CN-122024983-A - Diagnosis method and system for electronic medical record and DRG rule based on large language model

CN122024983ACN 122024983 ACN122024983 ACN 122024983ACN-122024983-A

Abstract

The application discloses a diagnosis method and a system of an electronic medical record and a DRG rule based on a large language model, and relates to the technical field of medical information; the method comprises the steps of inputting electronic medical record data into a pre-trained large language model, outputting candidate diagnosis items by the large language model, wherein the candidate diagnosis items comprise candidate disease diagnoses and/or candidate operation diagnoses, receiving confirmation or correction instructions of a user on the candidate diagnosis items, updating and determining a final diagnosis item set based on the instructions, wherein the diagnosis item set comprises disease diagnoses and/or operation diagnoses, inputting the diagnosis item set into a DRG rule engine, performing compliance screening, combination optimization and weight evaluation on items in the set, and generating and outputting optimal diagnosis item combinations. The application solves the technical problems of incomplete diagnosis information extraction and difficult DRG rule adaptation in the traditional manual coding mode, and realizes the improvement of diagnosis accuracy and compliance of medical insurance payment.

Inventors

JIANG HUA
GU WEIRONG
WANG JUE
LU JIAQI
QIN KAIZHOU
LI JIA
LI NANA
YU BIN

Assignees

复旦大学附属妇产科医院

Dates

Publication Date: 20260512
Application Date: 20260204

Claims (10)

1. A method for diagnosing an electronic medical record and DRG rules based on a large language model, comprising: Acquiring electronic medical record data of a patient; inputting the electronic medical record data into a pre-trained large language model, carrying out semantic understanding and information extraction by the large language model, and outputting candidate diagnosis items, wherein the candidate diagnosis items comprise candidate disease diagnosis and/or candidate operation diagnosis; receiving a confirmation or correction instruction of a user on the candidate diagnosis items, and updating and determining a final diagnosis item set based on the instruction, wherein the diagnosis item set comprises disease diagnosis and/or operation diagnosis; inputting the diagnosis item set to a DRG rule engine, carrying out compliance screening, combination optimization and weight evaluation on items in the diagnosis item set based on a preset DRG grouping rule, and generating and outputting an optimal diagnosis item combination meeting the DRG grouping requirement.
2. The method of claim 1, wherein the candidate disease diagnosis is formed by extracting disease record content from the electronic medical record data by the large language model.
3. The method of claim 1, wherein the candidate surgical diagnosis is formed by extracting surgical record content from the electronic medical record data by the large language model.
4. The method for diagnosing a large language model based electronic medical record and DRG rules of claim 1, wherein said performing a combined optimization and weight evaluation of entries in the set of diagnostic entries includes: Generating candidate pairing combinations of multiple diseases and operations based on the disease diagnosis and the operation diagnosis in the diagnosis item set; Filtering invalid combinations of the plurality of candidate pairing combinations based on DRG grouping rules; Calculating the weight of each effective pairing combination; and traversing each effective pairing combination, and outputting the pairing combination with the optimal weight as the optimal diagnosis item combination.
5. The method for diagnosing a DRG rule with an electronic medical record based on a large language model as recited in claim 4, the method is characterized in that the traversal of each effective pairing combination adopts a round robin algorithm.
6. The method of claim 4, wherein the optimal diagnosis entry combination includes at least one of a primary disease diagnosis, a secondary disease diagnosis, a primary surgical diagnosis, and a secondary surgical diagnosis.
7. The method for diagnosing a large language model based electronic medical record and DRG rules as recited in claim 1, wherein after the step of receiving the user's correction instruction for the candidate diagnostic entry, further comprises: and taking the correction instruction and the corresponding electronic medical record data fragment as training data to finely adjust the large language model.
8. A diagnostic system for electronic medical records and DRG rules based on a large language model, comprising: A data acquisition module configured to perform acquisition of electronic medical record data of a patient; A large language model processing module configured to perform inputting the electronic medical record data into a pre-trained large language model, semantic understanding and information extraction by the large language model, and output candidate diagnosis items, wherein the candidate diagnosis items comprise candidate disease diagnoses and/or candidate operation diagnoses; a human-machine interaction module configured to perform receiving user confirmation or correction instructions for the candidate diagnostic items, updating and determining a final set of diagnostic items based on the instructions, the set of diagnostic items including disease diagnosis and/or surgical diagnosis; The DRG rule engine module is configured to input the diagnosis item set into the DRG rule engine, carry out compliance screening, combination optimization and weight evaluation on items in the diagnosis item set based on a preset DRG grouping rule, and generate and output an optimal diagnosis item combination meeting the DRG grouping requirement.
9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the large language model based diagnostic method of electronic medical records and DRG rules of any one of claims 1-7.
10. A computer program product comprising a computer program which, when executed by a processor, implements the large language model based electronic medical record and DRG rules diagnostic method of any one of claims 1-7.

Description

Diagnosis method and system for electronic medical record and DRG rule based on large language model Technical Field The application relates to the technical field of medical information, in particular to a diagnosis method and a diagnosis system for electronic medical records and DRG rules based on a large language model. Background With the deepening of the national medical insurance payment system, disease diagnosis related groups (DRGs for short) have been comprehensively advanced in various places. In this mode, the payment standard of the medical insurance fund is directly hooked with the DRG of the patient group, and the core requirement is that the clinical diagnosis and the DRG grouping rule are accurately matched. However, in practice, medical institutions face various problems in achieving fine admission, such as clinical diagnostic information being widely distributed in unstructured text of electronic medical records (such as complaints, current medical history, operation records, disease course records) and various examination test reports, and are diverse in form, specialized in expression and strong in implication. The traditional mode of relying on manual reading and coding is low in efficiency, and is more prone to missing or misjudgment of key diagnosis features due to knowledge limitation or fatigue of a coder, and the key diagnosis features such as complications and complications cause incomplete and inaccurate extraction of original diagnosis information. In addition, the DRG grouping device has complex rules and continuously updated rules, clinicians have various writing habits, and the encoder can have deviation on the understanding of the rules, so that the finally submitted diagnosis codes cannot be mapped to the DRG group with the best matching resource consumption and the most reasonable medical insurance payment, the mismatching rate of the diagnosis and the optimal DRG group is high, the risk of insufficient medical insurance payment or illegal regulation is caused, and the economic operation of a hospital is directly influenced. The prior art is based on rules or auxiliary tools of traditional natural language processing, and is difficult to solve the contradiction between unstructured text semantic understanding of the electronic medical record and DRG dynamic rule adaptation, and the prior art cannot complete the generation of optimal diagnosis combination from the electronic medical record to the DRG rules. Disclosure of Invention The application aims to provide a diagnosis method and a diagnosis system for electronic medical records and DRG rules based on a large language model, wherein candidate diagnosis in the electronic medical records is automatically extracted through the large language model, and after human-computer collaborative confirmation, a DRG rule engine is utilized to carry out compliance screening, combination optimization and weight evaluation on diagnosis items, and finally an optimal diagnosis combination meeting the requirements of medical insurance grouping is generated. In order to achieve the above object, the present application provides the following solutions: The application provides a diagnosis method of electronic medical records and DRG rules based on a large language model, which comprises the steps of obtaining electronic medical record data of a patient, inputting the electronic medical record data into the large language model which is pre-trained, carrying out semantic understanding and information extraction by the large language model, outputting candidate diagnosis items, wherein the candidate diagnosis items comprise candidate disease diagnosis and/or candidate operation diagnosis, receiving confirmation or correction instructions of a user on the candidate diagnosis items, updating and determining a final diagnosis item set based on the instructions, wherein the diagnosis item set comprises disease diagnosis and/or operation diagnosis, inputting the diagnosis item set into a DRG rule engine, carrying out compliance screening, combination optimization and weight evaluation on items in the diagnosis item set based on a preset DRG grouping rule, and generating and outputting optimal diagnosis item combinations meeting DRG grouping requirements. Optionally, the candidate disease diagnosis is formed by the large language model by extracting disease record content in the electronic medical record data. Optionally, the candidate surgical diagnosis is formed by the large language model by extracting surgical record content in the electronic medical record data. Optionally, the combined optimization and weight evaluation of the items in the diagnosis item set comprises the steps of generating candidate pairing combinations of various diseases and operations based on disease diagnosis and operation diagnosis in the diagnosis item set, filtering invalid combinations in the candidate pairing combinations based on a DRG grouping rule, calculatin