the drugs, so that the drug vigilance system can perform drug vigilance on the drugs. For example, the corresponding adverse reaction information can be supplemented or modified in the instructions for subsequent batches of drugs, or relevant departments and the public can be notified by issuing relevant announcements so that drug users are aware of possible adverse reactions of the drugs.

[0031] At present, when conducting drug vigilance, the drug vigilance system mainly relies on obtaining adverse reaction information from the adverse reaction center database. When the adverse reaction center database does not include the atypical reactions of the drug, it is difficult to know the atypical reactions of the drug in a timely manner, which makes it inconvenient for the drug vigilance system to timely know the atypical reactions of the drug after analyzing the drug vigilance data, and it is also not convenient to share information on the atypical reactions of the drug.

[0032] It should be noted that all actions of acquiring signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where they are located, and with the authorization given by the owner of the corresponding device.

[0033] In view of the above technical problems, the present application provides a data analysis method and system based on a pharmacovigilance system. Figure 1 is a flow chart of a data analysis method based on a pharmacovigilance system according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps.

[0034] In step S101 , multiple pharmacovigilance data sets for the same drug from multiple manufacturers are obtained.

[0035] The same pharmacovigilance data set includes multiple data points of target drugs produced by the same manufacturer, and the same data point includes multiple sub-data corresponding to multiple adverse reaction monitoring dimensions; in order to facilitate comparison of data points in different pharmacovigilance data sets, multiple pharmacovigilance data sets may include the same number of data points; for example, multiple pharmacovigilance data sets may each include 500 data points.

[0036] The same data point may include multiple sub-data under the adverse reaction monitoring dimension, such as the user's age, gender, single medication dose, medication frequency, cumulative medication duration, adverse reaction symptoms, adverse reaction frequency, and cumulative adverse reaction duration. Different sub-data correspond to different adverse reaction monitoring dimensions, so as to comprehensively evaluate the performance of the drug.

[0037] Multiple drug vigilance data sets can be obtained by testing the same drug produced by multiple manufacturers. For example, when testing the same drug, information such as the test subject's age, gender, single dosage, frequency of dosage, cumulative duration of dosage, adverse reaction symptoms, adverse reaction frequency, and cumulative duration of adverse reactions can be recorded to determine the performance of the drug in different test subjects, or determine the performance of the drug at different dosages of the test subject, or determine the adverse reactions that may occur in the test subject under the same medication situation.

[0038] The possible frequency and cumulative duration of adverse reactions may vary depending on age, gender, single dosage, frequency of use, and cumulative duration of use. In order to more accurately determine the possible atypical reactions of drugs, the diverse data in the drug vigilance system can be analyzed to determine the possible scenarios in which atypical reactions may occur, so as to avoid more people using drugs in scenarios where atypical reactions may occur, thereby avoiding more people from suffering losses caused by atypical reactions.

[0039] Alternatively, multiple pharmacovigilance data sets can also be obtained from multiple drug usage reports. For example, users who take the drugs or other users can upload performance information after taking the drugs to the pharmacovigilance system. The pharmacovigilance system can generate a drug usage report based on the performance information after taking the drugs, so as to obtain multiple pharmacovigilance data sets based on these drug trial reports. Different pharmacovigilance data sets correspond to different manufacturers of the same drug.

[0040] In step S102 , for a candidate data point of a target pharmacovigilance dataset among a plurality of pharmacovigilance datasets, a first parameter value of the candidate data point is determined.

[0041] The target pharmacovigilance dataset may be any one of the multiple pharmacovigilance datasets, and the candidate data point may be any one of the target pharmacovigilance datasets; the first parameter value is used to characterize the degree of similarity between the candidate data point and data points in other pharmacovigilance datasets except the target pharmacovigilance dataset.

[0042] The same drug produced by different manufacturers has a high degree of consistency. In order to avoid the limitations of drug vigilance for drugs produced by a single manufacturer and to increase the diversity of drug vigilance samples, the medication feedback of the same drug produced by multiple manufacturers can be analyzed to determine possible atypical reactions to the drug.

[0043] In order to determine the performance of candidate data points in the target pharmacovigilance dataset, the candidate data points can be compared with data points in other pharmacovigilance datasets except the target pharmacovigilance dataset. Since different pharmacovigilance datasets correspond to different manufacturers, the medication performance of the drug produced by the manufacturer corresponding to the candidate data point can be compared with the same drug produced by other manufacturers.

[0044] While drugs produced by different manufacturers have a high degree of consistency, there may also be slight differences in production processes, production raw materials, and ingredients other than the main ingredients. These slight differences may cause differences in the medication performance of the drugs. Comparing the medication performance of the drug produced by the manufacturer corresponding to the candidate data point with the same drug produced by other manufacturers will help to more accurately determine whether the candidate data point is a data point with an atypical reaction.

[0045] In one embodiment, the first parameter value of the candidate data point is determined in the following manner: , F is the first parameter value of the candidate data point, N is the number of pharmacovigilance data sets, norm is the normalization function, is the similarity value between the target pharmacovigilance dataset and the ath other pharmacovigilance dataset, L is the number of adverse reaction monitoring dimensions, M is the number of data points in the pharmacovigilance dataset, is the sub-data of the candidate data point in the bth adverse reaction monitoring dimension, is the sub-data of the i-th data point in the a-th other pharmacovigilance dataset in the b-th adverse reaction monitoring dimension, To take the absolute value.

[0046] The similarity value between the target drug vigilance dataset and other drug vigilance datasets can characterize the overall similarity in adverse reactions between the drugs produced by the manufacturer corresponding to the target drug vigilance dataset and the same drugs produced by the manufacturers corresponding to other target drug vigilance datasets; the greater the similarity value between two different drug vigilance datasets, the greater the similarity of the adverse reactions of the same drug produced by the two manufacturers, or the closer the ages of patients who experience adverse reactions after taking the same drug produced by the two manufacturers.

[0047] In this way, by comparing the sub-data of different adverse reaction monitoring dimensions in the candidate data points with the sub-data of different adverse reaction monitoring dimensions of the data points in other drug vigilance data sets, the adverse reaction manifestations corresponding to the candidate data points can be compared with the adverse reaction manifestations of the data points in other drug vigilance data sets, so as to determine the atypical reactions of the drug.

[0048] The determined first parameter value of the candidate data point can characterize the degree of similarity between the performance of the candidate data point in the adverse reaction monitoring dimension and the performance of the data points in other drug vigilance data sets in the adverse reaction monitoring dimension; the smaller the first parameter value of the candidate data point, the greater the difference between the candidate data set and the data points in other drug vigilance data sets in the adverse reaction monitoring dimension; conversely, the larger the first parameter value of the candidate data point, the smaller the difference between the candidate data set and the data points in other drug vigilance data sets in the adverse reaction monitoring dimension.

[0049] While drugs produced by different manufacturers have a high degree of consistency, there may also be slight differences in production processes, raw materials and excipients. These slight differences may cause differences in adverse reaction manifestations of the same drug from different manufacturers. When different drug vigilance data sets are analyzed in the same multidimensional data space, the data points corresponding to atypical reactions from two different drug vigilance data sets may overlap or cluster, which may lead to the misjudgment of atypical reactions as conventional adverse reactions.

[0050] For example, when two clinical patients of the same age and gender take the same drug from different manufacturers at the same single dose, frequency of use and cumulative duration of use, the adverse reactions, frequency of adverse reactions and cumulative duration of adverse reactions may be the same. Since the two patients who took the same drug from different manufacturers experienced the same medication reactions, they might mistakenly believe that the adverse reactions experienced by these two patients were merely routine adverse reactions, thereby missing out on atypical reactions to the drug.

[0051] For example, when three clinical patients of the same age and gender take the same drug from different manufacturers with similar single dosage, frequency of use and cumulative duration of use, the adverse reactions, frequency of adverse reactions and cumulative duration of adverse reactions may be similar. Since the three patients who took the same drug from different manufacturers had similar medication reactions, they might mistakenly believe that the adverse reactions of these three patients were merely routine adverse reactions, thus causing the omission of atypical reactions to the drug.

[0052] When analyzing data points in multiple pharmacovigilance data sets using the same multidimensional data space, since data points from different pharmacovigilance data sets may overlap or cluster, the first parameter value of the candidate data point can characterize the degree of similarity between the candidate data point and data points in other pharmacovigilance data sets other than the target pharmacovigilance data set. The first parameter value of the candidate data point helps to pay more attention to data points from different pharmacovigilance data sets with similar performance, that is, the first parameter value of the candidate data point helps to pay more attention to test subjects who have taken the same drug from different manufacturers and have similar adverse reaction performance, so as to determine atypical reactions of the drug.

[0053] In one embodiment, the similarity value between the target pharmacovigilance dataset and other pharmacovigilance datasets is determined in the following manner: , is the similarity value between the target pharmacovigilance dataset and the ath other pharmacovigilance dataset, L is the number of adverse reaction monitoring dimensions, norm is the normalization function, is the median of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the median of the sub-data of the data point in the ath other pharmacovigilance dataset in the bth adverse reaction monitoring dimension.

[0054] The median of the sub-data of the data points in the pharmacovigilance dataset in the same adverse reaction monitoring dimension can better reflect the general level of the sub-data of the data points in the pharmacovigilance dataset in the same adverse reaction monitoring dimension. Compared with the average value of the sub-data of the data points in the pharmacovigilance dataset in the same adverse reaction monitoring dimension, the median of the sub-data of the data points in the pharmacovigilance dataset in the same adverse reaction monitoring dimension can better reflect the characteristics of the pharmacovigilance dataset, so as to compare the characteristics of the target pharmacovigilance dataset with other pharmacovigilance datasets.

[0055] The normalization function can be used to The absolute value of is normalized to the range of 0 to 1 to facilitate comparison between numerical values, thereby characterizing the similarity between the target pharmacovigilance dataset and other pharmacovigilance datasets through the similarity value.

[0056] In this way, the median of the sub-data of the data points in the target drug vigilance dataset in the same adverse reaction monitoring dimension is compared with the median of the sub-data of the data points in other drug vigilance datasets in the same adverse reaction monitoring dimension. The similarity value obtained can better characterize the similarity between the target drug vigilance dataset and the other drug vigilance datasets.

[0057] In step S103, the second parameter value of the candidate data point is determined.

[0058] The second parameter value is used to characterize the degree of difference between the candidate data point and other data points in the target pharmacovigilance dataset.

[0059] Compared with the first parameter value of the candidate data point, the second parameter value of the candidate data point can determine the characteristics of the candidate data point from another dimension; wherein, the first parameter value of the candidate data point is used to characterize the degree of similarity between the candidate data point and other drug vigilance data sets except the target drug vigilance data set; the second parameter value of the candidate data point is used to characterize the degree of difference between the candidate data point and other data points within the target drug vigilance data set.

[0060] The larger the second parameter value of the candidate data point, the greater the difference between the candidate data point and other data points in the target drug vigilance data set, and the candidate data point has a higher probability of being a data point generated by an atypical reaction; conversely, the smaller the second parameter value of the candidate data point, the smaller the difference between the candidate data point and other data points in the target drug vigilance data set, and the candidate data point has a lower probability of being a data point generated by an atypical reaction.

[0061] In one embodiment, the second parameter value of the candidate data point is determined in the following manner: , where D is the second parameter value of the candidate data point, L is the number of adverse reaction monitoring dimensions, is the normalization function, is the median of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the average value of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the sub-data of the candidate data point in the bth adverse reaction monitoring dimension.

[0062] The median of the sub-data of the data points in the drug vigilance dataset in the same adverse reaction monitoring dimension can better reflect the general level of the sub-data of the data points in the drug vigilance dataset in the same adverse reaction monitoring dimension; comparing the candidate data points with the median of the sub-data of the data points in the target drug vigilance dataset in the same adverse reaction monitoring dimension can reflect the overall difference between the candidate data and the target drug vigilance dataset.

[0063] Comparing the median of the sub-data of the data points in the target pharmacovigilance data set in the adverse reaction monitoring dimension with the average of the sub-data of the data points in the target pharmacovigilance data set in the adverse reaction monitoring dimension can reflect the symmetry of the distribution of the sub-data of the data points in the target pharmacovigilance data set in the adverse reaction monitoring dimension; The smaller the absolute value of , the higher the symmetry of the distribution of data points in the target drug vigilance dataset in the sub-data of the bth adverse reaction monitoring dimension, and the lower the probability that there are extreme data points in the sub-data of the bth adverse reaction monitoring dimension of the target drug vigilance dataset.

[0064] on the contrary, The larger the absolute value of is, the lower the symmetry of the distribution of the data points in the target drug vigilance dataset in the sub-data of the bth adverse reaction monitoring dimension is, and the higher the probability that there are extreme data points in the sub-data of the bth adverse reaction monitoring dimension of the target drug vigilance dataset is. Therefore, the second parameter value of the candidate data point can also reflect the distribution of the data points in the target drug vigilance dataset.

[0065] In this way, the sub-data of the candidate data point in the adverse reaction monitoring dimension are compared with the median of the sub-data of the data point in the target drug vigilance dataset in the adverse reaction monitoring dimension, and the median and average of the sub-data of the data point in the target drug vigilance dataset in the adverse reaction monitoring dimension are compared. The second parameter value can better characterize the degree of difference between the candidate data point and other data points in the target drug vigilance dataset, and can characterize the probability of the existence of extreme data points in the sub-data of the adverse reaction monitoring dimension of the target drug vigilance dataset, so as to better determine the data points corresponding to atypical reactions in the drug vigilance dataset.

[0066] In step S104, the inverse of the product of the second parameter value and the first parameter value of the candidate data point is used as the weight value of the candidate data point, and the weight value of the candidate data point and the outlier local factor LOF algorithm are used to determine the local deviation of the candidate data point in the target drug vigilance data set.

[0067] The local deviation is used to characterize the degree of isolation of data points. The higher the degree of isolation of data points, for example, the more unique the adverse reaction symptoms that patients experience after taking the drug, or the more unique the age of the patients who experience adverse reaction symptoms after taking the drug, or the more unique the frequency of adverse reactions that patients experience after taking the drug, or the more unique the cumulative duration of adverse reactions that patients experience after taking the drug.

[0068] The data points corresponding to these more unique manifestations are more likely to be caused by atypical drug reactions. Therefore, determining the local deviation of candidate data points in the target drug vigilance data set will help to more comprehensively discover possible atypical drug reactions.

[0069] Here, the smaller the weight value of the candidate data point is, the higher the level of attention the candidate data point requires, and the greater the increase in the local deviation of the candidate data point caused by the weight value corresponding to the candidate data point, so that the candidate data points requiring higher levels of attention can be more easily screened out from the target drug vigilance data set.

[0070] In one embodiment, the weight value of the candidate data point and the outlier local factor (LOF) algorithm are used to determine the local deviation of the candidate data point, including: mapping the data points in multiple drug vigilance data sets to a multidimensional data space; different dimensions of the multidimensional data space correspond to multiple adverse reaction monitoring dimensions; using the outlier local factor (LOF) algorithm to determine the initial local reachable density value of the candidate data point in the multidimensional data space; using the product of the initial local reachable density value of the candidate data point and the weight value as the target local reachable density value of the candidate data point; and determining the local deviation of the candidate data point according to the target local reachable density value of the candidate data point and other data points in the neighborhood of the candidate data point in the multidimensional data space.

[0071] By mapping data points in multiple pharmacovigilance data sets into a multidimensional data space, it is possible to use the multidimensional data space to integrate pharmacovigilance data from different manufacturers, thereby increasing the diversity of pharmacovigilance monitoring samples for the same drug, and helping to more comprehensively implement pharmacovigilance for the same drug, so as to more comprehensively discover atypical reactions to the drug.

[0072] The LOF (Local Outlier Factor) algorithm is an unsupervised learning algorithm used to identify outliers in a data set. For any data point in a data set, the 5 other data points closest to the data point can be selected based on the distances to the data point from a specified number of other data points closest to the data point. For example, the specified number can be 5. The closer the distance between two data points, the closer the performance of the two data points. The more other data points close to a data point, the more likely the data point is to be a common data point in the data set, and the lower the probability that the data point is a rare data point in the data set.

[0073] In a multidimensional data space, since different dimensions of the multidimensional data space correspond to different adverse reaction monitoring dimensions, data points located in the same local range in the multidimensional data space have high similarity in at least one adverse reaction monitoring dimension. For example, the frequency, symptoms and duration of adverse reactions are relatively close. Since these relatively similar data points are more common, the probability that the relatively similar data points belong to atypical reactions is lower.

[0074] The initial local reachable density value of a data point in a multidimensional data space is used to characterize the degree of aggregation of other data points within the K-neighborhood of the data point; the larger the initial local reachable density value of a data point, the greater the degree of aggregation of other data points within the K-neighborhood of the data point, and the data point has a lower probability of corresponding to an atypical drug reaction; conversely, the smaller the initial local reachable density value of a data point, the smaller the degree of aggregation of other data points within the K-neighborhood of the data point, and the data point has a higher probability of corresponding to an atypical drug reaction.

[0075] The product of the initial local reachable density value and the weight value of the candidate data point is used as the target local reachable density value of the candidate data point. Since the weight value of the candidate data point is equal to the inverse of the product of the first parameter value and the second parameter value of the candidate data point, the weight value of the candidate data point is negatively correlated with the degree of attention required for the candidate data point, so that candidate data points that require a higher degree of attention can be assigned a lower target local reachable density value. Since data points with lower density values ​​have a higher probability of atypical drug reactions, the target local reachable density value can better realize the screening of candidate data points that require attention.

[0076] In this way, the local deviation of the candidate data point is determined according to the target local reachable density value of the candidate data point and other data points in the neighborhood of the candidate data point in the multidimensional data space, so that more attention can be paid to the candidate data points that are clustered with different drug vigilance data sets and the candidate data points that are more different from other data points within the drug vigilance data set, so as to screen out the data points of atypical reactions to the corresponding drugs.

[0077] In one embodiment, the initial local reachable density value of the candidate data point in the multidimensional data space is determined by: determining the reachable distance of the data point in the drug vigilance data set to other data points in the K-neighborhood in the multidimensional data space; the K-neighborhood includes K other data points closest to the data point; and taking the reciprocal of the average of the reachable distances of the candidate data point to other data points in the neighborhood in the multidimensional data space as the initial local reachable density value of the candidate data point in the multidimensional data space.

[0078] When the candidate data points correspond to relatively common adverse reactions, the candidate data points are more clustered with other data points in the multidimensional data space, and these relatively common adverse reactions are more likely to be discovered by other regulatory agencies or drug manufacturers. Therefore, attention can be reduced on candidate data points that are highly clustered with other data points in the multidimensional data space.

[0079] When the candidate data points correspond to atypical reactions, the candidate data points have a lower degree of clustering with other data points in the multidimensional data space, and the probability of these atypical reactions being discovered by other regulatory agencies or drug manufacturers is lower. Therefore, attention can be paid to candidate data points with a lower degree of clustering with other data points in the multidimensional data space.

[0080] In this way, by determining the reachable distance of a data point in the drug vigilance data set to other data points in the K-neighborhood in the multidimensional data space, the greater the reachable distance of a data point to other data points in the K-neighborhood in the multidimensional data space, the lower the degree of aggregation of the data point with other data points; the reciprocal of the average value of the reachable distance of the candidate data point to other data points in the neighborhood in the multidimensional data space is used as the initial local reachable density value of the candidate data point in the multidimensional data space, and the initial local reachable density value can be used to better characterize the degree of aggregation of all data points within the K-neighborhood of the data point.

[0081] In one embodiment, the local deviation of the candidate data point is determined by: , W is the local deviation of the candidate data point, Q is the number of other data points in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the fth other data point in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the candidate data point.

[0082] In addition to comparing the distance between the candidate data point and other data points within the K-neighborhood in the multidimensional data space, the target local reachable density value of other data points within the K-neighborhood of the candidate data point can also be determined. By comparing the target local reachable density value of the candidate data point with other data points within the K-neighborhood, the degree of isolation of the candidate data point can be better characterized by the obtained local deviation, so as to characterize the probability of an atypical reaction of the candidate data point to the drug through the local deviation of the candidate data point.

[0083] In step S105, candidate data points whose local deviation is greater than a preset deviation threshold are used as target data points corresponding to atypical drug reactions.

[0084] Since the local deviation is used to characterize the degree of isolation of a data point, when the local deviation of a candidate data point is greater than the preset deviation threshold, it means that the candidate data point is relatively unique in the three-dimensional data space, and the candidate data point has a higher probability of corresponding to an atypical drug reaction. Moreover, since the data points in the three-dimensional data space correspond to multiple drug vigilance data sets of the same drug produced by multiple manufacturers, the local deviation of the candidate data point can more accurately evaluate the uniqueness of the adverse reaction manifestations after taking the drug.

[0085] Through the data analysis method based on the drug vigilance system provided in the embodiment of the present application, multiple drug vigilance data sets for the same drug from multiple manufacturers are obtained, the same drug vigilance data set includes multiple data points of the target drug produced by the same manufacturer, and the same data point includes multiple sub-data corresponding to multiple adverse reaction monitoring dimensions. By determining the local deviation of the candidate data points in the target drug vigilance data set, the candidate data points whose local deviation is greater than the preset deviation threshold have a higher probability of having atypical reactions, which can help monitoring personnel to more easily discover possible atypical reactions of the drug.

[0086] At the same time, since the local deviation of the candidate data point is determined based on the weight value of the candidate data point, the weight value of the candidate data point is obtained by comparing the candidate data point with other data points in the target drug vigilance data set where the candidate data point is located, and comparing the candidate data point with other data points in other drug vigilance data sets, these data points are obtained by monitoring the medication performance of the same drug produced by multiple manufacturers, and can more comprehensively evaluate the actual performance of the candidate data points, thereby screening out more accurate atypical reactions of the drug, and multiple drug vigilance data sets correspond to the same drug produced by multiple manufacturers, which can avoid the limitations of the performance of drugs produced by a single manufacturer, thereby avoiding the omission of atypical reactions of the drug.

[0087] In one embodiment, when the number of target data points corresponding to the atypical reactions of the drug is greater than a preset number threshold, all target data points can be sorted according to the specified priority order in multiple adverse reaction monitoring dimensions; the sorting results and the manufacturer information of the target data points are sent to a display device for display.

[0088] The specified priority can be specified by the user in advance or in real time. The specified priority is used to prioritize the various adverse reaction monitoring dimensions. For example, the order of priority from high to low can be adverse reaction frequency, cumulative duration of adverse reactions, single medication dose, medication frequency, cumulative duration of medication, age, and gender.

[0089] Compared with data points in multiple drug vigilance data sets, the target data points have a higher probability of corresponding atypical reactions to drugs. By sorting all data points according to the specified priority in multiple adverse reaction monitoring dimensions; sending the sorting results and the manufacturer information of the target data points to the display device for display, it is convenient for medical personnel to organize the sub-data in the target data points, further determine the causes of adverse reactions, or how to avoid the recurrence of atypical reactions, or perform component analysis on the same drug produced by different manufacturers.

[0090] In this way, the sorting results and the producer information of the target data points are sent to the display device for display, which helps relevant personnel to further investigate the causes of atypical drug reactions based on the sorting results and the producer information of the target data points, so as to avoid atypical reactions.

[0091] Figure 2 FIG. 1 is a schematic diagram of a data analysis system 1000 based on a drug vigilance system according to an exemplary embodiment. Figure 2 The data analysis system 1000 based on the drug vigilance system includes: a processor 1100 and a memory 1200, wherein the memory 1200 stores computer program instructions, and when the computer program instructions are executed by the processor 1100, all or part of the steps of the data analysis method based on the drug vigilance system in the present application are implemented.

[0092] It should be understood that the features of some embodiments of the various applications described herein may be combined with each other, unless specifically stated otherwise.

[0093] Although terms such as "first", "second", and "third" may be used herein to describe various components, parts, regions, layers, or sections, these components, parts, regions, layers, or sections are not limited to these terms. Instead, these terms are only used to distinguish one component, part, region, layer, or section from another component, part, region, layer, or section. Therefore, without departing from the teachings of the examples described herein, the first component, part, region, layer, or section mentioned in the examples may also be referred to as the second component, part, region, layer, or section.

[0094] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description herein, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0095] In addition, the word "exemplary" is used herein to mean serving as an example, instance, or diagram. Any aspect or design described herein as "exemplary" is not necessarily to be understood as being advantageous over other aspects or designs. Instead, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or".

[0096] Likewise, although the present application has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. With particular regard to the various functions performed by the components (e.g., elements, resources, etc.) described above, unless otherwise noted, the terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific functions of the described components, even if not structurally equivalent to the disclosed structures.

[0097] Additionally, while particular features of the present application may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and advantageous for any given or particular application.

[0098] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary technical means in the art that are not disclosed in the present application, and the specification and embodiments are only considered as exemplary.

[0099] It will be appreciated that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A data analysis method based on a pharmacovigilance system, characterized in that: include: Obtain multiple pharmacovigilance data sets of the same drug produced by multiple manufacturers, where the same pharmacovigilance data set includes multiple data points of the target drug produced by the same manufacturer, and the same data point includes multiple sub-data corresponding to multiple adverse reaction monitoring dimensions; For a candidate data point of a target pharmacovigilance data set in a plurality of pharmacovigilance data sets, a first parameter value of the candidate data point is determined, wherein the plurality of pharmacovigilance data sets include the same number of data points, and the first parameter value of the candidate data point is determined in the following manner: , F is the first parameter value of the candidate data point, N is the number of pharmacovigilance data sets, norm is the normalization function, is the similarity value between the target pharmacovigilance dataset and the ath other pharmacovigilance dataset, L is the number of adverse reaction monitoring dimensions, M is the number of data points in the pharmacovigilance dataset, is the sub-data of the candidate data point in the bth adverse reaction monitoring dimension, is the sub-data of the i-th data point in the a-th other pharmacovigilance dataset in the b-th adverse reaction monitoring dimension, is an absolute value; the first parameter value is used to characterize the similarity between the candidate data point and the data points in other pharmacovigilance data sets other than the target pharmacovigilance data set; The similarity value between the target pharmacovigilance dataset and other pharmacovigilance datasets is determined in the following way: , is the median of the sub-data of the data point in the bth adverse reaction monitoring dimension in the target pharmacovigilance dataset, is the median of the sub-data of the data point in the ath other pharmacovigilance dataset in the bth adverse reaction monitoring dimension; determining a second parameter value of the candidate data point, wherein the second parameter value is used to characterize a degree of difference between the candidate data point and other data points in the target pharmacovigilance dataset; The second parameter value of the candidate data point is determined in the following way: , where D is the second parameter value of the candidate data point, is the average value of the sub-data of the bth adverse reaction monitoring dimension of the data point in the target pharmacovigilance dataset; The inverse of the product of the second parameter value and the first parameter value of the candidate data point is used as the weight value of the candidate data point, and the local deviation of the candidate data point is determined by using the weight value of the candidate data point and the outlier local factor LOF algorithm; the local deviation is used to characterize the degree of isolation of the data point; The candidate data points whose local deviation is greater than a preset deviation threshold are used as target data points corresponding to the atypical reaction of the drug.

2. The data analysis method based on the pharmacovigilance system according to claim 1, characterized in that: The local deviation of the candidate data points is determined by using the weight values ​​of the candidate data points and the outlier local factor LOF algorithm, including: Mapping data points in multiple drug vigilance data sets to a multidimensional data space; different dimensions of the multidimensional data space correspond to multiple adverse reaction monitoring dimensions; using the outlier local factor LOF algorithm to determine the initial local reachable density value of the candidate data points in the multidimensional data space; The product of the initial local reachable density value of the candidate data point and the weight value is used as the target local reachable density value of the candidate data point; The local deviation of the candidate data point is determined according to the target local reachable density value of the candidate data point and other data points in the neighborhood of the candidate data point in the multidimensional data space.

3. The data analysis method based on the pharmacovigilance system according to claim 2, characterized in that: The initial local reachable density value of the candidate data point in the multidimensional data space is determined in the following way: Determine the reachable distance of a data point in a pharmacovigilance dataset to other data points in a K-neighborhood in a multidimensional data space; the K-neighborhood includes the K other data points closest to the data point; The reciprocal of the average of the reachable distances of the candidate data point to other data points in the neighborhood of the candidate data point in the multidimensional data space is used as the initial local reachable density value of the candidate data point in the multidimensional data space.

4. The data analysis method based on the pharmacovigilance system according to claim 2, characterized in that: The local deviation of a candidate data point is determined as follows: , W is the local deviation of the candidate data point, Q is the number of other data points in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the fth other data point in the neighborhood of the candidate data point in the multidimensional data space, is the target local reachable density value of the candidate data point.

5. The data analysis method based on the pharmacovigilance system according to claim 1, characterized in that: The method further comprises: When the number of target data points corresponding to the atypical reaction of the drug is greater than a preset number threshold, sorting all target data points according to a specified priority order in multiple adverse reaction monitoring dimensions; The sorting result and the producer information of the target data point are sent to a display device for display.

6. The data analysis method based on the pharmacovigilance system according to claim 1, characterized in that: Adverse reaction monitoring dimensions include: Age, gender, single dosage, frequency of medication, cumulative duration of medication, adverse reaction symptoms, frequency of adverse reactions, and cumulative duration of adverse reactions.

7. A data analysis system based on a pharmacovigilance system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the data analysis method based on the drug vigilance system according to any one of claims 1 to 6 is implemented.

Images