Multi-source medical outpatient service data comprehensive management method and platform based on RPA

Abstract

The invention provides a multi-source medical outpatient service data comprehensive management method and platform based on RPA, and the method comprises the steps: S1, designing a script based on robot process automation, and carrying out the data collection and storage of multiple systems; s2, constructing a knowledge graph # imgabs0 # according to the structured data set; s3, introducing a causal inference technology into the knowledge graph # imgabs1 #; s4, according to the abnormal mark knowledge graph # imgabs2, generating personalized medicine use optimization suggestions through joint learning of a graph structure and node attributes; and S5, implementing closed-loop management by executing optimization suggestions and dynamically updating system rules. According to the method, the defects of a traditional method in the aspects of data integration, drug use anomaly detection and optimization suggestion generation can be overcome.

Description

Technical Field

[0001] The present invention belongs to the field of RPA technology, and in particular relates to a multi-source medical outpatient data comprehensive management method and platform based on RPA. Background Art

[0002] In the medical industry, the monitoring and management of clinical medication information is crucial to patient safety and medical quality. Modern medical institutions usually need to integrate data from multiple medical information systems (such as electronic medical record systems, drug management systems, and laboratory testing systems) to monitor medication behavior. However, the data formats, interfaces, and standards of these systems are often incompatible with each other, resulting in the complexity of data integration. In addition, the design of medical information systems often focuses on meeting a single function (such as prescribing or recording medical records) and lacks the ability to link data across systems. This fragmented system architecture poses many challenges to medical institutions in monitoring medication data.

[0003] First of all, the traditional manual data integration method is inefficient and relies on manual login, query and download of data, which is not only time-consuming and labor-intensive, but also prone to incomplete or erroneous data due to human errors. At the same time, the heterogeneity of different systems increases the difficulty of data analysis. Medical staff often need to switch frequently between different platforms to extract and compare medication information, which not only reduces work efficiency, but also affects the real-time and accuracy of the data. In addition, although some existing automated integration tools can automate some processes, most of them rely on predefined fixed rules and are difficult to cope with complex and changeable medical scenarios. For example, when dealing with abnormal data or non-standardized data formats, these tools usually cannot provide effective repair or processing solutions, resulting in process interruptions.

[0004] Secondly, medication monitoring and anomaly detection technologies are also insufficient. Currently, many systems rely on rule-based anomaly detection methods, which cannot deeply analyze the complex relationship between medication behavior and patient characteristics. For example, the adverse reactions of certain drugs may only occur in specific patient groups, and the rule system often cannot capture such implicit associations. In addition, the lack of personalized medication optimization capabilities is also a major problem. Existing systems focus more on anomaly detection, and lack support for how to provide intelligent medication optimization recommendations, which limits the application value of the system in actual clinical decision-making.

[0005] In this context, the development of an automated management system that can efficiently integrate multi-source medication data, accurately detect medication anomalies, and provide personalized optimization suggestions has become an urgent need in the medical industry. As an efficient data processing and process automation tool, Robotic Process Automation (RPA) can automatically perform cross-system tasks and reduce manual intervention. However, relying solely on RPA for process automation still cannot overcome the deep-seated problems in medication anomaly detection and optimization. Summary of the invention

[0006] The purpose of the present invention is to design a comprehensive management method and platform for multi-source medical outpatient data based on RPA. Combining causal inference and knowledge graph technology, the present invention can solve the defects of traditional methods in data integration, abnormal medication detection and optimization suggestion generation.

[0007] In order to achieve the above object, a first aspect of the present invention provides a method for comprehensive management of multi-source medical outpatient data based on RPA, comprising the following steps:

[0008] S1. Design a script based on robotic process automation to collect and store data from multiple systems, then classify the data into structured, semi-structured data and unstructured text data, perform semantic alignment and preprocessing on the structured, semi-structured and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set contains: patient ID, drug name, dosage, test index value, test time, and disease name;

[0009] S2. Construct a knowledge graph KG according to the structured data set;

[0010] S3. Introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field, and update the knowledge graph KG based on the abnormal drugs to generate an abnormal labeling knowledge graph KG updated ;

[0011] S4. Label the knowledge graph KG according to the anomaly updated ,generate personalized medication optimization recommendations by jointly learning graph structure and node attributes;

[0012] S5. Knowledge graph KG with medication optimization suggestions and abnormal labeling updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules;

[0013] The structure of the knowledge graph KG is as follows: the node set represents the entity, the edge set represents the association relationship between the entities, and each node contains a context attribute vector representing the semantic feature.

[0014] Preferably, for the structured and semi-structured data, a recursive parser is designed to traverse the nested fields layer by layer and extract semantically related target fields; for the unstructured text data, a regular parsing module based on medical semantics is designed to extract key entities;

[0015] The semantic alignment specifically includes: unifying fields in different systems into target semantics through a domain knowledge alignment table;

[0016] The preprocessing specifically includes:

[0017] For numerical fields, the weighted average of neighboring samples is used for filling;

[0018] Perform unit conversion on numeric fields: unify different units to the same unit.

[0019] Preferably, constructing a knowledge graph KG according to the structured data set specifically includes:

[0020] From each record r in the structured dataset i = <f 1 ,f 2 ,…,f m > Extract key fields and map them to nodes v in the knowledge graph KG k ;

[0021] Directly establish explicit relationships based on the field logic of structured data sets, and discover implicit relationships by statistically analyzing the joint distribution of entities in structured data sets. kl ;

[0022] For each node v k Generate attribute vector x k , represents the context information of the node;

[0023] According to the implicit relationship e kl Periodically verify node v k and v l The co-occurrence probability P(v k ,v l ) is significantly lower than the initial threshold τ to remove low-quality edges and cluster the node names based on character similarity to merge nodes that represent the same entity but with different spellings;

[0024] Finally, the knowledge graph KG is output.

[0025] Preferably, the hidden relationship e is found by statistically analyzing the joint distribution of entities in the structured data set. kl , specifically:

[0026]

[0027] Among them, P(v k ,v l ) represents node v k and v l The co-occurrence probability of klis a noise correction term based on domain knowledge; τ is the threshold for generating relationships;

[0028] For each node v k Generate attribute vector x k , indicating that the context information of the node is embedded in a linear model, and then f j The multidimensional attributes of are mapped to fixed dimensions.

[0029] Preferably, the causal graph is a directed acyclic graph extended on the basis of the knowledge graph KG, which is used to describe the causal relationship between drug A, patient characteristics X and clinical test indicators Y; wherein nodes represent random variables, including drugs, patient characteristics and test indicators; edges E represent causal influences; the causal graph is constructed as follows:

[0030] Extract clear edges E from the knowledge graph KG and construct the initial causal graph;

[0031] Combine medical field knowledge to supplement implicit causal edges;

[0032] In order to deal with unobserved confounding variables, a hypothetical confounding variable U is introduced to connect patient characteristics X and test indicators Y to form a potential influence path X←U→Y.

[0033] Preferably, the effect of drug A on the test index Y is quantified, and the expected value of the index after intervention P(Y|do(A)) is calculated:

[0034] P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx

[0035] Where P(Y|A,X=x) is the joint probability distribution of drug A and patient characteristic X; P(X=x) is the prior distribution of patient characteristics, and x is the current patient x;

[0036] Introducing an innovative regularization term Δ confound , correcting for the bias caused by unobserved confounding variables:

[0037] P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx+Δ confound

[0038] Among them, Δ confound The construction of is based on graph structure inference and external statistical data estimation, for example, the potential impact of unobserved confounding on Y is obtained through Bayesian inference;

[0039] Compare the calculated P(Y|do(A)) with the medical safety standard Y safe, through a binary function to determine whether drug A causes abnormal indicators: 1 means there is an abnormality, 0 means there is no abnormality;

[0040] For each abnormal drug A, mark the relevant nodes and edges in the knowledge graph KG:

[0041] Update node v A and v Y Attributes to record the source of the exception;

[0042] Mark abnormal edges e aY , to support subsequent reasoning optimization.

[0043] Preferably, the abnormal labeling knowledge graph KG updated , through joint learning of graph structure and node attributes, personalized medication optimization suggestions are generated, including:

[0044] Get the abnormal labeling knowledge graph KG updated ; Among them, the abnormal labeling knowledge graph KG updated Contains a node set V and an edge set E. Each node v k ∈V has a context attribute vector x k , and the abnormal edge e AY ;

[0045] Constructing anomaly labeling knowledge graph KG updated The graph embedding update rule jointly models the graph structure information and node attributes:

[0046]

[0047] in, For node v k The embedding representation at the tth layer is initialized with the node attribute vector x k ; N(v k ) represents node v k The neighbor set of w jk For edge e jk The weight of ; W is the shared transformation matrix, b is the bias vector, and ReLU is the activation function; the embedding update process captures the global semantic information of the graph and integrates the contextual characteristics of each node;

[0048] Combined with the initial edge weight, the edge weight w is adjusted by abnormal marking jk ;

[0049] For patient node v patient , based on its final embedding representation h patient , calculated by the cosine similarity of the embedding vector with the drug node v drug The recommendation score, where the highest score indicates the highest recommendation priority;

[0050] The recommended result R recommend Write back to the anomaly tag knowledge graph KG updated , update the recommended attributes of the drug node;

[0051] The final output recommendation result R recommend , the recommended dosage and precautions for each drug and the results of the global risk assessment for the patient.

[0052] Preferably, a robotic process automation tool is used to convert the recommendation result R recommend Automatically transmit to medical information system, and perform the following tasks:

[0053] According to the recommended results R recommend Submit medication prescriptions, including drug name, recommended dosage, and instructions for use;

[0054] The optimization suggestions are stored in the patient's medical record and labeled as "recommended solution";

[0055] Based on the usage of recommended drugs, the actual usage data of drugs is monitored in real time and collected as feedback data. feedback ;

[0056] Collect feedback data feedback , calculate the execution effect of the optimization suggestion:

[0057]

[0058] Among them, E is the effect evaluation value, Indicates the actual effect of the i-th indicator; is the target value; g is the effect scoring function, defined as w i is the indicator weight, reflecting the importance of the indicator; range is the normal value range of the medical indicator, which is used for standard deviation calculation; the calculation result of E will be used as the quantitative evaluation value of the optimization effect to guide the subsequent dynamic adjustment;

[0059] According to the effect evaluation value E and feedback data F feedback , dynamically adjust the abnormal labeling knowledge graph KG updated and recommended rules;

[0060] Finally, according to the abnormal labeling knowledge graph KG updated Automatically generate feedback and adjustment reports based on recommended rules, including:

[0061] The overall execution effect evaluation value E of the optimization suggestion;

[0062] Specific content of dynamic rule updates;

[0063] Current stability evaluation to predict whether further manual intervention is needed.

[0064] Preferably, the effect evaluation value E and the feedback data F feedback , dynamically adjust the abnormal labeling knowledge graph KG updated And the recommended rules are:

[0065] If a drug triggers abnormalities multiple times or has poor effects, that is, the effect evaluation value E<τ, τ is the threshold, the priority of the drug node in the graph is reduced and its edge weight w is updated jk ;

[0066] If a drug performs well, its priority will be increased to promote recommendation.

[0067] In a second aspect of the present invention, a multi-source medical outpatient data integrated management platform based on RPA is provided, the platform comprising:

[0068] The RPA data acquisition subsystem is used to design scripts based on robotic process automation, collect and store data from multiple systems, and then classify the data into structured, semi-structured data, and unstructured text data, and semantically align and preprocess the structured, semi-structured, and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set contains: patient ID, drug name, dosage, test index value, test time, and disease name;

[0069] A knowledge graph construction subsystem, used to construct a knowledge graph KG according to the structured data set;

[0070] The knowledge graph update subsystem is used to introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field. The knowledge graph KG is updated based on the abnormal drugs to generate an abnormal labeling knowledge graph KG. updated ;

[0071] The medication recommendation subsystem is used to label the knowledge graph KG according to the abnormality updated ,generate personalized medication optimization recommendations by jointly learning graph structure and node attributes;

[0072] RPA closed-loop management subsystem for medication optimization suggestions and abnormal tagging knowledge graph KG updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules;

[0073] The structure of the knowledge graph KG is as follows: the node set represents the entity, the edge set represents the association relationship between the entities, and each node contains a context attribute vector representing the semantic feature.

[0074] The beneficial technical effects of the present invention are at least as follows:

[0075] (1) The present invention uses RPA technology to achieve automatic extraction, cleaning and standardization of cross-system data. Compared with traditional manual integration or regularization tools, the present invention can dynamically adapt to the interfaces and data formats of different systems and solve the integration problems caused by the heterogeneity of multi-source data. In addition, the built-in anomaly detection mechanism can identify and repair abnormal data in real time during data processing, ensuring the continuity and accuracy of the process.

[0076] (2) The present invention addresses the defect that existing systems that rely on rule-based methods are difficult to capture complex medication anomalies. The present invention introduces causal inference technology to construct a causal model of medication behavior and patient characteristics, and distinguish between correlation and causality. Through causal effect calculation, the system can accurately identify hidden medication risks. For example, for the adverse reactions of a certain drug, the system can infer its potential impact based on the patient's specific physical signs or medical history, avoiding missed detection or misjudgment due to a single rule.

[0077] (3) The present invention integrates the complex relationships between drugs, patient characteristics, and clinical data by constructing a medication knowledge graph. On this basis, the graph neural network technology is used for multi-hop reasoning to generate personalized medication optimization recommendations. For example, the system can recommend safer or more effective alternative medication plans based on the patient's current test results and medical records, thereby providing intelligent support for doctors' decision-making.

[0078] (4) By combining RPA with intelligent algorithms, the present invention realizes the full process automation management of data integration, anomaly detection and optimization suggestion generation. At the same time, the system can dynamically adjust the process rules according to the detection results and optimization suggestions to adapt to the changing clinical scenarios. For example, when a drug is identified as abnormal, the system can automatically update the medication process to avoid potential risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0080] Figure 1 This is a flow chart of a comprehensive management method for multi-source medical outpatient data based on RPA in the present invention.

[0081] Figure 2This is a framework diagram of a multi-source medical outpatient data comprehensive management platform based on RPA in the present invention. DETAILED DESCRIPTION

[0082] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0083] In one or more embodiments, Figure 1 As shown, a method for comprehensive management of multi-source medical outpatient data based on RPA is disclosed, and the method includes the following steps S1-S5:

[0084] S1. Design a script based on robotic process automation to collect and store data from multiple systems, then classify the data into structured data, semi-structured data and unstructured text data, perform semantic alignment and preprocessing on the structured data, semi-structured data and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set contains: patient ID, drug name, dosage, test index value, test time, and disease name.

[0085] Specifically, we designed scripts based on Robotic Process Automation (RPA) to automatically log into multiple systems, extract data, and store it.

[0086] Among them, the data extraction script F for each data system (such as EMR system, laboratory system, and drug management system) i It consists of the following parts:

[0087] Automatically log in to the system and obtain an authorization token.

[0088] Locate key data modules (such as medication records, test results) based on preset rules.

[0089] Download the target data and store it in the initial dataset D raw , which may be in JSON, XML, or CSV format.

[0090] Furthermore, after data collection is completed, D raw Recorded as:

[0091]

[0092] Where M represents the number of data sources, S i is the i-th data source.

[0093] Furthermore, for structured and semi-structured data (such as JSON and XML), a recursive parser is designed to traverse the nested fields layer by layer and extract semantically related target fields (such as “drug name”, “dosage”).

[0094] For unstructured text data, a regular parsing module based on medical semantics is designed to extract key entities. For example, "drug name: aspirin; dosage: 100mg" is parsed from the text description and mapped to the standard field "drug = aspirin, dosage = 100mg".

[0095] Furthermore, a domain knowledge alignment table is used to unify the fields in different systems to the target semantics:

[0096] f j =Mapping(SourceField i )

[0097] Among them, SourceField i represents a source field, Mapping is a semantic alignment rule, and f j is the target field.

[0098] Furthermore, missing values ​​are filled:

[0099] For numerical fields (such as dose and test value), the weighted average of neighboring samples is used for filling:

[0100]

[0101] Among them, k is the number of neighboring samples, w i is the weight associated with the sample distance, x miss is the field of the sample where the missing value is located, and ∈ is a small amount to avoid the denominator being zero.

[0102] Further, outlier detection:

[0103] The detection logic consists of two parts:

[0104] Distribution detection: flags values ​​that deviate from the normal distribution by more than 3 standard deviations.

[0105] Logical rules: For example, for the dose field, exclude records that exceed the recommended upper dose limit.

[0106] Furthermore, unit conversion is performed on numerical fields (such as dosage): different units (such as g, mg, μg) are unified to "mg". The conversion formula is:

[0107] x std = x·u

[0108] Wherein, x is the original value, and u is the unit conversion coefficient (such as u=1000 when converting from g to mg).

[0109] Use standard codes (such as ICD-10) for categorical fields (such as disease names).

[0110] Time fields are uniformly formatted according to the ISO8601 standard.

[0111] Furthermore, all standardized records are stored as a unified dataset D cleaned Each record r i Contains the following standard fields: patient ID (anonymized); drug name (standardized); dosage (mg); test index value (numeric value); test time (ISO8601 format); disease name (ICD-10 code).

[0112] Furthermore, the structured form of the data set is:

[0113] D cleaned = {r 1 ,r 2 ,…,r n},r i = <f 1 ,f 2 ,…,f m >

[0114] where f j Represents the standardized field, and n is the number of records.

[0115] S2. Construct a knowledge graph KG based on the structured data set; wherein the structure of the knowledge graph KG is as follows: the node set represents the entity, the edge set represents the association relationship between the entities, and each node contains a context attribute vector representing the semantic feature.

[0116] Specifically, from D cleaned Each record of r i = <f 1 ,f 2 ,…,f m > Extract key fields and map them to nodes v in the graph k :

[0117] f 1 (Patient ID): mapped to the patient entity node.

[0118] f 2 (drug name) and f 3 (Dose): Mapped to drug entity node and its attributes.

[0119] f 4 (test index) and f 5(Verification time): Mapped to the verification node and its timestamp.

[0120] f 6 (Disease name): mapped to disease node.

[0121] Furthermore, for multi-valued fields (such as f 2 contains multiple drug names) into multiple nodes, and retain the association between fields. i If both "aspirin" and "ibuprofen" are included, two drug nodes are generated respectively and associated with the patient node.

[0122] Preferably, a context enhancement module is designed for each node, by combining f j The field information of the entity is used to generate an attribute vector to capture the semantic features of the entity.

[0123] Further, direct rule relationship: According to D cleaned The field logic directly establishes an explicit relationship. For example:

[0124] f 1 (patient) and f 6 (Disease) Establishes a "suffering from" relationship.

[0125] f 2 (Drug) and f 6 (Disease) Establish a "treatment" relationship.

[0126] Further, implicit statistical relationship: through statistical analysis D cleaned The joint distribution of entities in , discovers implicit relationships. Introducing a special correction term Δ ij , to measure the impact of potential noise to ensure that the generated relationship is more accurate:

[0127]

[0128] Among them, P(v k ,v l ) represents node v k and v l The co-occurrence probability of kl is a noise correction term based on domain knowledge; τ is the threshold for generating relationships. kl The use of domain prior knowledge definition aims to reduce the impact of entities with high noise but weak semantic association on the graph (such as misspellings in drug names causing false associations).

[0129] Furthermore, for each node v k Generate attribute vector x k , represents the context information of the node. Using the linear embedding model, f j The multidimensional attributes are mapped to fixed dimensions:

[0130] x k =W·z+b

[0131] Among them, z is the node field f j W is the embedding matrix; b is the bias vector. The attribute vector not only describes the original features of the node, but also captures the characteristics related to the target task through field weighting. For example, for the drug node, x k Special emphasis will be placed on information on dosage and administration.

[0132] Further, the graph is optimized:

[0133] Relationship cleanup: For implicit relationships kl , introduce a dynamic elimination mechanism and regularly verify P(v k ,v l ) is significantly lower than the initial threshold τ to remove low-quality edges.

[0134] Node merging: Cluster node names based on character similarity and merge nodes that represent the same entity but have different spellings (such as "aspirin" and "ASPIRIN").

[0135] Furthermore, the output knowledge graph KG = (V, E) contains the following contents:

[0136] The node set V represents entities such as patients, drugs, diseases, and tests;

[0137] The edge set E represents the association relationship between entities;

[0138] Each node v k Contains the context attribute vector x k , indicating semantic features.

[0139] It can be understood that the noise correction term Δ of the implicit relationship is introduced kl and the contextual augmentation vector x k , specifically optimized for fuzzy relations and entity semantics in medical data.

[0140] S3. Introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field, and update the knowledge graph KG based on the abnormal drugs to generate an abnormal labeling knowledge graph KG updated .

[0141] Specifically, the cause-effect diagram C graph It is a directed acyclic graph (DAG) extended on the basis of the knowledge graph KG, which is used to describe the causal relationship between drugs, patient characteristics and clinical test indicators:

[0142] Nodes represent random variables, including drugs (A), patient characteristics (X, such as age, medical history), and test indicators (Y, such as blood sugar, liver function).

[0143] Edges represent causal influences (e.g., the effect of drug A on indicator Y, and the moderating effect of patient characteristic X on A).

[0144] Further, the specific construction process:

[0145] Extract explicit edges E from the KG and construct the initial causal graph. For example, the edge “drugs treat diseases” is mapped to the causal relationship A→Y.

[0146] Combine medical domain knowledge to supplement implicit causal relationships. For example, some drugs may affect liver function indicators through metabolic processes. This indirect relationship is defined by domain experts as A→M→Y, where M is a hidden variable (metabolite).

[0147] In order to deal with unobserved confounding variables (such as lifestyle habits), a hypothetical confounding variable U is introduced to connect patient characteristics X and test indicators Y to form a potential influence path X←U→Y.

[0148] Furthermore, in order to quantify the effect of drug A on the test index Y, the expected value of the index after intervention P(Y|do(A)) is calculated:

[0149] P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx

[0150] Among them, P(Y|A,X=x) is the joint probability distribution of drug A and patient characteristic X; P(X=x) is the prior distribution of patient characteristics.

[0151] Introducing an innovative regularization term Δ confound , correcting for the bias caused by unobserved confounding variables:

[0152] P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx+Δ confound

[0153] Δ confound The construction of is based on graph structure inference and external statistical data estimation, for example, the potential impact of unobserved confounding on Y is obtained through Bayesian inference.

[0154] This correction term specifically addresses the problem of bias in medical data caused by insufficient data collection and ensures the reliability of causal effect calculations.

[0155] Furthermore, the calculated P(Y|do(A)) is compared with the medical safety standard Y safe, determine whether drug A causes abnormal indicators:

[0156]

[0157] Among them, Y safe Defined by domain experts or medical guidelines. For example, the normal range of blood sugar is 3.9-6.1mmol / L. Anomaly (A) is a binary function, 1 indicates an abnormality and 0 indicates no abnormality.

[0158] Furthermore, for each abnormal drug A, mark the relevant nodes and edges in the knowledge graph KG:

[0159] Update node v A (drug) and v Y (Test indicator) attributes, record the source of the anomaly;

[0160] Mark abnormal edges e AY , to support subsequent reasoning optimization.

[0161] Further, generate an updated knowledge graph KG updated , contains the following extensions:

[0162] Each node (such as drug node v A ) Added the anomaly flag field AnomalyFlag;

[0163] Abnormal edge AY Contains causal path information for raising an exception.

[0164] KG updated It will serve as the input for the next step of optimization reasoning and directly guide the adjustment of the medication regimen.

[0165] S4. Label the knowledge graph KG according to the anomaly updated , personalized medication optimization recommendations are generated by jointly learning graph structure and node attributes.

[0166] Specifically, KG updated It contains a node set V (such as patients, drugs, diseases, and test indicators) and an edge set E (indicating the relationship between nodes). Each node v k ∈V has a context attribute vector x k .

[0167] Furthermore, we construct graph embedding update rules to jointly model graph structure information and node attributes:

[0168]

[0169] in, For node v kThe embedding representation at the tth layer is initialized with the node attribute vector x k ; Represents node v k The neighbor set of w jk For edge e jk The embedding update process captures the global semantic information of the graph and integrates the contextual characteristics of each node.

[0170] Furthermore, to process drug nodes marked as abnormal (such as v A ) affects the inference results and adjusts the edge weight w jk :

[0171]

[0172] in, is the initial edge weight; AnomalyFlag∈{0,1} represents the abnormal flag; α is the correction coefficient, which is used to control the degree of weakening of the abnormal edge weight (such as α=1.5). This correction mechanism reduces the impact of abnormal drug nodes on graph structure learning and gives priority to recommending risk-free drugs.

[0173] Furthermore, for patient node v patient , based on its final embedding representation h patient , calculate and drug node v drug Recommended score:

[0174] s(v patient ,v drug )=cos(h patient ,h drug )

[0175] Among them, cos represents the cosine similarity of the embedding vector, and the higher the score, the higher the recommendation priority. Sort all candidate drugs and generate the recommendation list R recommend , and additional specific recommendations in conjunction with node attributes such as dose.

[0176] Furthermore, the recommendation result R recommend Write back to the knowledge graph KG updated , update the recommended attributes of the drug node (such as priority, recommended dosage).

[0177] The resulting output includes:

[0178] Recommended medication list recommend , sorted by priority;

[0179] The recommended dosage and precautions for use of each drug;

[0180] Global risk assessment results for patients.

[0181] S5. Knowledge graph KG with medication optimization suggestions and abnormal labeling updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules.

[0182] Specifically, the robot process automation (RPA) tool is used to convert the recommendation results R recommend Automatically transmit to medical information system (such as electronic medical record system). Specific execution includes the following tasks:

[0183] Furthermore, according to R recommend Submit medication prescriptions, including drug name, recommended dosage, and instructions for use;

[0184] The optimization suggestions are stored in the patient's medical record and labeled as "recommended plan".

[0185] Based on the usage of recommended drugs, real-time monitoring of actual drug usage data (such as patient response and treatment effect) is performed and collected as feedback data. feedback .

[0186] Furthermore, collect F feedback , including patients’ actual medication data (such as medication dosage), changing trends of test indicators (such as blood sugar levels), and doctors’ manual adjustment information (such as drug replacement).

[0187] Calculate the execution effect of the optimization suggestion:

[0188]

[0189] in, Indicates the actual effect of the i-th indicator; is the target value (such as the medical standard range); g is the effect scoring function, defined as w i is the indicator weight, reflecting the importance of the indicator (such as liver function weight is higher than the general test indicators). The calculation result of E will be used as a quantitative evaluation value of the optimization effect to guide subsequent dynamic adjustments.

[0190] Further, according to the evaluation value E and the feedback data F feedback , dynamically adjust the knowledge graph KG updated And the recommended rules:

[0191] If a drug triggers abnormalities multiple times or has poor effects (such as E<τ, τ is the threshold), the priority of the drug node in the graph is reduced and its edge weight w is updated. jk :

[0192]

[0193] Where β is the update coefficient and PenaltyFlag∈{0,1} represents the penalty flag.

[0194] If a drug performs well, its priority is increased, facilitating its recommendation.

[0195] Furthermore, after the update, the graph state changes from KG updated Transformed to KG final , to support the next optimization reasoning.

[0196] Furthermore, a feedback and adjustment report is automatically generated, including the following:

[0197] The overall execution effect evaluation value E of the optimization suggestion;

[0198] Specific content of dynamic rule updates (e.g., which drugs are downgraded or upgraded in priority);

[0199] Evaluate the current stability of the system and predict whether further human intervention is needed.

[0200] In one or more embodiments, Figure 2 As shown, a multi-source medical outpatient data comprehensive management platform based on RPA is disclosed, and the platform includes:

[0201] The RPA data acquisition subsystem 301 is used to design a script based on robotic process automation, collect and store data from multiple systems, and then classify the data into structured data, semi-structured data, and unstructured text data, and semantically align and preprocess the structured data, semi-structured data, and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set includes: patient ID, drug name, dosage, test index value, test time, and disease name;

[0202] The knowledge graph construction subsystem 302 is used to construct a knowledge graph KG according to the structured data set; wherein the structure of the knowledge graph KG is as follows: a node set represents an entity, an edge set represents an association relationship between entities, and each node contains a context attribute vector representing a semantic feature;

[0203] The knowledge graph updating subsystem 303 is used to introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field, and update the knowledge graph KG based on the abnormal drugs to generate an abnormal labeling knowledge graph KG updated ;

[0204] Medication recommendation subsystem 304 is used to label the knowledge graph KG according to the abnormality updated,generate personalized medication optimization recommendations by jointly learning graph structure and node attributes;

[0205] RPA closed-loop management subsystem 305 is used to optimize medication recommendations and anomaly tag knowledge graph KG updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules.

[0206] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A comprehensive management method for multi-source medical outpatient data based on RPA, characterized in that: The following steps are involved: S1. Design a script based on robotic process automation to collect and store data from multiple systems, then classify the data into structured, semi-structured data and unstructured text data, perform semantic alignment and preprocessing on the structured, semi-structured and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set contains: patient ID, drug name, dosage, test index value, test time, and disease name; S2. Construct a knowledge graph KG according to the structured data set; S3. Introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field, and update the knowledge graph KG based on the abnormal drugs to generate an abnormal labeling knowledge graph KG updated ; S4. Label the knowledge graph KG according to the anomaly updated ,generate personalized medication optimization recommendations by jointly learning graph structure and node attributes; S5. Knowledge graph KG with medication optimization suggestions and abnormal labeling updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules; The structure of the knowledge graph KG is as follows: the node set represents the entity, the edge set represents the association relationship between the entities, and each node contains a context attribute vector representing the semantic feature.

2. According to claim 1, a comprehensive management method for multi-source medical outpatient data based on RPA is characterized in that: For the structured and semi-structured data, a recursive parser is designed to traverse the nested fields layer by layer and extract semantically related target fields; For the unstructured text data, a regular parsing module based on medical semantics is designed to extract key entities; The semantic alignment specifically includes: unifying fields in different systems into target semantics through a domain knowledge alignment table; The preprocessing specifically includes: For numerical fields, the weighted average of neighboring samples is used for filling; Perform unit conversion on numeric fields: unify different units into the same unit; use standard encoding for categorical fields and format time fields to the ISO8601 standard.

3. According to the RPA-based multi-source medical outpatient data comprehensive management method of claim 1, it is characterized in that: The step of constructing a knowledge graph KG according to the structured data set specifically includes: From each record r in the structured dataset i = <f1,f2,…,f m > Extract key fields and map them to nodes v in the knowledge graph KG k ; Directly establish explicit relationships based on the field logic of structured data sets, and discover implicit relationships by statistically analyzing the joint distribution of entities in structured data sets. kl ; For each node v k Generate attribute vector x k , represents the context information of the node; According to the implicit relationship e kl Periodically verify node v k and v l The co-occurrence probability P(v k ,v l ) is significantly lower than the initial threshold τ to remove low-quality edges and cluster the node names based on character similarity to merge nodes that represent the same entity but with different spellings; Finally, the knowledge graph KG is output.

4. According to claim 3, a comprehensive management method for multi-source medical outpatient data based on RPA is characterized in that: The method statistically analyzes the joint distribution of entities in structured data sets and discovers implicit relationships. kl , specifically: Among them, P(v k ,v l ) represents node v k and v l The co-occurrence probability of kl is a noise correction term based on domain knowledge; τ is the threshold for generating relationships; For each node v k Generate attribute vector x k , indicating that the context information of the node is embedded in a linear model, and then f j The multidimensional attributes of are mapped to fixed dimensions.

5. According to claim 1, a comprehensive management method for multi-source medical outpatient data based on RPA is characterized in that: The causal graph is a directed acyclic graph extended on the basis of the knowledge graph KG, which is used to describe the causal relationship between drug A, patient characteristics X and clinical test indicators Y; wherein nodes represent random variables, including drugs, patient characteristics and test indicators; edges E represent causal influences; the causal graph is constructed as follows: Extract clear edges E from the knowledge graph KG and construct the initial causal graph; Combine medical field knowledge to supplement implicit causal edges; In order to deal with unobserved confounding variables, a hypothetical confounding variable U is introduced to connect patient characteristics X and test indicators Y to form a potential influence path X←U→Y.

6. According to claim 5, a method for comprehensive management of multi-source medical outpatient data based on RPA is characterized in that: Quantify the effect of drug A on the test indicator Y and calculate the expected value of the indicator after intervention P(Y|do(A)): P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx Where P(Y|A,X=x) is the joint probability distribution of drug A and patient characteristic X; P(X=x) is the prior distribution of patient characteristics, and x is the current patient x; Introducing an innovative regularization term Δ confound , correcting for the bias caused by unobserved confounding variables: P(Y|do(A))=∫ x P(Y|A,X=x)P(X=x)dx+Δ confound Among them, Δ confound The construction of is based on graph structure inference and external statistical data estimation, for example, the potential impact of unobserved confounding on Y is obtained through Bayesian inference; Compare the calculated P(Y|do(A)) with the medical safety standard Y safe , through a binary function to determine whether drug A causes abnormal indicators: 1 means there is an abnormality, 0 means there is no abnormality; For each abnormal drug A, mark the relevant nodes and edges in the knowledge graph KG: Update node v A and v Y Attributes to record the source of the exception; Mark abnormal edges e aY , to support subsequent reasoning optimization.

7. According to claim 6, a method for comprehensive management of multi-source medical outpatient data based on RPA, characterized in that: The abnormal labeling knowledge graph KG updated , through joint learning of graph structure and node attributes, personalized medication optimization suggestions are generated, including: Get the abnormal labeling knowledge graph KG updated ; Among them, the abnormal labeling knowledge graph KG updated Contains a node set V and an edge set E. Each node v k ∈V has a context attribute vector x k , and the abnormal edge e AY ; Constructing anomaly labeling knowledge graph KG updated The graph embedding update rule jointly models the graph structure information and node attributes: in, For node v k The embedding representation at the tth layer is initialized with the node attribute vector x k ; N(v k ) represents node v k The neighbor set of w jk For edge e jk The weight of ; W is the shared transformation matrix, b is the bias vector, and ReLU is the activation function; the embedding update process captures the global semantic information of the graph and integrates the contextual characteristics of each node; Combined with the initial edge weight, the edge weight w is adjusted by abnormal marking jk ; For patient node v patient , based on its final embedding representation h patient , calculated by the cosine similarity of the embedding vector with the drug node v drug The recommendation score, where the highest score indicates the highest recommendation priority; The recommended result R recommend Write back to the anomaly tag knowledge graph KG updated , update the recommended attributes of the drug node; The final output recommendation result R recommend , the recommended dosage and precautions for each drug and the results of the global risk assessment for the patient.

8. According to claim 7, a comprehensive management method for multi-source medical outpatient data based on RPA is characterized in that: Use robotic process automation tools to convert the recommended results into recommend Automatically transmit to medical information system, and perform the following tasks: According to the recommended results R recommend Submit medication prescriptions, including drug name, recommended dosage, and instructions for use; The optimization suggestions are stored in the patient's medical record and marked as "recommended solution"; Based on the usage of recommended drugs, the actual usage data of drugs is monitored in real time and collected as feedback data. feedback ; Collect feedback data feedback , calculate the execution effect of the optimization suggestion: Among them, E is the effect evaluation value, Indicates the actual effect of the i-th indicator; is the target value; g is the effect scoring function, defined as w i is the indicator weight, reflecting the importance of the indicator; range is the normal value range of the medical indicator, which is used for standard deviation calculation; the calculation result of E will be used as the quantitative evaluation value of the optimization effect to guide the subsequent dynamic adjustment; According to the effect evaluation value E and feedback data F feedback , dynamically adjust the abnormal labeling knowledge graph KG updated and recommended rules; Finally, according to the abnormal labeling knowledge graph KG updated Automatically generate feedback and adjustment reports based on recommended rules, including: The overall execution effect evaluation value E of the optimization suggestion; Specific content of dynamic rule updates; Current stability evaluation to predict whether further manual intervention is needed.

9. According to claim 8, a comprehensive management method for multi-source medical outpatient data based on RPA is characterized in that: According to the effect evaluation value E and the feedback data F feedback , dynamically adjust the abnormal labeling knowledge graph KG updated And the recommended rules are: If a drug triggers abnormalities multiple times or has poor effects, that is, the effect evaluation value E<τ, τ is the threshold, the priority of the drug node in the graph is reduced and its edge weight w is updated jk ; If a drug performs well, its priority will be increased to promote recommendation.

10. A multi-source medical outpatient data comprehensive management platform based on RPA, characterized in that: The platform includes: The RPA data acquisition subsystem is used to design scripts based on robotic process automation, collect and store data from multiple systems, and then classify the data into structured, semi-structured data, and unstructured text data, and semantically align and preprocess the structured, semi-structured, and unstructured text data to obtain a preprocessed structured data set; wherein each record in the data set contains: patient ID, drug name, dosage, test index value, test time, and disease name; A knowledge graph construction subsystem, used to construct a knowledge graph KG according to the structured data set; The knowledge graph update subsystem is used to introduce causal inference technology into the knowledge graph KG, establish a causal graph, calculate the causal effect of medication behavior on clinical indicators, and detect potential abnormal drugs in combination with safety standards in the medical field. The knowledge graph KG is updated based on the abnormal drugs to generate an abnormal labeling knowledge graph KG. updated ; The medication recommendation subsystem is used to label the knowledge graph KG according to the abnormality updated ,generate personalized medication optimization recommendations by jointly learning graph structure and node attributes; RPA closed-loop management subsystem for medication optimization suggestions and abnormal tagging knowledge graph KG updated As input, closed-loop management is achieved by executing optimization suggestions and dynamically updating system rules; The structure of the knowledge graph KG is as follows: the node set represents the entity, the edge set represents the association relationship between the entities, and each node contains a context attribute vector representing the semantic feature.

Images