Medicine list processing method, electronic equipment, storage medium and program product

Abstract

The invention provides a medicine list processing method, electronic equipment, a storage medium and a program product, and relates to the technical field of medical insurance auditing, the processing method comprises the following steps: obtaining a medicine list corresponding to a target user, the medicine list comprising a plurality of medicines; taking the plurality of drugs as input, and determining a disease prediction result corresponding to the target user through a pre-constructed drug-disease association scoring model; and classifying the plurality of medicines according to the disease prediction result, and outputting a classification result. The possible diseases of the target user can be accurately predicted through the input drug list, the drugs are classified according to the disease prediction result, and efficient and reliable data support is provided for medical insurance auditing.

Description

Technical Field

[0001] The present application relates to the technical field of medical insurance review, and in particular to a method for processing a drug list, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the development of society and the increase in demand for medical insurance, the types and number of medical claims are increasing significantly. In this case, it is more difficult to judge the relationship between drugs and diseases. The traditional medical insurance review process relies on manual comparison of drug and disease indications. This method is inefficient, has a high error rate, and has the risk of human error, resulting in a long claims process. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method for processing a drug list, an electronic device, a storage medium, and a program product, which can accurately predict the possible diseases of the target user through an input drug list, and classify the drugs according to the disease prediction results, thereby providing efficient and reliable data support for medical insurance review.

[0004] To achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for processing a drug list, comprising the following steps:

[0006] Obtain a drug list corresponding to a target user, the drug list including a plurality of drugs; use the plurality of drugs as input, and determine a disease prediction result corresponding to the target user through a pre-built drug-disease association scoring model; classify the plurality of drugs according to the disease prediction result, and output a classification result.

[0007] In one embodiment, the plurality of drugs are used as input, and a pre-built drug-disease association scoring model is used to determine a disease prediction result corresponding to the target user, including:

[0008] For each drug in the drug list, the alternative disease corresponding to each drug and the scoring result between the drug and the alternative disease are determined according to a pre-constructed drug-disease association scoring model; based on the scoring results of the alternative diseases corresponding to multiple drugs, the comprehensive score of each alternative disease is determined; based on the comprehensive score of each alternative disease, the disease prediction result corresponding to the target user is determined.

[0009] In one embodiment, determining the disease prediction result corresponding to the target user according to the comprehensive score of each candidate disease includes:

[0010] The threshold is dynamically adjusted according to the comprehensive score of each candidate disease; the candidate disease with a comprehensive score greater than the threshold is used as the disease prediction result corresponding to the target user.

[0011] In one embodiment, a comprehensive score for each alternative disease is determined based on the scoring results of the alternative diseases corresponding to multiple drugs, including: if there are multiple drugs corresponding to the same alternative disease, then the comprehensive score of the alternative disease is the sum of the scoring results between the multiple drugs and the alternative disease; according to the hierarchical relationship of the disease, if there are multiple alternative diseases with the same parent disease code, then the comprehensive score of the multiple alternative diseases is the comprehensive score of the parent disease code, and the comprehensive score of the parent disease code is the sum of the scoring results of the multiple alternative diseases.

[0012] In one embodiment, the multiple drugs are classified according to the disease prediction results and the classification results are output, including: according to the drug-disease association scoring model, the drugs in the drug list that are associated with the disease prediction results are marked as related drugs for output; the drugs in the drug list that are not associated with the disease prediction results are marked as irrelevant drugs for output; according to the irresponsible medication data set, the irresponsible medications in the drug list are marked and output.

[0013] In one embodiment, before taking the multiple drugs as input and determining the disease prediction result corresponding to the target user through a pre-constructed drug-disease association scoring model, the method also includes: constructing a drug-disease knowledge graph based on drug indication instructions and disease types; determining an initial score of the drug based on the number of disease types associated with each drug in the drug-disease knowledge graph; and constructing the drug-disease association scoring model based on the initial score of the drug.

[0014] In one embodiment, the initial score of the drug is determined according to the number of disease types associated with each drug in the drug-disease knowledge graph, and is calculated using the following formula:

[0015] score=exp(-α×(num_diseases-1))

[0016] Among them, score is the initial score, exp is the exponential function, α is the adjustment parameter, and num_diseases represents the number of diseases associated with any drug.

[0017] In one embodiment, the drug-disease association scoring model is constructed based on the initial score of the drug, including: counting the association frequency between different drugs and different diseases in real case medication data; adjusting the initial score based on the association frequency between the different drugs and different diseases to obtain a final enhanced score for characterizing the degree of association between different drugs and different diseases; and constructing the drug-disease association scoring model based on the final enhanced score between different drugs and different diseases.

[0018] In one embodiment, the initial score is adjusted according to the association frequency between the different drugs and the different diseases to obtain a final enhanced score for characterizing the degree of association between the different drugs and the different diseases, which is calculated using the following formula:

[0019] enhanced_score=original_score×(1+λ1×dominance_factor)×(1+λ2×sigmoid_scaling)

[0020] Among them, enhanced_score is the final enhanced score, dominance_factor is the advantage factor, λ1 and λ2 are adjustment parameters used to control the weight of each factor, original_score is the initial association score, and sigmoid_scaling is the scaling factor.

[0021] The dominance factor is calculated using the following formula:

[0022] dominance_factor=drug_icd10_count / drug_count

[0023] Among them, drug_icd10_count is the number of times any drug matches any disease code in the real case medication data, and drug_count is the total number of times any drug is used in the real case medication data.

[0024] Among them, the scaling factor sigmoid_scaling is calculated using the following formula:

[0025] sigmoid_scaling=1 / (1+exp(-k×(drug_icd10_count-c)))

[0026] Among them, exp is the exponential function, k is an adjustable parameter used to control the center position and change rate of the Sigmoid function, and c is the median of the frequency data of drug-treated diseases.

[0027] In one embodiment, the multiple drugs are classified according to the disease prediction results. Before outputting the classification results, the method also includes: identifying abnormal medication situations and marking unreasonable medication based on the constructed gender-disease data set, age-disease data set and accidental disease data set, the patient's gender, age, insurance type and the disease prediction results input by the user.

[0028] According to a second aspect of the present application, there is provided an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the method for processing a drug list as described in the above embodiment.

[0029] According to a third aspect of the present application, a computer-readable storage medium is provided, wherein the storage medium stores a computer program, and the computer program can be executed by a processor to complete the method for processing a drug list as described in the above embodiment.

[0030] According to a fourth aspect of the present application, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the method for processing a drug list as described in the above embodiment.

[0031] The drug list processing method, electronic device, storage medium and product provided in the above embodiments can accurately predict the possible diseases of the target user through the input drug list by using the drug-disease association scoring model, thereby improving the accuracy of disease prediction, reducing the workload of manual review, reducing the possibility of human error, and improving the efficiency of medical insurance review. By intelligently classifying drugs, the rationality evaluation of drug use is optimized, which is conducive to the refined management of medical insurance. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A flowchart of a method for processing a drug list provided in an embodiment of the present application;

[0034] Figure 2 A flowchart of a pre-built drug-disease association scoring model provided in an embodiment of the present application;

[0035] Figure 3A flowchart of constructing a drug-disease association scoring model based on an initial drug score provided in an embodiment of the present application;

[0036] Figure 4 A flowchart for predicting diseases based on a drug list provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0038] Figure 1 A flowchart of a method for processing a drug list provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes the following steps S101-S103:

[0039] Step S101, obtaining a drug list corresponding to a target user, the drug list including multiple drugs;

[0040] Step S102: taking multiple drugs as input, and determining the disease prediction result corresponding to the target user through a pre-built drug-disease association scoring model;

[0041] Step S103: classify multiple drugs according to the disease prediction results and output the classification results.

[0042] Specifically, when obtaining the drug list corresponding to the target user in step S101, the disease information and drug information in the medical insurance policy are automatically parsed and identified through the large language model (LLM), and the input natural language description, such as the doctor's notes or medical record summary, is processed to improve the accuracy of disease prediction. In this process, the role of the large language model is not limited to parsing the drug name, it can also extract the patient's clinical information, medical history, and related symptoms. With this information, a multi-dimensional reasoning input can be formed to generate more accurate and personalized algorithm prediction results. The drug list can be one or more.

[0043] Figure 2 A flowchart of a pre-built drug-disease association scoring model provided in the embodiments of the present application, such as Figure 2 As shown, in some embodiments, before step S102, the following steps S201 to S203 are also included:

[0044] Step S201: construct a drug-disease knowledge graph based on drug indication instructions and disease types.

[0045] Specifically, collect the public drug indication instructions to determine the applicability of the drug. Each disease can be represented by the code in the 10th revised edition of the medical insurance version of ICD (International Classification of Diseases, International Classification of Diseases) (referred to as ICD-10). In one embodiment, real case medication data can also be used to clean and standardize the collected drug names and disease ICD-10 codes to ensure the consistency and accuracy of the drug names and ICD-10 versions, such as unified drug names. Then, integrate the drug-disease ICD10 coding data, connect the drugs and diseases in a structured form, store the association between the drugs and their applicable diseases, and build a comprehensive drug-disease ICD10 coding knowledge graph to provide a basis for scoring calculations and ensure the scientific nature of disease prediction results.

[0046] In order to ensure that the knowledge graph reflects the latest medical research and drug regulatory changes, the knowledge graph will also be updated regularly. When processing new data, the drug-disease matching model can be continuously optimized to adapt to the ever-changing medical practice and medical insurance policy requirements, so as to improve the accuracy and efficiency of the review process and provide data-driven decision support for the refined management of medical insurance.

[0047] Step S202: Determine the initial score of the drug according to the number of disease types associated with each drug in the drug-disease knowledge graph.

[0048] If a certain drug can treat a certain disease, then the drug and the disease are associated in the drug-disease knowledge graph. The initial score is used to characterize the degree of association between the drug and its associated disease.

[0049] In some embodiments, specifically, step S202 includes: based on the applicable scope of the drug, the initial score is calculated using the following formula:

[0050] score=exp(-α×(num_diseases-1))

[0051] Among them, score is the initial score, exp is the exponential function, α is the adjustment parameter, and num_diseases represents the number of diseases associated with any drug. The more diseases a drug is applicable to, the lower the initial score of the drug-ICD10 code, indicating that the drug is less targeted, has a wider applicability, and accounts for a lower proportion in disease prediction.

[0052] Step S203: construct a drug-disease association scoring model based on the initial drug scores.

[0053] Figure 3A flowchart of constructing a drug-disease association scoring model based on the initial score of the drug provided in the embodiment of the present application is as follows: Figure 3 As shown, in order to improve the accuracy of the algorithm in real clinical scenarios, in some embodiments, specifically, step S203 includes the following steps S301-S303:

[0054] Step S301: Count the association frequencies between different drugs and different diseases in real case medication data.

[0055] The actual clinical usage of drugs was analyzed based on the association frequency, and the results were used to enhance the calculation of the initial score of drug-ICD10 coding to improve the accuracy of the algorithm in real clinical scenarios.

[0056] Step S302: According to the association frequencies between different drugs and different diseases, the initial score is adjusted to obtain a final enhanced score for characterizing the association degree between different drugs and different diseases.

[0057] The association frequency may be the ratio of the number of times any drug matches any disease code in the real case medication data to the total number of times any drug is used in the real case medication data.

[0058] In some embodiments, the final enhanced score of the drug-ICD10 code can be calculated by combining the initial score, the dominance factor (ie, the association frequency) and the scaling factor sigmoid_scaling, and is calculated using the following formula:

[0059] enhanced_score=original_score×(1+λ1×dominance_factor)×(1+λ2×sigmoid_scaling)

[0060] Among them, enhanced_score is the final enhanced score, dominance_factor is the dominance factor; λ1 and λ2 are adjustment parameters used to control the weights of each factor to ensure the optimal model performance; original_score is the initial association score; sigmoid_scaling is the scaling factor;

[0061] Calculate the dominance factor to measure how common a drug is in treating a particular disease. It is calculated using the following formula:

[0062] dominance_factor=drug_icd10_count / drug_count,

[0063] Among them, drug_icd10_count is the number of times any drug matches any disease code in the real case medication data, and drug_count is the total number of times any drug is used in the real case medication data;

[0064] Among them, the scaling factor sigmoid_scaling is used to smooth the frequency data to avoid the influence of extreme values ​​on the score, and it is calculated using the following formula:

[0065] sigmoid_scaling=1 / (1+exp(-k×(drug_icd10_count-c)))

[0066] Among them, exp is the exponential function, k is an adjustable parameter used to control the center position and change rate of the Sigmoid function, and c is the median of the frequency data of drug-treated diseases.

[0067] Step S303: constructing a drug-disease association scoring model according to the final enhanced scores between different drugs and different diseases.

[0068] The drug-disease association scoring model can be used to characterize the degree of association between each drug and each disease, which can be represented by the final enhanced score. If the input list includes drug A, and the final enhanced score between drug A and a certain disease X is greater than a threshold, it can be considered that the candidate diseases of the target user include disease X. Through these steps, a comprehensive, accurate, and updatable drug-disease association scoring model is constructed, providing a solid foundation for intelligent drug list processing and disease prediction.

[0069] Figure 4 A flowchart for predicting a disease based on a drug list provided in an embodiment of the present application is as follows: Figure 4 As shown, in some embodiments, step S102 takes multiple drugs as input, and determines the disease prediction result corresponding to the target user through a pre-built drug-disease association scoring model, which specifically includes the following steps S401-S403:

[0070] Step S401: for each drug in the drug list, determine the candidate disease corresponding to each drug and the score result between the drug and the candidate disease according to the pre-constructed drug-disease association scoring model.

[0071] Step S402: Determine a comprehensive score for each candidate disease based on the scoring results of the candidate diseases corresponding to the multiple drugs.

[0072] In some embodiments, step S402 specifically includes:

[0073] If there are multiple drugs corresponding to the same candidate disease, the comprehensive score of the candidate disease is the sum of the scores between the multiple drugs and the candidate diseases. That is, for the ICD-10 disease code directly corresponding to the drug, the scores of all the same ICD-10 disease codes are accumulated to reflect the possibility of the disease under a certain drug combination.

[0074] According to the hierarchical relationship of diseases, if there are multiple alternative diseases with the same parent disease code, the comprehensive score of the multiple alternative diseases is the comprehensive score of the parent disease code, and the comprehensive score of the parent disease code is the sum of the scoring results of the multiple alternative diseases.

[0075] Specifically, the ICD-10 coding system is hierarchical, in which each code can have one or more sub-codes. In order to improve the accuracy and coverage of prediction, the hierarchical structure of disease ICD-10 codes needs to be considered. For example, if disease 1 (A01) and disease 2 (A02) are both subclasses of disease 3 (A03), this means that A01 and A02 are both specific classifications of A03.

[0076] For each drug in the drug list, a pre-built drug-disease association scoring model is used to determine the candidate disease (sub-class disease) corresponding to the drug. The candidate disease can be represented by the ICD-10 code. When multiple disease ICD-10 codes have the same parent code, the parent disease ICD-10 code is retained and the sub-class ICD-10 code is discarded. The score of the parent ICD-10 code is the sum of the sub-class ICD-10 code scores, but each drug can only contribute once at most.

[0077] For example, if the scores of disease 1 and disease 2 corresponding to drug A are 0.8 and 1 respectively, and disease 1 and disease 2 are both subcategories of disease 3, then the score of disease 3 is 1 point instead of 1.8 points.

[0078] When the disease ICD-10 codes corresponding to the drugs have a parent-child relationship, the sub-category ICD-10 codes are discarded and the score calculation is the same as above.

[0079] The scores of ICD-10 codes for diseases that do not match the patient's gender and age are set to 0.

[0080] Step S403: Determine the disease prediction result corresponding to the target user according to the comprehensive score of each candidate disease.

[0081] In some embodiments, step S403 specifically includes: dynamically adjusting the threshold according to the comprehensive score of each candidate disease; and taking the candidate disease with a comprehensive score greater than the threshold as the disease prediction result corresponding to the target user.

[0082] Exemplarily, in the process of determining the disease prediction results, the threshold is dynamically adjusted according to the comprehensive score of each candidate disease. If the score of a disease ICD-10 code is greater than 1, the threshold is set to 1; if the scores of all disease ICD-10 codes are less than 1, the threshold is set to the score of the disease ICD10 code with the highest score, so as to dynamically adjust the threshold. This makes the threshold adjustment more flexible, able to adapt to different situations, and improves the adaptability and accuracy of the prediction. Then, based on the disease ICD-10 code score corresponding to each drug, the final score of the predicted disease ICD-10 code is summarized and calculated, and the disease ICD-10 codes above the threshold are screened out to generate a predicted disease list.

[0083] In some embodiments, step S103 classifies multiple drugs according to the disease prediction results and outputs the classification results, which specifically includes:

[0084] According to the drug-disease association scoring model, the drugs in the drug list that are associated with the disease prediction results are marked as related drugs for output; the drugs in the drug list that are not associated with the disease prediction results are marked as irrelevant drugs for output; according to the irresponsible medication dataset, the irresponsible medications in the drug list are marked and output.

[0085] Specifically, in the process of determining the disease prediction results, the system will dynamically adjust the threshold according to the comprehensive score of each alternative disease. When the drug score is higher than the threshold, the system will identify it as a drug that contributes to the disease prediction results, and these drugs are classified as "related drugs." On the contrary, when the drug score is higher than the threshold, drugs that fail to contribute to the score are classified as "irrelevant drugs." If the ICD-10 code for the disease corresponding to a drug is in the non-responsible medication data set, the drug will be classified as non-responsible medication and the corresponding non-responsible disease will be attached to it. Non-responsible medication refers to those drugs that are not covered by medical insurance.

[0086] This classification method makes drug review more refined, helps to quickly identify related and irrelevant drugs, as well as irresponsible medications, and labels and classifies them during the drug list processing process. It can help medical insurance auditors quickly identify and process these drugs, ensure the rational use of medical insurance funds, improve audit efficiency, and provide a clear basis for subsequent audits and decisions.

[0087] In some embodiments, before classifying the multiple drugs according to the disease prediction results in step S103 and outputting the classification results, the process further includes:

[0088] Based on the constructed gender-disease dataset, age-disease dataset and accidental disease dataset, the patient's gender, age, insurance type and disease prediction results entered by the user are used to identify abnormal medication use and mark unreasonable medication use.

[0089] Specifically, the gender-disease dataset stores the ICD-10 disease codes for male- and female-specific diseases, the age-disease dataset stores the ICD-10 disease codes for diseases of special age groups and their corresponding age ranges, and the accidental disease dataset stores medical information related to accidents.

[0090] Specifically, this process is achieved by introducing multi-level algorithms to ensure in-depth analysis of drug use. The medical knowledge graph is used to evaluate the indications of drugs, and combined with clinical guidelines and medical insurance policies, drugs that do not match the patient's condition are automatically identified, and irrational drug use can be classified based on specific standards. This identification process covers multiple dimensions, such as gender inconsistency, age contraindications, and drugs that are not covered by medical insurance.

[0091] For example, a female patient is prescribed a drug designed for men, or a young patient is prescribed a drug that is usually only suitable for the elderly. These situations will be marked as potential abnormal medication use and labeled as irrational medication use.

[0092] For example, if the insurance policy stipulates that it only covers medical needs caused by accidents, then hypertension drugs may be identified as not covered by accident insurance claims, that is, the drug may be labeled as unreasonable use of drugs.

[0093] When the model generates output results, all drugs identified as irrational will be marked with specific classification reasons. This classification not only helps medical insurance reviewers quickly identify problematic drugs, but also provides a transparent decision-making basis to support subsequent review and intervention measures.

[0094] Furthermore, the method can also identify non-standard medications and specially mark them to distinguish them from conventional treatment options. For example, although some drugs were originally approved for specific diseases, doctors may use them to treat other diseases based on clinical experience and research evidence. Intelligent identification of non-standard medications helps medical insurance auditors quickly identify problematic drugs to support subsequent audits and interventions.

[0095] In some embodiments, it also includes combining the latest medical research and clinical guidelines, using knowledge graphs to provide medical basis for prediction results, and generating explanatory reports through large language models.

[0096] These reports not only show the results of the algorithm's predictions, but also explain the reasons in natural language to help auditors understand the basis and reasons for the predictions. In the final output text, a large language model is used to explain all drugs in detail, and natural language descriptions are generated based on the drug's indications and main diseases to explain its role in clinical treatment and related diseases. For example, if multiple drugs are used to treat hypertension, it will be judged whether this is a hypertension treatment process based on the weight and the patient's specific situation, and it will be labeled accordingly.

[0097] This professional explanation will provide medical insurance auditors with a deeper understanding and help them make more accurate and reasonable judgments during the audit process. In this way, auditors can not only clearly identify the drugs related to the patient's condition, but also effectively avoid reimbursement problems caused by drug misuse, thereby improving the accuracy and efficiency of medical insurance audits.

[0098] In some embodiments, for new drugs that are not yet covered in the database or drugs with different Chinese spellings, prompts will be generated to mark these drugs as needing to be added to the database, so as to achieve intelligent prompts and additions for unidentified drugs.

[0099] In some embodiments, before outputting the results of predicting the disease, the results are also reordered, merged, and filtered. Specifically, common diseases are sorted in descending order according to the prevalence of the disease, and rare or major diseases are displayed first to improve the value of clinical decision-making. Similar diseases coded in ICD-10 are merged to simplify the display of the prediction results and enhance readability and applicability. The prediction results are further screened according to the patient's age, gender, and other health information to ensure that the results are consistent with the individual patient.

[0100] In some embodiments, the output of the predicted disease results includes structured output and user-friendly display. Specifically, the prediction results will be returned in the form of a structured dictionary and summarized by a large language model (LLM) to generate a clear natural language report. Through natural language description, the readability and operability of the information are improved to help medical insurance reviewers quickly understand key information and achieve user-friendly display.

[0101] In one embodiment, the specific process of implementing the drug list processing is as follows:

[0102] First, a drug list is entered for a male patient, including drug A, drug B, drug C, and drug D, as well as the patient's gender (male), age (45 years old), and insurance type (comprehensive health insurance). The system receives the drug list and extracts the drug name and related information.

[0103] Then, the knowledge graph of drug-disease-ICD-10 code is queried to determine the association between drugs A, B, C and D and possible diseases. Based on the drug-ICD-10 code association scoring model, the patient's possible diseases are predicted, such as disease X (ICD-10 code: A01) and disease Z (ICD-10 code: B02). This step involves the calculation of the score of the association between drugs and diseases, as well as the processing of the comprehensive score and hierarchical relationship of the candidate diseases.

[0104] Finally, the input drugs are classified and marked before irrational drug use. For example, drugs A and B are classified as related drugs related to diseases X and Z, and drug C is classified as an irrelevant drug because it is not related to the predicted disease. Based on the irresponsible drug use dataset, the system identifies that drug D is not covered by comprehensive health insurance, so it is marked as irresponsible drug use. The system further analyzes the patient's gender, age, and disease prediction results, and confirms that although drug A is a related drug, it is also marked as irrational drug use because it is not usually used for male patients. The classification results are as follows:

[0105] Related drugs: (sex abnormality) drug A, drug B (related to diseases X and Z);

[0106] Unrelated drugs: Drug C (not related to the predicted disease);

[0107] Non-responsible medication: Drug D (not covered by insurance).

[0108] The structured prediction results (including the annotation information of predicted diseases, related drugs, irrelevant drugs, and irresponsible drug use) are input into the large language model to generate an explanation report in natural language. The output report shows that the patient may have diseases X and Z, drugs A and B are related to these diseases, drug C is irrelevant to the predicted disease, drug D is not covered by insurance, and drug A is also marked as irrational drug use. The report is presented in natural language to facilitate user understanding and review.

[0109] This output not only provides predicted disease outcomes, but also provides users with detailed explanations about drug use, helping users understand the relationship between drugs and diseases, and the basis for insurance audits.

[0110] An embodiment of the present application also provides an electronic device, which may include a processor and a memory for storing instructions executable by the processor, and the processor is configured to implement the steps of the processing method of any of the above embodiments.

[0111] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. The computer program can be executed by a processor to complete the steps of the drug list processing method of any of the above embodiments.

[0112] An embodiment of the present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the method for processing a drug list of any of the above embodiments.

[0113] In summary, the technical solution provided by the embodiment of the present invention has the following advantages: by combining a large language model with real case medication data, intelligent drug list processing and disease prediction are realized, and the accuracy and clinical relevance of the prediction are improved.

[0114] The automated process significantly improves the efficiency of medical insurance review, while providing data support to assist decision makers in formulating reasonable medical insurance policies and optimizing resource allocation.

[0115] Taking into account the individual characteristics of patients and combining the ICD-10 coding hierarchy, personalized disease prediction and management are carried out to enhance the scientificity and applicability of the prediction results.

[0116] Improve the transparency and explainability of audit results through natural language reports, enhancing user trust and satisfaction.

[0117] The system has self-learning capabilities and can adapt to changes in medical practice and medical insurance policies, continuously optimize disease prediction and drug classification models, and maintain the system's advancement and adaptability.

[0118] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative.

[0119] It should be noted that if the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can essentially be embodied in the form of a software product, or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM) random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0120] The above are only embodiments of the present application and are not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for processing a drug list, characterized in that: The following steps are involved: Obtaining a drug list corresponding to a target user, wherein the drug list includes multiple drugs; Taking the multiple drugs as input, determining the disease prediction result corresponding to the target user through a pre-built drug-disease association scoring model; The multiple drugs are classified according to the disease prediction results, and the classification results are output.

2. The method according to claim 1, characterized in that The method of using the multiple drugs as input and determining the disease prediction result corresponding to the target user through a pre-built drug-disease association scoring model includes: For each drug in the drug list, determine the candidate disease corresponding to each drug and the score result between the drug and the candidate disease according to a pre-constructed drug-disease association scoring model; Determine a comprehensive score for each candidate disease based on the scoring results of the candidate diseases corresponding to the multiple drugs; According to the comprehensive score of each candidate disease, the disease prediction result corresponding to the target user is determined.

3. The method according to claim 2, characterized in that Determining the disease prediction result corresponding to the target user according to the comprehensive score of each candidate disease includes: The threshold is adjusted dynamically according to the comprehensive score of each candidate disease; The candidate diseases with the comprehensive scores greater than the threshold are used as the disease prediction results corresponding to the target users.

4. The method according to claim 2, characterized in that: Determining the comprehensive score of each candidate disease according to the scoring results of the candidate diseases corresponding to the multiple drugs includes: If there are multiple drugs corresponding to the same candidate disease, the comprehensive score of the candidate disease is the sum of the score results between the multiple drugs and the candidate disease; According to the hierarchical relationship of diseases, if there are multiple alternative diseases with the same parent disease code, the comprehensive score of the multiple alternative diseases is the comprehensive score of the parent disease code, and the comprehensive score of the parent disease code is the sum of the scoring results of the multiple alternative diseases.

5. The method according to claim 1, characterized in that: Classifying the multiple drugs according to the disease prediction results and outputting classification results, including: According to the drug-disease association scoring model, marking the drugs in the drug list that are associated with the disease prediction result as related drugs for output; Marking the drugs in the drug list that are not related to the disease prediction result as irrelevant drugs for output; According to the irresponsible medication dataset, irresponsible medication in the medication list is marked and output.

6. The method according to claim 1, characterized in that Before taking the multiple drugs as input and determining the disease prediction result corresponding to the target user through the pre-built drug-disease association scoring model, the method further includes: Construct a drug-disease knowledge graph based on drug indications and disease types; Determining an initial score of the drug according to the number of disease types associated with each drug in the drug-disease knowledge graph; The drug-disease association scoring model is constructed based on the initial score of the drug.

7. The method according to claim 6, characterized in that The initial score of the drug is determined according to the number of disease types associated with each drug in the drug-disease knowledge graph, and is calculated using the following formula: score=exp(_α×(num_diseases_1)) Among them, score is the initial score, exp is the exponential function, α is the adjustment parameter, and num_diseases represents the number of diseases associated with any drug.

8. The method according to claim 6, characterized in that The step of constructing the drug-disease association scoring model according to the initial score of the drug comprises: Count the association frequencies between different drugs and different diseases in real case medication data; According to the association frequencies between the different drugs and the different diseases, the initial scores are adjusted to obtain final enhanced scores for characterizing the association degrees between the different drugs and the different diseases; According to the final enhancement scores between different drugs and different diseases, the drug-disease association scoring model is constructed.

9. The method according to claim 8, characterized in that The initial score is adjusted according to the association frequency between the different drugs and the different diseases to obtain a final enhanced score for characterizing the degree of association between the different drugs and the different diseases, which is calculated using the following formula: enhanced_score=original_score×(1+λ1×dominance_factor)×(1+λ2×sigmoid_scaling) Among them, enhanced_score is the final enhanced score, dominance_factor is the dominance factor, λ1 and λ2 are adjustment parameters used to control the weight of each factor, original_score is the initial association score, and sigmoid_scaling is the scaling factor; The dominance factor is calculated using the following formula: dominance_factor=drug_icd10_count / drug_count, Among them, drug_icd10_count is the number of times any drug matches any disease code in the real case medication data, and drug_count is the total number of times any drug is used in the real case medication data; Among them, the scaling factor sigmoid_scaling is calculated using the following formula: sigmoid_scaling=1 / (1+exp(_k×(drug_icd10_count_ c))) Among them, exp is the exponential function, k is an adjustable parameter used to control the center position and change rate of the Sigmoid function, and c is the median of the frequency data of drug-treated diseases.

10. The method according to claim 1, characterized in that Before classifying the plurality of drugs according to the disease prediction result and outputting the classification result, the method further includes: Based on the constructed gender-disease dataset, age-disease dataset and accidental disease dataset, the patient's gender, age, insurance type and the disease prediction results entered by the user, abnormal medication use is identified and unreasonable medication use is marked.

11. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to execute the method for processing a drug list as described in any one of claims 1-10.

12. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program can be executed by a processor to complete the method for processing a drug list as described in any one of claims 1-10.

13. A computer program product, characterized in that The invention comprises a computer program / instruction, which, when executed by a processor, implements the method for processing a drug list as described in any one of claims 1 to 10.

Images