Search

CN-121983216-A - Electronic medical record examination method, system, equipment and medium

CN121983216ACN 121983216 ACN121983216 ACN 121983216ACN-121983216-A

Abstract

The application discloses a method, a system, equipment and a medium for inspecting electronic medical records, which relate to the technical field of computers and comprise the steps that a computer device acquires original medical data from different sources, and the original medical data is preprocessed to obtain preprocessed data; the method comprises the steps of determining a rule chain corresponding to an inspection task based on an inspection rule or a large model manually configured by a user, wherein the inspection rule comprises a target rule based on a structured query language and a large model semantic rule based on prompt word engineering, executing the inspection task based on preprocessed data according to the rule chain of the inspection task after triggering the inspection task to obtain an inspection result, identifying the type of a last rule in the rule chain, processing the inspection result based on the identified type to obtain a processed result, and integrating the processed result corresponding to each inspection task with an electronic medical record to obtain a structured inspection report. Therefore, the application can automatically and batchly assist the staff to complete the task of reviewing the electronic medical records.

Inventors

  • FU JIANFEI
  • LI ZHEN

Assignees

  • 宁波大学附属第一医院(宁波市第一医院)

Dates

Publication Date
20260505
Application Date
20260408

Claims (10)

  1. 1. An electronic medical record review method, applied to a computer device, comprising: Acquiring original medical data of different sources, and preprocessing the original medical data to obtain preprocessed data; Determining a rule chain corresponding to an inspection task based on inspection rules or a large model manually configured by a user, wherein the inspection rules comprise target rules based on a structured query language and large model semantic rules based on prompt word engineering; After triggering the examination task, executing the examination task based on the preprocessed data according to a rule chain of the examination task to obtain a corresponding examination result; identifying the type of the last rule in the rule chain, and correspondingly processing the examination result based on the identified type to obtain a processed result; And integrating the processed results corresponding to each examination task with the electronic medical record to obtain a corresponding structured examination report.
  2. 2. The electronic medical record review method of claim 1, wherein the obtaining raw medical data from different sources comprises: And extracting the original medical data from different sources according to the data extraction task and the configuration information which are created in advance by the user, wherein the configuration information comprises a source database, data extraction sentences, data previews and application programming interface parameters of the original medical data.
  3. 3. The electronic medical record review method of claim 1, wherein triggering the review task comprises: triggering the auditing task based on a monitored manual operation of a user or a preset time threshold.
  4. 4. The electronic medical record review method of claim 1, wherein the preprocessing the raw medical data to obtain preprocessed data comprises: performing data cleaning on the original medical data to obtain cleaned medical data; Carrying out format standardization treatment on the medical data after cleaning to obtain standardized medical data; Storing the standardized medical data into a target table structure according to a logical relationship to verify the integrity and accuracy of the standardized medical data; And storing the standardized medical data into a target data table according to the corresponding verification result to obtain preprocessed data.
  5. 5. The electronic medical record review method of claim 1, wherein the determining a rule chain corresponding to a review task based on a review rule or a large model manually configured by a user comprises: Determining machine-executable censoring rules manually configured by a user within a rules engine configuration unit; sorting the inspection rules to determine rule chains corresponding to the inspection tasks according to the corresponding sorted rules; or decomposing the natural language input by the user into a plurality of examination tasks by utilizing the target large model, and decomposing each examination task into an executable rule chain based on the target large model.
  6. 6. The electronic medical record review method of claim 5, wherein the identifying the type of the last rule in the rule chain and correspondingly processing the review result based on the identified type to obtain a processed result comprises: If the last rule in the rule chain is the target rule, directly displaying the rule name and the examination result to obtain a processed result; If the last rule in the rule chain is a big model semantic rule, analyzing an examination result returned by the big model into structural data, and determining a processed result according to target information in the structural data, wherein the target information comprises a medical record list, hit number, confidence coefficient and evidence.
  7. 7. The electronic medical record review method according to any one of claims 1 to 6, wherein the performing the review task based on the preprocessed data according to the rule chain of the review task to obtain the corresponding review result includes: and sequentially executing the inspection tasks according to the sequence of each rule in the rule chain of the inspection tasks so as to acquire a target result output by the last rule, and determining the target result as the inspection result.
  8. 8. An electronic medical record review system, characterized by being applied to a computer device, comprising: The data preprocessing module is used for acquiring original medical data from different sources, and preprocessing the original medical data to obtain preprocessed data; the rule chain determining module is used for determining a rule chain corresponding to the examination task based on examination rules or a large model manually configured by a user, wherein the examination rules comprise target rules based on a structured query language and large model semantic rules based on prompt word engineering; The task execution module is used for executing the examination task based on the preprocessed data according to a rule chain of the examination task after triggering the examination task so as to obtain a corresponding examination result; the examination result processing module is used for identifying the type of the last rule in the rule chain and correspondingly processing the examination result based on the identified type to obtain a processed result; and the examination report acquisition module is used for integrating the processed results corresponding to each examination task with the electronic medical record so as to obtain a corresponding structured examination report.
  9. 9. An electronic device, comprising: A memory for storing a computer program; A processor for executing a computer program to perform the steps of the electronic medical record review method of any one of claims 1 to 7.
  10. 10. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, which when executed by a processor, implements the steps of the electronic medical record review method according to any one of claims 1 to 7.

Description

Electronic medical record examination method, system, equipment and medium Technical Field The present invention relates to the field of computer technologies, and in particular, to a method, a system, an apparatus, and a medium for electronic medical record inspection. Background The medical government administration needs to review the electronic medical records according to the knowledge base such as the system, the specification and the like to determine whether the medical behaviors or the medical record documents are in compliance. Such tasks are widely distributed in a plurality of departments such as medical records management, medical management, quality management, hospital sensing prevention and control, public health, medical insurance management, medicine consumption management and the like. In order to ensure that the tasks are successfully completed, medical institutions need to invest huge manpower and time cost, and the working capacity, knowledge level and the like of different personnel are different, so that heterogeneity exists in measurement standards, understanding depth and auditing results. The examination work involves multiple departments of medical management, quality management, hospital infection prevention and control, public health, medical insurance management, medicine consumption management, medical records management and the like. 4 common ways to complete the audit task: A. when the storage mode of the electronic medical record is a structured field, the electronic medical record can be queried by adopting a database query method, because the range of the value range is fixed. For example, looking up all patients diagnosed with "cholera" by ICD (International classification of diseases, international disease classification) for a defined period of time, the ICD code may be defined as "a00" at the time of database query. The method has the characteristics of being limited and incapable of checking tasks aiming at texts at semantic understanding level. B. For long text format electronic medical record content, workers can use manual review methods, which are the most commonly used methods. C. The long text format electronic medical record can also be computer-aided understood and read using natural language processing and methods of building proprietary medical models. The method has the characteristics of strong model specificity, weak generalization capability, poor portability and no support for long-context reasoning. D. In recent years, large language model technology has developed rapidly, and some hospitals assist in the task of screening by deploying large language models. The method has the characteristics of strong generalization capability of the model, strong portability and strong portability in support of long context reasoning, but weak professional capability specificity. With iterative updating of the technology in the field, the problem of weak professional ability can be optimally solved by prompting mechanisms such as engineering optimization, context reinforcement mechanism, small sample tuning, parameter optimization and the like. In actual work, in order to improve the working efficiency, a worker usually completes the task by combining multiple examination modes. For example, when task 1 "when diagnosing < malignancy > and < pleural effusion > are present at the same time, it is necessary to verify whether < pleural effusion > is < malignancy pleural effusion >", the general implementation is: (1) And in the structural searching stage, a structural database is searched, and the conditions of the structural database are simultaneously satisfied, wherein the partial searching can be automatically performed by adopting a computer, for example, a relational database is adopted to apply SQL sentence writing rules. Such structured search is referred to as a "mechanical rule". (2) And in the manual verification stage, the electronic medical records are manually opened and verified one by one according to the retrieved medical records list, and the method is called as manual verification. For task 1, it is manually necessary to review the pathology report, test results and disease course record, find relevant "malignancy" relevant evidence, and mark the result of "yes/no" for the task. Currently, most post-electronic medical record application systems use a "mechanical rules+manual verification" mode, which consumes more manpower and time. However, mechanical rules are hardly useful for semantic understanding class tasks. The current possibilities are not careful in reliability and stress for complex medical scenarios. In summary, existing electronic medical record review systems rely heavily on structured data with limited unstructured text processing capabilities. Therefore, how to improve the working efficiency and quality of coping with complex medical scenes, reduce task heterogeneity, and reduce labor and time costs is a current urgent problem to be s