Method, device, related equipment and program product for auditing rationality of drg grouping

Abstract

The application discloses a DRG group combination rationality auditing method and device, related equipment and program product. The application obtains a target case to be audited and a target group prepared for the target case; obtains auditing logic corresponding to the target group, the auditing logic being obtained by summarizing diagnosis and treatment path data corresponding to the target group through a large model configured, and including a plurality of auditing rules; obtains an auditing subtask corresponding to each auditing rule corresponding to the target group, and calls a medical insurance auditing large model to execute each auditing subtask on the target case to obtain an auditing result corresponding to each auditing subtask, the medical insurance auditing large model being obtained by training a base large model with training data of a plurality of auditing subtasks; and the application summarizes the auditing results of the auditing subtasks to obtain a final auditing result. The application completes the formulation of the auditing logic and the group combination rationality auditing by calling the large model capacity, reduces the demand for expert resources, and is not affected by the personal experience and judgment of experts.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and more particularly, to a DRG grouping rationality auditing method and device, related equipment and program product. BACKGROUND

[0002] With the continuous improvement of medical security system, the medical insurance payment method is upgraded from project payment to disease payment, among which, the DRG (Diagnosis Related Groups) payment method has become the mainstream payment method under the disease payment method. The DRG payment method is a system that groups inpatients according to diagnosis, treatment process, and disease severity, and pays medical insurance according to the predetermined payment standard of the group. This payment method helps to alleviate the over-diagnosis and treatment behavior under the project payment, and promotes medical institutions to actively control costs and standardize diagnosis and treatment behavior. Although the DRG payment method has the above advantages, it also has the problem of unreasonable grouping and other audit risks, which requires comprehensive grouping rationality supervision.

[0003] At present, the method for auditing the rationality of DRG grouping mainly relies on the auditing rules formulated by expert teams. These experts usually include medical insurance management experts, information technology experts, statistical analysts, medical record management experts and clinical medicine experts, etc. They combine their professional knowledge and rich clinical experience to jointly formulate a rule system aimed at ensuring the quality and efficiency of the audit. This auditing method still has some defects:

[0004] Firstly, the expert resources are limited and the cost is high, which is difficult to cover all subfields and regions. Secondly, the audit standards may be affected by the personal experience and judgment of the experts, and there is subjectivity. SUMMARY

[0005] In view of the above problems, the present application is proposed to provide a DRG grouping rationality auditing method, device, related equipment and program product, to solve some or all of the defects existing in the method of relying on the auditing rules formulated by experts. The specific scheme is as follows:

[0006] In a first aspect, a DRG grouping rationality auditing method is provided, comprising:

[0007] obtaining a target case to be audited and a target disease group to which the target case is intended to be grouped;

[0008] obtaining the configured audit logic corresponding to the target disease group, the audit logic being an audit logic obtained by configuring a large model to summarize the diagnosis and treatment path data corresponding to the target disease group, the audit logic including one or more audit rules;

[0009] For the audit logic corresponding to the target disease group, an audit subtask corresponding to each of the audit rules is obtained, and a configured medical insurance audit large model is called to instruct the medical insurance audit large model to perform each of the audit subtasks on the target case to obtain an audit result corresponding to each of the audit subtasks, the medical insurance audit large model being obtained by training a base large model using training data of multiple audit subtasks;

[0010] The audit results of the audit subtasks in the audit logic are summarized to obtain an audit result representing DRG group rationality.

[0011] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the process of obtaining the audit logic by summarizing the diagnosis and treatment path data corresponding to the target disease group through the configured large model includes:

[0012] The first prompt instruction prompt is sent to the configured large model to obtain the audit logic of the target disease group output by the large model;

[0013] The first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic of the target disease group.

[0014] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the first prompt instruction prompt adopts a thinking chain structure and is used to instruct the large model to sequentially perform the following steps:

[0015] First, the diagnosis and treatment path data corresponding to the target disease group is analyzed to identify key elements therein;

[0016] Further, the audit logic of the target disease group is formulated according to the key elements.

[0017] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, for the audit logic corresponding to the target disease group, the process of obtaining an audit subtask corresponding to each of the audit rules includes:

[0018] The correspondence between the configured audit rules and the audit subtasks is queried to obtain an audit subtask corresponding to each of the audit rules in the audit logic corresponding to the target disease group.

[0019] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, before the configured medical insurance audit large model is called to instruct the medical insurance audit large model to perform each of the audit subtasks on the target case, the process further includes:

[0020] For each of the audit rules in the audit logic:

[0021] determine a task complexity level based at least on explicitness of the audit rule;

[0022] if the task complexity level indicates that a set complexity threshold is exceeded, invoke the medical insurance audit large model to instruct the medical insurance audit large model to perform an audit subtask corresponding to the audit rule for the target case.

[0023] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the method further includes:

[0024] if the task complexity level indicates that the set complexity threshold is not exceeded, invoke a configured rule engine to perform the audit subtask corresponding to the audit rule for the target case to obtain an audit result corresponding to the audit subtask.

[0025] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the process of determining the task complexity level based at least on the explicitness of the audit rule includes:

[0026] determine the task complexity level based on the explicitness of the audit rule, complexity of the target disease group, and structured degree of the target case;

[0027] wherein the less explicit the audit rule is, the higher the complexity of the target disease group is, and the lower the structured degree of the target case is, the higher the corresponding task complexity level is.

[0028] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the explicitness of the audit rule, the complexity of the target disease group, and the structured degree of the target case are all obtained by invoking the medical insurance audit large model.

[0029] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the training process of the medical insurance audit large model includes:

[0030] construct training data corresponding to each audit subtask in the audit logic of each disease group, the training data including task instruction description corresponding to the audit subtask, case sample, and result label, the result label indicating an audit result obtained by performing the audit subtask on the case sample;

[0031] train a configured base large model using the training data corresponding to each audit subtask to obtain the trained medical insurance audit large model.

[0032] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the process of constructing the training data corresponding to each audit subtask includes:

[0033] Obtaining the training data samples corresponding to each audit subtask manually labeled to form a sample library;

[0034] Iteratively extracting the training data samples from the sample library and assembling them into a second prompt instruction prompt, the second prompt instruction prompt being used to instruct the large model to refer to the training data samples to generate new case samples and corresponding result labels;

[0035] Sending the second prompt instruction prompt into the large model to obtain the case samples and result labels output by the large model, and combining the task instruction description in the training data samples into new training data;

[0036] Adding the new training data to the training data set corresponding to the audit subtask.

[0037] In a possible design, in another implementation manner of the first aspect of the embodiment of the present application, the second prompt instruction prompt, on the basis of instructing the large model to refer to the training data samples to generate new case samples and corresponding result labels, is also used to instruct the large model to output explanation content, the explanation content being the explanatory content of the generated new case sample for the audit result of the task instruction description being the result label.

[0038] Secondly, a DRG grouping reasonableness auditing device is provided, which includes:

[0039] A to-be-audited data acquisition unit is configured to acquire a target case to be audited and a target disease group to which the target case is to be grouped;

[0040] An auditing logic acquisition unit is configured to acquire an auditing logic corresponding to the target disease group configured, the auditing logic being an auditing logic obtained by a configured large model summarizing diagnosis and treatment path data corresponding to the target disease group, the auditing logic including one or more auditing rules;

[0041] A subtask acquisition unit is configured to acquire, for the auditing logic corresponding to the target disease group, an audit subtask corresponding to each of the auditing rules;

[0042] A medical insurance auditing large model calling unit is configured to call a configured medical insurance auditing large model to instruct the medical insurance auditing large model to perform each of the audit subtasks on the target case to obtain an audit result corresponding to each of the audit subtasks, the medical insurance auditing large model being obtained by training a base large model with training data of multiple audit subtasks.

[0043] a result aggregation unit configured to aggregate the audit results of the audit sub-tasks in the audit logic to obtain an audit result representing the DRG grouping rationality.

[0044] In a third aspect, an electronic device is provided, comprising: a memory and a processor;

[0045] The memory is configured to store a program.

[0046] The processor is configured to execute the program to implement the steps of the DRG grouping rationality auditing method described in any one of the preceding first aspects.

[0047] In a fourth aspect, a readable storage medium is provided, having a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the DRG grouping rationality auditing method described in any one of the preceding first aspects.

[0048] In a fifth aspect, a computer program product is provided, comprising a computer program, and the computer program, when executed by a processor, implements the steps of the DRG grouping rationality auditing method described in any one of the preceding first aspects.

[0049] Through the above technical solution, the present application uses the ability of a large model to summarize the abstract diagnosis and treatment path data to obtain the audit logic of the target disease group, which includes one or more audit rules. That is, the present application does not rely on experts to formulate audit rules based on clinical experience, but starts from the diagnosis and treatment path data as a clinical treatment guideline, and uses the natural language understanding and objectification summarization ability of a large model to summarize the audit logic of the target disease group from the diagnosis and treatment path data, thereby reducing the demand for expert resources and being more objective without being affected by the personal experience and judgment of experts.

[0050] Further, in the auditing of the DRG grouping rationality of the target case, the auditing sub-tasks corresponding to each auditing rule in the auditing logic corresponding to the target group are obtained, that is, for each auditing rule in the auditing logic, a corresponding auditing sub-task is constructed, which is used to verify whether the case data meets the requirements of the corresponding auditing rule. Through the construction of the above-mentioned auditing sub-tasks, the complex auditing process is divided into several auditing sub-tasks. On this basis, the configured medical insurance auditing large model can be called to sequentially execute each auditing sub-task for the target case to obtain the auditing result of each auditing sub-task. The medical insurance auditing large model can be obtained by training the base large model with the training data of multiple auditing sub-tasks. Through the training of multiple tasks, the medical insurance auditing large model can better understand and generate medical auditing related texts, and on this basis, the medical insurance auditing large model can better execute each auditing sub-task to obtain more accurate auditing results. Finally, by aggregating the auditing results of each auditing sub-task, the DRG grouping rationality auditing result of the target case can be obtained. Since the complex auditing logic of the target group is disassembled into several auditing sub-tasks, and the medical insurance auditing large model is called to execute each auditing sub-task, the explainability of the final auditing result is stronger, which matches the high requirement of transparency of the decision-making process in the medical field. BRIEF DESCRIPTION OF DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference made to the accompanying drawings. The drawings are for purposes of illustration only and are not intended to be limiting of the present application. Like reference numerals are used to refer to like elements throughout. In the drawings:

[0052] Figure 1 An embodiment system architecture schematic diagram of the DRG grouping rationality auditing method provided by the embodiments of the present application;

[0053] Figure 2 A terminal structure schematic diagram provided by the embodiments of the present application;

[0054] Figure 3 A server structure schematic diagram provided by the embodiments of the present application;

[0055] Figure 4 An embodiment system architecture schematic diagram of the DRG grouping rationality auditing method provided by the embodiments of the present application;

[0056] Figure 5 An embodiment system architecture schematic diagram of the DRG grouping rationality auditing method provided by the embodiments of the present application;

[0057] Figure 6 An embodiment system architecture schematic diagram of the DRG grouping rationality auditing method provided by the embodiments of the present application;

[0058] Figure 7 A structure schematic diagram of a DRG group combination rationality auditing device provided for an embodiment of the present application;

[0059] Figure 8 A structure schematic diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0060] Before introducing the scheme of the present application, first, the related concepts involved in this paper are explained:

[0061] DRG group entry: refers to determining the disease group to which the target case is to be divided according to the DRG payment method.

[0062] Diagnosis and treatment path: the diagnosis and treatment path is a standardized diagnosis and treatment plan formulated for a specific disease or surgery, aiming to standardize medical service behavior, reduce resource waste, and ensure that patients receive appropriate medical care services.

[0063] prompt: instruction. When interacting with AI (such as artificial intelligence models), instructions need to be sent to AI, which can be a text description, such as "please recommend a popular song" when you interact with AI, or a parameter description in a certain format, such as asking AI to draw according to a certain format, and describing the relevant drawing parameters.

[0064] Large model: in the field of artificial intelligence, large model usually refers to large-scale pre-training model. Such models are called "large" because they have achieved pre-training on a large amount of data and can be transferred to multiple downstream tasks. Their English full name is Large Pre-Trained Models or Large-Scale Pre-Training Models. The characteristics of large models are large scale, containing tens of billions or even more parameters to help them learn complex patterns in data. The emerging capabilities of large models include but are not limited to: context learning, instruction following, code generation, sequential reasoning, etc. Large models can include large language models (Large language model, LLM), as well as multi-modal large models. Large language models are mainly used to process text modal data, and multi-modal large models further integrate multi-modal capabilities on the basis of large language models, and can process information in multiple modalities such as images, texts, audios, etc.

[0065] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0066] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0067] This application provides a method for rational review of DRG inclusion factors, which can be applied to, for example... Figure 1 The system architecture shown may include a terminal 100 and a server 200. The server 200 may include one or more servers (…). Figure 1 (This example uses a server as an illustration).

[0068] Either terminal 100 or server 200 can be used independently to execute the DRG input rationality review method provided in this application embodiment. Alternatively, terminal 100 and server 200 can also be used collaboratively to execute the DRG input rationality review method provided in this application embodiment.

[0069] The following description Figure 1 The product form of the mid-terminal 100;

[0070] The terminal 100 in this application embodiment can be a mobile phone, tablet computer, interactive robot, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc., and this application embodiment does not impose any restrictions on it.

[0071] Figure 2 A schematic diagram of an optional hardware structure for terminal 100 is shown.

[0072] refer to Figure 2As shown, the terminal 100 can include a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160, a speaker 161, a microphone 162, a headphone jack 163 (optional), a processor 170, an external interface 180, a power supply 190, and the like. Those skilled in the art can understand that Figure 2 The terminal or multifunctional device is merely an example and does not constitute a limitation on the terminal or multifunctional device, and can include more or fewer components than shown, or combine certain components, or different components.

[0073] The input unit 130 can be used to receive inputted digital or character information, and to generate key signal input related to user settings and function control of the portable multifunctional device. Specifically, the input unit 130 can include a touch screen 131 and / or other input devices 132. The touch screen 131 can provide an input interface and an output interface between the terminal 100 and a user. In addition to the touch screen 131, the input unit 130 can also include other input devices. Specifically, the other input devices 132 can include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, on / off keys, etc.), trackballs, mice, joysticks, etc.

[0074] Among them, the other input devices 132 can receive inputted data and the like.

[0075] The display unit 140 can be used to display information inputted by a user or provided to a user, various menus of the terminal 100, interactive interfaces, file display, and / or playing of any kind of multimedia file. In the embodiments of the present application, the display unit 140 can be used to display various interactive interfaces, processing results, etc. in the DRG group combination rationality auditing method.

[0076] The memory 120 can be used to store instructions and data. The memory 120 can mainly include a storage instruction area and a storage data area. The storage data area can store various data, such as multimedia files, texts, etc. The storage instruction area can store software units required by at least one function, such as an operating system, an application, instructions, etc., or their subsets, expanded sets. It can also include a non-volatile random access memory; to provide the processor 170 with software and applications that include managing hardware, software, and data resources in a computing processing device, supporting control. It is also used for the storage of multimedia files, and the storage of running programs and applications.

[0077] The processor 170 is the control center of the terminal 100, which connects each part of the entire terminal 100 through various interfaces and lines, executes various functions of the terminal 100 and processes data by running or executing instructions stored in the memory 120 and calling data stored in the memory 120, thereby performing overall control of the terminal device.

[0078] The memory 120 can be configured to store software codes related to the DRG group combination rationality auditing method, and the processor 170 can execute the steps of the DRG group combination rationality auditing method, and can also dispatch other units (for example, the input unit 130 and the display unit 140) to realize corresponding functions.

[0079] The radio frequency unit 110 (optional) can be configured to receive and send signals in the process of receiving or calling information, for example, receiving the downlink information of the base station and processing by the processor 170; in addition, sending the designed uplink data to the base station. In the embodiments of the present application, the radio frequency unit 110 can send data to the server 200 and receive the processing result sent by the server 200. For example, the radio frequency unit 110 sends the target case to be audited and the target disease group of the proposed DRG group to the server 200, the server 200 audits and judges the rationality of the DRG group combination, obtains the auditing result and feeds back to the terminal 100.

[0080] It should be understood that the radio frequency unit 110 is optional, which can be replaced by other communication interfaces, for example, can be a network interface.

[0081] Although not shown, the terminal 100 can also include a flash, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which will not be described here. Part or all of the methods described below can be applied in the terminal 100 as shown in Figure 2 .

[0082] Next, the product form of the server 200 in Figure 1 will be described;

[0083] Figure 3 A structural diagram of the server 200 is provided, as shown in Figure 3 , the server 200 includes a bus 201, a processor 202, a communication interface 203 and a memory 204. The processor 202, the memory 204 and the communication interface 203 communicate through the bus 201.

[0084] The bus 201 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 3 , only one thick line is used, but it does not mean that there is only one bus or one type of bus.

[0085] The processor 202 can be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor, a microcontroller, a digital signal processor (DSP), or the like.

[0086] The memory 204 can include volatile memory, such as random access memory (RAM), and non-volatile memory, such as read-only memory (ROM), a floppy disk drive, a hard disk drive, or a solid-state drive.

[0087] The memory 204 can be configured to store software code related to the DRG group entry rationality auditing method, and the processor 202 can execute the steps of the DRG group entry rationality auditing method or schedule other units to implement corresponding functions.

[0088] It should be understood that the terminal 100 and the server 200 described above can be centralized or distributed devices, and the processors (such as the processor 170 and the processor 202) in the terminal 100 and the server 200 can be hardware circuits (such as application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), general-purpose processors, digital signal processors (DSPs), microprocessors, or microcontrollers) or combinations of these hardware circuits. For example, the processor can be a hardware system with an execution instruction function, such as a CPU, a DSP, or the like, or a hardware system without an execution instruction function, such as an ASIC, an FPGA, or the like, or a combination of the hardware system without the execution instruction function and the hardware system with the execution instruction function.

[0089] Currently, the method for auditing the rationality of DRG grouping mainly relies on the auditing rules established by expert teams. Experts, with their professional knowledge and rich clinical experience, jointly develop a rule system aimed at ensuring the quality and efficiency of the audit. First, the expert resources are limited and the cost is high, which is difficult to cover all subfields and regions. Second, the auditing standards may be influenced by the personal experience and judgment of experts, and there is subjectivity.

[0090] Therefore, a possible solution can use a traditional deep learning model to perform the rationality audit of DRG grouping. The deep learning model is trained to identify and predict the rationality of DRG grouping. The specific steps include data preprocessing, feature engineering, model training, model verification, and finally the audit decision. The trained model is applied to the actual DRG grouping audit, and the model can automatically score or classify new cases to assist auditors in making decisions.

[0091] However, the training of traditional deep learning models requires a large amount of labeled data, and the acquisition and labeling of medical data is costly. The deep learning model has limited capabilities, and its output accuracy is easily limited when facing some complex case data. In addition, deep learning models are often considered "black box" models, and their decision-making process lacks transparency and interpretability, which conflicts with the high requirements for transparency of decision-making process in the medical field. These limiting factors hinder the widespread application of deep learning models in the rationality audit of DRG grouping.

[0092] In summary, the present application provides a solution to achieve the rationality audit of DRG grouping.

[0093] The embodiments of the present application provide a DRG grouping rationality audit method. The method is applied to a computer device, which can be a terminal 100 in Figure 1 or a system composed of a terminal 100 and a server 200. Referring to Figure 4 , the DRG grouping rationality audit method specifically includes the following steps:

[0094] Step S100, obtaining a target case to be audited and a target disease group to be grouped in the DRG.

[0095] Specifically, the present application can achieve the audit of the rationality of DRG grouping. First, the target case information of the patient to be audited and the target disease group to be grouped in the DRG need to be obtained. The DRG grouping rationality audit of the present application is to review the rationality of dividing the target case into the target disease group.

[0096] Step S110, obtain the audit logic corresponding to the configured target disease group, the audit logic being an audit logic obtained by a large model configured to summarize the diagnosis and treatment path data corresponding to the target disease group, the audit logic including one or more audit rules.

[0097] Specifically, in order to solve the problem that the conventional scheme relies on experts to formulate the audit rules of the disease group, the present application adopts a large model-based audit logic formulation scheme.

[0098] Before performing the DRG group entry rationality audit task, the present application can pre-aid large model powerful natural language understanding and concretization induction ability, and the diagnosis and treatment path data corresponding to each disease group is summarized and summarized, so that the large model generates the audit logic of each disease group based on the diagnosis and treatment path data, which can include one or more audit rules.

[0099] Among them, the diagnosis and treatment path data corresponding to each disease group can refer to the standard diagnosis and treatment plan file formulated in the medical field, which specifies the standard diagnosis and treatment process for each type of disease in the disease group. By calling the large model, the diagnosis and treatment path data corresponding to the disease group is summarized, and the audit logic of the disease group can be formulated based on the standard diagnosis and treatment process. Compared with the way that experts rely on clinical experience and personal judgment, the audit logic formulated by the present application scheme is more standardized and objective.

[0100] The process of formulating the audit logic of the disease group by means of the large model in this embodiment can use various types of general large models or medical field vertical large models.

[0101] For each disease group corresponding to the pre-constructed audit logic, it can be saved accordingly. In the process of performing the DRG group entry rationality audit task, the corresponding relationship can be queried for the target disease group to be audited to obtain its corresponding audit logic.

[0102] An example of the audit logic of HT1 disease group is provided as follows, which includes 5 audit rules:

[0103] ①

Diagnosis

[0104] ②

Hospital orders, charge list

[0105] ③

Medical record

[0106] ④ The [Medical Records and Test Results] indicate a Marshall score ≥2, which proves organ dysfunction. These indicators include systolic blood pressure, creatinine, and oxygenation index.

[0107] ⑤

Examination Results

[0108] Step S120: For the review logic corresponding to the target disease group, obtain the review sub-task corresponding to each review rule.

[0109] Specifically, in this embodiment, the DRG inclusion rationality review task is executed by calling the medical insurance review model. To improve the accuracy of the review results, this application breaks down the complex review process into several review sub-tasks, and then calls the medical insurance review model to execute each review sub-task separately. The DRG inclusion rationality review process is mainly based on review logic. Therefore, a corresponding review sub-task can be constructed for each review rule in the review logic. The review sub-task is used to verify whether the case data meets the requirements of the corresponding review rule.

[0110] The process of obtaining the audit subtasks corresponding to the audit rules in this step can be done by temporarily generating the audit subtasks for each audit rule, or by pre-constructing the corresponding audit subtasks for each audit rule before executing the DRG group rationality audit task. Then, during the execution of the DRG group rationality audit task, the correspondence between the configured audit rules and audit subtasks can be queried to obtain the audit subtasks corresponding to each audit rule in the audit logic for the target disease group.

[0111] Depending on the different review rules, the review subtasks can be of various types, including but not limited to:

[0112] Numerical comparison tasks: Large models compare the numerical values ​​of different test results, such as blood glucose and blood pressure, to determine whether they conform to the normal range or the trend of improvement / deterioration specified in the clinical pathway. These tasks help assess patients' health status and treatment effectiveness.

[0113] Examination-related tasks: The large model learns to interpret various medical examination reports, including imaging and laboratory tests, to determine whether the results support the current diagnosis or whether further investigation is needed. This task is crucial for ensuring patients receive appropriate examinations and treatments.

[0114] Medical order-related tasks: The large model analyzes doctors' medical orders, including medication prescriptions, treatment plans, and follow-up instructions, to ensure they align with the patient's condition and treatment pathway. This helps improve the personalization and compliance of treatment.

[0115] Drug-related tasks: The large model evaluates whether the drug use is reasonable, including drug selection, dosage, frequency of drug use, and length of treatment course, as well as drug interactions. Through this task, the large model can provide more accurate review for clinical drug treatment.

[0116] Expense-related tasks: The large model reviews the reasonableness of medical expenses, including diagnosis and treatment expenses, drug expenses, and examination expenses, etc., to ensure that the expenses are consistent with the medical services provided and meet the medical insurance payment standards.

[0117] In this step, for the target disease group, the audit sub-tasks corresponding to each audit rule in the audit logic are obtained, that is, the complex audit process for the target disease group is decomposed into several audit sub-tasks. Through the audit of the audit sub-tasks, the audit process for the target disease group is completed.

[0118] Step S130, calling the configured medical insurance audit large model to instruct the medical insurance audit large model to execute each audit sub-task for the target case, and obtaining the audit result corresponding to each audit sub-task.

[0119] The medical insurance audit large model is obtained by training the base large model using multiple audit sub-task training data. The base large model can be a general large model or a medical field large model. By training the base large model using multiple audit sub-task training data, the trained medical insurance audit large model can better understand and generate medical audit related text, that is, to promote the medical insurance audit large model to better complete the audit sub-task and obtain more accurate audit results.

[0120] In one possible implementation, when the medical insurance audit large model is called to execute the audit sub-task in this embodiment, a multi-round dialogue interactive audit mechanism can be used between the user and the system,

[0121] During the multi-round dialogue process, the medical insurance audit large model can output the audit result of each audit sub-task and its corresponding audit rule, so that the user can clearly understand the audit basis, and the final audit result has stronger explainability. In addition, during the multi-round dialogue process, the system is also allowed to obtain more information by further asking the user when encountering uncertain or complex situations, so as to support the medical insurance audit large model to give more accurate audit decisions.

[0122] Step S140, aggregating the audit results of each audit sub-task in the audit logic of the target disease group to obtain an audit result representing the reasonableness of DRG grouping.

[0123] After aggregating the audit results of each audit sub-task in the audit logic of the target disease group, the final audit result of the reasonableness of DRG grouping can be determined, that is, the result of whether it is reasonable to divide the target case into the target disease group is determined.

[0124] In a possible implementation, if the audit results of each audit subtask in the audit logic all indicate that the corresponding audit rule is met, it can be determined that the DRG group entry is reasonable, otherwise, it is determined that the DRG group entry is unreasonable.

[0125] The DRG group entry reasonableness auditing method provided in this embodiment is based on abstract diagnosis and treatment path data, and the audit logic of the target disease group is obtained by calling the summarization capability of the large model. Instead of relying on experts to formulate audit rules based on clinical experience, the audit logic of the target disease group is summarized from the diagnosis and treatment path data by means of the natural language understanding and objectification summarization capability of the large model, thereby reducing the demand for expert resources and being more objective and less affected by personal experience and judgment of experts.

[0126] Further, the complex audit process is divided into a plurality of audit subtasks in the method. On this basis, the configured medical insurance audit large model can be called to execute each audit subtask for the target case in sequence to obtain the audit result of each audit subtask. Finally, the reasonableness auditing result of the DRG group entry of the target case can be obtained by summarizing the audit results of the audit subtasks. Since the complex audit logic of the target disease group is disassembled into a plurality of audit subtasks and the medical insurance audit large model is called to execute each audit subtask, the explainability of the final audit result is stronger, which matches the high requirement of transparency of the decision-making process in the medical field.

[0127] The method can realize an automatic audit process. On the other hand, when the medical insurance audit large model is called for auditing, a multi-round dialogue interactive mechanism can be used to simulate natural language dialogue, so that the auditing process is more transparent and easy to understand.

[0128] In some embodiments of the present application, the process of calling the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic involved in the foregoing embodiments is introduced.

[0129] In order to call the large model to execute the audit logic generation task, a prompt instruction prompt is provided in this embodiment, which is defined as a first prompt instruction prompt. The first prompt instruction prompt is specifically used to instruct the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic of the target disease group.

[0130] On this basis, the first prompt instruction prompt can be sent to the large model to obtain the audit logic of the target disease group output by the large model.

[0131] In some possible implementations, in order to better stimulate the large model to perform the audit logic generation task, a thought chain type prompt design scheme is provided in this embodiment, that is, the first prompt instruction prompt adopts a thought chain type structure design, and the audit logic generation task is divided into a plurality of sub-tasks executed in sequence, which can specifically include:

[0132] First, analyze the diagnosis and treatment path data corresponding to the target patient group, and identify the key elements therein.

[0133] Among them, the key elements include but are not limited to: diagnosis code, treatment item, drug use, surgical operation, examination result, etc.

[0134] Further, according to the identified key elements, the audit logic of the target patient group is formulated.

[0135] As follows, an optional example of the first prompt instruction prompt is provided:

[0136] "Suppose you are a medical audit expert, please summarize the audit logic according to the given diagnosis and treatment path, and summarize according to the following steps:

[0137] First, identify the key elements in the diagnosis and treatment path:

[0138] 1. Please analyze the diagnosis and treatment path data provided by the patient group, and identify the key elements in the diagnosis and treatment path.

[0139] 2. Key elements include but are not limited to diagnosis code, treatment item, drug use, surgical operation, examination result, etc.

[0140] 3. For each element, extract the relevant data field and identify its role and importance in the diagnosis and treatment path.

[0141] Then, formulate the patient group audit logic:

[0142] 1. According to the identified key elements of the diagnosis and treatment path, formulate a set of patient group audit logic.

[0143] 2. The audit logic should include verification rules for key elements, such as the consistency of diagnosis code and treatment item, the rationality of drug use, etc.

[0144] 3. Please consider the characteristics of the patient group, and formulate the corresponding audit process and standards to ensure that the audit logic not only meets the medical standards, but also efficiently assists ordinary doctors in medical record auditing.

[0145] Finally, output the audit logic of the patient group.".

[0146] In this embodiment, by adopting the prompt design scheme of the thinking chain, the large model can be guided to identify the key element information in the diagnosis and treatment path first, and then the audit logic is formulated based on the key element information. The complete audit logic formulation task is divided into several logically related sub-tasks, which can better guide the large model to complete the formulation of the audit logic step by step, and improve the accuracy of the output results of the large model.

[0147] The application introduces the ability of large model instantiation induction, solves the problem of wasting manpower and low efficiency existing in the traditional way of formulating audit rules relying on expert resources. The large model has strong knowledge memory ability, can remember multi-dimensional knowledge of different diseases through pre-training of massive medical books and electronic medical records, and can analyze and find the internal rules and patterns of data through artificial guidance. This feature makes it better to complete the formulation of the disease group audit logic.

[0148] In some embodiments of the application, another implementation scheme of DRG group combination rationality audit is provided. In this embodiment, both the medical insurance audit large model and the traditional rule engine are provided, and the audit capability can be selected according to the complexity and intelligence of the audit rules. Finally, the audit results of the two audit capabilities are summarized to obtain the final audit result. Referring to Figure 5 As shown, the method can specifically include the following steps:

[0149] Step S200, obtaining a target case to be audited and a target disease group to which the target case is intended to be grouped.

[0150] Step S210, obtaining the audit logic corresponding to the target disease group configured, the audit logic being obtained by configuring the large model to summarize the diagnosis and treatment path data corresponding to the target disease group, and the audit logic including one or more audit rules.

[0151] Step S220, for the audit logic corresponding to the target disease group, obtaining the audit sub-tasks corresponding to each audit rule.

[0152] Steps S200-S220 in this embodiment correspond one by one to steps S100-S120 in the foregoing embodiments, and details are referred to the foregoing description, which will not be repeated here.

[0153] Step S230, calculating the task complexity for each audit sub-task.

[0154] Specifically, for each audit rule in the audit logic corresponding to the target disease group, the task complexity of the audit subtask corresponding to the audit rule is determined based at least on the explicitness of the audit rule. The task complexity represents the complexity of executing the audit subtask. It can be understood that the higher the task complexity, the higher the task processing capability of the engine required to execute the task, at which point the medical insurance audit large model can be called to execute such audit subtasks. The lower the task complexity, the lower the processing difficulty of this type of audit subtask, which can be processed by standardized audit processing through pre-audit rules, and therefore a traditional rule engine can be called to execute such audit subtasks.

[0155] In some possible implementations, it is considered that the task complexity can be related not only to the explicitness of the audit rule, but also to other factors, such as: the complexity of the target disease group, the structured degree of the target case, etc. Therefore, one or more of the influencing factors such as the explicitness of the audit rule, the load of the target disease group, and the structured procedure of the target case can be considered to determine the task complexity of the audit subtask.

[0156] For example, the explicitness of the audit rule, the complexity of the target disease group, and the structured degree of the target case can be considered to determine the task complexity.

[0157] Wherein, the less explicit the audit rule, the higher the complexity of the target disease group, and the lower the structured degree of the target case, the higher the corresponding task complexity.

[0158] The task complexity score is defined as C, the explicitness score of the audit rule is defined as R m , the complexity score of the target disease group is defined as R c , and the structured degree score of the target case is defined as R s , then the following formula is satisfied:

[0159] C = w m × (1-R m ) + w s × (1-R s ) + w c × R c

[0160] Wherein, R m ranges from 0 to 1, 1 represents complete explicitness, and 0 represents complete inexplicitness; R s ranges from 0 to 1, 1 represents complete structuring, and 0 represents complete non-structuring, R c ranges from 0 to 1, 1 represents very complex, and 0 represents very simple. w m is the weight of the explicitness of the audit rule, w s is the weight of the structured degree of the target case, and w cTo weight the complexity of the target disease group, three weights can be adjusted according to actual conditions to reflect the influence of different factors on the complexity of the task.

[0161] Step S240, determining whether the complexity of the task exceeds the set complexity threshold, if yes, executing step S250, otherwise, executing step S260.

[0162] Specifically, if the complexity of an audit subtask exceeds the set complexity threshold, it indicates that the audit subtask is too complex and cannot be processed by the traditional rule engine, and thus the medical insurance audit large model needs to be called to process the audit subtask. If the complexity of an audit subtask does not exceed the set complexity threshold, it indicates that the audit subtask is relatively simple and can be processed by the traditional rule engine.

[0163] Step S250, calling the configured medical insurance audit large model to instruct the medical insurance audit large model to execute the audit subtask for the target case, and obtaining the audit result corresponding to the audit subtask.

[0164] Step S260, calling the configured rule engine to execute the audit subtask for the target case, and obtaining the audit result corresponding to the audit subtask.

[0165] The rule engine can be configured by using existing technologies or other methods. The rule engine can implement the audit subtask through rule matching for cases with relatively clear rules, high data structure, and low disease group complexity.

[0166] Step S270, summarizing the audit results of each audit subtask in the audit logic to obtain the audit result representing the rationality of DRG grouping.

[0167] Specifically, by summarizing the output of the medical insurance audit large model and the output of the rule engine, the audit results of each audit subtask corresponding to the target disease group can be obtained, and the final audit result representing the rationality of DRG grouping can be determined.

[0168] The method provided in the embodiment simultaneously provides two processing cores of the medical insurance audit large model and the rule engine, and can intelligently select the processing core of each audit subtask according to factors such as rule explicitness, to ensure the comprehensiveness and efficiency of the audit through the cooperative work of the two processing cores.

[0169] In some embodiments of the present application, the calculation process of the explicitness of the above-mentioned audit rule, the complexity of the target disease group, and the structure degree of the target case is described.

[0170] In one possible implementation, the explicitness of the review rule, the complexity of the target case group, and the degree of structuring of the target case can be predicted by calling the medical insurance review large model.

[0171] Specifically, the medical insurance review large model can be instructed to evaluate the explicitness of the review rule, the complexity of the target case group, and the degree of structuring of the target case by prompting instructions. In the prompt instructions, further requirement information can be added to instruct the large model to evaluate according to the specified requirements.

[0172] In some possible implementations, to improve the explainability and accuracy of the medical insurance review large model evaluation output of the explicitness of the review rule, the complexity of the target case group, and the degree of structuring of the target case, the medical insurance review large model can also be required to output the explanation of the current evaluation result in the prompt instructions, that is, to explain why the current evaluation result is given.

[0173] Next, an example of prompt is provided for the explicitness of the review rule, the complexity of the target case group, and the degree of structuring of the target case:

[0174] The prompt example for calling the medical insurance review large model to evaluate the explicitness of the review rule is as follows:

[0175] "Suppose you are an evaluation expert, and you can evaluate the explicitness of the review rule R m according to the requirements.

[0176]

Requirements

[0177] Please carefully read and analyze the provided medical record review rules.

[0178] 2. For a given rule, evaluate whether its expression is clear, specific, and easy to understand and execute.

[0179] 3. Consider whether there is ambiguous or ambiguous language in the rule, and whether the rule can be consistently applied to different medical record data.

[0180] 4. Evaluate whether the rule provides sufficient contextual information so that the reviewer can accurately apply them.

[0181] 5. According to the above factors, assign an explicitness score (Rm) to the given rule, with the score ranging from 0 (completely unclear) to 1 (completely clear).

[0182] 6. For a given rule, provide a specific score and explain why this score is given.

[0183]

Medical Record Review Rules

[0184] [Output] R m : Reason:.

[0185] Call the healthcare audit large model to evaluate the structured degree of the case R s The prompt example is as follows:

[0186] "Assuming you are an evaluation expert, you can evaluate and analyze the structured degree score R s of the medical record according to the requirements.

[0187] [Requirements]

[0188] 1. Please conduct in-depth analysis on the provided medical record data to determine the structured degree of the data.

[0189] 2. Evaluate whether the field content in the medical record follows consistent formats and standards, and whether it can be easily read and parsed by machines.

[0190] 3. Evaluate whether there are unstructured or semi-structured elements in the medical record, such as free text or non-standard format data.

[0191] 4. Evaluate whether the fields in the medical record are complete, missing, and whether there is enough information to support accurate interpretation and use of the data.

[0192] 5. According to the consistency, parsability and completeness of the field content of the medical record, assign a structured degree score (R s ) to the entire medical record, with the score ranging from 0 (completely unstructured) to 1 (completely structured).

[0193] 6. Provide a specific score and explain why this score is given.

[0194] [Medical record data] {}

[0195] [Output] Rs: Reason:.

[0196] Call the healthcare audit large model to evaluate the disease group complexity of the case R c The prompt example is as follows:

[0197] "Assuming you are an evaluation expert, you can evaluate and analyze the disease group complexity score R c of the disease group according to the requirements.

[0198] [Requirements]

[0199] 1. Please analyze the disease group information in a specific disease group to evaluate its complexity.

[0200] 2. Consider the diversity of the disease group information, including the complexity of diagnosis, the complexity of surgery, the difference in the number of diagnoses and the number of surgeries.

[0201] 3. Evaluate the complexity of elements within the disease group, as well as their interactions and dependencies.

[0202] 4. Consider the difficulty of disease group processing, including complex decisions during the diagnosis, treatment, and review process of the case.

[0203] 5. According to the diversity, complexity and processing difficulty of the disease group information, assign a complexity score (R c ) to the disease group, with a score range from 0 (very simple) to 1 (very complex).

[0204] 6. Provide a specific score and explain why this score is given.

[0205]

Disease group information

[0206]

Output result

[0207] In this embodiment, by calling the medical insurance review large model, the clarity of the review rules, the complexity of the target disease group, and the structured degree of the target case can be evaluated. With the help of the capabilities of the medical insurance review large model, more accurate evaluation results can be obtained, thereby facilitating the accurate determination of the task complexity of the review subtask, and further realizing the selection of intelligent review capabilities.

[0208] In some embodiments of the present application, the training process of the medical insurance review large model is further introduced.

[0209] In this embodiment, a high-efficiency training strategy for a multi-task large model is provided.

[0210] In order to enable the base large model to better understand and generate medical review related text, thereby adapting to subsequent review tasks, in this embodiment, training data corresponding to each review subtask of each review rule of a disease group can be constructed, that is, a plurality of pieces of training data corresponding to the review subtasks are obtained. The training data includes task instruction description corresponding to the review subtask, case sample and result label, and the result label represents the review result obtained by reviewing the case sample according to the review subtask.

[0211] After the training data is constructed, the base large model is trained using the training data corresponding to each review subtask to obtain a trained medical insurance review large model.

[0212] Through this training method, the medical insurance review large model can better understand and generate medical review related text, and improve the quality and efficiency of the review.

[0213] In a possible implementation, to reduce the resource consumption of data labeling work, in constructing the training data corresponding to each auditing subtask, a large model can be called in this embodiment to assist in generating the training data.

[0214] Specifically, a few-shot strategy can be adopted, a small amount of training data samples corresponding to each auditing subtask are manually labeled to form a sample library.

[0215] Then, the training data samples are iteratively extracted from the sample library and assembled into a second prompt instruction prompt, which is used to instruct the large model to refer to the training data samples to generate new case samples and corresponding result labels.

[0216] The second prompt instruction prompt is sent to the large model to obtain the case samples and result labels output by the large model, and combined with the task instruction description in the training data samples to form new training data, which is added to the training data set corresponding to the auditing subtask.

[0217] The large model here can be a general large model or a medical field large model.

[0218] In this embodiment, the large model is called to generate the training data, and only a small amount of training data samples are manually labeled for the large model to refer to, thereby reducing the labeling resource consumption.

[0219] In a possible implementation, to improve the explainability of the case samples and result labels generated by the large model, the second prompt instruction prompt is used to instruct the large model to output explanation content on the basis of instructing the large model to refer to the training data samples to generate new case samples and corresponding result labels, the explanation content being a generated explanation content of the new case sample for the task instruction description, and the auditing result being the result label.

[0220] An optional example of the second prompt instruction prompt is provided as follows:

[0221] "Instruction: Assume you are a medical expert, please generate a medical record text according to the following requirements:

[0222] Given task instruction description: {instruction}

[0223] Given case sample example: {content}

[0224] Given result label: {true_label}

[0225] Wherein, for the given task instruction, the result label of the case sample is {true_label}.

[0226] Now ask to generate a medical record sample similar to the example.

[0227] The language style of the generated medical record sample is similar to the medical record sample example, and the language is required to be formal and not colloquial.

[0228] The number of tokens in the generated text is required to be between 1500 and 2000.

[0229] In addition to the medical record sample, the explanation of the medical record sample that meets the task instruction is also required.

[0230] The explanation part only contains two steps, that is, first extract the relevant original medical record fragments according to the given task instruction, and give the analysis of the relevant original medical record fragments that meet the task instruction, and then give the conclusion that the result is: "{true_label}".

[0231] The overall output format is: text: \n explanation.

[0232] In some possible implementations, in order to improve the adaptability and accuracy of the large model, a multi-role playing training mode can be introduced for different audit sub-tasks in the process of calling the large model to generate training data for different audit sub-tasks in the above embodiments.

[0233] In the multi-role playing training mode, the large model needs to simulate the behavior and language style of a specific character according to the given role setting. This training method requires the large model not only to understand the text content, but also to be able to respond appropriately according to the role background and scene.

[0234] For example, according to different audit sub-tasks, the role is constructed, such as for the numerical comparison type audit task, the role set for the large model can be "medical numerical verification expert", such role construction makes the large model more focused on specific field audit tasks. For example, for drug audit tasks, the role set for the large model can be "pharmaceutical expert", etc., which will not be listed one by one here.

[0235] Referring to Figure 6 It illustrates a flowchart of a DRG entry rationality audit method.

[0236] The entire DRG entry rationality audit scheme can be divided into two stages: stage 1 and stage 2.

[0237] Stage 1 is a preprocessing process, that is, it can be completed before the DRG entry rationality audit of the case to be audited.

[0238] Stage 2 is the processing process of the DRG entry rationality audit of the case to be audited.

[0239] The stage 1 specifically includes:

[0240] The large model is called to concretize and summarize the diagnosis and treatment path data of each disease group to obtain the corresponding audit logic. In the implementation process, a prompt design method similar to a thinking chain can be used. First, the large model is used to concretize the diagnosis and treatment path data into identifiable key elements. Then, the large model is used to develop the audit logic of the disease group based on the identified key elements.

[0241] Based on the sample library, the large model generates training data corresponding to multiple different audit sub-tasks.

[0242] The sample library includes a small amount of training data samples for each audit sub-task labeled by humans. That is, for each audit rule in the audit logic of each disease group, a corresponding audit sub-task can be constructed. Then, a small amount of training data samples for each audit sub-task are labeled by humans.

[0243] When generating training data for audit sub-tasks using a large model, training data samples for audit sub-tasks extracted from the sample library can be added to the corresponding prompt instructions to instruct the large model to generate training data based on the training data samples. In this way, the large model can generate training data corresponding to multiple different audit sub-tasks.

[0244] On this basis, the medical insurance audit large model can be trained using multi-task training data.

[0245] The stage 2 specifically includes:

[0246] For the case to be audited and the target disease group to be prepared, the audit logic corresponding to the target disease group and the audit sub-task corresponding to each audit rule in the audit logic can be obtained.

[0247] For each audit sub-task, the task complexity of the audit sub-task can be calculated based on the explicitness of the audit rule, the complexity of the target disease group, and the degree of structuring of the case to be audited. According to the task complexity, the corresponding audit capability can be intelligently selected. Specifically, when the task complexity is high (exceeding a set task complexity threshold), the medical insurance audit large model can be selected to execute the current audit sub-task; when the task complexity is low (not exceeding the set task complexity threshold), the traditional rule engine can be selected to execute the current audit sub-task.

[0248] Finally, the audit results of the medical insurance audit large model and the audit results of the rule engine are coupled to form the final audit results according to the audit logic, and a detailed audit report is output, including the audit logic, the audit process, and the audit basis.

[0249] In summary, the DRG grouping rationality auditing method provided by the present application has the following advantages:

[0250] 1. Concrete identification of diagnosis and treatment path:

[0251] The present application improves the accuracy and operability of the auditing process by concretizing the abstract diagnosis and treatment path into identifiable elements such as diagnosis codes, treatment items, and drug use. Through these elements, the large model is further guided to develop a set of auditing logic to ensure the rationality of DRG grouping, thereby reducing the waste of medical resources and unreasonable expenditure of medical insurance funds.

[0252] 2. Efficient training of medical insurance auditing large model:

[0253] The multi-task training method proposed by the present application uses a small number of sample cases to generate training data, effectively reducing the cost and workload of data labeling. The training task covers multiple auditing sub-task dimensions, including but not limited to test value comparison, examination type, medical order type, drug dimension, and cost type, comprehensively improving the accuracy and efficiency of auditing.

[0254] 3. Explainable auditing decision mechanism:

[0255] The present application proposes an explainable auditing decision mechanism, enhancing the transparency and credibility of the auditing results. The automatic selection mechanism of auditing capability can automatically determine whether to use the traditional rule engine or the medical insurance auditing large model for auditing according to the characteristics of the disease group, achieving optimal auditing effect. This auditing mechanism improves the automation level of the auditing process, enabling automatic auditing, reducing manual intervention, and improving auditing speed and consistency.

[0256] 4. Improve the level of intelligence:

[0257] Through the application of the medical insurance auditing large model multi-round dialogue interactive auditing, the intelligent level of medical insurance DRG grouping rationality auditing is improved, making the auditing process more efficient and accurate. The introduction of the medical insurance auditing large model enables the system to understand and interpret key elements in the diagnosis and treatment path, improving the analysis capability of complex medical data.

[0258] 5. Save costs and resources:

[0259] Through automated and intelligent auditing processes, the dependence on professional personnel is reduced, saving human resources and auditing costs. Reducing medical disputes and medical insurance fund losses caused by auditing errors improves the efficiency of medical insurance fund use.

[0260] 6. Flexibility and scalability:

[0261] The method of the present application has good flexibility and scalability, and can adapt to the changing medical policies and medical insurance regulations. With the accumulation of medical data and the continuous optimization of the model, the accuracy and efficiency of the audit system will be further improved.

[0262] 7. Promote medical quality improvement:

[0263] Through accurate DRG group audit, it promotes medical service providers to improve service quality and efficiency, so as to improve the overall medical level.

[0264] The DRG group rationality audit device provided by the embodiment of the present application is described below. The DRG group rationality audit device described below can be referred to in conjunction with the DRG group rationality audit method described above.

[0265] Referring to Figure 7 , Figure 7 A structural schematic diagram of a DRG group rationality audit device disclosed by the embodiment of the present application is shown.

[0266] As Figure 7 shown, the device can include:

[0267] A to-be-audited data acquisition unit 11 is configured to acquire a target case to be audited and a target disease group of the target case to be audited;

[0268] An audit logic acquisition unit 12 is configured to acquire an audit logic corresponding to the target disease group configured by the target disease group, wherein the audit logic is obtained by configuring a large model to summarize and summarize the diagnosis and treatment path data corresponding to the target disease group, and the audit logic includes one or more audit rules;

[0269] A subtask acquisition unit 13 is configured to acquire an audit subtask corresponding to each audit rule in the audit logic corresponding to the target disease group;

[0270] A medical insurance audit large model calling unit 14 is configured to call a configured medical insurance audit large model to instruct the medical insurance audit large model to execute each audit subtask for the target case, and obtain an audit result corresponding to each audit subtask, wherein the medical insurance audit large model is obtained by training a base large model using training data of multiple audit subtasks;

[0271] A result aggregation unit 15 is configured to aggregate the audit results of each audit subtask in the audit logic to obtain an audit result representing the rationality of DRG group.

[0272] In one possible implementation, the device of the present application can further include:

[0273] The audit logic generation unit is configured to generate audit logic of the target disease group by summarizing the diagnosis and treatment path data corresponding to the target disease group by using the large model.

[0274] The first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic of the target disease group.

[0275] The first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic of the target disease group.

[0276] In one possible implementation, the first prompt instruction prompt adopts a thought chain structure, and is used to instruct the large model to sequentially perform the following steps:

[0277] First, analyze the diagnosis and treatment path data corresponding to the target disease group to identify key elements therein;

[0278] Further, according to the key elements, the audit logic of the target disease group is formulated.

[0279] In one possible implementation, the subtask acquisition unit acquires the audit subtask corresponding to each audit rule in the audit logic corresponding to the target disease group, including:

[0280] Query the correspondence between the configured audit rules and the audit subtasks, and acquire the audit subtask corresponding to each audit rule in the audit logic corresponding to the target disease group.

[0281] In one possible implementation, the device of the present application can further include:

[0282] The task complexity calculation unit is configured to, before the medical insurance audit large model calling unit calls the configured medical insurance audit large model to instruct the medical insurance audit large model to perform each audit subtask for the target case, determine the task complexity degree based on at least the explicitness of each audit rule in the audit logic; and if the task complexity degree exceeds the set complexity threshold, call the medical insurance audit large model through the medical insurance audit large model calling unit to instruct the medical insurance audit large model to perform the audit subtask corresponding to the audit rule for the target case.

[0283] Further optionally, the device of the present application can further include:

[0284] The traditional rule engine calling unit is configured to, if the task complexity degree does not exceed the set complexity threshold, call the configured rule engine to perform the audit subtask corresponding to the audit rule for the target case to obtain the audit result corresponding to the audit subtask.

[0285] In a possible implementation, the task complexity calculation unit determines the task complexity based at least on explicitness of the review rule, and the process includes:

[0286] determining the task complexity based on the explicitness of the review rule, complexity of the target disease group, and structured degree of the target case;

[0287] wherein the more implicit the review rule, the higher the complexity of the target disease group, and the lower the structured degree of the target case, the higher the corresponding task complexity.

[0288] In a possible implementation, the explicitness of the review rule, the complexity of the target disease group, and the structured degree of the target case are all obtained by calling the medical insurance review large model.

[0289] In a possible implementation, the device of the application can further include:

[0290] a medical insurance review large model training unit configured to train a medical insurance review large model, and the corresponding training process includes:

[0291] constructing training data corresponding to each review subtask in the review logic of each disease group, the training data including task instruction description, case sample, and result label corresponding to the review subtask, and the result label representing the review result obtained by reviewing the case sample according to the review subtask;

[0292] training the configured base large model using the training data corresponding to each review subtask to obtain the trained medical insurance review large model.

[0293] In a possible implementation, the medical insurance review large model training unit constructs the training data corresponding to each review subtask, and the process includes:

[0294] obtaining artificially annotated training data samples corresponding to each review subtask to form a sample library;

[0295] iteratively extracting the training data samples from the sample library and assembling them into a second prompt instruction prompt, the second prompt instruction prompt being used to instruct the large model to refer to the training data samples to generate new case samples and corresponding result labels;

[0296] sending the second prompt instruction prompt into the large model to obtain the case samples and result labels output by the large model, and combining the task instruction description in the training data samples into new training data;

[0297] Add the new training data to the training dataset corresponding to the audit subtask.

[0298] In one possible implementation, the second prompt instruction, in addition to instructing the large model to generate new case samples and corresponding result labels by referring to the training data samples, is also used to instruct the large model to output explanatory content. The explanatory content is a description of the task instruction, and the review result of the generated new case sample is the inference explanation content of the result label.

[0299] This application also provides an electronic device in its embodiments. (See reference...) Figure 8 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, tablets, desktop computers, medical diagnostic equipment, interactive robots, etc. Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0300] like Figure 8 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603, to implement the DRG input rationality review method of the foregoing embodiments of this application. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0301] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, memory cards, hard drives, etc.; and communication devices 609. Communication device 609 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0302] The embodiment of the present application further provides a computer program product comprising computer readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the DRG group combination rationality auditing methods provided by the embodiments of the present application.

[0303] The embodiment of the present application further provides a computer readable storage medium, which carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement any of the DRG group combination rationality auditing methods provided by the embodiments of the present application.

[0304] In addition, it should be noted that the apparatus embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments. In addition, the connection relationship between the modules in the apparatus embodiments provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines.

[0305] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware, and of course can also be realized by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. Generally, functions completed by computer programs can be easily realized by corresponding hardware, and specific hardware structures for realizing the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better embodiment. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., including a plurality of instructions to make a computer device (which can be a personal computer, training device, or network device, etc.) execute the methods described in various embodiments of the present application.

[0306] In the above embodiments, all or part can be realized by software, hardware, firmware or any combination thereof. When realized by software, it can be realized in the form of a computer program product in whole or in part.

[0307] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the flow or function described in the embodiments of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, training device or data center to another website, computer, training device or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0308] The various embodiments in the specification are described in a progressive manner, each embodiment focuses on the difference from other embodiments, the various embodiments can be combined as needed, and the same or similar parts refer to each other.

Claims

1. A DRG grouping rationality auditing method, characterized in that, The method comprises the following steps: obtaining a target case to be audited and a target disease group to which the target case is to be assigned; obtaining the audit logic corresponding to the target disease group configured by the large model, wherein the audit logic is obtained by summarizing the diagnosis and treatment path data corresponding to the target disease group by the large model configured, and the audit logic comprises one or more audit rules; for the audit logic corresponding to the target disease group, obtaining an audit subtask corresponding to each audit rule in the audit logic, and calling the medical insurance audit large model configured to instruct the medical insurance audit large model to execute each audit subtask on the target case to obtain an audit result corresponding to each audit subtask, wherein the medical insurance audit large model is obtained by training a base large model using training data of multiple audit subtasks; summarizing the audit results of each audit subtask in the audit logic to obtain an audit result representing the rationality of DRG assignment; wherein the training process of the medical insurance audit large model comprises: constructing training data corresponding to each audit subtask in the audit logic of each disease group, wherein the training data comprises task instruction description, case sample and result label corresponding to the audit subtask, and the result label represents the audit result obtained by auditing the case sample according to the audit subtask; training the base large model configured using the training data corresponding to each audit subtask to obtain the trained medical insurance audit large model; the process of constructing the training data corresponding to each audit subtask comprises: obtaining artificially annotated training data samples corresponding to each audit subtask to form a sample library; iteratively extracting the training data samples from the sample library and assembling them into a second prompt instruction prompt, wherein the second prompt instruction prompt is used to instruct the large model to refer to the training data samples to generate new case samples and corresponding result labels; sending the second prompt instruction prompt into the large model to obtain the case samples and result labels output by the large model, and combining them with the task instruction description in the training data samples to form new training data; adding the new training data to the training data set corresponding to the audit subtask.

2. The method of claim 1, wherein, the process of obtaining the audit logic by summarizing the diagnosis and treatment path data corresponding to the target disease group by the large model configured comprises: sending a first prompt instruction prompt into the large model configured to obtain the audit logic of the target disease group output by the large model; wherein the first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment path data corresponding to the target disease group to obtain the audit logic of the target disease group.

3. The method of claim 2, wherein, The first prompt instruction prompt adopts a thought chain structure and is used to instruct the large model to perform the following steps in sequence: first, analyze the diagnosis and treatment path data corresponding to the target disease group to identify the key elements therein; further, according to the key elements, formulate the audit logic of the target disease group.

4. The method of claim 1, wherein, for the audit logic corresponding to the target disease group, the process of obtaining an audit subtask corresponding to each audit rule in the audit logic comprises: The query configuration audits the rules and the corresponding relationship between the audit sub-tasks, obtains the audit logic corresponding to the target disease group, and each audit rule corresponds to an audit sub-task in the audit logic.

5. The method of claim 1, wherein, Before calling the configured medical insurance audit large model to instruct the medical insurance audit large model to perform each audit sub-task on the target case, the method further comprises: For each audit rule in the audit logic: Determine the task complexity level based at least on the explicitness of the audit rule; If the task complexity level indicates that it exceeds the set complexity threshold, call the medical insurance audit large model to instruct the medical insurance audit large model to perform the audit sub-task corresponding to the audit rule on the target case.

6. The method of claim 5, wherein, Further comprising: If the task complexity level indicates that it does not exceed the set complexity threshold, call the configured rule engine to perform the audit sub-task corresponding to the audit rule on the target case to obtain the audit result corresponding to the audit sub-task.

7. The method of claim 5, wherein, The process of determining the task complexity level based at least on the explicitness of the audit rule comprises: Determine the task complexity level based on the explicitness of the audit rule, the complexity of the target disease group, and the structured degree of the target case; Wherein, the less explicit the audit rule is, the higher the complexity of the target disease group is, and the lower the structured degree of the target case is, the higher the corresponding task complexity level is.

8. The method of claim 7, wherein, The explicitness of the audit rule, the complexity of the target disease group, and the structured degree of the target case are all obtained by calling the medical insurance audit large model.

9. The method of claim 1, wherein, The second prompt instruction prompt is used to instruct the large model to generate new case samples and corresponding result labels based on the training data samples, and is also used to instruct the large model to output explanation content, which is the explanation content of the generated audit result of the new case sample for the task instruction description. The result label is the result label.

10. A DRG grouping rationality auditing device, characterized by, It comprises: A to-be-audited data acquisition unit is configured to acquire a target case to be audited and a target disease group to which the target case is intended to belong; An audit logic acquisition unit is configured to acquire configured audit logic corresponding to the target disease group, the audit logic being audit logic obtained by configuring a large model to summarize diagnosis and treatment path data corresponding to the target disease group, the audit logic comprising one or more audit rules; A sub-task acquisition unit is configured to acquire, for the audit logic corresponding to the target disease group, an audit sub-task corresponding to each audit rule in the audit logic; The medical insurance audit large model calling unit is configured to call a configured medical insurance audit large model to instruct the medical insurance audit large model to perform each audit subtask on the target case to obtain an audit result corresponding to each audit subtask. The medical insurance audit large model is obtained by training a base large model using training data of multiple audit subtasks. The training process of the medical insurance audit large model includes: constructing training data corresponding to each audit subtask in the audit logic of each disease group, wherein the training data includes task instruction description corresponding to the audit subtask, a case sample, and a result label, and the result label represents an audit result obtained by auditing the case sample according to the audit subtask; The base large model is trained using the training data corresponding to each audit subtask to obtain a trained medical insurance audit large model; The process of constructing the training data corresponding to each audit subtask includes: Obtaining artificially labeled training data samples corresponding to each audit subtask to form a sample library; Iteratively extracting the training data samples from the sample library and assembling them into a second prompt instruction prompt, wherein the second prompt instruction prompt is used to instruct the large model to refer to the training data samples to generate new case samples and corresponding result labels; The second prompt instruction prompt is sent to the large model to obtain case samples and result labels output by the large model, and the case samples and the result labels are combined with the task instruction description in the training data samples to form new training data; The new training data is added to the training data set corresponding to the audit subtask; A result aggregation unit is configured to aggregate the audit results of each audit subtask in the audit logic to obtain an audit result representing the DRG grouping rationality.

11. An electronic device, comprising: It includes: A memory and a processor; The memory is configured to store a program; The processor is configured to execute the program to implement each step of the DRG grouping rationality audit method according to any one of claims 1-9.

12. A readable storage medium, having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement each step of the DRG grouping rationality audit method according to any one of claims 1-9.

13. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement each step of the DRG grouping rationality audit method according to any one of claims 1-9.

Images