Reasonable checking method and device for DRG entry combination, related equipment and program product

Abstract

The invention discloses a DRG entry rational auditing method and device, related equipment and a program product. The method comprises the steps of obtaining a to-be-audited target case and a planned target disease group thereof; an auditing logic corresponding to the target disease group is obtained, the auditing logic is obtained by summarizing diagnosis and treatment path data corresponding to the target disease group through a configured large model, and the auditing logic comprises a plurality of auditing rules; an auditing sub-task corresponding to each auditing rule corresponding to the target disease group is obtained, a medical insurance auditing large model is called to execute each auditing sub-task for the target case, an auditing result corresponding to each auditing sub-task is obtained, and the medical insurance auditing large model is obtained by training a base large model through training data of multiple auditing sub-tasks; and summarizing the auditing results of the auditing sub-tasks to obtain a final auditing result. By calling the large model capability, the formulation of the auditing logic and the rational auditing of the combination are completed, the requirement for expert resources is reduced, and the influence of the personal experience and judgment of experts is avoided.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method, device, related equipment and program product for rational review of DRG admission. Background Art

[0002] With the continuous improvement of the medical security system, the medical insurance payment method has been comprehensively upgraded from payment by project to payment by disease type. Among them, the DRG (Diagnosis Related Groups) payment method has become the mainstream payment method under the payment method by disease type. The DRG payment method is a system that groups hospitalized patients according to factors such as diagnosis, treatment process, and severity of the disease, and pays medical insurance according to the payment standards predetermined by the group. This payment method helps to alleviate excessive diagnosis and treatment under payment by project, and promotes medical institutions to actively control costs and standardize diagnosis and treatment behaviors. Although the DRG payment method has the above advantages, it also has audit risks such as unreasonable grouping, and requires comprehensive and rational supervision of grouping.

[0003] At present, the method for rational review of DRG enrollment mainly relies on the review rules formulated by the expert team. These experts usually include medical insurance management experts, information technology experts, statistical analysts, medical record management experts and clinical medicine experts. They combine their professional knowledge and rich clinical experience to jointly formulate a set of rules to ensure the quality and efficiency of the review. This review method still has some defects:

[0004] First, expert resources are limited and costly, making it difficult to cover all sub-sectors and regions. Second, the audit standards may be influenced by the personal experience and judgment of experts and are subjective. Summary of the invention

[0005] In view of the above problems, this application is proposed to provide a DRG admission rational review method, device, related equipment and program product to solve some or all of the defects of the method of relying on the review rules formulated by experts. The specific plan is as follows:

[0006] First, a rational review method for DRG admission is provided, including:

[0007] Obtain the target cases to be reviewed and the target disease groups for their proposed DRG enrollment;

[0008] Acquire the audit logic corresponding to the configured target disease group, wherein the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured big model, and the audit logic includes more than one audit rule;

[0009] For the audit logic corresponding to the target disease group, the audit subtask corresponding to each of the audit rules is obtained, and the configured medical insurance audit big model is called to instruct the medical insurance audit big model to execute each of the audit subtasks for the target case, and the audit result corresponding to each of the audit subtasks is obtained. The medical insurance audit big model is obtained by training the base big model with the training data of multiple audit subtasks;

[0010] Summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

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

[0012] Sending the first prompt instruction prompt to the configured big model to obtain the audit logic of the target disease group output by the big model;

[0013] Among them, the first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment pathway data corresponding to the target disease group to obtain the review logic of the target disease group.

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

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

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

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

[0019] In a possible design, in another implementation of the first aspect of the embodiment of the present application, before calling the configured medical insurance review big model to instruct the medical insurance review big model to perform each of the review subtasks for the target case, it also includes:

[0020] For each audit rule in the audit logic:

[0021] Determine the complexity of the task based at least on the clarity of the audit rules described;

[0022] If the complexity of the task indicates that it exceeds the set complexity threshold, the medical insurance review model is called to instruct the medical insurance review model to execute the review subtask corresponding to the review rule for the target case.

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

[0024] If the task complexity indicates that it does not exceed the set complexity threshold, then for the target case, the configured rule engine is called to execute the audit subtask corresponding to the audit rule to obtain the audit result corresponding to the audit subtask.

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

[0026] Determine the complexity of the task based on the clarity of the review rules, the complexity of the target disease group, and the degree of structure of the target case;

[0027] Among them, the less clear the review rules are, the higher the complexity of the target disease group is, and the lower the degree of structuring of the target cases is, the higher the corresponding task complexity is.

[0028] In one possible design, in another implementation of the first aspect of the embodiment of the present application, the clarity of the review rules, the complexity of the target disease group and the degree of structuring of the target cases are all predicted by calling the medical insurance review big model.

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

[0030] For the audit subtasks corresponding to each audit rule in the audit logic of each disease group, construct training data corresponding to each audit subtask, wherein the training data includes a task instruction description, a case sample and a 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;

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

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

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

[0034] Iteratively extract the training data samples from the sample library and assemble 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;

[0035] Send the second prompt instruction prompt to the big model to obtain the case sample and result label output by the big model, and combine them with the task instruction description in the training data sample to form new training data;

[0036] The new training data is added to the training data set corresponding to the audit subtask.

[0037] In one possible design, in another implementation 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 sample to generate a new case sample and a corresponding result label, is also used to instruct the large model to output an explanation content, wherein the explanation content is the reasoning explanation content of the result label for the review result of the new case sample generated for the task instruction description.

[0038] In a second aspect, a DRG admission rationality review device is provided, including:

[0039] The data acquisition unit to be reviewed is used to obtain the target cases to be reviewed and the target disease groups to be included in the proposed DRG;

[0040] An audit logic acquisition unit, used to acquire the audit logic corresponding to the configured target disease group, wherein the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured big model, and the audit logic includes more than one audit rule;

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

[0042] A medical insurance review model calling unit is used to call the configured medical insurance review model to instruct the medical insurance review model to execute each of the review subtasks for the target case and obtain the review result corresponding to each of the review subtasks. The medical insurance review model is obtained by training the base model with the training data of multiple review subtasks.

[0043] A result summary unit is used to summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

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

[0045] The memory is used to store programs;

[0046] The processor is used to execute the program to implement the various steps of the DRG admission rational review method described in any one of the first aspects of the present application.

[0047] In a fourth aspect, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the various steps of the DRG admission rational review method described in any one of the first aspects of the present application are implemented.

[0048] In a fifth aspect, a computer program product is provided, including a computer program, which, when executed by a processor, implements the various steps of the DRG admission group rational review method described in any one of the aforementioned first aspects of the present application.

[0049] Through the above technical solution, this application uses the ability of the big model to summarize the abstract diagnosis and treatment pathway data and obtain the audit logic of the target disease group, which includes more than one audit rule. That is, this application does not rely on experts to formulate audit rules based on clinical experience, but starts from the diagnosis and treatment pathway data as a clinical treatment guide, and uses the natural language understanding and concrete inductive ability of the big model to summarize the audit logic of the target disease group from the diagnosis and treatment pathway data, reducing the demand for expert resources, and will not be affected by the personal experience and judgment of experts, which is more objective.

[0050] Furthermore, when the rationality of DRG grouping of the target case is reviewed, the review subtask corresponding to each review rule in the review logic corresponding to the target disease group is obtained, that is, for each review rule in the review logic, a corresponding review subtask is constructed, which is used to verify whether the case data meets the requirements of the corresponding review rule. Through the construction of the above review subtasks, the complex review process is divided into several review subtasks. On this basis, the configured medical insurance review big model can be called to execute each review subtask in turn for the target case to obtain the review result of each review subtask. Among them, the medical insurance review big model can be pre-trained with the training data of multiple review subtasks to train the base big model. Through multi-task training, the medical insurance review big model can better understand and generate text related to medical review. On this basis, the medical insurance review big model can be used to better execute each review subtask and obtain more accurate review results. Finally, by summarizing the review results of each review subtask, the rationality review result of the DRG grouping of the target case can be obtained. Since the complex audit logic of the target disease group is broken down into several audit subtasks, the medical insurance audit model is called to execute each audit subtask separately, making the final audit results more interpretable, which matches the high requirements of the medical field for transparency of the decision-making process. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0052] Figure 1 A schematic diagram of an implementation system architecture of a rational review method for DRG admission provided in an embodiment of the present application;

[0053] Figure 2 A schematic diagram of a terminal structure provided in an embodiment of the present application;

[0054] Figure 3 A schematic diagram of a server structure provided in an embodiment of the present application;

[0055] Figure 4 A schematic diagram of a rational review method for DRG admission provided in an embodiment of the present application;

[0056] Figure 5 A schematic diagram of another DRG group admission rational review method provided in an embodiment of the present application;

[0057] Figure 6 A flowchart of another method for rational review of DRG admission provided in an embodiment of the present application;

[0058] Figure 7 A schematic diagram of the structure of a DRG admission rational review device provided in an embodiment of the present application;

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

[0060] Before introducing this application solution, the relevant concepts involved in this article are first explained:

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

[0062] Treatment pathway: A treatment pathway is a standardized treatment plan developed for a specific disease or surgery, which aims to standardize medical service behaviors, reduce resource waste, and ensure that patients receive appropriate medical care services.

[0063] prompt: Instructions. When interacting with AI (such as an artificial intelligence model), you need to send instructions to AI. It can be a text description, such as "Please recommend me a pop song" when you interact with AI, or it can be a parameter description in a certain format. For example, if you want AI to draw in a certain format, you need to describe the relevant drawing parameters.

[0064] Large models: In the field of artificial intelligence, large models usually refer to large-scale pre-trained models. Such models are called "large" because they can be pre-trained on a large amount of data and can be transferred to a variety of downstream tasks. Their full English name is Large Pre-Trained Models or Large-Scale Pre-Training Models. Large models are characterized by their large scale and contain billions or even more parameters, which help them learn complex patterns in the data. The emerging capabilities of large models include but are not limited to: contextual learning, instruction following, code generation, step-by-step reasoning capabilities, etc. Large models can include large language models (LLMs) and multimodal large models. Large language models are mainly used to process text modal data. Multimodal large models further integrate multimodal capabilities on the basis of large language models and can process information in multiple modalities, such as images, text, audio, etc.

[0065] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

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

[0067] This application provides a DRG admission rational review method that can be applied to Figure 1 The system architecture shown in FIG. 1 may include a terminal 100 and a server 200. The server 200 may include one or more servers ( Figure 1 A server is included as an example for explanation).

[0068] The terminal 100 or the server 200 can be used alone to execute the DRG admission rational audit method provided in the embodiment of the present application. In addition, the terminal 100 and the server 200 can also be used in conjunction to execute the DRG admission rational audit method provided in the embodiment of the present application.

[0069] Next describe Figure 1 The product form of the mid-terminal 100;

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

[0071] Figure 2 An optional hardware structure diagram of the terminal 100 is shown.

[0072] refer to Figure 2As shown, the terminal 100 may 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, an earphone jack 163 (optional), a processor 170, an external interface 180, a power supply 190 and other components. Those skilled in the art will appreciate that Figure 2 These are merely examples of terminals or multi-function devices and do not constitute limitations on the terminals or multi-function devices, which may include more or fewer components than those shown in the figures, or combinations of certain components, or different components.

[0073] The input unit 130 can be used to receive input digital or character information, and generate key signal input related to the user settings and function control of the portable multi-function device. Specifically, the input unit 130 may 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 the user. In addition to the touch screen 131, the input unit 130 may also include other input devices. Specifically, the other input devices 132 may include, but are not limited to, one or more of a physical keyboard, a function key (such as a volume control key, a switch key, etc.), a trackball, a mouse, a joystick, etc.

[0074] Among them, other input devices 132 can receive input data and so on.

[0075] The display unit 140 can be used to display information input by the user or provided to the user, various menus of the terminal 100, interactive interfaces, file display and / or playback of any multimedia file. In the embodiment of the present application, the display unit 140 can be used to display various interactive interfaces, processing results, etc. in the DRG entry rational review method.

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

[0077] Processor 170 is the control center of terminal 100. It uses various interfaces and lines to connect various parts of the entire terminal 100, and executes various functions of terminal 100 and processes data by running or executing instructions stored in memory 120 and calling data stored in memory 120, thereby controlling the terminal device as a whole.

[0078] Among them, the memory 120 can be used to store software codes related to the DRG group entry rational review method, the processor 170 can execute the steps of the DRG group entry rational review method, and can also schedule other units (such as the above-mentioned input unit 130 and display unit 140) to implement corresponding functions.

[0079] The RF unit 110 (optional) can be used for receiving and sending information or receiving and sending signals during a call. For example, after receiving the downlink information of the base station, it is sent to the processor 170 for processing; in addition, the designed uplink data is sent to the base station. Among them, in an embodiment of the present application, the RF unit 110 can send data to the server 200 and receive the processing results sent by the server 200. Exemplarily, the RF unit 110 sends the received target cases to be reviewed and the target disease groups to be included in the DRG to the server 200, and the server 200 rationally reviews and judges the DRG inclusion, obtains the review results and feeds them back to the terminal 100.

[0080] It should be understood that the radio frequency unit 110 is optional and can be replaced by other communication interfaces, such as a network port.

[0081] Although not shown, the terminal 100 may also include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which are not described in detail here. Some or all of the methods described below may be applied in the following embodiments. Figure 2 In the terminal 100 shown.

[0082] Next describe Figure 1 The product form of the server 200;

[0083] Figure 3 A structural diagram of a server 200 is provided, such as Figure 3 As shown, 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 with each other via the bus 201.

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

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

[0086] The memory 204 may include a volatile memory, such as a random access memory (RAM). The memory 204 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0087] Among them, the memory 204 can be used to store software codes related to the DRG group admission rational review method, the processor 202 can execute the steps of the DRG group admission rational review method, and can also schedule other units to implement corresponding functions.

[0088] It should be understood that the above-mentioned terminal 100 and server 200 can be centralized or distributed devices, and the processors in the above-mentioned terminal 100 and server 200 (such as processor 170 and processor 202) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processor (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

[0089] Considering that the current method for rational review of DRG admissions mainly relies on the review rules formulated by the expert team. Experts combine their own professional knowledge and rich clinical experience to jointly formulate a set of rules to ensure the quality and efficiency of the review. First, expert resources are limited and the cost of funds is high, making it difficult to cover all sub-sectors and regions. Second, the review standards may be affected by the personal experience and judgment of experts and are subjective.

[0090] To this end, a possible solution is to use traditional deep learning models to perform DRG enrollment rationality review. Deep learning models are trained to identify and predict the rationality of DRG enrollment. The specific steps include: data preprocessing, feature engineering, model training, model validation, and finally audit decision-making. The trained model is applied to the actual DRG enrollment review. 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 expensive. The capabilities of deep learning models are limited, and the accuracy of their output results is easily limited when faced with some complex case data. In addition, deep learning models are often considered "black box" models, and their decision-making process lacks transparency and explainability, which conflicts with the high requirements for transparency of the decision-making process in the medical field. These limitations have hindered the widespread application of deep learning models in the rational review of DRG admission.

[0092] In summary, this application provides a solution to achieve rational review of DRG admission.

[0093] The present application embodiment provides a DRG entry rational review method, which is illustrated by applying the method to a computer device. The computer device may be Figure 1 The terminal 100 or the system consisting of the terminal 100 and the server 200. Figure 4 The rational review method for DRG entry group specifically includes the following steps:

[0094] Step S100, obtaining the target cases to be reviewed and the target disease groups for their planned DRG inclusion.

[0095] Specifically, the present application can realize the rational review of DRG grouping. First, it is necessary to obtain the target case information of the patient to be reviewed and the target disease group for DRG grouping. The rational review of DRG grouping in the present application is to examine the rationality of classifying the target case into the target disease group.

[0096] Step S110, obtaining the audit logic corresponding to the configured target disease group, the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured large model, and the audit logic includes more than one audit rule.

[0097] Specifically, in order to solve the problem that the traditional solution relies on experts to formulate the review rules for disease groups, this application adopts an review logic formulation solution based on a large model.

[0098] Before executing the DRG admission rational review task, the present application can make use of the powerful natural language understanding and concrete induction capabilities of the big model to concretely summarize the diagnosis and treatment pathway data corresponding to each disease group, thereby obtaining the review logic for each disease group generated by the big model based on the diagnosis and treatment pathway data, which can include more than one review rule.

[0099] Among them, the diagnosis and treatment pathway data corresponding to each disease group can refer to the standardized diagnosis and treatment plan documents developed in the medical field, which stipulate the standardized diagnosis and treatment process for each type of disease in the disease group. This application uses a large model to summarize the diagnosis and treatment pathway data corresponding to the disease group, and can formulate the review logic of the disease group based on the standardized diagnosis and treatment process. Compared with the way experts rely on clinical experience and personal judgment, the review logic formulated by this application is more standardized and objective.

[0100] In this embodiment, the process of formulating the audit logic of the disease group with the help of a big model, the big model used can be a general big model of various types or a vertical big model in the medical field.

[0101] The pre-built audit logic corresponding to each disease group can be saved accordingly. When executing the DRG group entry rational audit task, the corresponding relationship can be queried for the target disease group to be audited to obtain its corresponding audit logic.

[0102] The following is an example of the audit logic for the HT1 disease group, which contains 5 audit rules:

[0103] ① [Diagnosis] includes a diagnosis of severe complications, such as sepsis, acute respiratory distress syndrome, etc. Any one of these diagnoses is considered sufficient;

[0104] ② The [Hospitalization Orders and Charges List] contains orders and charges for treatment of serious complications, such as CRRT, plasma exchange, etc. Any one of them is considered satisfactory;

[0105] ③ The medical records contain descriptions of abnormal mental status of the patient, such as coma, shock, etc.; and descriptions of special signs of severe pancreatitis, such as Cullen's sign;

[0106] ④ [Medical records and test results] indicate that the Marshall score is ≥ 2 points, indicating organ dysfunction. These indicators include systolic blood pressure, creatinine, and oxygenation index;

[0107] ⑤

Examination results

[0108] Step S120: For the audit logic corresponding to the target disease group, obtain the audit subtask corresponding to each audit rule therein.

[0109] Specifically, in this embodiment, the rational audit task of DRG admission is performed by calling the medical insurance audit model. In order to improve the accuracy of the audit results, this application divides the complex audit process into several audit subtasks, and then calls the medical insurance audit model to execute each audit subtask separately. The rational audit process of DRG admission is mainly completed based on the audit logic. Therefore, a corresponding audit subtask can be constructed for each audit rule in the audit logic. The audit subtask is used to verify whether the case data meets the requirements of the corresponding audit rules.

[0110] The process of obtaining the audit subtasks corresponding to the audit rules in this step can be to temporarily generate the audit subtasks corresponding to each audit rule. It can also be before executing the DRG admission group rational audit task, that is, before constructing the corresponding audit subtasks for each audit rule. Then, in the process of executing the DRG admission group rational audit task, the correspondence between the configured audit rules and the audit subtasks can be queried to obtain the audit subtasks corresponding to each audit rule in the audit logic corresponding to the target disease group.

[0111] According to different audit rules, the audit subtasks constructed can be of various types, including but not limited to:

[0112] Test value comparison tasks: The large model compares the values ​​of different test results, such as blood sugar, blood pressure and other indicators, to determine whether they meet the normal range or improvement / deterioration trend specified in the clinical pathway. This type of task helps to evaluate the patient's health status and treatment effect.

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

[0114] Doctor’s orders: The big model analyzes doctors’ orders, including medication prescriptions, treatment plans, and follow-up instructions, to ensure that they are consistent with the patient’s condition and treatment path. This helps improve treatment personalization and compliance.

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

[0116] Cost tasks: The large model reviews the rationality of medical expenses, including medical expenses, drug expenses, and examination expenses, 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 subtasks corresponding to each audit rule in the audit logic are obtained, that is, the complex audit process of the target disease group is broken down into several audit subtasks. By auditing the audit subtasks, the audit process of the target disease group is completed.

[0118] Step S130: call the configured medical insurance review big model to instruct the medical insurance review big model to execute each review subtask for the target case and obtain the review result corresponding to each review subtask.

[0119] Among them, the medical insurance review big model is obtained by training the base big model with the training data of multiple review subtasks. The base big model can be a general big model or a medical field big model. By training the base big model with the training data of multiple review subtasks, the trained medical insurance review big model can better understand and generate text related to medical review, that is, it can promote the medical insurance review big model to better complete the review subtasks and obtain more accurate review results.

[0120] In a possible implementation, in this embodiment, when calling the medical insurance audit model to perform the audit subtask, a multi-round interactive audit mechanism with the user can be adopted.

[0121] During the multi-round dialogue process, the medical insurance audit model can output the audit results of each audit subtask and its corresponding audit rules, so that users can clearly understand the audit basis and make the final audit results more explainable. In addition, during the multi-round dialogue process, the system is also allowed to obtain more information by further asking users when encountering uncertain or complex situations, so as to support the medical insurance audit model to make more accurate audit decisions.

[0122] Step S140, summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

[0123] After summarizing the audit results of each audit subtask in the audit logic of the target disease group, the final audit result of the rationality of DRG entry can be determined, that is, whether it is reasonable to divide the target case into the target disease group.

[0124] In one possible implementation, if the audit results of each audit subtask in the audit logic indicate that the requirements of the corresponding audit rules are met, it can be determined that the DRG inclusion is reasonable, otherwise, it is determined that the DRG inclusion is unreasonable.

[0125] The DRG admission rational review method provided in this embodiment is based on the abstract diagnosis and treatment pathway data, and summarizes it by calling the capabilities of the big model to obtain the review logic of the target disease group. Instead of relying on experts to formulate review rules based on clinical experience, it starts from the diagnosis and treatment pathway data as a clinical treatment guide, and uses the natural language understanding and concrete induction capabilities of the big model to summarize the review logic of the target disease group from the diagnosis and treatment pathway data, thereby reducing the demand for expert resources, and will not be affected by the personal experience and judgment of experts, making it more objective.

[0126] Furthermore, the method of this embodiment divides the complex audit process into several audit subtasks. On this basis, the configured medical insurance audit model can be called to execute each audit subtask in turn for the target case to obtain the audit result of each audit subtask. Finally, by summarizing the audit results of each audit subtask, the rationality audit result of the DRG entry of the target case can be obtained. Since the complex audit logic of the target disease group is broken down into several audit subtasks, the medical insurance audit model is called to execute each audit subtask separately, so that the final audit result is more interpretable, which matches the high requirements for transparency of the decision-making process in the medical field.

[0127] The method of this embodiment can realize an automated audit process. On the other hand, when calling the medical insurance audit model for audit, a multi-round dialogue interactive mechanism can be used to simulate natural language dialogue to make the audit process more transparent and easy to understand.

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

[0129] In order to call the big model to perform the audit logic generation task, a prompt instruction 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 big model to summarize the diagnosis and treatment pathway 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 big model to obtain the review logic of the target disease group output by the big model.

[0131] In some possible implementations, in order to better stimulate the large model to perform the audit logic generation task, this embodiment provides a thought chain prompt design scheme, that is, the first prompt instruction prompt adopts a thought chain structure design, and divides the audit logic generation task into several subtasks to be executed in sequence, which may specifically include:

[0132] First, analyze the diagnosis and treatment pathway data corresponding to the target disease group and identify the key elements.

[0133] Among them, key elements include but are not limited to: diagnosis codes, treatment items, drug use, surgical operations, examination results, etc.

[0134] Furthermore, based on the identified key factors, the audit logic for the target disease group is developed.

[0135] An optional example of the first prompt instruction prompt is provided as follows:

[0136] “Assuming you are a medical review expert, please summarize the review logic based on the given diagnosis and treatment pathways and follow the steps below:

[0137] First, identify the key elements of the treatment pathway:

[0138] 1. Please analyze the provided disease group-related diagnosis and treatment pathway data and identify the key elements in the diagnosis and treatment pathway.

[0139] 2. Key elements include but are not limited to diagnosis codes, treatment items, drug use, surgical procedures, examination results, etc.

[0140] 3. For each element, extract relevant data fields and identify their role and importance in the diagnosis and treatment pathway.

[0141] Then, formulate the disease group review logic:

[0142] 1. Develop a disease group review logic based on the key elements of the identified diagnosis and treatment pathway.

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

[0144] 3. Please consider the characteristics of the patient group and develop corresponding review processes and standards to ensure that the review logic not only meets medical standards but can also efficiently assist general doctors in conducting medical record reviews.

[0145] Finally, the audit logic of this disease group is output. ".

[0146] In this embodiment, by adopting the prompt design scheme of thinking chain, the big model can be guided to first identify the key element information in the diagnosis and treatment path, and then formulate the audit logic based on the key element information. Breaking down the complete audit logic formulation task into several logically related subtasks can better guide the big model to gradually complete the formulation of the audit logic, and improve the accuracy of the output results of the big model.

[0147] This application introduces the ability of large models to visualize and summarize, solving the problem of waste of manpower and inefficiency in the traditional way of relying on expert resources to formulate audit rules. The large model has a strong knowledge memory ability and can memorize the multi-dimensional knowledge of different diseases through pre-training of massive medical books and electronic medical records. And it can analyze and discover the inherent laws and patterns of the data through manual guidance. This feature enables it to better complete the formulation of disease group audit logic.

[0148] In some embodiments of the present application, another implementation scheme for rational review of DRG admission is provided. In this embodiment, two review capabilities are provided: a large model for medical insurance review and a traditional rule engine. The review capability can be selected intelligently according to the complexity of the review rules, and the review results of the two review capabilities are finally summarized to obtain the final review result. Figure 5 As shown, the following steps may be specifically included:

[0149] Step S200, obtaining the target cases to be reviewed and the target disease groups for their proposed DRG inclusion.

[0150] Step S210, obtaining the audit logic corresponding to the configured target disease group, the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured large model, and the audit logic includes more than one audit rule.

[0151] Step S220: For the audit logic corresponding to the target disease group, obtain the audit subtask corresponding to each audit rule therein.

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

[0153] Step S230: For each audit subtask, calculate the task complexity.

[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, at least based on the clarity of the audit rule. The task complexity indicates 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 is required to perform the task. At this time, the medical insurance audit model can be called to perform such audit subtasks. The lower the task complexity, the lower the difficulty of processing this type of audit subtask, and the standardized audit processing can be performed through the pre-review rules. Therefore, the traditional rule engine can be called to perform such audit subtasks.

[0155] In some possible implementations, the task complexity may be related not only to the clarity of the audit rules, but also to other factors, such as the complexity of the target disease group, the structured degree of the target case, etc. Therefore, the task complexity of the audit subtask may be determined by considering one or more of the influencing factors such as the clarity of the audit rules, the load of the target disease group, and the structured procedure of the target case.

[0156] For example, the complexity of the task can be determined by simultaneously considering the clarity of the audit rules, the complexity of the target disease group, and the degree of structuring of the target cases.

[0157] Among them, the less clear the review rules are, the higher the complexity of the target disease group is, and the lower the degree of structuring of the target cases is, the higher the corresponding task complexity is.

[0158] The complexity score of the defined task is represented by C, and the clarity score of the audit rules is represented by R. m The complexity score of the target disease group is expressed as R c The structural degree score of the target case is expressed as R s Expressed as follows, it satisfies the following formula:

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

[0160] Among them, R m The value range is from 0 to 1, 1 means completely clear, 0 means completely unclear; R s The value range is from 0 to 1, 1 means completely structured, 0 means completely unstructured, R c The value range is from 0 to 1, 1 means very complex, 0 means very simple. m is the weight of the clarity of the review rules, w s is the weight of the structured degree of the target case, w care the weights of the complexity of the target disease group. The three weights can be adjusted according to the actual situation to reflect the impact of different factors on the complexity of the task.

[0161] Step S240, determine whether the complexity of the task exceeds the set complexity threshold, if so, execute step S250, otherwise, execute step S260.

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

[0163] Step S250: call the configured medical insurance review big model to instruct the medical insurance review big model to execute the review subtask for the target case and obtain the review result corresponding to the review subtask.

[0164] Step S260: for the target case, call the configured rule engine to execute the review subtask and obtain the review result corresponding to the review subtask.

[0165] The rule engine can be configured using existing technologies or other methods. For cases with clear rules, highly structured data, and low complexity of disease groups, the rule engine can implement the review subtask by matching rules.

[0166] Step S270, summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

[0167] Specifically, by summarizing the output of the medical insurance audit 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 then the final audit result representing the rationality of DRG entry can be determined.

[0168] The method provided in this embodiment is equipped with two processing cores: a large model of medical insurance review and a rule engine. It can intelligently select the processing core for each review subtask based on factors such as rule clarity. The collaborative work of the two processing cores ensures the comprehensiveness and efficiency of the review.

[0169] In some embodiments of the present application, the calculation process of the clarity of the above-mentioned review rules, the complexity of the target disease group and the degree of structuring of the target case is explained.

[0170] In one possible implementation, the clarity of the audit rules, the complexity of the target disease groups, and the degree of structuring of the target cases can all be predicted by calling the medical insurance audit big model.

[0171] Specifically, the medical insurance review model can be instructed through prompt instructions to evaluate the clarity of the review rules, the complexity of the target disease group, and the structured degree of the target case. In the prompt instructions, requirement information can be further added to instruct the model to perform the evaluation according to the specified requirements.

[0172] In some possible implementations, in order to improve the clarity of the audit rules output by the medical insurance audit big model, the complexity of the target disease groups and the interpretability and accuracy of the structured degree of the target cases, the medical insurance audit big model can also be required to synchronously output the explanation of the current evaluation results in the prompt instruction, that is, to explain why the current evaluation results are given.

[0173] Next, an example of prompt is provided for the clarity of the review rules, the complexity of the target disease group, and the degree of structure of the target case:

[0174] The prompt example for calling the medical insurance audit model to evaluate the clarity of audit rules is as follows:

[0175] “Assuming you are an assessment expert, you can clarify the audit rules according to the requirements. m Conduct evaluation and analysis.

[0176]

Require

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

[0178] 2. For a given rule, assess whether it is clear, specific, and easy to understand and implement.

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

[0180] 4. Evaluate whether the rules provide enough contextual information so that auditors can apply them accurately.

[0181] 5. Based on the above factors, assign a clarity score (Rm) to the given rule, ranging from 0 (completely unclear) to 1 (completely clear).

[0182] 6. For a given rule, provide a specific score and explain why you gave that score.

[0183]

Medical Record Review Rules

[0184] Output: R m :\nExplain the reason:\n".

[0185] Calling the medical insurance review model to evaluate the structural degree of the case R s The prompt example is as follows:

[0186] “Suppose you are an assessment expert who can score the medical record structured degree R according to the requirements. s Conduct evaluation and analysis.

[0187]

Require

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

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

[0190] 3. Assess the presence of unstructured or semistructured elements in the medical record, such as free text or data in nonstandard formats.

[0191] 4. Evaluate whether fields in medical records are complete, missing, and contain sufficient information to support accurate interpretation and use of the data.

[0192] 5. Assign a structuredness score (R) to the entire medical record based on the consistency, interpretability, and completeness of the fields in the medical record content. s ), with scores ranging from 0 (completely unstructured) to 1 (completely structured).

[0193] 6. Provide a specific score and explain why it was given.

[0194]

Medical record data

[0195]

Output result

[0196] Calling the medical insurance audit model to evaluate the complexity of the disease group R c The prompt example is as follows:

[0197] “Suppose you are an assessment expert who can assess the complexity score R of the disease group according to the requirements. c Conduct evaluation and analysis.

[0198]

Require

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

[0200] 2. Consider the diversity of patient group information, including differences in diagnostic complexity, surgical complexity, number of diagnoses, and number of surgeries.

[0201] 3. Assess the complexity of the elements within the disease group, as well as the interactions and dependencies between them.

[0202] 4. Consider the difficulty of group management, including complex decision-making during diagnosis, treatment, and review of cases.

[0203] 5. Assign a complexity score (R) to each disease group based on the diversity, complexity, and difficulty of processing the disease group information. c ), with scores ranging from 0 (very easy) to 1 (very complex).

[0204] 6. Provide a specific score and explain why it was given.

[0205]

Disease group information

[0206]

Output result

[0207] In this embodiment, the medical insurance review big model is called to evaluate the clarity of the review rules, the complexity of the target disease group and the degree of structuring of the target case. With the help of the capabilities of the medical insurance review big model, more accurate evaluation results can be obtained, thereby facilitating the accurate determination of the complexity of the review subtasks, and further realizing the selection of intelligent review capabilities.

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

[0209] This embodiment provides a multi-task large model efficient training strategy.

[0210] In order to enable the base model to better understand and generate text related to medical audits, so as to adapt to subsequent audit tasks, in this embodiment, the audit subtasks corresponding to each audit rule of each disease group can be used to construct training data corresponding to the audit subtasks, that is, to obtain training data corresponding to multiple audit subtasks. The training data includes the task instruction description, case samples and result labels corresponding to the audit subtasks, and the result labels represent the audit results obtained by auditing the case samples according to the audit subtasks.

[0211] After constructing the training data, the training data corresponding to each audit subtask is used to train the base model to obtain the trained medical insurance audit model.

[0212] Through this training method, the medical insurance review model can better understand and generate texts related to medical reviews, thereby improving the quality and efficiency of reviews.

[0213] In one possible implementation, in order to reduce resource consumption in data annotation work, when constructing training data corresponding to each audit subtask, this embodiment can call a large model to assist in generating training data.

[0214] Specifically, a few-shot strategy can be adopted to manually label a small number of training data samples corresponding to each audit subtask to form a sample library.

[0215] Then, training data samples are iteratively extracted from the sample library and assembled into the second prompt instruction prompt. 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.

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

[0217] The large model here can be a general large model, a large model in the medical field, etc.

[0218] In this embodiment, the training data is generated by calling the large model, and only a small number of training data samples need to be manually labeled for reference by the large model, thereby reducing the consumption of labeling resources.

[0219] In one possible implementation, in order to improve the interpretability of 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 based on instructing the large model to refer to training data samples to generate new case samples and corresponding result labels. The explanation content is the reasoning explanation content of the result label for the description of the task instruction, and the review result of the new case sample generated is generated.

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

[0221] “Instructions: Assuming you are a medical expert, please generate medical record text according to the following requirements:

[0222] The given task instruction description: {instruction}

[0223] Example of a given medical record sample: {content}

[0224] Given result label: {true_label}

[0225] Among them, for a given task instruction, the result label of the medical record sample is {true_label}.

[0226] Now we are asked to generate a sample of medical records similar to this example.

[0227] The language style of the generated medical record samples is required to be similar to that of the medical record sample examples. The language must be rigorous and not colloquial.

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

[0229] In addition to the medical record samples, the output also includes explanations of why the case samples meet the task instructions.

[0230] The explanation part only includes two steps, namely, first extracting the relevant original medical record fragments according to the given task instructions, and giving an analysis of whether the relevant original medical record fragments meet the task instructions, and then giving the conclusion that the result is: "{true_label}".

[0231] The overall output format is: text:\nexplanation".

[0232] In some possible implementations, the above embodiments call the big model to generate training data for different audit subtasks. In order to improve the adaptability and accuracy of the big model, a multi-role-playing training mode can be introduced for different audit subtasks.

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

[0234] In the following example, roles are constructed according to different audit subtasks. For example, for the audit task of comparing test values, the role set for the big model can be "medical numerical test expert". Such role construction enables the big model to focus more on audit tasks in specific fields. For another example, for the audit task of drugs, the role set for the big model can be "pharmacy expert", etc., which will not be listed here one by one.

[0235] Reference Figure 6 , which illustrates a flow chart of a rational review method for DRG admission.

[0236] The entire DRG admission group rational review plan can be divided into two stages: Stage 1 and Stage 2.

[0237] Phase 1 is a pre-processing process, which means it can be completed before the DRG admission rational review of the cases to be reviewed.

[0238] Phase 2 is the process of conducting a rational review of the DRG admissions for the cases to be reviewed.

[0239] Phase 1 specifically includes:

[0240] The big model is called to visualize and summarize the diagnosis and treatment pathway data of each disease group, and the audit logic corresponding to the disease group is summarized. The specific implementation process can adopt the prompt design method of thinking chain. First, let the big model visualize the diagnosis and treatment pathway data into identifiable key elements, and then let the big model formulate the audit logic of the disease group based on the identified key elements.

[0241] Based on the sample library, the large model is used to generate training data corresponding to multiple different audit subtasks.

[0242] The sample library includes a small amount of training data samples corresponding to each audit subtask that are manually annotated. That is, for each audit rule in the audit logic of each disease group, a corresponding audit subtask can be constructed. Then, a small amount of training data samples corresponding to each audit subtask are manually annotated.

[0243] When the training data of the audit subtask is generated by the large model, the training data sample of the audit subtask extracted from the sample library can be added to the corresponding prompt instruction prompt to instruct the large model to generate training data with reference to the training data sample. In this way, the training data corresponding to multiple different audit subtasks generated by the large model can be obtained.

[0244] On this basis, multi-task training data can be used to train a large medical insurance review model.

[0245] Phase 2 specifically includes:

[0246] For the cases to be reviewed and the proposed target disease groups, we can first obtain the review logic corresponding to the target disease groups and the review subtasks corresponding to each review rule in the review logic.

[0247] For each audit subtask, the task complexity of the audit subtask can be calculated by combining the clarity of the audit rules, the complexity of the target disease group, and the degree of structuring of the cases to be audited. According to the size of the task complexity, the corresponding audit capability can be selected intelligently. Specifically, when the task complexity is high (exceeding the set task complexity threshold), the medical insurance audit large model can be selected to perform the current audit subtask; when the task complexity is low (not exceeding the set task complexity threshold), the traditional rule engine can be selected to perform the current audit subtask.

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

[0249] From the above embodiments, it can be seen that the DRG admission rational review method provided by this application has the following advantages:

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

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

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

[0253] The multi-task training method proposed in this application uses a small number of sample (few-shot) cases to generate training data, effectively reducing the cost and workload of data annotation. The training tasks cover multiple audit subtask dimensions, including but not limited to test value comparison, inspection, doctor's order, drug dimension, cost, etc., which comprehensively improves the accuracy and efficiency of the audit.

[0254] 3. Explainable audit decision mechanism:

[0255] This application proposes an explainable audit decision mechanism to enhance the transparency and credibility of the audit results. The automatic selection mechanism of audit capability can automatically decide whether to use the traditional rule engine audit method or the medical insurance audit model for audit according to the characteristics of the disease group to achieve the best audit effect. This audit mechanism improves the automation of the audit process, can realize automatic audit, reduce manual intervention, and improve audit speed and consistency.

[0256] 4. Improve the level of intelligence:

[0257] Through the application of multi-round interactive audit of the medical insurance audit big model, the intelligent level of rational audit of medical insurance DRG enrollment has been improved, making the audit process more efficient and accurate. The introduction of the medical insurance audit big model enables the system to understand and interpret the key elements in the diagnosis and treatment pathway, and improves the ability to analyze complex medical data.

[0258] 5. Save costs and resources:

[0259] Through the automated and intelligent audit process, the reliance on professionals is reduced, human resources and audit costs are saved, medical disputes and medical insurance fund losses caused by audit errors are reduced, and the efficiency of the use of medical insurance funds is improved.

[0260] 6. Flexibility and scalability:

[0261] The method of this application has good flexibility and scalability, and can adapt to the ever-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 the improvement of medical quality:

[0263] Through accurate DRG inclusion review, medical service providers are encouraged to improve service quality and efficiency, thereby improving the overall medical level.

[0264] The following is a description of the DRG admission group rational review device provided in an embodiment of the present application. The DRG admission group rational review device described below and the DRG admission group rational review method described above can be referenced to each other.

[0265] See also Figure 7 , Figure 7 This is a schematic diagram of the structure of a DRG admission rational review device disclosed in an embodiment of the present application.

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

[0267] The pending data acquisition unit 11 is used to acquire the pending target cases and the target disease groups for which DRGs are planned to be included;

[0268] An audit logic acquisition unit 12 is used to acquire the audit logic corresponding to the configured target disease group, wherein the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured large model, and the audit logic includes more than one audit rule;

[0269] A subtask acquisition unit 13 is used to acquire the audit subtask corresponding to each of the audit rules for the audit logic corresponding to the target disease group;

[0270] The medical insurance review model calling unit 14 is used to call the configured medical insurance review model to instruct the medical insurance review model to execute each of the review subtasks for the target case and obtain the review result corresponding to each of the review subtasks. The medical insurance review model is obtained by training the base model with the training data of multiple review subtasks.

[0271] The result summarizing unit 15 is used to summarize the audit results of each audit subtask in the audit logic to obtain the audit results that characterize the rationality of DRG grouping.

[0272] In a possible implementation, the device of the present application may further include:

[0273] The audit logic generation unit is used to summarize the diagnosis and treatment pathway data corresponding to the target disease group through the configured large model to obtain the audit logic, and the process includes:

[0274] Sending the first prompt instruction prompt to the configured big model to obtain the audit logic of the target disease group output by the big model;

[0275] Among them, the first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment pathway data corresponding to the target disease group to obtain the review logic of the target disease group.

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

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

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

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

[0280] The correspondence between the configured audit rules and the audit subtasks is queried to obtain the audit subtask corresponding to each audit rule in the audit logic corresponding to the target disease group.

[0281] In a possible implementation, the device of the present application may further include:

[0282] The task complexity calculation unit is used for determining the task complexity for each audit rule in the audit logic before the medical insurance audit big model calling unit calls the configured medical insurance audit big model to instruct the medical insurance audit big model to execute each audit subtask for the target case: at least based on the clarity of the audit rule; if the task complexity indicates that it exceeds the set complexity threshold, the medical insurance audit big model is called through the medical insurance audit big model calling unit to instruct the medical insurance audit big model to execute the audit subtask corresponding to the audit rule for the target case.

[0283] Further optionally, the device of the present application may also include:

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

[0285] In a possible implementation, the task complexity calculation unit determines the task complexity based on at least the clarity of the audit rule, including:

[0286] Determine the complexity of the task based on the clarity of the review rules, the complexity of the target disease group, and the degree of structure of the target case;

[0287] Among them, the less clear the review rules are, the higher the complexity of the target disease group is, and the lower the degree of structuring of the target cases is, the higher the corresponding task complexity is.

[0288] In a possible implementation, the clarity of the review rules, the complexity of the target disease group, and the degree of structuring of the target cases are all predicted by calling the medical insurance review model.

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

[0290] The medical insurance review model training unit is used to train the medical insurance review model. The corresponding training process includes:

[0291] For the audit subtasks corresponding to each audit rule in the audit logic of each disease group, construct training data corresponding to each audit subtask, wherein the training data includes a task instruction description, a case sample and a 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;

[0292] The training data corresponding to each audit subtask is used to train the configured base large model to obtain the trained medical insurance audit large model.

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

[0294] Obtain manually labeled training data samples corresponding to each audit subtask to form a sample library;

[0295] Iteratively extract the training data samples from the sample library and assemble 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;

[0296] Send the second prompt instruction prompt to the big model to obtain the case sample and result label output by the big model, and combine them with the task instruction description in the training data sample to form new training data;

[0297] The new training data is added to the training data set corresponding to the audit subtask.

[0298] In one possible implementation, the second prompt instruction prompt, on the basis of instructing the large model to refer to the training data sample to generate a new case sample and a corresponding result label, is also used to instruct the large model to output an explanation content, wherein the explanation content is the reasoning explanation content of the result label for the description of the task instruction, and the review result of the new case sample generated is generated.

[0299] The present application also provides an electronic device in an embodiment. Figure 8 As shown, it shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present application. The electronic device in the embodiment of the present application may include but is not limited to fixed terminals such as mobile phones, tablet computers, desktop computers, medical diagnostic equipment, interactive robots, etc. Figure 8 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0300] like Figure 8 As shown, the electronic device may include a processing device (such as 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 to a random access memory (RAM) 603, so as to implement the DRG entry rational audit method of the aforementioned embodiment of the present application. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0301] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0302] Also provided in an embodiment of the present application is a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the DRG group entry rational review methods provided in the embodiments of the present application.

[0303] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When one or more computer programs are executed by an electronic device, the electronic device can implement any one of the DRG admission rational review methods provided in the embodiment of the present application.

[0304] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0305] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement 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 implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

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

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

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

Claims

1. A rational review method for DRG admission, characterized in that: include: Obtain the target cases to be reviewed and the target disease groups for their proposed DRG enrollment; Acquire the audit logic corresponding to the configured target disease group, wherein the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured big model, and the audit logic includes more than one audit rule; For the audit logic corresponding to the target disease group, the audit subtask corresponding to each of the audit rules is obtained, and the configured medical insurance audit big model is called to instruct the medical insurance audit big model to execute each of the audit subtasks for the target case, and the audit result corresponding to each of the audit subtasks is obtained. The medical insurance audit big model is obtained by training the base big model with the training data of multiple audit subtasks; Summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

2. The method according to claim 1, characterized in that The process of summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured large model to obtain the audit logic includes: Sending the first prompt instruction prompt to the configured big model to obtain the audit logic of the target disease group output by the big model; Among them, the first prompt instruction prompt is used to instruct the large model to summarize the diagnosis and treatment pathway data corresponding to the target disease group to obtain the review logic of the target disease group.

3. The method according to claim 2, characterized in that The first prompt instruction prompt adopts a thought chain structure, which is used to instruct the large model to perform the following steps in sequence: First, the diagnosis and treatment pathway data corresponding to the target disease group is analyzed to identify the key elements therein; Further based on the key elements, the audit logic of the target disease group is formulated.

4. The method according to claim 1, characterized in that: For the audit logic corresponding to the target disease group, the process of obtaining the audit subtask corresponding to each of the audit rules includes: The correspondence between the configured audit rules and the audit subtasks is queried to obtain the audit subtask corresponding to each audit rule in the audit logic corresponding to the target disease group.

5. The method according to claim 1, characterized in that Before calling the configured medical insurance review model to instruct the medical insurance review model to execute each review subtask for the target case, it also includes: For each audit rule in the audit logic: Determine the complexity of the task based at least on the clarity of the audit rules described; If the complexity of the task indicates that it exceeds the set complexity threshold, the medical insurance review model is called to instruct the medical insurance review model to execute the review subtask corresponding to the review rule for the target case.

6. The method according to claim 5, characterized in that Also includes: If the task complexity indicates that it does not exceed the set complexity threshold, then for the target case, the configured rule engine is called to execute the audit subtask corresponding to the audit rule to obtain the audit result corresponding to the audit subtask.

7. The method according to claim 5, characterized in that A process for determining the complexity of the task, based at least on the clarity of the review rules, including: Determine the complexity of the task based on the clarity of the review rules, the complexity of the target disease group, and the degree of structure of the target case; Among them, the less clear the review rules are, the higher the complexity of the target disease group is, and the lower the degree of structuring of the target cases is, the higher the corresponding task complexity is.

8. The method according to claim 7, characterized in that The clarity of the review rules, the complexity of the target disease group and the degree of structuring of the target cases are all predicted by calling the medical insurance review model.

9. The method according to claim 1, characterized in that: The training process of the medical insurance review model includes: For the audit subtasks corresponding to each audit rule in the audit logic of each disease group, construct training data corresponding to each audit subtask, wherein the training data includes a task instruction description, a case sample and a 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; The training data corresponding to each audit subtask is used to train the configured base large model to obtain the trained medical insurance audit large model.

10. The method according to claim 9, characterized in that The process of constructing training data corresponding to each audit subtask includes: Obtain manually labeled training data samples corresponding to each audit subtask to form a sample library; Iteratively extract the training data samples from the sample library and assemble 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; Send the second prompt instruction prompt to the big model to obtain the case sample and result label output by the big model, and combine them with the task instruction description in the training data sample to form new training data; The new training data is added to the training data set corresponding to the audit subtask.

11. The method according to claim 10, characterized in that The second prompt instruction prompt is used to instruct the large model to refer to the training data sample to generate a new case sample and a corresponding result label, and is also used to instruct the large model to output an explanation content. The explanation content is the reasoning explanation content of the result label for the description of the task instruction, and the audit result of the new case sample generated is generated.

12. A DRG entry rational review device, characterized in that: include: The data acquisition unit to be reviewed is used to obtain the target cases to be reviewed and the target disease groups to be included in the proposed DRG; An audit logic acquisition unit, used to acquire the audit logic corresponding to the configured target disease group, wherein the audit logic is the audit logic obtained by summarizing the diagnosis and treatment pathway data corresponding to the target disease group through the configured big model, and the audit logic includes more than one audit rule; A subtask acquisition unit, configured to acquire, for the audit logic corresponding to the target disease group, an audit subtask corresponding to each of the audit rules; A medical insurance review model calling unit is used to call the configured medical insurance review model to instruct the medical insurance review model to execute each of the review subtasks for the target case and obtain the review result corresponding to each of the review subtasks. The medical insurance review model is obtained by training the base model with the training data of multiple review subtasks. A result summary unit is used to summarize the audit results of each audit subtask in the audit logic to obtain an audit result that represents the rationality of DRG group entry.

13. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; The processor is used to execute the program to implement the various steps of the DRG admission rational review method as described in any one of claims 1 to 11.

14. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the DRG admission rational review method as described in any one of claims 1 to 11 is implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, each step of the DRG admission rational review method as described in any one of claims 1 to 11 is implemented.

Images