Outpatient prescription compliance evaluation method and system based on multi-dimensional behavior modeling

Abstract

The invention provides an outpatient prescription compliance assessment method and system based on multi-dimensional behavior modeling, and the method comprises the steps: determining a doctor group corresponding to a to-be-assessed prescription, and enabling the doctor group to be obtained through the division based on the medication characteristics in the historical prescription data of a plurality of doctors; the prescription data of the to-be-evaluated prescription are modeled in multiple preset dimensions, a multi-dimensional behavior vector corresponding to the to-be-evaluated prescription is obtained, and the preset dimensions comprise cost, medication and a diagnosis result; and inputting the multi-dimensional behavior vector corresponding to the prescription to be evaluated into a trained anomaly detection model to obtain a compliance evaluation result of the prescription to be evaluated, the anomaly detection model being obtained based on training of a plurality of sample prescriptions to be evaluated, and the sample prescriptions to be evaluated being consistent with a doctor group corresponding to the prescription to be evaluated. According to the invention, the accuracy of the prescription violation detection result can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a clinic prescription compliance evaluation method and system based on multi-dimensional behavior modeling. BACKGROUND

[0002] A prescription is a medication plan issued by a doctor based on a diagnosis. Existing prescription supervision methods detect repeated prescriptions and dose conflicts in real time through a rule base, but only rely on surface information of the prescription (such as drug name, dose), and cannot detect hidden prescription violations. The prescription violation detection result is not accurate. SUMMARY

[0003] The present application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling, which solves the defect that the prescription violation detection result is not accurate in the prior art, and achieves the effect of improving the accuracy of the prescription violation detection result.

[0004] The present application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling, comprising: determining a doctor group corresponding to a prescription to be evaluated, the doctor group being obtained by dividing based on medication characteristics in historical prescription data of a plurality of doctors; modeling prescription data of the prescription to be evaluated in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, the preset dimensions including cost, medication, and diagnosis result; inputting the multi-dimensional behavior vector corresponding to the prescription to be evaluated into a trained anomaly detection model to obtain a compliance evaluation result of the prescription to be evaluated, the anomaly detection model being trained based on a plurality of sample prescriptions to be evaluated, the sample prescriptions to be evaluated being consistent with the doctor group corresponding to the prescription to be evaluated.

[0005] According to the clinic prescription compliance evaluation method based on multi-dimensional behavior modeling provided by the present application, before determining the doctor group corresponding to the prescription to be evaluated, comprising: constructing a heterogeneous graph network based on medication data in the historical prescription data; classifying the plurality of doctors based on the heterogeneous graph network to determine a plurality of doctor groups; wherein the heterogeneous graph network includes a plurality of first nodes and a plurality of second nodes, the first nodes corresponding to doctors, and the second nodes corresponding to drugs; the heterogeneous graph network includes a first edge and a second edge, the first edge connecting the first nodes and the second nodes, the first edge indicating that the prescriptions issued by the doctors corresponding to the first nodes include the drugs corresponding to the second nodes, and the second edge connecting two second nodes, the second edge indicating that the drugs corresponding to the second nodes appear together in the same prescription.

[0006] The application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling. extracting statistical features of the cost corresponding to the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the cost dimension; performing semantic embedding on the text of the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the medication dimension; performing one-hot encoding on the diagnosis category corresponding to the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the diagnosis dimension.

[0007] The application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling. The application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling. inputting the multi-dimensional behavior vector corresponding to the prescription to be evaluated into the corresponding autoencoder to obtain a feature vector output by the autoencoder; determining the compliance evaluation result corresponding to the prescription to be evaluated based on the distance between the feature vector and a preset group center, wherein the group center is obtained by training together with the anomaly detection model.

[0008] The application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling. The training process of the anomaly detection model comprises: inputting the multi-dimensional behavior vector corresponding to the sample prescription to be evaluated into the anomaly detection model to obtain a sample feature vector corresponding to the sample prescription to be evaluated; determining a training loss based on the sample feature vectors corresponding to the plurality of sample prescriptions to be evaluated, the compliance evaluation result labels corresponding to the second sample in the plurality of sample prescriptions to be evaluated, and the current group center; updating the anomaly detection model and the group center based on the training loss.

[0009] According to the outpatient prescription compliance evaluation method based on multi-dimensional behavior modeling provided by the application, the training loss is determined based on the sample feature vectors corresponding to the plurality of sample prescriptions to be evaluated, the compliance evaluation result labels corresponding to the second sample in the plurality of sample prescriptions to be evaluated, and the current population center, comprising: A first loss is determined based on the distance between the sample feature vectors corresponding to the plurality of sample prescriptions to be evaluated and the population center. A second loss is determined based on the similarity between the sample feature vectors between the sample prescriptions to be evaluated corresponding to different compliance evaluation result labels. The training loss is determined based on the first loss and the second loss.

[0010] The application further provides an outpatient prescription compliance evaluation system based on multi-dimensional behavior modeling, comprising: A population division module is configured to determine a doctor population corresponding to a prescription to be evaluated, wherein the doctor population is divided based on the medication characteristics in the historical prescription data of a plurality of doctors. A behavior modeling module is configured to model the prescription data of the prescription to be evaluated in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, wherein the preset dimensions include cost, medication, and diagnosis results. An anomaly detection module is configured to input the multi-dimensional behavior vector corresponding to the prescription to be evaluated into a trained anomaly detection model to obtain a compliance evaluation result of the prescription to be evaluated, wherein the anomaly detection model is trained based on a plurality of sample prescriptions to be evaluated, and the sample prescriptions to be evaluated are consistent with the doctor population corresponding to the prescription to be evaluated.

[0011] The application further provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the outpatient prescription compliance evaluation method based on multi-dimensional behavior modeling according to any one of the above.

[0012] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the outpatient prescription compliance evaluation method based on multi-dimensional behavior modeling according to any one of the above.

[0013] The application further provides a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the outpatient prescription compliance evaluation method based on multi-dimensional behavior modeling according to any one of the above.

[0014] The application provides a clinic prescription compliance evaluation method based on multi-dimensional behavior modeling, which realizes division of a doctor group by clustering doctors with consistent medication modes, and further models prescription data of a prescription to be evaluated based on multiple dimensions including cost, medication and diagnosis result, to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, and uses an abnormality detection model trained by sample prescription to be evaluated belonging to the same doctor group to evaluate the multi-dimensional behavior vector of the prescription to be evaluated, so that multi-dimensional deep mining of information contained in the prescription to be evaluated can be realized, and the accuracy of prescription violation detection result is improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort.

[0016] Figure 1 is a flowchart of the clinic prescription compliance evaluation method based on multi-dimensional behavior modeling provided by the application.

[0017] Figure 2 is a flowchart of the specific implementation steps of the clinic prescription compliance evaluation method based on multi-dimensional behavior modeling provided by the application.

[0018] Figure 3 is a structural schematic diagram of the clinic prescription compliance evaluation system based on multi-dimensional behavior modeling provided by the application.

[0019] Figure 4 is a structural schematic diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the application more clear, the technical solutions in the application will be described clearly and completely in the following with reference to the drawings in the application. Obviously, the described embodiments are some embodiments of the application, but not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative effort belong to the protection scope of the application.

[0021] It should be understood that when used in the specification and the appended claims, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0022] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0024] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0025] The following combination Figures 1-2 The present invention describes the outpatient prescription compliance assessment method based on multi-dimensional behavior modeling. Figure 1 As shown, the outpatient prescription compliance assessment method based on multidimensional behavior modeling includes the following steps: S110, determining a doctor group corresponding to the prescription to be evaluated, where the doctor group is divided based on medication characteristics in historical processing data of multiple doctors; S120, modeling the prescription data of the prescription to be evaluated in multiple preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated; S130. Input the multi-dimensional behavior vector corresponding to the prescription to be evaluated into the anomaly detection model corresponding to the trained doctor group to obtain the compliance evaluation result of the prescription to be evaluated. The anomaly detection model is trained based on multiple sample prescriptions to be evaluated, and the sample prescriptions to be evaluated are consistent with the doctor group corresponding to the prescription to be evaluated.

[0026] The method provided by the present invention realizes the division of doctor groups by clustering doctors with consistent medication patterns, and further models the prescription data of the prescription to be evaluated based on multiple dimensions including cost, medication and diagnosis results to obtain the multi-dimensional behavior vector corresponding to the prescription to be evaluated. The anomaly detection model trained on sample prescriptions to be evaluated belonging to the same doctor group is used to evaluate the multi-dimensional behavior vector of the prescription to be evaluated, which can realize multi-dimensional in-depth mining of the information contained in the prescription to be evaluated, which is conducive to the detection of hidden prescription violation information and improves the accuracy of prescription violation detection results.

[0027] A prescription is a medication plan prescribed by a doctor based on a diagnosis result, and prescription violations can include various forms, such as diagnosis-drug logic contradiction (for example, prescribing a drug that raises blood sugar under a diagnosis of diabetes), improper drug combination (for example, high-frequency combination of aspirin + clopidogrel, a risky drug combination), off-list medication, single-day cost exceeding the limit, and splitting prescriptions using policy loopholes, etc. The method provided by the present application aims to detect whether the prescription to be evaluated is in violation, and output a compliance evaluation result reflecting whether the prescription to be evaluated is in violation. It is worth noting that the method provided by the present application does not involve a separate judgment on the diagnosis result, but a judgment on the prescription data of the prescription to be evaluated, which includes the diagnosis result and the medication data prescribed based on the diagnosis result.

[0028] In existing prescription supervision methods, doctors are divided into groups based on static rules such as departments and titles, for example, doctors in the cardiology department are classified as the "cardiovascular treatment group". This simple grouping ignores the differences in individual treatment preferences of doctors, such as the difference in medication patterns between interventional treatment and drug conservative treatment among doctors in the same department, and cannot identify hidden behaviors such as cross-department medication. Due to the lack of data-driven, traditional methods cannot dynamically capture changes in doctors' treatment habits, resulting in the inability to establish accurate group behavior benchmarks, affecting the relevance of subsequent anomaly detection.

[0029] Different doctors have different medication patterns, such as conservative treatment or interventional treatment, and there are significant differences in drugs between these medication patterns. The method provided by the present application classifies doctors to avoid prescription violation evaluation errors caused by differences in medication patterns, classifies doctors based on their medication characteristics, obtains multiple doctor groups, and trains corresponding anomaly detection models for different doctor groups. When a prescription to be evaluated needs to be evaluated for violations, first determine the doctor group corresponding to the prescription to be evaluated, that is, which doctor group the doctor who prescribed the prescription to be evaluated belongs to, and detect the prescription to be evaluated based on the anomaly detection model corresponding to the doctor group, to avoid the influence of medication pattern differences on prescription violation detection.

[0030] As shown in Figure 2 , medical data is collected by a data collection module for subsequent doctor group division and training data for training anomaly detection models. Specifically, structured data such as prescription details, diagnosis codes, and cost amounts can be automatically captured through integrated hospital systems, medical insurance settlement platforms, and electronic medical record systems, and unstructured data such as diagnosis descriptions and medication reasons in electronic medical records can be parsed to form a multi-dimensional outpatient prescription data set. In the data cleaning stage, a combination of rules and statistical methods is used to remove abnormal records and align cross-system data through standardized coding.

[0031] Specifically, before determining the doctor group corresponding to the prescription to be evaluated, the following should be included: Construct a heterogeneous graph network based on medication data in historical prescription data; Based on the heterogeneous graph, multiple doctors are classified and multiple doctor groups are identified; Among them, the heterogeneous graph network includes multiple first nodes and multiple second nodes, the first node corresponds to a doctor, and the second node corresponds to a drug. The heterogeneous graph network includes a first edge and a second edge. The first edge connects the first node and the second node. The first edge indicates that the prescription issued by the doctor corresponding to the first node includes the drug corresponding to the second node. The second edge connects two of the second nodes. The second edge indicates that the drugs corresponding to the second nodes appear together in the same prescription.

[0032] In the method provided by the present invention, a doctor-drug heterogeneous network is constructed based on historical prescription data, and a weighted edge is established between the doctor node and the drug node through the prescription frequency. The doctor-drug network is a heterogeneous network. Indicates that Represent the vertices, edges, weights and vertex types of the network respectively. Vertex types can be divided into two categories: doctors and medicines .exist There are two kinds of edges in : Indicates doctor The treatment plan includes drugs ; Indicates drugs and co-occur in the same prescription or usage record. In this heterogeneous network, the weight between two entities is defined as the frequency of association between the two entities. Defined as a doctor Use of medicines on their prescriptions frequency; It is defined as the frequency of drugs appearing together in prescriptions. , using community discovery algorithms such as Louvain or LPA to divide doctors with similar prescription characteristics into the same group.

[0033] The method provided by the present invention can achieve effective division of doctor groups by constructing a heterogeneous graph network and performing group division based on the network, retaining the diversity of medication and making the medication patterns within the group consistent.

[0034] As explained above, the prescription violation may exist more hidden violations, the prior art relies on rule base and statistical threshold to identify the violation, and the generalization ability is poor for hidden violation mode, and the static rule base cannot adapt to the evolution of hidden behaviors such as "long-term low-dose repeated drug use", and needs to be updated by artificial rules, which cannot meet the real-time supervision demand. The method provided by the present application extracts features of the prescription to be evaluated in multiple preset dimensions, integrates three-dimensional data of cost (self-payment ratio, daily average amount), drug (Word2Vec semantic vector), and diagnosis (ICD code grouping one-hot feature) to construct a behavior vector. Compared with single-dimensional features, this vector can significantly improve the detection effect, especially the ability to identify hidden violation behaviors such as cross-department drug use.

[0035] Specifically, the prescription data of the prescription to be evaluated is modeled in multiple preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, including: extracting the statistical features of the cost corresponding to the prescription to be evaluated to obtain the behavior vector of the prescription to be evaluated in the cost dimension; performing semantic embedding on the text of the prescription to be evaluated to obtain the behavior vector of the prescription to be evaluated in the drug dimension; one-hot encoding the diagnosis category corresponding to the prescription to be evaluated to obtain the behavior vector of the prescription to be evaluated in the diagnosis dimension.

[0036] The method provided by the present application generates a multi-dimensional prescription behavior vector by fusing three-dimensional data of cost, drug, and diagnosis. The cost dimension uses statistical analysis method to extract statistical features such as daily average prescription amount, self-payment ratio, and medical insurance reimbursement ratio, which are recorded as ; the drug dimension performs semantic embedding on the prescription text based on Word2Vec technology to generate a high-dimensional semantic vector to represent the drug mode, which is recorded as ; the diagnosis dimension performs grouping one-hot encoding based on ICD code, converts the diagnosis category into a sparse binary vector after grouping according to disease systems (such as cardiovascular system, endocrine system), and records it as . Through multi-dimensional modeling, the expression ability of the prescription behavior features can be significantly improved, and the single-dimensional features can more accurately reflect the doctor's drug mode and compliance difference.

[0037] After the multi-dimensional behavior vector corresponding to the to-be-evaluated prescription is extracted, the multi-dimensional behavior vector is input into the anomaly detection model, and the anomaly detection model is used to obtain the compliance evaluation result of the to-be-evaluated prescription, which reflects whether the to-be-evaluated prescription is in violation. The anomaly detection model used to determine the compliance evaluation result of the to-be-evaluated prescription is trained based on historical data of prescriptions issued by a doctor group corresponding to the to-be-evaluated prescription, that is, for each doctor group, an anomaly detection model corresponding to the doctor group is trained in advance using historical prescription data in the doctor group, and when subsequent detection of a to-be-evaluated prescription is required, the anomaly detection model of the same doctor group corresponding to the to-be-evaluated prescription is selected for detection.

[0038] Specifically, the anomaly detection model includes a self-encoder corresponding to each preset dimension; the multi-dimensional behavior vector corresponding to the to-be-evaluated prescription is input into the trained anomaly detection model to obtain the compliance evaluation result of the to-be-evaluated prescription, including: The multi-dimensional behavior vector corresponding to the to-be-evaluated prescription is input into the corresponding self-encoder to obtain a feature vector output by the self-encoder; Based on the distance between the feature vector and a preset group center, a compliance evaluation result corresponding to the to-be-evaluated prescription is determined, and the group center is trained together with the anomaly detection model.

[0039] For the same group, a multi-self-encoder parallel architecture is used, and a self-encoder is constructed for each of the three dimensions of cost, medication, and diagnosis, and the parallel processing of different perspective features enhances the ability of the model to capture complex violation patterns.

[0040] In actual application, there is relatively little labeled data in the field of prescription detection, and the method provided by the present application trains the anomaly detection model in a semi-supervised manner. Specifically, the sample to-be-evaluated prescriptions used to train the anomaly detection model include a plurality of first samples and a plurality of second samples, the first samples do not have corresponding compliance evaluation result labels, the second samples have corresponding compliance evaluation result labels, and in practice, the number of first samples is greater than the number of second samples. The training process of the anomaly detection model includes: The multi-dimensional behavior vector corresponding to the sample to-be-evaluated prescription is input into the anomaly detection model to obtain a sample feature vector corresponding to the sample to-be-evaluated prescription; Based on the sample feature vectors corresponding to the plurality of sample to-be-evaluated prescriptions, the compliance evaluation result labels corresponding to the second samples in the plurality of sample to-be-evaluated prescriptions, and the current group center, a training loss is determined; The anomaly detection model and the group center are updated based on the training loss.

[0041] The training loss includes a center loss, i.e., by calculating the Euclidean distance between the encoded feature vector and the population center, the same population feature is forced to converge to the center, realizing the intra-class compact optimization. Further, the training loss also includes a multi-view contrastive loss, by comparing the latent feature representations of different dimensional autoencoders, the consistency between the features of normal prescriptions is maximized, while the similarity of the features of abnormal prescriptions is minimized, realizing the effect of significantly improving the discrimination ability of the model to the violation mode. That is, based on the sample feature vectors corresponding to the multiple sample prescriptions to be evaluated, the compliance evaluation result labels corresponding to the second sample in the multiple sample prescriptions to be evaluated, and the current population center, the training loss is determined, including: determining a first loss based on the distance between the sample feature vectors corresponding to the multiple sample prescriptions to be evaluated and the population center; determining a second loss based on the similarity between the sample feature vectors between the sample prescriptions to be evaluated corresponding to different compliance evaluation result labels; determining the training loss based on the first loss and the second loss.

[0042] The first loss corresponds to the center loss, and the loss function formula of the first loss is as follows: ; wherein, represents an autoencoder, which functions to compress high-dimensional input data (behavior vector) into a low-dimensional feature vector, extracting the core information of the data. represents the behavior vector of the i-th sample prescription to be evaluated in dimension v, represents the number of second samples and first samples input to the model. represents the compliance evaluation result label corresponding to the second sample, -1 represents a violation sample, and 1 represents a normal sample. is a pre-defined population center, and the closer the sample is to the center, the smaller the violation risk.

[0043] The second loss corresponds to the multi-view contrastive loss. Given a normal sample, K violation samples are selected to form a set, and the following contrastive loss function can be defined as the loss function of the second loss: ; wherein is used to measure the similarity of two vectors, and distance, cosine similarity, etc. can be used. represents the number of normal samples, represents the h-th normal sample, represents the k-th violation sample combined with ; and represent autoencoders of different dimensions.

[0044] By jointly training the autoencoder through the center loss and the contrast loss, the feature expression capability and the abnormal pattern generalization of the model can be improved, so that the model can accurately identify abnormal prescriptions and locate the violation types.

[0045] The outpatient prescription compliance evaluation system based on multi-dimensional behavior modeling provided by the present application is described below, and the outpatient prescription compliance evaluation system based on multi-dimensional behavior modeling described below can be correspondingly referred to the outpatient prescription compliance evaluation method based on multi-dimensional behavior modeling described above. As shown in Figure 3 The outpatient prescription compliance evaluation system based on multi-dimensional behavior modeling provided by the present application includes a group division module 310, a behavior modeling module 320, and an anomaly detection module 330. Among them: The group division module 310 is used to determine the doctor group corresponding to the prescription to be evaluated, and the doctor group is divided based on the medication characteristics in the historical prescription data of a plurality of doctors; The behavior modeling module 320 is used to model the prescription data of the prescription to be evaluated in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, and the preset dimensions include cost, medication, and diagnosis results; The anomaly detection module 330 is used to input the multi-dimensional behavior vector corresponding to the prescription to be evaluated into the trained anomaly detection model to obtain the compliance evaluation result of the prescription to be evaluated, and the anomaly detection model is trained based on a plurality of sample prescriptions to be evaluated. The sample prescription to be evaluated is consistent with the doctor group corresponding to the prescription to be evaluated.

[0046] Figure 4 An example of an entity structure diagram of an electronic device is shown in Figure 4As shown, the electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 complete communication with each other through the communications bus 440. The processor 410 can invoke a logic instruction in the memory 430 to execute a method for outpatient prescription compliance evaluation based on multi-dimensional behavior modeling. The method for outpatient prescription compliance evaluation based on multi-dimensional behavior modeling includes: determining a doctor group corresponding to a to-be-evaluated prescription, the doctor group being obtained based on medication characteristics in historical prescription data of a plurality of doctors; modeling prescription data of the to-be-evaluated prescription in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the to-be-evaluated prescription, the preset dimensions including cost, medication, and diagnosis result; inputting the multi-dimensional behavior vector corresponding to the to-be-evaluated prescription into a trained anomaly detection model to obtain a compliance evaluation result of the to-be-evaluated prescription, the anomaly detection model being trained based on a plurality of sample to-be-evaluated prescriptions, the sample to-be-evaluated prescriptions being consistent with the doctor group corresponding to the to-be-evaluated prescription.

[0047] In addition, the logic instruction in the memory 430 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0048] In another aspect, the present application also provides a computer program product comprising a computer program, which can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable a computer to perform the method of outpatient prescription compliance evaluation based on multi-dimensional behavior modeling. The method of outpatient prescription compliance evaluation based on multi-dimensional behavior modeling comprises: determining a doctor group corresponding to a to-be-evaluated prescription, the doctor group being obtained based on medication characteristics in historical prescription data of a plurality of doctors; modeling prescription data of the to-be-evaluated prescription in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the to-be-evaluated prescription, the preset dimensions including cost, medication, and diagnosis result; and inputting the multi-dimensional behavior vector corresponding to the to-be-evaluated prescription into a trained anomaly detection model to obtain a compliance evaluation result of the to-be-evaluated prescription, the anomaly detection model being trained based on a plurality of sample to-be-evaluated prescriptions, the sample to-be-evaluated prescriptions being consistent with the doctor group corresponding to the to-be-evaluated prescription.

[0049] In another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, and the computer program can be executed by a processor to implement the method of outpatient prescription compliance evaluation based on multi-dimensional behavior modeling. The method of outpatient prescription compliance evaluation based on multi-dimensional behavior modeling comprises: determining a doctor group corresponding to a to-be-evaluated prescription, the doctor group being obtained based on medication characteristics in historical prescription data of a plurality of doctors; modeling prescription data of the to-be-evaluated prescription in a plurality of preset dimensions to obtain a multi-dimensional behavior vector corresponding to the to-be-evaluated prescription, the preset dimensions including cost, medication, and diagnosis result; and inputting the multi-dimensional behavior vector corresponding to the to-be-evaluated prescription into a trained anomaly detection model to obtain a compliance evaluation result of the to-be-evaluated prescription, the anomaly detection model being trained based on a plurality of sample to-be-evaluated prescriptions, the sample to-be-evaluated prescriptions being consistent with the doctor group corresponding to the to-be-evaluated prescription.

[0050] The device embodiments described above are merely illustrative, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0051] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0052] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for evaluating outpatient prescription compliance based on multidimensional behavioral modeling, characterized in that: include: Determine the doctor group corresponding to the prescription to be evaluated, where the doctor group is divided based on the medication characteristics in the historical prescription data of multiple doctors; Modeling the prescription data of the prescription to be evaluated in multiple preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, wherein the preset dimensions include cost, medication, and diagnosis results; The multi-dimensional behavior vector corresponding to the prescription to be evaluated is input into the trained anomaly detection model to obtain the compliance evaluation result of the prescription to be evaluated. The anomaly detection model is trained based on multiple sample prescriptions to be evaluated, and the sample prescriptions to be evaluated are consistent with the doctor group corresponding to the prescription to be evaluated.

2. The outpatient prescription compliance assessment method based on multidimensional behavior modeling according to claim 1 is characterized in that: Before determining the doctor group corresponding to the prescription to be evaluated, the following steps are included: constructing a heterogeneous graph network based on the medication data in the historical prescription data; Classifying the multiple doctors based on the heterogeneous graph network to determine multiple doctor groups; The heterogeneous graph network includes a plurality of first nodes and a plurality of second nodes, wherein the first nodes correspond to doctors and the second nodes correspond to medicines; The heterogeneous graph network includes a first edge and a second edge, the first edge connects the first node and the second node, the first edge indicates that the prescription issued by the doctor corresponding to the first node includes the medicine corresponding to the second node, and the second edge connects two second nodes, and the second edge indicates that the medicines corresponding to the second nodes appear together in the same prescription.

3. The outpatient prescription compliance assessment method based on multidimensional behavior modeling according to claim 1 is characterized in that: The step of modeling the prescription data of the prescription to be evaluated in multiple preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated includes: Extracting statistical features of the cost corresponding to the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the cost dimension; Performing semantic embedding on the text of the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the medication dimension; Perform one-hot encoding on the diagnosis category corresponding to the prescription to be evaluated to obtain a behavior vector of the prescription to be evaluated in the diagnosis dimension.

4. The outpatient prescription compliance assessment method based on multidimensional behavior modeling according to claim 1 is characterized in that: The anomaly detection model includes an autoencoder corresponding to each of the preset dimensions; The step of inputting the multi-dimensional behavior vector corresponding to the prescription to be evaluated into the trained anomaly detection model to obtain a compliance evaluation result of the prescription to be evaluated includes: Inputting the multi-dimensional behavior vectors corresponding to the prescription to be evaluated into the corresponding autoencoders respectively, and obtaining the feature vectors output by the autoencoders; The compliance assessment result corresponding to the prescription to be assessed is determined based on the distance between the feature vector and a preset group center, and the group center is trained together with the anomaly detection model.

5. The outpatient prescription compliance assessment method based on multidimensional behavior modeling according to claim 4 is characterized in that: The sample prescriptions to be evaluated include a plurality of first samples and a plurality of second samples, and the second samples correspond to compliance evaluation result labels; The training process of the anomaly detection model includes: Inputting the multi-dimensional behavior vector corresponding to the sample prescription to be evaluated into the anomaly detection model to obtain a sample feature vector corresponding to the sample prescription to be evaluated; determining a training loss based on the sample feature vectors corresponding to the plurality of sample prescriptions to be evaluated, the compliance assessment result labels corresponding to the second samples in the plurality of sample prescriptions to be evaluated, and the current group center; The anomaly detection model and the group center are updated based on the training loss.

6. The outpatient prescription compliance assessment method based on multidimensional behavior modeling according to claim 5 is characterized in that: The determining of the training loss based on the sample feature vectors corresponding to the plurality of sample prescriptions to be evaluated, the compliance assessment result label corresponding to the second sample in the plurality of sample prescriptions to be evaluated, and the current group center includes: determining a first loss based on distances between sample feature vectors corresponding to a plurality of sample prescriptions to be evaluated and the group center; determining a second loss based on similarities between sample feature vectors of the sample prescriptions to be evaluated corresponding to different compliance assessment result labels; The training loss is determined based on the first loss and the second loss.

7. An outpatient prescription compliance assessment system based on multi-dimensional behavior modeling, characterized by: include: A group segmentation module is used to determine the doctor group corresponding to the prescription to be evaluated, where the doctor group is divided based on the medication characteristics in the historical prescription data of multiple doctors; A behavior modeling module is used to model the prescription data of the prescription to be evaluated in multiple preset dimensions to obtain a multi-dimensional behavior vector corresponding to the prescription to be evaluated, wherein the preset dimensions include cost, medication, and diagnosis results; An anomaly detection module is used to input the multi-dimensional behavior vector corresponding to the prescription to be evaluated into a trained anomaly detection model to obtain a compliance evaluation result of the prescription to be evaluated. The anomaly detection model is trained based on multiple sample prescriptions to be evaluated, and the sample prescriptions to be evaluated are consistent with the doctor group corresponding to the prescription to be evaluated.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the outpatient prescription compliance assessment method based on multidimensional behavior modeling as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the outpatient prescription compliance assessment method based on multi-dimensional behavior modeling as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the outpatient prescription compliance assessment method based on multi-dimensional behavior modeling as claimed in any one of claims 1 to 6 is implemented.

Images