Medical reimbursement auditing method, related device, equipment and storage medium

Abstract

The invention discloses a medical reimbursement auditing method, a related device, equipment and a storage medium, and the method comprises the steps: carrying out the analysis based on a reimbursement condition of a to-be-reimbursed medical object, obtaining a plurality of sub-conditions needing to be judged when checking whether the reimbursement condition is met or not, a plurality of medical record chapters needing to be concerned when the sub-conditions are judged, and a logical relationship between the sub-conditions; based on a plurality of medical record chapters needing to be concerned when the sub-conditions are judged, extracting reference information needing to be concerned when the sub-conditions are judged from the medical record information; performing analysis on the basis of reference information needing to be concerned when the sub-conditions are judged to obtain sub-judgment results of the sub-conditions; wherein the sub-judgment result comprises whether a sub-condition is met or not; and determining a reimbursement audit result of the to-be-reimbursed medical object based on the respective sub-judgment results of the sub-conditions and the logical relationship between the sub-judgment results. According to the scheme, the auditing efficiency and auditing precision of medical reimbursement auditing can be improved.

Description

Technical Field

[0001] The present application relates to the field of natural language processing technology, and in particular to a medical reimbursement review method and related devices, equipment and storage media. Background Art

[0002] Medical reimbursement review refers to the review of reimbursement of relevant medical drugs, medical devices, etc. according to specific restrictions in the reimbursement catalog. If the specific restrictions are met, they can be reimbursed, otherwise they will not be reimbursed. For example, the restriction condition of the medical drug "Proline Henggliflozin Tablets" in the 2024 drug catalog is "limited to adult patients with type 2 diabetes."

[0003] At present, whether it is traditional methods such as manual review or new methods such as artificial intelligence models, they are all limited by the complex logic of restrictive conditions and the long text of medical records, which makes it difficult to ensure the review efficiency and accuracy. In view of this, how to improve the review efficiency and accuracy of medical reimbursement review has become an urgent problem to be solved. Summary of the invention

[0004] The main technical problem solved by this application is to provide a medical reimbursement review method and related devices, equipment and storage media, which can improve the review efficiency and review accuracy of medical reimbursement review.

[0005] In order to solve the above-mentioned technical problems, the first aspect of the present application provides a medical reimbursement review method, including: parsing based on the reimbursement conditions of the medical object to be reimbursed, obtaining several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition; based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, extracting reference information that needs to be paid attention to when judging the sub-conditions from the medical record information; analyzing based on the reference information that needs to be paid attention to when judging the sub-conditions, and obtaining sub-judgment results of the sub-conditions; wherein the sub-judgment results include whether the sub-conditions are met; based on the sub-judgment results of each sub-condition and the logical relationship between them, determining the reimbursement review result of the medical object to be reimbursed.

[0006] In order to solve the above-mentioned technical problems, the second aspect of the present application provides a medical reimbursement review device, including: a condition parsing module, an information extraction module, a condition judgment module and a comprehensive analysis module, the condition parsing module is used to parse based on the reimbursement conditions of the medical object to be reimbursed, and obtain a number of sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, a number of medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition; the information extraction module is used to extract reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on the number of medical record chapters that need to be paid attention to when judging the sub-conditions; the condition judgment module is used to analyze based on the reference information that needs to be paid attention to when judging the sub-conditions, and obtain a sub-judgment result of the sub-condition; wherein the sub-judgment result includes whether the sub-condition is met; the comprehensive analysis module is used to determine the reimbursement review result of the medical object to be reimbursed based on the sub-judgment results of each sub-condition and the logical relationship between them.

[0007] In order to solve the above technical problems, the third aspect of the present application provides an electronic device, which at least includes a memory and a processor coupled to each other, the memory at least stores program instructions, and the processor is used to execute the program instructions to implement the medical reimbursement review method in the above first aspect.

[0008] In order to solve the above technical problems, the fourth aspect of the present application provides a computer-readable storage medium storing program instructions that can be executed by a processor, and the program instructions are used to implement the medical reimbursement review method of the first aspect.

[0009] The above scheme is based on the analysis of the reimbursement conditions of the object to be reimbursed, and obtains several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions. Based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, the reference information that needs to be paid attention to when judging the sub-conditions is extracted from the medical record information, so as to analyze the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment results of the sub-conditions, and the sub-judgment results include whether the sub-conditions are met, and then based on the respective sub-judgment results of each sub-condition and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined. Therefore, on the one hand, due to reliance on manual review, the review efficiency can be improved to a certain extent. On the one hand, by disassembling the reimbursement conditions, it is helpful to break the reimbursement conditions into parts, form several logically simple sub-conditions as much as possible, and judge each sub-condition separately, which can further improve the audit efficiency. On the other hand, each sub-condition also parses out several medical record chapters that need to be paid attention to when judging, so as to selectively extract reference information from relevant medical record chapters when judging sub-conditions, which helps to break the limitations of long text on semantic understanding as much as possible, and can improve the audit accuracy to a certain extent. On the other hand, the logical relationship between each sub-condition is also parsed, and finally the logical relationship between the sub-conditions constrains the sub-judgment results of each sub-condition to determine the reimbursement audit results of the medical object to be reimbursed, which can further improve the audit accuracy. Therefore, the audit efficiency and audit accuracy of medical reimbursement audit can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a flowchart of an embodiment of the medical reimbursement review method; Figure 2a is a schematic diagram of a process of extracting reference information according to an embodiment; Figure 2b It is a schematic diagram of the process of an embodiment of sub-condition judgment; Figure 2c This is a process diagram of an embodiment of the medical reimbursement review method; Figure 3 It is a schematic diagram of the framework of an embodiment of the medical reimbursement review device of the present application; Figure 4 It is a schematic diagram of the framework of an embodiment of the electronic device of the present application; Figure 5 It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0011] The scheme of the embodiment of the present application is described in detail below in conjunction with the drawings of the specification.

[0012] In the following description, for the purpose of explanation rather than limitation, specific details such as specific system structures, interfaces, and technologies are provided to facilitate a thorough understanding of the present application.

[0013] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the fragment " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0014] See also Figure 1 , Figure 1 This is a flowchart of an embodiment of the medical reimbursement review method of the present invention. Specifically, it may include the following steps: Step S11: Analyze the reimbursement conditions of the medical object to be reimbursed to obtain several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions.

[0015] In one implementation scenario, medical objects to be reimbursed may include but are not limited to: medicines to be reimbursed, equipment to be reimbursed, treatments to be reimbursed (such as therapeutic items such as transplants), services to be reimbursed (such as non-therapeutic items such as physical examinations and beauty treatments), etc. The specific categories and contents of the reimbursement objects are not limited here.

[0016] In an implementation scenario, the reimbursement conditions may specifically be the restriction conditions of the object to be reimbursed in the relevant catalog. For example, the reimbursement conditions of the object to be reimbursed can be obtained from catalogs such as drug catalogs, diagnosis and treatment catalogs, and medical service facilities, and their specific contents are not limited here. For example, taking the medical drug "Proline Henggliflozin Tablets" as an example, its reimbursement conditions may include but are not limited to "limited to adult patients with type 2 diabetes". In the case where the object to be reimbursed is other objects, it can be deduced by analogy, and no examples are given one by one here.

[0017] In an implementation scenario, the sub-conditions may represent the audit elements (i.e., audit points) contained in the reimbursement conditions, such as but not limited to: medical concepts such as "type 2 diabetes" and "chronic renal failure", basic concepts such as "adult" and "female", or clear audit standards such as "a maximum of 9 devices per eye" and "payment for no more than 5 days".

[0018] In an implementation scenario, the medical record chapters may include, but are not limited to: admission record, first medical course, daily medical course, examination report, test report, etc., without any limitation on the medical record chapters.

[0019] In an implementation scenario, the logical relationship can represent that each sub-condition can be judged whether the reimbursement condition is met through a specific logical combination. It should be noted that the logical relationship may include but is not limited to: and, or, not, etc., and the specific combination of the logical relationship is not limited. For example, if the reimbursement condition is parsed to obtain three sub-conditions A, B, and C, the logical relationship between the three can be: the logical combination of A, B, and C, that is, if all three sub-conditions are judged to be true, it can be determined that the reimbursement condition is met; or, the logical relationship between the three can be: the logical combination of A and (B or C), that is, at least one of sub-condition B and sub-condition C is true and A must be true, then it can be determined that the reimbursement condition is met. Of course, the above examples are only a few possible examples of logical relationships in actual applications, and the specific content of the logical relationship is not limited here.

[0020] In an implementation scenario, as a possible implementation method, the reimbursement conditions can be parsed using traditional network models such as BERT to obtain several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition. In order to improve the parsing accuracy of the traditional network model, the traditional network model can be trained before this. For example, several sample reimbursement conditions can be collected, and the sample reimbursement conditions can be annotated with: several sample sub-conditions that need to be judged when reviewing whether the sample reimbursement conditions are met, several sample medical record chapters that need to be paid attention to when judging the sample sub-conditions, and the logical relationship between each sample sub-condition. On this basis, the sample reimbursement conditions are parsed based on the above-mentioned traditional network model to obtain several predicted sub-conditions that need to be judged when reviewing whether the sample reimbursement conditions are met, several predicted medical record chapters that need to be paid attention to when judging the predicted sub-conditions, and the logical relationship between each predicted sub-condition, and other prediction information, and then the network parameters of the above-mentioned traditional network model can be adjusted based on the difference between the annotated information and the predicted information.

[0021] In another implementation scenario, different from the aforementioned implementation, as another possible implementation, a first instruction can also be constructed based on the reimbursement conditions and the instruction template for indicating the condition parsing, and the first instruction is at least used to instruct the large language model to parse the reimbursement conditions, and then the first instruction is input into the first large model to obtain several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition. It should be noted that the large language model may include but is not limited to: open source large models such as LLAMA, Bloom, etc., and may also include but is not limited to large language models obtained by fine-tuning the parameters of open source large models based on specific corpora, and may also include but is not limited to custom large models. The specific source of the large language model is not limited here. Compared with the audit scheme based on structured manual annotation resources, the above method automatically extracts relevant content such as sub-conditions through the large model and determines the audit logic by itself, which saves the process of manual annotation resources and greatly improves the system development efficiency and iteration speed.

[0022] In a specific implementation scenario, as a possible example, the instruction template for indicating conditional parsing may include but is not limited to the following: Large model input template: If a certain reimbursement object restricts the use of [{reimbursement conditions}] patients, please design a specific The judgment steps of the restriction condition "{reimbursement conditions}" and the possible medical records involved are listed according to the judgment steps. Surround.

[0023] Large model output template: Based on the restriction condition "{reimbursement conditions}", the key audit elements that need to be paid attention to are: {sub-condition 1, sub-condition 2,…} The following judgment steps can be obtained: {Logical relationship between sub-conditions} At the same time, the above judgment steps may involve the following medical records: {Several medical record chapters involved in sub-condition 1, several medical record chapters involved in sub-condition 2, …}.

[0024] It should be noted that the above instruction template for indicating conditional parsing is only a possible example in actual application, and does not limit other possible instruction templates. The specific content of the instruction template is not limited here.

[0025] In a specific implementation scenario, as a possible example, taking the reimbursement condition "limited to adult type 2 diabetes patients" as an example, the first instruction input into the first large model may include but is not limited to the following: If a certain reimbursement object is limited to "only adult patients with type 2 diabetes" The judgment steps for the restriction condition "patients with type 2 diabetes" are listed according to the judgment steps. Capacity range.

[0026] On this basis, the output content of the first model may include but is not limited to the following: According to the restriction condition "{limited to adult patients with type 2 diabetes}", the key audit elements that need to be paid attention to are: {Adult, type 2 diabetes} The following judgment steps can be obtained: {Step 1: Determine whether the patient is an adult. If so, continue to determine. Otherwise, the patient is deemed not to meet the reimbursement conditions. Item; Step 2: Determine whether the patient has type 2 diabetes. If yes, continue to determine if the patient does not meet the criteria for reporting. Sales conditions; Step 3: Determine whether the patient meets the above conditions at the same time. If so, the patient is considered to meet the reimbursement conditions. Otherwise The patient is considered not to meet the reimbursement conditions} At the same time, the above judgment steps may involve the following medical records: {The first step may involve the medical record cover, admission record, first course of illness, and discharge record; The second step may involve the medical record cover, admission record, first course of illness, discharge record, test results, and daily course of illness}.

[0027] That is to say, when the reimbursement condition is "limited to adult patients with type 2 diabetes", several sub-conditions include: "adults", "type 2 diabetes", and the logical relationship between the sub-condition "adults" and the sub-condition "type 2 diabetes" is "and", that is, they need to be met at the same time to determine that the reimbursement condition is met. The several medical record chapters involved in the sub-condition "adults" include: medical record homepage, admission record, first course of illness, discharge record, and the medical record chapters involved in the sub-condition "type 2 diabetes" include: medical record homepage, admission record, first course of illness, discharge record, test results, and daily course of illness. In addition, it should be noted that the above output result of the first large model is only a possible example in the actual application process, and other possible output results are not limited thereto. The specific content of the output result of the first large model is not limited here.

[0028] Step S12: Based on the several medical record sections that need to be paid attention to when judging the sub-conditions, reference information that needs to be paid attention to when judging the sub-conditions is extracted from the medical record information.

[0029] In an implementation scenario, as a possible example, for each sub-condition, several medical record chapters and their contents that need to be paid attention to when judging the sub-condition can be selected as reference information that needs to be paid attention to when judging the sub-condition. Still taking the aforementioned reimbursement condition "limited to adult type 2 diabetes patients" as an example, for the sub-condition "adult", the chapter contents of the four medical record chapters "medical record homepage", "admission record", "first course of illness", and "discharge record" can be selected as reference information that needs to be paid attention to when judging the sub-condition "adult"; similarly, for the sub-condition "type 2 diabetes", the chapter contents of the six medical record chapters "medical record homepage", "admission record", "first course of illness", "discharge record", "test results", and "daily course of illness" can be selected as reference information that needs to be paid attention to when judging the sub-condition "type 2 diabetes". Of course, the above example is only a possible example of reference information when taking the reimbursement condition "limited to adult type 2 diabetes patients" as an example, and other possible situations will not be given examples one by one here.

[0030] In another implementation scenario, please refer to Figure 2a , Figure 2a FIG. 1 is a schematic diagram of a process of extracting reference information according to an embodiment. Figure 2a As shown, as another possible example, different from the aforementioned implementation, for each sub-condition, it is also possible to obtain the medical record content in each medical record chapter that needs to be paid attention to when judging the sub-condition, and extract the key information in the chapter content for judging the sub-condition. On this basis, based on the key information in each medical record chapter that needs to be paid attention to when judging the sub-condition, a medical record summary can be integrated as reference information that needs to be paid attention to when judging the sub-condition. In the above method, for each sub-condition, by extracting key information from the relevant medical record chapters and integrating the information to form a medical record summary, the amount of text when judging the sub-condition can be simplified as much as possible, and only retaining a small amount of key content can effectively prevent irrelevant information from interfering with the judgment of the sub-condition.

[0031] In a specific implementation scenario, as a possible example, a traditional network model such as BERT can be used to extract key information from the chapter content. For details, please refer to the technical details of traditional network models such as BERT, which will not be repeated here; or, as another possible example, a large language model can also be used to extract key information from the chapter content. For example, for each sub-condition, a second instruction can be constructed based on the sub-condition, the chapter content involved in the sub-condition, and the instruction template for information extraction, and the second instruction is used to instruct the large language model to extract key information from the chapter content for judging the sub-condition, and then input the second instruction to the second large model to obtain the key information in the chapter content for judging the sub-condition. As a possible example, the instruction template for information extraction may include but is not limited to the following: Large model input template: Please follow the above judgment steps to identify the possible The scope of medical record content is as follows: {Chapter content}.

[0032] It should be noted that the above instruction template for information extraction is only a possible example in the actual application process, and does not limit the specific content of the instruction template. No examples are given here. Taking the sub-condition "type 2 diabetes" as an example, when the medical record section involved includes "daily course of disease on a single day", the following second instruction can be input into the second largest model: Identify the range of medical records that may be involved in the judgment of [{Type 2 diabetes}] and extract the Key information, complete medical record content is as follows: {Daily Course of Disease Today, I accompanied Chief Physician Zhou xx on the ward rounds. The patient had no obvious special discomfort, his spirit, appetite and sleep were good, his bowel movements were normal, and his small There is a small amount of foam in the stool. Physical examination: temperature: 36.5℃, pulse: 76 times / min, respiration: 19 times / min, blood pressure: 116 / 76mmHg, development Normal, well-nourished, cooperative on physical examination. No superficial lymph nodes or thyroid gland enlargement were palpable. Cardiac, lung, and abdominal examination No obvious abnormality was found. No pigmentation was found on the skin of the whole body. No edema was found in both lower limbs, and the dorsalis pedis artery pulsation was still strong. Limb muscle strength 5 Grade, muscle tone is normal. Knee tendon and Achilles tendon reflexes are normal, biceps and triceps tendon reflexes are normal, bilateral Babins sign, Oppenham Negative for sputum, Gordon's sign. No obvious abnormality was found in liver and kidney function and infectious disease series. Auxiliary examinations: blood routine, liver and kidney function, coagulation function No obvious abnormality was found. Yesterday's blood sugar monitoring: 11.9mmol / l before breakfast, 12.9mmol / l after breakfast, 2 hours after lunch Blood sugar was 13.4mmol / L, 2 hours after dinner, 13.3mmol / L, and before bedtime, 10.3mmol / L. Instructions: 1. The patient's blood sugar levels before and after meals were still high yesterday, so sitagliptin combined with metformin was used to strengthen the blood sugar-lowering treatment. The patient should follow a balanced diet for diabetes and continue to monitor blood sugar and make adjustments based on blood sugar. 2. The patient's back pain has not recurred, so it is considered that the stones have been expelled. The urine sediment will be rechecked tomorrow to confirm the presence of occult blood in the urine. The rest of the treatment is the same as before.

[0033] After inputting the second instruction to the second largest model, as a possible example, the second largest model may output the following content as key information of the sub-condition "type 2 diabetes" in the daily course of the disease: According to the judgment requirements of "type 2 diabetes", the key information extracted from the daily course of disease is: [The patient's mental state, appetite and sleep are good, bowel movements are normal, urine has a small amount of foam, blood sugar monitoring: before breakfast yesterday 11.9mmol / l, 12.9mmol / l after breakfast, 13.4mmol / L 2 hours after lunch, 2 hours after dinner 13.3mmol / L, blood sugar before bedtime 10.3mmol / L. After seeing the patient, Director Zhou instructed: 1. The patient's blood sugar before and after meals was monitored yesterday. If the blood sugar level is still high, sitagliptin combined with metformin is added to strengthen the hypoglycemic treatment. The patient is advised to have a balanced diet and continue to monitor blood sugar levels. Sugar, adjusted according to blood sugar].

[0034] Still taking the sub-condition "type 2 diabetes" as an example, when the medical record section involved includes "test results", the second instruction can be constructed with reference to the above example and input to the second largest model, so as to use the following output content of the second largest model as the key information of the sub-condition "type 2 diabetes" in the test results: According to the need to judge type 2 diabetes, the key information extracted from the test results is:

【Blood lipid four items Blood lipid four items 125 Glucose 7.29mmol / L 7.29↑ Glycated hemoglobin determination Glycated hemoglobin determination 3103 Glycated hemoglobin 6.40% 6.40↑].

[0035] It should be noted that the above examples are only possible examples of extracting key information from "daily course of disease" and "test results" based on the sub-condition "type 2 diabetes", and other possible situations will not be given one by one here.

[0036] In a specific implementation scenario, after extracting the key information, as a possible example, a traditional network model such as BERT can be used to integrate the key information in each medical record chapter that needs to be paid attention to when judging the sub-conditions to obtain a medical record summary. For details, please refer to the technical details of traditional network models such as BERT, which will not be repeated here; or, as another possible example, a medical record summary can also be formed based on a large language model. For each sub-condition, a third instruction can be constructed based on the key information in each medical record chapter and the instruction template for summary integration, and the third instruction is used to instruct the large language model to integrate the key information to form a medical record summary, and input the third instruction to the third large model to obtain the medical record summary that needs to be paid attention to when judging the sub-condition. Still taking the aforementioned sub-condition "type 2 diabetes" as an example, the following third instruction can be constructed, including but not limited to: Please summarize the key information extracted above to form a medical record summary for the judgment of [Type 2 Diabetes] want.

[0037] On this basis, the third instruction can be input into the third model to obtain the medical record summary of the sub-condition "type 2 diabetes" as reference information for judging the sub-condition "type 2 diabetes", which may specifically include but is not limited to the following: Based on the need for judgment of type 2 diabetes, the summarized medical records are as follows:

【Admission Record Current medical history: The patient developed dizziness and loss of balance without obvious cause 1 year ago, and was unable to move after falling. The patient felt that the above symptoms had worsened, accompanied by headache, nasal congestion, and general discomfort. He came to our hospital for further treatment. Sequelae, cerebral infarction, hypertension grade 3 (extremely high risk), type 2 diabetes, upper respiratory tract disease” were admitted to our department.

[0038] Past medical history: He had a history of diabetes for more than 8 years, and usually took metformin hydrochloride sustained-release tablets and acarbose capsules orally. Blood sugar control was unknown.

[0039] Western medicine diagnosis: sequelae of cerebral hemorrhage, cerebral infarction; hypertension grade 3 (extremely high risk); type 2 diabetes; upper respiratory tract disease sick.

[0040] Blood cell analysis: Neutrophil ratio: 70.9%; B-type natriuretic peptide precursor 299.4 pg / ml. Glycated hemoglobin White 6.4% urobilin +; heparin binding protein 46.74ng / mL Blood biochemistry series: glucose 7.29 mmol / L, high-density lipoprotein 1.01 mmol / L, erythrocyte sedimentation rate 55 mm / h; electrolysis Quality: K 4.33 mmol / L Na 137.6mmol / L (This hospital 2022-03-29) Daily course of illness The patient's mental state, appetite and sleep are good, bowel movements are normal, and there is a small amount of foam in the urine. Blood sugar monitoring: before breakfast yesterday 11.9mmol / l, 12.9mmol / l after breakfast, 13.4mmol / L 2 hours after lunch, 2 hours after dinner 13.3mmol / L, blood sugar before bedtime 10.3mmol / L. After seeing the patient, Director Zhou instructed: 1. The patient's blood sugar before and after meals was monitored yesterday. If the blood sugar level is still high, sitagliptin combined with metformin is added to strengthen the hypoglycemic treatment. The patient is advised to have a balanced diet and continue to monitor blood sugar levels. Sugar, adjusted according to blood sugar.

[0041] Test results Blood lipid four items Blood lipid four items 125 Glucose 7.29mmol / L 7.29↑ Glycated hemoglobin determination Glycated hemoglobin determination 3103 Glycated hemoglobin 6.40% 6.40↑].

[0042] It should be noted that the above example is only a possible example of the medical record summary of the sub-condition "type 2 diabetes", and other possible situations will not be given one by one here.

[0043] Step S13: Analyze the reference information that needs to be paid attention to when judging the sub-conditions to obtain the sub-judgment results of the sub-conditions.

[0044] In the embodiment of the present disclosure, the sub-judgment result includes whether the sub-condition is satisfied. Exemplarily, the sub-judgment result may include but is not limited to the following possible situations: satisfying the sub-condition, not satisfying the sub-condition, and being unable to determine whether the sub-condition is satisfied. The actual satisfaction of the sub-condition is not specifically limited here.

[0045] In one implementation scenario, as a possible example, for each sub-condition, a prompt instruction can be constructed based on the sub-condition, the reference information that needs to be paid attention to when judging the sub-condition, and the instruction template used for condition judgment, and the prompt instruction is used to instruct the large language model to judge whether the sub-condition is met with the assistance of the reference information, and input the prompt instruction to the fifth large model to obtain the sub-judgment result of the sub-condition.

[0046] In another embodiment, as another possible example, different from the above embodiment, before analyzing in combination with reference information, it is also possible to refine based on sub-conditions to obtain judgment rules of sub-conditions. It should be noted that the judgment rules of sub-conditions may specifically include but are not limited to: quantitative indicators of sub-conditions. Still taking the sub-condition "type 2 diabetes" as an example, its judgment rules may include but are not limited to quantitative indicators of "type 2 diabetes" (such as how many values ​​of blood sugar are reached, etc.); or, still taking the sub-condition "adult" as an example, its judgment rules may include but are not limited to quantitative indicators of "adult" (such as how many years old the age is). On this basis, it is possible to analyze based on the sub-conditions and their judgment rules and the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment results of the sub-conditions. In the above manner, by refining the sub-conditions to obtain their judgment rules, and then combining the sub-conditions and their judgment rules and the reference information that needs to be paid attention to when judging the sub-conditions, the sub-judgment results of the sub-conditions are obtained, which can save a lot of manpower compared to manual labeling, and refine the sub-conditions to obtain their judgment rules, so that the sub-conditions can be judged as objectively as possible.

[0047] In a specific implementation scenario, please refer to Figure 2b , Figure 2b Schematic diagram of the process of sub-condition judgment of an embodiment. Figure 2b As shown, a fourth instruction can be constructed based on the sub-condition and the instruction template for condition refinement, and the fourth instruction is used to instruct the large language model to refine the judgment rule of the sub-condition, and then the fourth instruction is input to the fourth large model to obtain the judgment rule of the sub-condition. As a possible example, the instruction template for condition refinement may include but is not limited to the following: Large model input template: Please refine each of the following sub-conditions based on medical knowledge and clinical experience. The sub-conditions are as follows: {Sub-condition 1, Sub-condition 2,...}.

[0048] It should be noted that the above example is only a possible example of an instruction template for condition refinement, and the specific contents of the instruction template for condition refinement will not be given one by one. For example, still taking the above reimbursement condition "limited to adult type 2 diabetes patients" as an example, the following fourth instruction can be constructed: Please refine the basis for each of the following sub-conditions based on your medical knowledge and clinical experience and give specific The sub-conditions are as follows: {Adults, type 2 diabetes, …}.

[0049] On this basis, the fourth instruction can be input into the fourth model to obtain the judgment rules of each sub-condition. Taking the above reimbursement condition "limited to adult type 2 diabetes patients" as an example, the following judgment rules can be obtained: Based on the information given, combined with medical knowledge and clinical experience, the following results can be drawn: 1. Rules for determining whether the patient is an adult: the patient is 18 years old or older.

[0050] 2. Rules for determining whether a patient has type 2 diabetes: 1. The patient has a diagnosis of type 2 diabetes; 2. Related indicators (such as fasting blood sugar, 2-hour postprandial blood sugar, glycosylated hemoglobin, etc.) exceeds the threshold; 3. The patient has a history of type 2 diabetes or typical Symptoms, etc. If any of the above conditions are met, the patient can be considered to be a type 2 diabetic patient.

[0051] In a specific implementation scenario, please continue to refer to Figure 2b After obtaining the judgment rule of the sub-condition, the fifth instruction can be obtained based on the sub-condition and its judgment rule, the reference information that needs to be paid attention to when judging the sub-condition, and the instruction template for conditional judgment, and the fifth instruction is at least used to instruct the large language model to follow the judgment rule with the assistance of the reference information to determine whether the sub-condition is met, and then input the fifth instruction to the fifth large model to obtain the sub-judgment result of the sub-condition. As a possible example, the instruction template for conditional judgment may include but is not limited to the following: Large model input template: If a certain reimbursement item restricts {sub-conditions} patient use, you need to consider: {sub-conditions} If the condition is met, the patient can be considered as a patient with {sub-condition}. According to the above judgment logic, Based on the following medical record summary, determine whether the patient is a {sub-condition} patient and explain the reason. For information that does not appear, do not When answering, you must provide an analysis and end with "In summary, based on the medical records, we It can be concluded that the patient is / is not a patient with {sub-condition}." End. The medical record summary is as follows: [{Medical record summary}].

[0052] It should be noted that the above instruction template for conditional judgment is only a possible example in the actual application process, and the specific content of the instruction template is not limited here. Taking the sub-condition "adult" as an example, the following fifth instruction can be constructed by combining the above judgment rules and the above instruction template: If a reimbursement item is restricted to {adult} patients, you need to consider: {1. Patient age is greater than or equal to 18 If the condition is met, the patient can be considered an {adult} patient. History summary, determine whether the patient is an {adult} patient and explain the reason. Do not mention the information that does not appear in your answer. When answering, you must give an analysis and end with "In summary, based on the medical record information, we can conclude that the patient Conclusion for the patient is / is not {adult}." The medical record summary is as follows:

【{Medical Record Home Page Gender: Male Age: 72.0 Admission time: 2022-03-28T11:12:34 Discharge time: 2022-04-08T10:45:18}].

[0053] On this basis, the fifth instruction can be input into the fifth model to obtain the sub-judgment result of the sub-condition "adult": Based on the medical records, we can draw the following conclusions: 1. The patient is 72 years old, which exceeds the threshold of 18 years old and is therefore classified as an adult patient.

[0054] To sum up, based on the medical record information, we can conclude that the patient is an adult patient.

[0055] It should be noted that the above example is only a possible example of the fifth instruction and its sub-judgment result, taking the sub-condition "adult" as an example, and the specific content of the fifth instruction and the sub-judgment result is not limited here. Taking the sub-condition "type 2 diabetes" as an example, the following fifth instruction can be constructed by combining the above judgment rules and the above instruction template: If a reimbursement item restricts its use by patients with type 2 diabetes, then the following should be considered: 1. The patient has type 2 diabetes 1. The patient has 2. If the patient meets the following conditions, the patient can be considered as having type 2 diabetes. Combined with the above judgment logic and the following medical record summary, determine whether the patient is a {type 2 diabetes} patient and explain the reasons. Do not mention information that does not appear in your answer. When giving an answer, you must give an analysis and end with "In summary, Based on the medical record information, we can conclude that the patient is / is not a patient with {Type 2 Diabetes}. ” End. The medical record summary is as follows:

【{Admission Record Current medical history: The patient developed dizziness and loss of balance without obvious cause 1 year ago, and was unable to move after falling. The patient felt that the above symptoms had worsened, accompanied by headache, nasal congestion, and general discomfort. He came to our hospital for further treatment. Sequelae, cerebral infarction, hypertension grade 3 (extremely high risk), type 2 diabetes, upper respiratory tract disease” were admitted to our department.

[0056] Past medical history: He had a history of diabetes for more than 8 years, and usually took metformin hydrochloride sustained-release tablets and acarbose capsules orally. Blood sugar control was unknown.

[0057] Western medicine diagnosis: sequelae of cerebral hemorrhage, cerebral infarction; hypertension grade 3 (extremely high risk); type 2 diabetes; upper respiratory tract disease sick.

[0058] Blood cell analysis: Neutrophil ratio: 70.9%; B-type natriuretic peptide precursor 299.4 pg / ml. Glycated hemoglobin White 6.4% urobilin +; heparin binding protein 46.74ng / mL Blood biochemistry series: glucose 7.29 mmol / L, high-density lipoprotein 1.01 mmol / L, erythrocyte sedimentation rate 55 mm / h; electrolysis Quality: K 4.33 mmol / L Na 137.6mmol / L (This hospital 2022-03-29) Daily course of illness The patient's mental state, appetite and sleep are good, bowel movements are normal, and there is a small amount of foam in the urine. Blood sugar monitoring: before breakfast yesterday 11.9mmol / l, 12.9mmol / l after breakfast, 13.4mmol / L 2 hours after lunch, 2 hours after dinner 13.3mmol / L, blood sugar before bedtime 10.3mmol / L. After seeing the patient, Director Zhou instructed: 1. The patient's blood sugar before and after meals was monitored yesterday. If the blood sugar level is still high, sitagliptin combined with metformin is added to strengthen the hypoglycemic treatment. The patient is advised to have a balanced diet and continue to monitor blood sugar levels. Sugar, adjusted according to blood sugar.

[0059] Test results Blood lipid four items Blood lipid four items 125 Glucose 7.29mmol / L 7.29↑ Glycated hemoglobin determination Glycated hemoglobin determination 3103 Glycated hemoglobin 6.40% 6.40↑}】.

[0060] On this basis, the fifth instruction can be input into the fifth model to obtain the sub-judgment result of the sub-condition "type 2 diabetes": Based on the medical records, we can draw the following conclusions: 1. The patient has a diagnosis of type 2 diabetes: The medical record mentions that the patient has a diagnosis of "type 2 diabetes", which is clear Diagnosis of type 2 diabetes.

[0061] 2. The patient's fasting blood sugar and 2h postprandial blood sugar both exceeded the threshold: The medical record mentioned that the fasting blood sugar was 7.29mmol / L, and the blood sugar levels after lunch and dinner were 13.4mmol / L and 13.3mmol / L, both exceeding the threshold. value.

[0062] In summary, based on the medical record information, we can conclude that the patient is a type 2 diabetic.

[0063] It should be noted that the above example is only a possible example of the fifth instruction and its sub-judgment results, taking the sub-condition "type 2 diabetes" as an example, and the specific contents of the fifth instruction and the sub-judgment results are not limited here.

[0064] Step S14: Determine the reimbursement review result of the medical object to be reimbursed based on the respective sub-judgment results of each sub-condition and the logical relationship between them.

[0065] In an implementation scenario, as a possible example, an expression for judging the reimbursement condition can be obtained based on the logical relationship between the sub-conditions, and then the sub-judgment result of the sub-condition is substituted into the expression to obtain the reimbursement review result of the object to be reimbursed. Exemplarily, still taking the reimbursement condition "limited to adult type 2 diabetes" as an example, since the logical relationship between its sub-conditions "adult" and "type 2 diabetes" is "and", the expression "adult AND type 2 diabetes" for judging the reimbursement condition "limited to adult type 2 diabetes" can be obtained. Combined with the above-mentioned sub-condition "adult" sub-judgment result is "true", and the sub-judgment result of the sub-condition "type 2 diabetes" is "true", after substituting into the above expression, the reimbursement review result of the object to be reimbursed can be obtained as "true", which means that the reimbursement condition is determined to be met. Of course, the above example is only a possible example of the reimbursement review result when the reimbursement condition is "limited to adult type 2 diabetes", and other possible situations are not given one by one here.

[0066] In another implementation scenario, different from the aforementioned implementation, as another possible example, a comprehensive judgment can also be made based on the large language model in combination with the respective sub-judgment results of each sub-condition and the logical relationship between them. Specifically, the sixth instruction can be obtained based on the respective sub-judgment results of each sub-condition, the logical relationship between each sub-condition and the instruction template for determining the conclusion, and the sixth instruction is used to instruct the large language model to determine the reimbursement review result in combination with the respective sub-judgment results of each sub-condition and the logical relationship between them, and then input the sixth instruction into the sixth large model to obtain the reimbursement review result of the medical object to be reimbursed. The above method uses a large language model to make a comprehensive judgment in combination with the respective sub-judgment results of each sub-condition and the logical relationship between them. Compared with making a comprehensive judgment by constructing a logical expression, it is helpful to make a comprehensive judgment through the general understanding ability of the large language model.

[0067] In a specific implementation scenario, as a possible example, the instruction template for determining the conclusion may include but is not limited to the following specific contents: Large model input template: If you want to determine whether the patient meets the [{reimbursement conditions}], you need to consider: {Logical relationship between sub-conditions} According to the above logic, refer to the following information to determine whether the patient meets the [{reimbursement conditions}] and give a conclusion. Exam information: {Sub-judgment results of each sub-condition} Large model output template: {Large model analysis process}, therefore, {Large model analysis process}, {Audit violation / Audit No violation / unable to judge}.

[0068] It should be noted that the above example is only a possible example of an instruction template for determining a conclusion. The specific content of the instruction template for determining a conclusion is not limited here, and examples are not given one by one here.

[0069] In a specific implementation scenario, still taking the aforementioned reimbursement condition "limited to adult type 2 diabetes" as an example, combined with the above instruction template for determining the conclusion, the following sixth instruction can be constructed: If you want to determine whether a patient meets the criteria [limited to adult patients with type 2 diabetes], you need to consider: Step 1: Determine whether the patient is an adult. If so, continue to determine. Otherwise, the patient is deemed not to meet the restriction conditions. Step 2: Determine whether the patient has type 2 diabetes. If yes, continue to determine. Otherwise, the patient is considered not to meet the criteria. Control conditions; Finally, determine whether the patient meets the above conditions at the same time. If so, the patient is considered to meet the restriction conditions. Otherwise, The patient does not meet the restriction conditions.

[0070] According to the above logic, refer to the following information to determine whether the patient meets the requirements of [limited to adult type 2 diabetes patients] and give Draw a conclusion. Reference information: 1. Based on the medical record information, we can conclude that the patient is an adult patient.

[0071] 2. Based on the medical record information, we can conclude that the patient is a type 2 diabetic.

[0072] On this basis, the sixth instruction can be input into the sixth model to obtain the reimbursement review result of the object to be reimbursed, "Proline Henggliflozin Tablets": According to the information given, it can be known that the patient is an adult patient and also a patient with type 2 diabetes. The patient meets the following two conditions: 1. Adult and 2. Suffering from type 2 diabetes.

[0073] Therefore, the patient meets the requirements for adult patients with type 2 diabetes and the review does not violate regulations.

[0074] That is to say, under this circumstance, the reimbursement audit result "no violation in the audit" can be obtained. Of course, the above example is only a possible example in the actual application process, and does not limit other possible situations. Other possible situations will not be given one by one here. In addition, it should be noted that the aforementioned first large model, second large model, third large model, fourth large model, fifth large model, and sixth large model can be the same large language model or different large language models, and can also be part of the same large language model and part of different large language models, which is not limited here.

[0075] In an implementation scenario, please refer to Figure 2c , Figure 2c This is a schematic diagram of the process of an embodiment of the medical reimbursement review method. Figure 2cAs shown, after the reimbursement conditions of the object to be reimbursed are analyzed, several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition can be obtained; based on this, the chapter content of each medical record chapter in the medical record information can be obtained, and then the key information used to judge the sub-conditions in the chapter content can be extracted, so as to integrate the key information in each medical record chapter that needs to be paid attention to when judging the sub-conditions to obtain the medical record summary; finally, the sub-conditions can be refined to obtain the judgment rules of the sub-conditions, and the sub-conditions and their judgment rules and the reference information that needs to be paid attention to when judging the sub-conditions can be analyzed to obtain the sub-judgment results of the sub-conditions, and then the reimbursement review results of the object to be reimbursed can be determined based on the sub-judgment results of each sub-condition and the logical relationship between each sub-condition.

[0076] The above scheme is based on the analysis of the reimbursement conditions of the object to be reimbursed, and obtains several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions. Based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, the reference information that needs to be paid attention to when judging the sub-conditions is extracted from the medical record information, so as to analyze the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment results of the sub-conditions, and the sub-judgment results include whether the sub-conditions are met, and then based on the respective sub-judgment results of each sub-condition and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined. Therefore, on the one hand, due to reliance on manual review, the review efficiency can be improved to a certain extent. On the one hand, by disassembling the reimbursement conditions, it is helpful to break the reimbursement conditions into parts, form several logically simple sub-conditions as much as possible, and judge each sub-condition separately, which can further improve the audit efficiency. On the other hand, each sub-condition also parses out several medical record chapters that need to be paid attention to when judging, so as to selectively extract reference information from relevant medical record chapters when judging sub-conditions, which helps to break the limitations of long text on semantic understanding as much as possible, and can improve the audit accuracy to a certain extent. On the other hand, the logical relationship between each sub-condition is also parsed, and finally the logical relationship between the sub-conditions constrains the sub-judgment results of each sub-condition to determine the reimbursement audit results of the medical object to be reimbursed, which can further improve the audit accuracy. Therefore, the audit efficiency and audit accuracy of medical reimbursement audit can be improved.

[0077] See also Figure 3 , Figure 3It is a schematic diagram of the framework of an embodiment of the medical reimbursement review device of the present application. The medical reimbursement review device 30 includes: a condition parsing module 31, an information extraction module 32, a condition judgment module 33 and a comprehensive analysis module 34. The condition parsing module 31 is used to parse based on the reimbursement conditions of the medical object to be reimbursed, and obtain a number of sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, a number of medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between each sub-condition; the information extraction module 32 is used to extract reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on the number of medical record chapters that need to be paid attention to when judging the sub-conditions; the condition judgment module 33 is used to analyze based on the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment result of the sub-condition; wherein the sub-judgment result includes whether the sub-condition is met; the comprehensive analysis module 34 is used to determine the reimbursement review result of the medical object to be reimbursed based on the respective sub-judgment results of each sub-condition and the logical relationship between them.

[0078] In the above scheme, the medical reimbursement review device 30 analyzes the reimbursement conditions of the object to be reimbursed, obtains several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the various sub-conditions, and then extracts the reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, so as to analyze the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment results of the sub-conditions, and the sub-judgment results include whether the sub-conditions are met, and then based on the respective sub-judgment results of each sub-condition and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined. Therefore, on the one hand, due to reliance on manual review, it can be to a certain extent. Improve audit efficiency. On the other hand, by disassembling the reimbursement conditions, it is helpful to break the reimbursement conditions into parts, forming several logically simple sub-conditions as much as possible, and judging each sub-condition separately, which can further improve the audit efficiency. On the other hand, each sub-condition also parses out several medical record chapters that need to be paid attention to when judging, so as to selectively extract reference information from relevant medical record chapters when judging sub-conditions, which helps to break the limitations of long text on semantic understanding as much as possible, and can improve the audit accuracy to a certain extent. On the other hand, the logical relationship between each sub-condition is also parsed, and finally the logical relationship between the sub-conditions constrains the sub-judgment results of each sub-condition to determine the reimbursement audit results of the medical object to be reimbursed, which can further improve the audit accuracy. Therefore, the audit efficiency and audit accuracy of medical reimbursement audit can be improved.

[0079] In some disclosed embodiments, the condition parsing module 31 includes a first construction submodule, which is used to construct a first instruction based on the reimbursement condition and an instruction template for condition parsing; wherein the first instruction is at least used to instruct the large language model to parse the reimbursement condition; the condition parsing module 31 includes a second processing submodule, which is used to input the first instruction into the first large model to obtain a plurality of sub-conditions, a plurality of medical record chapters and the logical relationship between each sub-condition.

[0080] In some disclosed embodiments, the information extraction module 32 includes a content acquisition submodule for acquiring the chapter content of each medical record chapter in the medical record information; the information extraction module 32 includes a key extraction submodule for extracting key information in the chapter content for judging sub-conditions; the information extraction module 32 includes a summary integration submodule for integrating the medical record summary as reference information that needs to be paid attention to when judging sub-conditions based on the key information in each medical record chapter that needs to be paid attention to when judging sub-conditions.

[0081] In some disclosed embodiments, the key extraction submodule includes a second construction submodule, which is used to construct a second instruction based on the sub-conditions, chapter content and an instruction template for information extraction; wherein the second instruction is used to instruct the large language model to extract key information for judging the sub-conditions from the chapter content; the key extraction submodule includes a second processing submodule, which is used to input the second instruction into the second large model to obtain key information in the chapter content for judging the sub-conditions.

[0082] In some disclosed embodiments, the summary integration submodule includes a third construction submodule, which is used to construct a third instruction based on the key information in each medical record chapter and the instruction template for summary integration; wherein the third instruction is used to instruct the large language model to integrate the key information to form a medical record summary; the summary integration submodule includes a third processing submodule, which is used to input the third instruction into the third large model to obtain the medical record summary that needs to be paid attention to when judging the sub-conditions.

[0083] In some disclosed embodiments, the medical reimbursement review device 30 includes a condition refinement module, which is used to perform refinement based on sub-conditions to obtain judgment rules for sub-conditions; the condition judgment module 33 is specifically used to perform analysis based on the sub-conditions and their judgment rules and reference information that needs to be paid attention to when judging the sub-conditions to obtain sub-judgment results of the sub-conditions.

[0084] In some disclosed embodiments, the condition refinement module includes a fourth construction submodule, which is used to construct a fourth instruction based on the sub-condition and the instruction template for condition refinement; wherein the fourth instruction is used to indicate the judgment rule of the large language model to refine the sub-condition; the condition refinement module includes a fourth processing submodule, which is used to input the fourth instruction to the fourth large model to obtain the judgment rule of the sub-condition.

[0085] In some disclosed embodiments, the conditional judgment module 33 includes a fifth construction submodule, which is used to obtain a fifth instruction based on the sub-conditions and their judgment rules, reference information that needs to be paid attention to when judging the sub-conditions, and an instruction template for conditional judgment; wherein the fifth instruction is at least used to instruct the large language model to follow the judgment rules with the assistance of reference information to determine whether the sub-conditions are met; the conditional judgment module 33 includes a fifth processing submodule, which is used to input the fifth instruction to the fifth large model to obtain the sub-judgment result of the sub-condition.

[0086] In some disclosed embodiments, the comprehensive analysis module 34 includes a sixth construction submodule, which is used to obtain a sixth instruction based on the sub-judgment results of each sub-condition, the logical relationship between each sub-condition and the instruction template for determining the conclusion; wherein the sixth instruction is used to instruct the large language model to determine the reimbursement review result in combination with the sub-judgment results of each sub-condition and the logical relationship between them; the comprehensive analysis module 34 includes a sixth processing submodule, which is used to input the sixth instruction into the sixth large model to obtain the reimbursement review result of the medical object to be reimbursed.

[0087] See also Figure 4 , Figure 4 : is a schematic diagram of the framework of an embodiment of an electronic device of the present application. The electronic device 40 includes at least a memory 41 and a processor 42 coupled to each other, the memory 41 stores at least program instructions, and the processor 42 is used to execute the program instructions to implement the steps in any of the above-mentioned medical reimbursement review method embodiments. For details, please refer to the aforementioned disclosed embodiments, which will not be repeated here. As a possible example, the electronic device 40 may include but is not limited to a server, etc., and the specific type of the electronic device 40 is not limited here.

[0088] Specifically, the processor 42 is used to control itself and the memory 41 to implement the steps in any of the above-mentioned medical reimbursement review method embodiments. The processor 42 can also be called a CPU (Central Processing Unit). The processor 42 may be an integrated circuit chip with signal processing capabilities. The processor 42 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 42 can be implemented by an integrated circuit chip.

[0089] In the above scheme, the electronic device 40 parses the reimbursement conditions of the object to be reimbursed, obtains several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions, and then extracts the reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, so as to analyze based on the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment result of the sub-condition, and the sub-judgment result includes whether the sub-condition is met, and then based on the respective sub-judgment results of each sub-condition and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined. Therefore, on the one hand, due to reliance on manual review, it can improve to a certain extent. Audit efficiency, on the other hand, by disassembling the reimbursement conditions, it is helpful to break the reimbursement conditions into parts, form several logically simple sub-conditions as much as possible, and judge each sub-condition separately, which can further improve the audit efficiency. On the other hand, each sub-condition also parses out several medical record chapters that need to be paid attention to when judging, so as to selectively extract reference information from relevant medical record chapters when judging sub-conditions, which helps to break the limitations of long text on semantic understanding as much as possible, and can improve the audit accuracy to a certain extent. On the other hand, the logical relationship between each sub-condition is also parsed, and finally the logical relationship between the sub-conditions constrains the sub-judgment results of each sub-condition to determine the reimbursement audit results of the medical object to be reimbursed, which can further improve the audit accuracy. Therefore, the audit efficiency and audit accuracy of medical reimbursement audit can be improved.

[0090] See also Figure 5 , Figure 5 The schematic diagram is a schematic diagram of a computer-readable storage medium 50 of the present application. The computer-readable storage medium 50 stores program instructions 51 that can be executed by a processor, and the program instructions 51 are used to implement the steps in any of the above-mentioned medical reimbursement review method embodiments.

[0091] In the above scheme, the computer-readable storage medium 50 parses the reimbursement conditions of the object to be reimbursed, obtains several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions, and then extracts the reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on the several medical record chapters that need to be paid attention to when judging the sub-conditions, so as to analyze based on the reference information that needs to be paid attention to when judging the sub-conditions, and obtain the sub-judgment results of the sub-conditions, and the sub-judgment results include whether the sub-conditions are met, and then based on the respective sub-judgment results of each sub-condition and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined. Therefore, on the one hand, due to reliance on manual review, it can be to a certain extent. Improve audit efficiency. On the other hand, by disassembling the reimbursement conditions, it is helpful to break the reimbursement conditions into parts, forming several logically simple sub-conditions as much as possible, and judging each sub-condition separately, which can further improve the audit efficiency. On the other hand, each sub-condition also parses out several medical record chapters that need to be paid attention to when judging, so as to selectively extract reference information from relevant medical record chapters when judging sub-conditions, which helps to break the limitations of long text on semantic understanding as much as possible, and can improve the audit accuracy to a certain extent. On the other hand, the logical relationship between each sub-condition is also parsed, and finally the logical relationship between the sub-conditions constrains the sub-judgment results of each sub-condition to determine the reimbursement audit results of the medical object to be reimbursed, which can further improve the audit accuracy. Therefore, the audit efficiency and audit accuracy of medical reimbursement audit can be improved.

[0092] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0093] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0094] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0095] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0096] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0097] If the integrated unit is implemented in the form of a software functional unit 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 is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0098] If the technical solution of this application involves personal information, the product using the technical solution of this application has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using the technical solution of this application has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, the personal information processing rules are notified by obvious signs / information, and the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A medical reimbursement review method, characterized in that: include: Based on the reimbursement conditions of the medical object to be reimbursed, several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record chapters that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions are obtained; Based on the several medical record sections that need to be paid attention to when judging the sub-conditions, extracting reference information that needs to be paid attention to when judging the sub-conditions from the medical record information; Analyze the reference information that needs to be paid attention to when judging the sub-condition to obtain a sub-judgment result of the sub-condition; wherein the sub-judgment result includes whether the sub-condition is met; Based on the respective sub-judgment results of each of the sub-conditions and the logical relationship between them, the reimbursement review result of the medical object to be reimbursed is determined.

2. The method according to claim 1, characterized in that The reimbursement conditions based on the medical object to be reimbursed are parsed to obtain several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record sections that need to be paid attention to when judging the sub-conditions, and the logical relationship between the sub-conditions, including: Based on the reimbursement condition and the instruction template for condition parsing, construct a first instruction; wherein the first instruction is at least used to instruct the large language model to parse the reimbursement condition; Input the first instruction into the first large model to obtain the logical relationship between the plurality of sub-conditions, the plurality of medical record chapters and each of the sub-conditions.

3. The method according to claim 1, characterized in that The extracting reference information that needs to be paid attention to when judging the sub-condition from the medical record information based on the several medical record sections that need to be paid attention to when judging the sub-condition includes: Obtain the chapter content of each medical record chapter in the medical record information; Extracting key information from the chapter content for judging the sub-conditions; Based on the key information in each of the medical record chapters that needs to be paid attention to when judging the sub-conditions, a medical record summary is integrated as reference information that needs to be paid attention to when judging the sub-conditions.

4. The method according to claim 3, characterized in that The extracting key information from the chapter content for judging the sub-condition includes: Based on the sub-condition, the chapter content and the instruction template for information extraction, construct a second instruction; wherein the second instruction is used to instruct the large language model to extract key information for judging the sub-condition from the chapter content; Input the second instruction into the second large model to obtain key information in the chapter content for judging the sub-condition.

5. The method according to claim 3, characterized in that: The key information in each of the medical record chapters that needs to be paid attention to when judging the sub-conditions is integrated to obtain a medical record summary, including: Based on the key information in each of the medical record chapters and the instruction template for summary integration, construct a third instruction; wherein the third instruction is used to instruct the large language model to integrate the key information to form the medical record summary; Input the third instruction into the third large model to obtain the medical record summary that needs to be paid attention to when judging the sub-condition.

6. The method according to claim 1, characterized in that Before analyzing the reference information that needs to be paid attention to when judging the sub-condition to obtain the sub-judgment result of the sub-condition, the method further includes: Refining the sub-conditions to obtain judgment rules for the sub-conditions; The analyzing the reference information that needs to be paid attention to when judging the sub-condition to obtain the sub-judgment result of the sub-condition includes: An analysis is performed based on the sub-conditions and their judgment rules and reference information that needs to be paid attention to when judging the sub-conditions to obtain sub-judgment results of the sub-conditions.

7. The method according to claim 6, characterized in that The refining based on the sub-conditions to obtain the judgment rules of the sub-conditions includes: Based on the sub-condition and the instruction template for condition refinement, construct a fourth instruction; wherein the fourth instruction is used to instruct the large language model to refine the judgment rule of the sub-condition; The fourth instruction is input into the fourth large model to obtain the judgment rule of the sub-condition.

8. The method according to claim 6, characterized in that The sub-condition and its judgment rule and the reference information that needs to be paid attention to when judging the sub-condition are analyzed to obtain the sub-judgment result of the sub-condition, including: Based on the sub-condition and its judgment rule, the reference information that needs to be paid attention to when judging the sub-condition, and the instruction template for condition judgment, a fifth instruction is obtained; wherein the fifth instruction is at least used to instruct the large language model to follow the judgment rule with the assistance of the reference information to determine whether the sub-condition is satisfied; Input the fifth instruction to the fifth large model to obtain the sub-judgment result of the sub-condition.

9. The method according to claim 1, characterized in that: The determining of the reimbursement review result of the medical object to be reimbursed based on the respective sub-judgment results of the sub-conditions and the logical relationship therebetween includes: Based on the respective sub-judgment results of the sub-conditions, the logical relationships between the sub-conditions, and the instruction template for determining the conclusion, a sixth instruction is obtained; wherein the sixth instruction is used to instruct the large language model to determine the reimbursement review result in combination with the respective sub-judgment results of the sub-conditions and the logical relationships between them; Input the sixth instruction into the sixth model to obtain the reimbursement review result of the medical object to be reimbursed.

10. A medical reimbursement review device, characterized in that: include: A condition analysis module is used to analyze the reimbursement conditions of the medical object to be reimbursed, and obtain several sub-conditions that need to be judged when reviewing whether the reimbursement conditions are met, several medical record sections that need to be paid attention to when judging the sub-conditions, and the logical relationship between each of the sub-conditions; An information extraction module, for extracting reference information that needs to be paid attention to when judging the sub-conditions from the medical record information based on a number of medical record sections that need to be paid attention to when judging the sub-conditions; A condition judgment module, used to analyze the reference information that needs to be paid attention to when judging the sub-condition, and obtain a sub-judgment result of the sub-condition; wherein the sub-judgment result includes whether the sub-condition is met; The comprehensive analysis module is used to determine the reimbursement review result of the medical object to be reimbursed based on the sub-judgment results of each of the sub-conditions and the logical relationship between them.

11. An electronic device, characterized in that: The medical reimbursement review method comprises at least a memory and a processor coupled to each other, wherein the memory stores at least program instructions, and the processor is used to execute the program instructions to implement the medical reimbursement review method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: Program instructions that can be executed by a processor are stored, and the program instructions are used to implement the medical reimbursement review method according to any one of claims 1 to 9.

Images