Medical insurance fraud behavior analysis method and device, storage medium and electronic equipment

Abstract

The invention relates to a medical insurance fraud behavior analysis method and device, a storage medium and electronic equipment, and relates to the technical field of medical insurance fraud analysis, and the method comprises the steps: determining the reimbursement material type of each medical insurance reimbursement material picture based on an OCR technology, screening target reimbursement pictures of at least one target reimbursement material type from the medical insurance reimbursement material pictures; screening clinical diagnosis information and medication list information of the patient to be reimbursed from all picture information identification results; according to the clinical diagnosis information and the medication list information, determining a medication rationality result of the patient to be reimbursed through a preset large language model; and according to the picture information identification result, performing compliance analysis on the reimbursement material of the patient to be reimbursed to obtain a compliance identification result, and according to the compliance identification result and the medication rationality result, determining a medical insurance fraud behavior analysis result of the patient to be reimbursed. The method has the effect of improving the accuracy of analyzing the insurance fraud behaviors of the medical insurance.

Description

Technical Field

[0001] The present application relates to the technical field of medical insurance fraud analysis, and specifically to a method, device, storage medium and electronic device for analyzing medical insurance fraud behavior. Background Art

[0002] In recent years, with the popularization of the medical insurance system, medical insurance fraud has also increased, which has brought huge challenges to the management and use of medical insurance funds. In addition, medical insurance funds play an important role in protecting people's livelihood. In addition, the ways of medical insurance fraud are endless and difficult to prevent. Therefore, it is particularly important to objectively analyze medical insurance fraud and effectively combat and control medical insurance fraud. Among them, medical insurance fraud refers to the behavior of individuals or institutions illegally obtaining medical insurance funds through forgery, alteration, concealment and other means. This behavior not only endangers the security of medical insurance funds, but also damages the legitimate rights and interests of the people.

[0003] At present, the method usually adopted for analyzing medical insurance fraud is to identify the printed text, handwritten text and stamps in the pictures of medical insurance reimbursement materials based on optical character recognition (OCR), and finally manually review the recognition results to realize the analysis of medical insurance fraud. However, under this method, the accuracy of OCR in recognizing special parts such as handwritten text and stamps is low. At the same time, manual review is also prone to ignore the correlation between the medication in the medical insurance reimbursement materials and the diseases suffered by the medication users, resulting in poor accuracy in the final analysis of medical insurance fraud. Summary of the invention

[0004] In order to improve the accuracy of the analysis of medical insurance fraud behavior, the present application provides a medical insurance fraud behavior analysis method, device, storage medium and electronic device.

[0005] In a first aspect of the present application, a method for analyzing medical insurance fraud behavior is provided, which specifically includes: Obtain at least one picture of the medical insurance reimbursement materials of the patient to be reimbursed; Determine the reimbursement material type of each of the medical insurance reimbursement material images based on OCR technology, and select at least one target reimbursement material type from each of the medical insurance reimbursement material images, wherein the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials and / or the rationality of medication of the patient to be reimbursed; Input each of the target reimbursement pictures into a preset multimodal macro model to obtain corresponding picture information recognition results, and screen the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results; According to the clinical diagnosis information and the medication list information, determining the rationality of medication for the patient to be reimbursed through a preset large language model; According to the picture information recognition result, the reimbursement materials of the patient to be reimbursed are subjected to compliance analysis to obtain a compliance recognition result, and according to the compliance recognition result and the rationality result of medication, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined.

[0006] By adopting the above technical solution, after obtaining the medical insurance reimbursement material pictures of the patients to be reimbursed, the keywords in the medical insurance reimbursement material pictures are firstly identified by OCR technology, so as to determine the reimbursement material type of the medical insurance reimbursement material pictures, and then the target reimbursement pictures that can analyze the compliance of the reimbursement materials and / or the rationality of medication are selected, and the corresponding image information recognition results are more accurately extracted therefrom through the multimodal large model, and the rationality of medication for the patients to be reimbursed is analyzed according to the clinical diagnosis information and medication list information in the image information recognition results, and then the compliance of the reimbursement materials is analyzed according to other information in the image information recognition results, and finally, based on the compliance recognition results and the rationality of medication, combined with the two dimensions of rationality of medication and compliance of reimbursement materials, the medical insurance reimbursement materials of the patients to be reimbursed are analyzed more comprehensively, and the analysis results of the medical insurance fraud behavior of the patients to be reimbursed are determined, thereby improving the accuracy of the analysis of the medical insurance fraud behavior.

[0007] Optionally, the step of inputting each of the target reimbursement pictures into a preset multimodal macro model to obtain corresponding picture information recognition results specifically includes: Construct the first prompt instruction; Based on the first prompt instruction, the target reimbursement picture is input into the multimodal large model to obtain a corresponding picture information recognition result, and the first prompt instruction is a prompt instruction for at least one information recognition task for the target reimbursement picture.

[0008] By adopting the above technical solution, through the first prompt instruction, the multimodal large model is guided to identify different types of information for each target reimbursement picture, and the corresponding picture information recognition results are obtained. By adopting the multimodal large model to recognize the picture information, the recognition accuracy is higher.

[0009] Optionally, the determining the rationality of medication for the patient to be reimbursed through a preset large language model based on the clinical diagnosis information and the medication list information specifically includes: Matching the reimbursement drug instructions corresponding to each reimbursement drug in the medication list information from a preset drug database, wherein the drug database includes different drugs and corresponding drug instructions; Constructing a second prompt instruction, wherein the second prompt instruction is a prompt instruction for determining whether the corresponding reimbursement drug is effective for the disease in the clinical diagnosis information based on the reimbursement drug instructions; Input the second prompt instruction, the clinical diagnosis information, the medication list information and all reimbursement drug instructions into the large language model to obtain at least one output result, and if all the output results are valid, determine that the medication rationality result of the patient to be reimbursed is reasonable; If there is an invalid output result among the output results, it is determined that the rationality result of the medication of the patient to be reimbursed is unreasonable.

[0010] By adopting the above technical solution, through the second prompt instruction, the large language model is guided to determine whether the reimbursed drugs are effective for the diseases in the clinical diagnosis information according to the reimbursement drug instructions; if all output results are valid, it means that all reimbursed drugs in the medication list information are effective for the disease, then the medication rationality result of the patient to be reimbursed is determined to be reasonable; if there are invalid output results among the output results, it means that there are reimbursed drugs that are invalid for the disease in the medication list information, then the medication rationality result of the patient to be reimbursed is determined to be unreasonable, so as to facilitate the subsequent combination of the medication rationality results to more accurately determine the medical insurance fraud behavior analysis results of the patients to be reimbursed.

[0011] Optionally, the step of performing compliance analysis on the reimbursement materials of the patient to be reimbursed based on the image information recognition result to obtain a compliance recognition result specifically includes: Determine a target reimbursement material type target reimbursement material type image used to analyze the compliance of the reimbursement materials of the patient to be reimbursed as a compliance identification image, and determine whether there is a stamp in the compliance identification image based on the image information recognition result corresponding to the compliance identification image; If not, determining that the compliance identification result of the reimbursement materials of the patient to be reimbursed is non-compliant; If so, in the case where the stamp is forged, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, if there is no forgery of the stamp, and if there is no doctor's signature in the compliance identification picture, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, when there is no forgery in the stamp and there is a doctor's signature in the compliance identification picture, if there is a photo of a mobile phone screen and / or a photo of a computer screen in the compliance identification picture, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant.

[0012] By adopting the above technical solution, if there is no seal in the compliance identification picture, which is contrary to the normal situation where a seal should exist, then the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; if there is a seal in the compliance identification picture, it cannot be directly regarded as the reimbursement materials are compliant, and the authenticity of the seal information needs to be further judged. If the seal is forged, it means that the patient to be reimbursed has the possibility of forging the reimbursement materials, and the compliance identification result is determined to be non-compliant; if there is no forgery of the seal, if there is no doctor's signature, it means that the reimbursement materials of the patient to be reimbursed are still likely to be forged, and the compliance identification result is determined to be non-compliant. If there is a doctor's signature, in order to ensure the accuracy of the compliance identification result, it is determined whether there are also photos of the mobile phone screen and / or computer screen. If there are photos of the mobile phone screen and / or computer screen, this compliance identification picture may contain content taken from other people's mobile phone screens and / or computer screens, and the reimbursement materials have authenticity risks, and then the compliance identification result is determined to be non-compliant.

[0013] Optionally, determining the medical insurance fraud behavior analysis result of the patient to be reimbursed based on the compliance identification result and the medication rationality result specifically includes: When the compliance identification result is compliant, determining whether the medication rationality result is unreasonable; If so, the target reimbursement picture from which the clinical diagnosis information and the medication list information are derived is determined as the medication determination picture, and the medical insurance fraud behavior analysis results of the patient to be reimbursed are determined based on the picture clarity of the medication determination picture and whether there is handwritten text in the text of the clinical diagnosis information and the medication list information in the corresponding medication determination picture.

[0014] By adopting the above technical solution, when the compliance identification result is compliant, if the result of the rationality of medication is irrational, it means that not all the reimbursed drugs in the medication list information are effective for the disease, there is a possibility of over-prescribing and the risk of insurance fraud, and further, according to the picture clarity of the medication judgment picture and the presence of handwriting, the risk of extraction bias of clinical diagnosis information and medication list information is checked, so as to finally determine the analysis results of medical insurance fraud behavior when medication is irrational more accurately.

[0015] Optionally, the method further includes: When the compliance identification result is non-compliant, if there is no stamp or the stamp is forged or there is no doctor's signature in the compliance identification picture, then the medical insurance fraud analysis result of the patient to be reimbursed is determined to be an obvious insurance fraud; If there is a seal in the compliance identification picture, the seal is not forged and there is a doctor's signature, and if there is a mobile phone screen photo and / or a computer screen photo in the compliance identification picture, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be a suspected fraud.

[0016] By adopting the above technical solution, when the compliance identification result is non-compliant, if the compliance identification image lacks a seal or the seal is forged or the doctor's signature is missing, it means that the reimbursement materials are likely to be fictitious, and the medical insurance fraud analysis result of the patient to be reimbursed is directly determined to be an obvious fraud; on the contrary, if the compliance identification image contains a seal, the seal is not forged, and there is a doctor's signature, but there are also photos of mobile phone screens and / or computer screens, and there is only the possibility of taking pictures of the contents of other people's mobile phone screens and / or computer screens, it means that the reimbursement materials are likely to be fictitious, and the medical insurance fraud analysis result of the patient to be reimbursed is determined to be a suspected fraud, thereby making the medical insurance fraud analysis result more accurate.

[0017] Optionally, determining the analysis result of the medical insurance fraud behavior of the patient to be reimbursed according to the picture clarity of the medication determination picture and whether the clinical diagnosis information and the medication list information are handwritten in the text in the corresponding medication determination picture specifically includes: Determine whether the image clarity of the medication determination image is 1, and determine whether the text of the clinical diagnosis information and the medication list information in the corresponding medication determination image is handwritten. The image clarity takes a value of [0,1]. The greater the image clarity, the clearer the medication determination image; If there is no handwriting and the picture clarity of the medication determination picture is 1, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be an obvious insurance fraud behavior; If there is handwriting and the picture clarity of the medication determination picture is 1, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be a suspected insurance fraud behavior; If the picture clarity of the medication determination picture is not 1, it is determined that the medical insurance fraud analysis result of the patient to be reimbursed is a suspected insurance fraud.

[0018] By adopting the above technical solution, if there is no handwriting and the image clarity of the medication determination picture is 1, it means that the medication determination picture is relatively clear and the text corresponding to the clinical diagnosis information and medication list information therein is printed, and the accuracy of the clinical diagnosis information and medication list information extracted from the medication determination picture is high, and then the result of verifying the rationality of unreasonable medication is relatively accurate, and the possibility of misjudgment is small, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be an obvious fraud; if there is handwriting and the image clarity of the medication determination picture is 1, it means that the clinical diagnosis information and / or medication list information extracted are accurate. There may be deviations in the drug list information, which will affect the accuracy of the results of the rationality of medication use, that is, there may be errors in the unreasonable results of the rationality of medication use, so it is more reasonable to analyze the medical insurance fraud behavior as suspected fraud; if the image clarity of the medication determination image is not 1, it means that it is relatively unclear. Therefore, when extracting clinical diagnosis information and medication list information through the medication determination image, there is a risk of deviation in the extracted content, and there may be errors in the unreasonable results of the rationality of medication use, so it is determined that the analysis result of the medical insurance fraud behavior is suspected fraud, which makes the analysis of the medical insurance fraud behavior more accurate.

[0019] In a second aspect of the present application, a device for analyzing medical insurance fraud behavior is provided, which specifically includes: A picture acquisition module is used to acquire at least one picture of medical insurance reimbursement materials of the patient to be reimbursed; An image screening module is used to determine the reimbursement material type of each of the medical insurance reimbursement material images based on OCR technology, and screen at least one target reimbursement material type from each of the medical insurance reimbursement material images, wherein the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials of the patient to be reimbursed and / or the rationality of medication; An information screening module, used to input each of the target reimbursement pictures into a preset multimodal macro model, obtain corresponding picture information recognition results, and screen the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results; A rational determination module, used to determine the rationality of the medication of the patient to be reimbursed through a preset large language model according to the clinical diagnosis information and the medication list information; The insurance fraud analysis module is used to perform compliance analysis on the reimbursement materials of the patient to be reimbursed based on the image information recognition result, obtain a compliance recognition result, and determine the medical insurance fraud behavior analysis result of the patient to be reimbursed based on the compliance recognition result and the medication rationality result.

[0020] By adopting the above technical solution, the image acquisition module obtains at least one image of medical insurance reimbursement materials, and the image screening module screens at least one target reimbursement material type of target reimbursement image from various medical insurance reimbursement material images. Then the information screening module screens the clinical diagnosis information and medication list information of the patient to be reimbursed from all image information recognition results. Then the reasonable judgment module determines the rationality of medication for the patient to be reimbursed based on the clinical diagnosis information and medication list information through a preset large language model. Finally, the insurance fraud analysis module determines the medical insurance fraud behavior analysis results of the patient to be reimbursed based on the compliance recognition results and the rationality of medication.

[0021] In a third aspect of the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is loaded and executed by a processor, the method steps described in any one of the first aspects are performed.

[0022] In a fourth aspect of the present application, an electronic device is provided, specifically comprising: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the processor is used to load and execute the computer program stored in the memory so that the electronic device performs the method as described in any one of the first aspects.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: identifying keywords in medical insurance reimbursement material images through OCR technology, thereby determining the reimbursement material type of the medical insurance reimbursement material images, and then selecting target reimbursement images that can analyze the compliance of the reimbursement materials and / or the rationality of medication, and extracting the corresponding image information recognition results more accurately from them through a multimodal large model, analyzing the rationality of medication for patients to be reimbursed based on the clinical diagnosis information and medication list information in the image information recognition results, and then analyzing the compliance of the reimbursement materials based on other information in the image information recognition results, and finally, based on the compliance recognition results and the rationality of medication results, combined with the two-dimensional analysis of medication rationality and compliance of reimbursement materials, the analysis results of the medical insurance fraud behavior of patients to be reimbursed are more comprehensively determined, thereby improving the accuracy of the analysis of medical insurance fraud behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flowchart of a method for analyzing medical insurance fraud behavior provided in an embodiment of the present application; Figure 2 It is a flowchart of another medical insurance fraud behavior analysis method provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of a medical insurance fraud behavior analysis device provided in an embodiment of the present application; Figure 4It is a structural schematic diagram of another medical insurance fraud behavior analysis device provided in an embodiment of the present application.

[0025] Explanation of the accompanying drawings: 11. Image acquisition module; 12. Image screening module; 13. Information screening module; 14. Reasonable judgment module; 15. Insurance fraud analysis module; 16. Behavior analysis module. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0027] In the description of the embodiments of the present application, words such as "illustrative", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "illustrative", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "illustrative", "for example" or "for example" is intended to present related concepts in a concrete way.

[0028] In the description of the embodiments of the present application, the term "and / or" is only a kind of association relationship describing the associated objects, indicating that there may be three kinds of relationships, for example, A and / or B, which can represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0029] See also Figure 1 The present application embodiment discloses a flowchart of a method for analyzing medical insurance fraud behavior, which can be implemented by a computer program or run on a medical insurance fraud behavior analysis device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S101: Obtain at least one picture of medical insurance reimbursement materials of the patient to be reimbursed.

[0030] Specifically, in the embodiment of the present application, the patient to be reimbursed refers to a patient who has medical insurance reimbursement for the medical expenses incurred, and the medical expenses include but are not limited to hospitalization expenses, medication expenses, etc. during disease treatment. In other embodiments, the medical expenses may also be the diagnosis and treatment expenses related to fertility in the hospital. The medical insurance reimbursement material picture is a picture file of the medical insurance reimbursement material that the patient to be reimbursed needs to provide when making medical insurance reimbursement. Among them, the medical insurance reimbursement material refers to a series of documents and materials that need to be submitted to the medical insurance agency when making medical expenses reimbursement. These materials are mainly used to confirm the patient's identity information, medical expenses details and medical treatment conditions, and ensure the smooth progress of the reimbursement process. Different types of patients to be reimbursed have different types of medical insurance reimbursement and different medical insurance reimbursement materials to be provided. Exemplarily, patient A to be reimbursed is a pregnant woman, and the corresponding type of medical insurance reimbursement is maternity insurance reimbursement. Patient B to be reimbursed is a patient hospitalized for illness, and the corresponding type of medical insurance reimbursement is ordinary hospitalization reimbursement. The medical insurance reimbursement materials required for the two different types of medical insurance reimbursement are different.

[0031] In addition, the embodiment of the present application discloses a method for analyzing medical insurance fraud behavior, the execution subject of which is a server, the server is wirelessly connected to the terminal of the patient to be reimbursed, the terminal of the patient to be reimbursed is installed with a client or applet related to medical insurance reimbursement, the server can be a background server of the client or applet, the background server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. The terminal can be a smart phone or a personal computer.

[0032] A feasible way to obtain pictures of medical insurance reimbursement materials of patients to be reimbursed is: the patient to be reimbursed uploads the pictures of medical insurance reimbursement materials saved in the terminal through the client or applet in the terminal, and the server finally obtains at least one picture of the medical insurance reimbursement materials of the patient to be reimbursed.

[0033] S102: Determine the reimbursement material type of each medical insurance reimbursement material image based on OCR technology, and select at least one target reimbursement material type from each medical insurance reimbursement material image, where the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials and / or the rationality of medication of the patient to be reimbursed.

[0034] Specifically, obtain the medical insurance reimbursement material picture, and use the OCR technology to identify and extract the information in each medical insurance reimbursement material picture, mainly identifying and extracting the title information of the material in each medical insurance reimbursement material picture, because the title in the medical insurance reimbursement material is usually printed, and the accuracy of OCR recognition and extraction is higher. Further, according to the information extracted from the medical insurance reimbursement material picture, the reimbursement material type of each medical insurance reimbursement material picture is determined, wherein the reimbursement material type is divided into prescription notes, medical records, examination certificates, diagnosis certificates, discharge certificates, medicine boxes, etc. Exemplarily, if the title information extracted from the medical insurance reimbursement material picture is "XX Hospital Prescription Note", then it is determined that the corresponding reimbursement material type is a prescription note.

[0035] Further, a target reimbursement picture of at least one target reimbursement material type is screened out from each medical insurance reimbursement material, wherein the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials of the patient to be reimbursed and / or the rationality of the medication. In an embodiment of the present application, the compliance of the reimbursement materials of the patient to be reimbursed is analyzed mainly by analyzing a class of reimbursement materials that normally contain a real seal and / or a class of reimbursement materials that contain a doctor's signature. In other embodiments, a class of reimbursement materials that normally contain both a real seal and a doctor's signature can also be analyzed. The rationality of medication is analyzed mainly by analyzing a class of reimbursement materials that contain both medication list information and clinical diagnosis information, or a class of reimbursement materials that contain medication list information and clinical diagnosis information respectively, so as to determine whether the medication situation of the patient to be reimbursed is reasonable. Among them, the clinical diagnosis information includes the disease information of the patient to be reimbursed. The medication list information includes but is not limited to at least one therapeutic drug and corresponding usage and dosage information.

[0036] In the embodiment of the present application, the target reimbursement images corresponding to the two target reimbursement material types, medical records and prescription notes, are screened out. Medical records normally contain real stamps and doctor signatures, which are used to analyze the compliance of reimbursement materials, and prescription notes contain clinical diagnosis information and medication list information, which are used to analyze the rationality of medication. In other embodiments, the target reimbursement images corresponding to only the target reimbursement material type, prescription notes, can also be screened out, because the prescription notes of some hospitals normally have real stamps, doctor signatures, clinical diagnosis information, and medication list information at the same time.

[0037] S103: Input each target reimbursement picture into a preset multimodal large model to obtain the corresponding picture information recognition result, and filter the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results.

[0038] Specifically, after each target reimbursement picture is screened out, each target reimbursement picture is input into a preset multimodal large model to obtain a corresponding picture recognition result. In an embodiment of the present application, an achievable implementation method is: construct a first prompt instruction corresponding to the multimodal large model, wherein the prompt instruction is an input instruction that instructs the model to generate a specific output, which is usually in the form of a text string, used to tell the model the task to be performed or the content to be generated. In an embodiment of the present application, the first prompt instruction is a prompt instruction for at least one information recognition task for the target reimbursement picture, telling the multimodal large model the information that needs to be identified and extracted from the target reimbursement picture. In addition, a multimodal large model refers to an artificial intelligence model that can process and understand multiple types of data (such as text, images, audio, etc.), and specifically a MiniGPT-4 model or a ConvNext-L model can be used. Furthermore, the task description information of the information recognition task for the target reimbursement picture can be sent through the terminal of the medical insurance reimbursement staff, and the first prompt instruction is constructed according to the task description information, and then the target reimbursement picture and the first prompt instruction are input into the multimodal large model, and finally the corresponding picture information recognition result is obtained. It should be noted that the format of the picture information recognition result output by the multimodal large model is in JSON format. In addition, the information recognition task includes but is not limited to judging whether the mobile phone screen and computer screen exist in the picture, the clarity of the picture, etc. For the target reimbursement material type used to analyze the compliance of the reimbursement materials, the corresponding information recognition task also includes: judging whether there is a doctor's signature in the picture, whether there is a stamp information, and judging the authenticity of the stamp; for the target reimbursement material type used to analyze the rationality of medication, the corresponding information recognition task also includes: identifying the basic information in the picture, and the basic information includes but is not limited to clinical diagnosis information and medication list information.

[0039] For example, the first prompt instruction is: You are a medical auditor, and I need you to identify the information in the picture and complete the following tasks: 1. Determine the [Picture Category] in the picture. [Picture Category] includes [Prescription Note], [Medicine Box], [Medical Record], [Discharge Certificate], [Physical Examination Report], [Test Report], [Diagnosis Report], [Prescription Form], [Disease Identification Form], and [Drug Instructions]; 2. Identify the [stamp information] in the image. The [stamp information] includes two parts: [whether there is a stamp] and [stamp content]. Stamps do not include handwriting. 1 indicates that there is a stamp, and 0 indicates that there is no stamp; 3. Identify the [signature information] in the image, mainly to identify whether the [doctor], [physician], [auditor], [dispenser], [reviewer], [dispenser] and other roles have signatures. When identifying signatures, "handwritten words", "printed words or misprinted words" are all considered as signatures. Please keep the roles with signatures; [signature information] includes [whether there is a signature] and [signature role]; requirements: 1 means there is a signature, 0 means no signature: if there are multiple signature roles, they can be merged into one array; 4. Determine whether there is a mobile phone screen: whether the photo on the mobile phone screen is displayed in the picture, 1 represents yes, 0 represents no; 5. Determine whether there is a computer screen: whether the picture shows the photo on the computer screen, 1 means yes, 0 means no; 6. Determine [Image Clarity]: Determine the clarity of the image content. The clarity range is [0,1]. The higher the clarity, the clearer it is.

[0040] After the picture information recognition result corresponding to each target reimbursement picture is determined, the information corresponding to the fields [Clinical Diagnosis] and [Medication List] are filtered out from all the picture information recognition results, and finally the clinical diagnosis information and medication list information of the patient to be reimbursed are obtained.

[0041] S104: Based on the clinical diagnosis information and medication list information, the rationality of medication for the patient to be reimbursed is determined through a preset large language model.

[0042] Specifically, after the clinical diagnosis information and the medication list information are determined, it is necessary to determine whether the reimbursement drugs in the medication list information are reasonable for the clinical diagnosis information. In the embodiment of the present application, a feasible implementation method is: matching the reimbursement drug instructions corresponding to each reimbursement drug in the medication list information from the preset drug database, the reimbursement drug instructions contain the efficacy and symptomatic diseases of the corresponding reimbursement drugs, etc., and the drug database includes different drugs and corresponding drug instructions. Further, construct a second prompt instruction, that is, based on the reimbursement drug instructions, determine whether the corresponding reimbursement drug is effective for the disease in the clinical diagnosis information. The second prompt instruction, clinical diagnosis information, medication list information and all reimbursement drug instructions are input into the preset large language model, and finally the large language model outputs at least one output result, each output result characterizing whether a single reimbursement drug in the medication list information is effective for the disease in the clinical diagnosis information. Exemplarily, the output result can be expressed as: "Reimbursement drug name": "A", "Whether it is effective": true, indicating that reimbursement drug A is effective for the disease in the clinical diagnosis information.

[0043] Furthermore, if all output results are valid, it means that all reimbursed drugs in the medication list information are effective for the disease, and the medication rationality result of the patient to be reimbursed is determined to be reasonable; if there are invalid output results in each output result, it means that there are reimbursed drugs that are invalid for the disease in the medication list information, and the medication rationality result of the patient to be reimbursed is determined to be unreasonable. It should be noted that the large language model can use ChatGPT, and in other embodiments, the large language model can also use the BERT model.

[0044] S105: Based on the image information recognition results, the reimbursement materials of the patients to be reimbursed are subjected to compliance analysis to obtain compliance recognition results, and based on the compliance recognition results and the rationality of medication results, the analysis results of the medical insurance fraud behavior of the patients to be reimbursed are determined.

[0045] Specifically, after the rationality of the medication of the patient to be reimbursed is determined, the compliance analysis of the reimbursement materials of the patient to be reimbursed is performed according to the image information recognition results determined by the multimodal large model. In an embodiment of the present application, a feasible analysis method is: the target reimbursement material type used to analyze the compliance of the reimbursement materials in each target reimbursement picture is determined as a compliance identification picture. For example, under normal circumstances, the target reimbursement picture should correspond to the medical record with a real seal and doctor's signature. Furthermore, from the image information recognition result corresponding to the compliance identification image, information corresponding to fields such as "seal information" and "signature information" is obtained to determine whether there is a seal and a doctor's signature in the compliance identification image. If there is no seal in the compliance identification image, which is contrary to the normal situation where a seal should exist, then the compliance recognition result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; conversely, if there is a seal in the compliance identification image, it cannot be directly regarded as the reimbursement materials being compliant, and the authenticity of the seal information needs to be further judged. Then the preset seal authenticity prediction model is called to determine whether the existing seal is forged. The seal authenticity prediction model is a trained convolutional neural network model or a recurrent neural network model. The training process is: the real seal and the seal marked as "false seal" are input into the model as training samples for training until the model converges.

[0046] Furthermore, when the stamp is forged, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; conversely, when the stamp is not forged, if there is no doctor's signature in the compliance identification picture, it means that the reimbursement materials of the patient to be reimbursed do not meet the requirements of medical insurance reimbursement, and the compliance identification result is determined to be non-compliant. Further, if there is a doctor's signature in the compliance identification picture, in order to ensure the accuracy of the compliance identification result, it is necessary to determine whether there is a mobile phone screen photo and / or a computer screen photo in the compliance identification picture when there is a stamp, the stamp is not forged, and the doctor's signature is present according to the corresponding information of the "whether there is a mobile phone screen" and "whether there is a computer screen" fields in the corresponding picture information identification result. If there is a mobile phone screen photo and / or a computer screen photo, this compliance identification picture may contain content taken from other people's mobile phone screens and / or computer screens, and the reimbursement materials have authenticity risks, then the compliance identification result is determined to be non-compliant; conversely, if there is no mobile phone screen photo and / or computer screen photo, then the compliance identification result is determined to be compliant.

[0047] Further, according to the determined compliance identification results and medication rationality results, the analysis results of the medical insurance fraud behavior of the patient to be reimbursed are determined, and the medical insurance fraud behavior analysis results are divided into three categories: significant insurance fraud behavior, doubtful insurance fraud behavior, and non-existent insurance fraud behavior. The possibility of insurance fraud for significant insurance fraud behavior is higher than the possibility of insurance fraud for doubtful insurance fraud behavior. In an embodiment of the present application, a feasible way to determine the analysis results of medical insurance fraud behavior is: if the result of medication rationality is unreasonable, directly determine the analysis results of medical insurance fraud behavior as significant insurance fraud behavior; if the result of medication rationality is reasonable, when the compliance identification result is compliant, determine that the patient to be reimbursed is likely to not have insurance fraud behavior, and determine the analysis results of medical insurance fraud behavior as non-existent insurance fraud behavior; if the result of medication rationality is reasonable, when the compliance identification result is non-compliant, determine the analysis results of medical insurance fraud behavior as doubtful insurance fraud behavior.

[0048] See also Figure 2 The present application embodiment discloses a flowchart of another method for analyzing medical insurance fraud behavior, which can be implemented by a computer program or run on a medical insurance fraud behavior analysis device based on the von Neumann system. The computer program can be integrated into an application or run as an independent tool application, specifically including: S201: Obtain at least one picture of medical insurance reimbursement materials of the patient to be reimbursed.

[0049] S202: Determine the reimbursement material type of each medical insurance reimbursement material image based on OCR technology, and select at least one target reimbursement material type from each medical insurance reimbursement material image, where the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials and / or the rationality of medication of the patient to be reimbursed.

[0050] S203: Input each target reimbursement picture into a preset multimodal large model to obtain the corresponding picture information recognition result, and filter the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results.

[0051] S204: Based on the clinical diagnosis information and medication list information, the rationality of medication for the patient to be reimbursed is determined through a preset large language model.

[0052] S205: Based on the image information recognition result, compliance analysis is performed on the reimbursement materials of the patient to be reimbursed to obtain a compliance recognition result.

[0053] For details, please refer to steps S101-S105, which will not be described in detail here.

[0054] S206: When the compliance identification result is compliant, determine whether the medication rationality result is unreasonable.

[0055] S207: If so, the target reimbursement image from which the clinical diagnosis information and medication list information originate is determined as the medication determination image, and the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined based on the image clarity of the medication determination image and whether there is handwritten text in the clinical diagnosis information and medication list information in the corresponding medication determination image.

[0056] Specifically, if the compliance identification result is compliant, it means that the reimbursement materials of the patient to be reimbursed are compliant and no insurance fraud is found. Then, it is necessary to analyze the insurance fraud in combination with the medication rationality result. If the medication rationality result is unreasonable, it means that not all reimbursed drugs in the medication list information are effective for the disease, there is a possibility of over-prescription and a risk of insurance fraud. Then, the target reimbursement picture from which the clinical diagnosis information and medication list information come is determined as the medication determination picture. In this embodiment of the present application, the clinical diagnosis information and the medication list information come from the same target reimbursement picture. In other embodiments, the clinical diagnosis information and the medication list information may also come from different target reimbursement pictures.

[0057] Furthermore, information corresponding to the "Picture Clarity" field is filtered out from the picture information recognition result corresponding to the medication determination picture, and according to the information corresponding to the "Picture Clarity" field, it is determined whether the picture clarity of the medication determination picture is 1. At the same time, according to the information corresponding to the "Whether there is handwriting" field in the picture information recognition result corresponding to the medication determination picture, it is determined whether there is handwriting in the text of the clinical diagnosis information and the medication list information in the corresponding medication determination picture, wherein the picture clarity takes a value of [0,1], and the greater the picture clarity, the clearer the corresponding medication determination picture.

[0058] If there is no handwriting in the text of the clinical diagnosis information and the medication list information in the corresponding medication determination picture, and the picture clarity of the medication determination picture is 1, that is, it means that the medication determination picture is relatively clear and the text corresponding to the clinical diagnosis information and the medication list information therein is printed, the accuracy of the clinical diagnosis information and the medication list information extracted from the medication determination picture is relatively high, and then the result of verifying the rationality of unreasonable medication is relatively accurate, and the possibility of misjudgment is relatively small, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be an obvious fraud; if the picture clarity of the medication determination picture is 1, and the clinical diagnosis information and / or Or there is handwriting in the text corresponding to the medication list information, then it is easy to have deviations when extracting clinical diagnosis information and / or medication list information, thereby affecting the accuracy of the results of the rationality of medication, that is, the unreasonable rationality of medication may have errors, so it is more reasonable to analyze the medical insurance fraud behavior as suspected insurance fraud; if the picture clarity of the medication determination picture is not 1, it means that it is relatively unclear, so when extracting clinical diagnosis information and medication list information through the medication determination picture, there is a risk of deviation in the extracted content, and the unreasonable rationality of medication may have errors, so the analysis result of the medical insurance fraud behavior is determined to be suspected insurance fraud. In other embodiments, the analysis result of the medical insurance fraud behavior can also be determined synchronously based on the compliance recognition result and the rationality of medication. Exemplarily, when the compliance recognition result is non-compliant and the rationality of medication result is unreasonable, the analysis result of the medical insurance fraud behavior is determined to be an obvious insurance fraud.

[0059] S208: When the compliance identification result is non-compliant, if there is no seal in the compliance identification image or the seal is forged or there is no doctor's signature, then the medical insurance fraud analysis result of the patient to be reimbursed is determined to be an obvious insurance fraud.

[0060] S209: If there is a stamp in the compliance identification image, the stamp is not forged and there is a doctor's signature, then when there is a mobile phone screen photo and / or a computer screen photo in the compliance identification image, the medical insurance fraud analysis result of the patient to be reimbursed is determined to be a suspected insurance fraud.

[0061] Specifically, when the compliance identification result is non-compliant, if the compliance identification picture lacks a seal or the seal is forged or the doctor's signature is missing, it means that the reimbursement materials of the patient to be reimbursed are likely to be fictitious, and the medical insurance fraud analysis result of the patient to be reimbursed is directly determined to be obvious insurance fraud; if the compliance identification picture contains both a seal and a doctor's signature, and the seal is not forged, and at the same time there are photos of mobile phone screens and / or computer screens in the compliance identification picture, it means that the compliance identification picture may contain content from other people's mobile phone screens and / or computer screens, and there is a risk of stealing other people's content, and there is a possibility of insurance fraud, then at this time the medical insurance fraud analysis result of the patient to be reimbursed is determined to be suspected insurance fraud.

[0062] The implementation principle of a medical insurance fraud analysis method in an embodiment of the present application is as follows: keywords in a picture of medical insurance reimbursement materials are identified through OCR technology, thereby determining the reimbursement material type of the picture of the medical insurance reimbursement materials, and then a target reimbursement picture that can analyze the compliance of the reimbursement materials and / or the rationality of medication is selected therefrom, and the corresponding picture information recognition result is more accurately extracted therefrom through a multimodal large model, and the rationality of medication of the patient to be reimbursed is analyzed based on the clinical diagnosis information and medication list information in the picture information recognition result, and then the compliance of the reimbursement materials is analyzed based on other information in the picture information recognition result, and finally, based on the compliance recognition result and the rationality of medication, combined with the two-dimensional analysis of medication rationality and compliance of reimbursement materials, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is more comprehensively determined, thereby improving the accuracy of the analysis of the medical insurance fraud behavior.

[0063] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0064] See also Figure 3 , which is a schematic diagram of the structure of the medical insurance fraud behavior analysis device provided in the embodiment of the present application. The medical insurance fraud behavior analysis device can be implemented as all or part of the device through software, hardware or a combination of both. The device includes an image acquisition module 11, an image screening module 12, an information screening module 13, a reasonable determination module 14 and a fraud analysis module 15.

[0065] The picture acquisition module 11 is used to acquire at least one picture of medical insurance reimbursement materials of the patient to be reimbursed; The image screening module 12 is used to determine the reimbursement material type of each medical insurance reimbursement material image based on the OCR technology, and screen at least one target reimbursement material type from each medical insurance reimbursement material image, where the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials of the patient to be reimbursed and / or the rationality of medication; The information screening module 13 is used to input each target reimbursement picture into a preset multimodal macro model, obtain the corresponding picture information recognition result, and screen the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results; A rational determination module 14 is used to determine the rationality of medication for patients to be reimbursed through a preset large language model based on clinical diagnosis information and medication list information; The insurance fraud analysis module 15 is used to perform compliance analysis on the reimbursement materials of the patients to be reimbursed based on the image information recognition results, obtain compliance recognition results, and determine the medical insurance fraud behavior analysis results of the patients to be reimbursed based on the compliance recognition results and the rationality of medication results.

[0066] Optionally, the information screening module 13 is specifically used for: Construct the first prompt instruction; Based on the first prompt instruction, the target reimbursement picture is input into the multimodal large model to obtain the corresponding picture information recognition result. The first prompt instruction is a prompt instruction for at least one information recognition task for the target reimbursement picture.

[0067] Optionally, the reasonable determination module 14 is specifically used for: Matching the reimbursement drug instructions corresponding to each reimbursement drug in the medication list information from a preset drug database, the drug database including different drugs and corresponding drug instructions; Constructing a second prompt instruction, the second prompt instruction is a prompt instruction for determining whether the corresponding reimbursable drug is effective for the disease in the clinical diagnosis information based on the reimbursable drug instructions; Input the second prompt instruction, clinical diagnosis information, medication list information and all reimbursement drug instructions into the large language model to obtain at least one output result. If all output results are valid, the medication rationality result of the patient to be reimbursed is determined to be reasonable; If there is an invalid output result among the output results, the rationality result of the medication of the patient to be reimbursed is determined to be unreasonable.

[0068] Optionally, the insurance fraud analysis module 15 is specifically used for: Determine a target reimbursement material type target reimbursement material type image used to analyze the compliance of the reimbursement materials of the patient to be reimbursed as a compliance identification image, and determine whether there is a stamp in the compliance identification image based on the image information identification result corresponding to the compliance identification image; If not, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, in the case of forgery in the stamp, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, if there is no forgery in the stamping, and if there is no doctor's signature in the compliance identification image, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, when there is no forgery of the seal and the doctor's signature is present in the compliance identification picture, and if there is a photo of the mobile phone screen and / or the computer screen in the compliance identification picture, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant.

[0069] Optionally, the insurance fraud analysis module 15 is specifically used for: When the compliance identification result is compliant, determine whether the rationality of medication is unreasonable; If so, the target reimbursement image from which the clinical diagnosis information and medication list information originate is determined as the medication determination image, and the results of the medical insurance fraud analysis of the patient to be reimbursed are determined based on the image clarity of the medication determination image and whether there is handwritten text in the clinical diagnosis information and medication list information in the corresponding medication determination image.

[0070] Optional, such as Figure 4 As shown, the device also includes a behavior analysis module 16, which is specifically used to: When the compliance identification result is non-compliant, if there is no stamp in the compliance identification image or the stamp is forged or there is no doctor's signature, the medical insurance fraud analysis result of the patient to be reimbursed is determined to be an obvious insurance fraud; If there is a seal in the compliance identification picture, the seal is not forged and there is a doctor's signature, and if there is a mobile phone screen photo and / or a computer screen photo in the compliance identification picture, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be suspected insurance fraud.

[0071] Optionally, the insurance fraud analysis module 15 is specifically used for: Determine whether the image clarity of the medication determination image is 1, and determine whether the text of the clinical diagnosis information and medication list information in the corresponding medication determination image is handwritten. The image clarity value is [0,1]. The greater the image clarity, the clearer the medication determination image; If there is no handwriting and the picture clarity of the medication determination picture is 1, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be an obvious insurance fraud behavior; If there is handwriting and the image clarity of the medication determination image is 1, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be suspected insurance fraud; If the image clarity of the medication determination image is not 1, the medical insurance fraud analysis result of the patient to be reimbursed is determined to be suspected insurance fraud.

[0072] It should be noted that the medical insurance fraud behavior analysis device provided in the above embodiment only uses the division of the above functional modules as an example when executing the medical insurance fraud behavior analysis method. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the medical insurance fraud behavior analysis device provided in the above embodiment and the medical insurance fraud behavior analysis method embodiment belong to the same concept, and the implementation process thereof is detailed in the method embodiment, which will not be repeated here.

[0073] An embodiment of the present application further discloses a computer-readable storage medium, and the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, a medical insurance fraud behavior analysis method of the above embodiment is adopted.

[0074] Among them, the computer program can be stored in a computer-readable medium, the computer program includes computer program code, the computer program code can be in the form of source code, object code, executable file or certain middleware, etc. The computer-readable medium includes any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer-readable medium includes but is not limited to the above-mentioned components.

[0075] Among them, through this computer-readable storage medium, a medical insurance fraud behavior analysis method of the above embodiment is stored in a computer-readable storage medium, and is loaded and executed on a processor to facilitate the storage and application of the above method.

[0076] An embodiment of the present application also discloses an electronic device, in which a computer program is stored in a computer-readable storage medium. When the computer program is loaded and executed by a processor, the above-mentioned medical insurance fraud behavior analysis method is adopted.

[0077] The electronic device may be a desktop computer, a laptop computer, a cloud server or other electronic device, and the electronic device includes but is not limited to a processor and a memory. For example, the electronic device may also include input and output devices, a network access device, and a bus.

[0078] Among them, the processor can adopt a central processing unit (CPU). Of course, according to actual usage, other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be adopted. The general-purpose processor can adopt a microprocessor or any conventional processor, etc., and this application does not impose any restrictions on this.

[0079] Among them, the memory can be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device, or it can be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD) or a flash memory card (FC) equipped on the electronic device. Moreover, the memory can also be a combination of an internal storage unit and an external storage device of the electronic device. The memory is used to store computer programs and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or is to be output, and this application does not impose any restrictions on this.

[0080] Among them, through this electronic device, a medical insurance fraud behavior analysis method of the above embodiment is stored in the memory of the electronic device, and is loaded and executed on the processor of the electronic device for easy use.

[0081] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field that are not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for analyzing medical insurance fraud behavior. It is characterized in that The method comprises: Obtain at least one picture of the medical insurance reimbursement materials of the patient to be reimbursed; Determine the reimbursement material type of each of the medical insurance reimbursement material images based on OCR technology, and select at least one target reimbursement material type from each of the medical insurance reimbursement material images, wherein the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials and / or the rationality of medication of the patient to be reimbursed; Input each of the target reimbursement pictures into a preset multimodal macro model to obtain corresponding picture information recognition results, and screen the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results; According to the clinical diagnosis information and the medication list information, determining the rationality of medication for the patient to be reimbursed through a preset large language model; According to the picture information recognition result, the reimbursement materials of the patient to be reimbursed are subjected to compliance analysis to obtain a compliance recognition result, and according to the compliance recognition result and the rationality result of medication, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined.

2. The method for analyzing medical insurance fraud behavior according to claim 1, It is characterized in that The target reimbursement pictures are input into a preset multi-modal large model to obtain corresponding picture information recognition results, specifically including: Construct the first prompt instruction; Based on the first prompt instruction, the target reimbursement picture is input into the multimodal large model to obtain a corresponding picture information recognition result, and the first prompt instruction is a prompt instruction for at least one information recognition task for the target reimbursement picture.

3. The method for analyzing medical insurance fraud behavior according to claim 1, It is characterized in that The method of determining the rationality of medication for the patient to be reimbursed based on the clinical diagnosis information and the medication list information by using a preset large language model specifically includes: Matching the reimbursement drug instructions corresponding to each reimbursement drug in the medication list information from a preset drug database, wherein the drug database includes different drugs and corresponding drug instructions; Constructing a second prompt instruction, wherein the second prompt instruction is a prompt instruction for determining whether the corresponding reimbursement drug is effective for the disease in the clinical diagnosis information based on the reimbursement drug instructions; Input the second prompt instruction, the clinical diagnosis information, the medication list information and all reimbursement drug instructions into the large language model to obtain at least one output result, and if all the output results are valid, determine that the medication rationality result of the patient to be reimbursed is reasonable; If there is an invalid output result among the output results, it is determined that the rationality result of the medication of the patient to be reimbursed is unreasonable.

4. The method for analyzing medical insurance fraud behavior according to claim 1, It is characterized in that The step of performing compliance analysis on the reimbursement materials of the patient to be reimbursed based on the image information recognition result to obtain a compliance recognition result specifically includes: Determine a target reimbursement material type target reimbursement material type image used to analyze the compliance of the reimbursement materials of the patient to be reimbursed as a compliance identification image, and determine whether there is a stamp in the compliance identification image based on the image information recognition result corresponding to the compliance identification image; If not, determining that the compliance identification result of the reimbursement materials of the patient to be reimbursed is non-compliant; If so, in the case where the stamp is forged, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, if there is no forgery of the stamp, and if there is no doctor's signature in the compliance identification picture, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant; If so, when there is no forgery in the stamp and there is a doctor's signature in the compliance identification picture, if there is a photo of a mobile phone screen and / or a photo of a computer screen in the compliance identification picture, the compliance identification result of the reimbursement materials of the patient to be reimbursed is determined to be non-compliant.

5. The method for analyzing medical insurance fraud behavior according to claim 4, It is characterized in that Determining the analysis result of the medical insurance fraud behavior of the patient to be reimbursed based on the compliance identification result and the rationality of medication result specifically includes: When the compliance identification result is compliant, determining whether the medication rationality result is unreasonable; If so, the target reimbursement picture from which the clinical diagnosis information and the medication list information are derived is determined as the medication determination picture, and the medical insurance fraud behavior analysis results of the patient to be reimbursed are determined based on the picture clarity of the medication determination picture and whether there is handwritten text in the text of the clinical diagnosis information and the medication list information in the corresponding medication determination picture.

6. The method for analyzing medical insurance fraud behavior according to claim 5, It is characterized in that The method further comprises: When the compliance identification result is non-compliant, if there is no stamp or the stamp is forged or there is no doctor's signature in the compliance identification picture, then the medical insurance fraud analysis result of the patient to be reimbursed is determined to be an obvious insurance fraud; If there is a seal in the compliance identification picture, the seal is not forged and there is a doctor's signature, and if there is a mobile phone screen photo and / or a computer screen photo in the compliance identification picture, the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be a suspected fraud.

7. The method for analyzing medical insurance fraud behavior according to claim 5, It is characterized in that The analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined based on the clarity of the medication determination picture and whether the clinical diagnosis information and the medication list information are handwritten in the text in the corresponding medication determination picture, specifically including: Determine whether the image clarity of the medication determination image is 1, and determine whether the text of the clinical diagnosis information and the medication list information in the corresponding medication determination image is handwritten. The image clarity takes a value of [0,1]. The greater the image clarity, the clearer the medication determination image; If there is no handwriting and the picture clarity of the medication determination picture is 1, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be an obvious insurance fraud behavior; If there is handwriting and the picture clarity of the medication determination picture is 1, then the analysis result of the medical insurance fraud behavior of the patient to be reimbursed is determined to be a suspected insurance fraud behavior; If the picture clarity of the medication determination picture is not 1, it is determined that the medical insurance fraud analysis result of the patient to be reimbursed is a suspected insurance fraud.

8. A device for analyzing medical insurance fraud behavior, It is characterized in that include: A picture acquisition module (11), used to acquire at least one picture of medical insurance reimbursement materials of the patient to be reimbursed; An image screening module (12) is used to determine the reimbursement material type of each of the medical insurance reimbursement material images based on OCR technology, and to screen at least one target reimbursement material type from each of the medical insurance reimbursement material images, wherein the target reimbursement material type is a reimbursement material type used to analyze the compliance of the reimbursement materials of the patient to be reimbursed and / or the rationality of medication; An information screening module (13) is used to input each of the target reimbursement pictures into a preset multimodal macro model to obtain corresponding picture information recognition results, and screen the clinical diagnosis information and medication list information of the patient to be reimbursed from all the picture information recognition results; A rationality determination module (14), used to determine the rationality of the medication of the patient to be reimbursed through a preset large language model based on the clinical diagnosis information and the medication list information; The insurance fraud analysis module (15) is used to perform compliance analysis on the reimbursement materials of the patient to be reimbursed based on the image information recognition result, obtain a compliance recognition result, and determine the medical insurance fraud behavior analysis result of the patient to be reimbursed based on the compliance recognition result and the medication rationality result.

9. A computer-readable storage medium having a computer program stored therein. It is characterized in that When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 7 is adopted.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, It is characterized in that When the processor loads and executes the computer program, the method according to any one of claims 1 to 7 is adopted.

Images