CN-122025104-A - Intelligent monitoring and early warning system for medical adverse events

CN122025104ACN 122025104 ACN122025104 ACN 122025104ACN-122025104-A

Abstract

The application discloses an intelligent monitoring and early warning system for medical adverse events, which relates to the technical field of medical treatment, and comprises an ETL module, a medical adverse event prediction module, a medical adverse event model and a medical adverse event reporting module, wherein the ETL module is used for extracting medical text data from a clinical information system of a hospital, the medical adverse event prediction module is used for predicting the medical text data to obtain medical adverse event types and grades corresponding to the medical text data, the medical adverse event prediction module comprises a medical adverse event model obtained through training, the medical adverse event model is obtained through training the medical model according to historical medical adverse event samples, and the medical adverse event model can output corresponding medical adverse event types and grades according to the input medical text data and is used for outputting prompt information comprising the medical text data and the medical adverse event types and grades corresponding to the medical text data.

Inventors

CAO KUN
XIE JIAHAO
ZHU LINGFENG
Ying Qianshan
Zhu Jiakang

Assignees

台州恩泽医疗中心(集团)

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. An intelligent monitoring and early warning system for medical adverse events, which is characterized by comprising: The medical adverse event prediction system comprises an ETL module, a hospital clinical information system, a medical adverse event prediction module and a display module, wherein the medical adverse event prediction module comprises a medical adverse event model obtained through training, the medical adverse event model is obtained through training the medical model according to a historical medical adverse event sample, and the medical adverse event model can output corresponding medical adverse event types and grades according to input medical text data; The ETL module is used for extracting medical text data from the hospital clinical information system; the medical adverse event prediction module is used for predicting the medical text data and acquiring the class and grade of the medical adverse event corresponding to the medical text data; the display module is used for outputting prompt information, wherein the prompt information comprises medical text data and medical adverse event categories and grades corresponding to the medical text data.
2. The system of claim 1, further comprising: And the display module is also used for receiving a reporting instruction triggered by the user, and reporting the medical text data, the corresponding medical adverse event category and grade according to the reporting instruction.
3. The system of claim 1, further comprising a detection module; The detection module is used for detecting medical text data of which the predicted medical adverse event category and grade meet the preset requirements and sending the medical text data of which the medical adverse event category and grade meet the preset requirements and the corresponding medical adverse event category and grade to the display module; the display module is used for outputting prompt information, wherein the prompt information comprises medical text data required by the preset, and medical adverse event types and grades corresponding to the medical text data.
4. The system of claim 1, further comprising a training module; the training module is used for acquiring a historical medical adverse event sample, wherein the historical medical adverse event sample comprises a sample medical adverse event described, a true grade and a true category of the sample medical adverse event, and training a medical model according to the historical medical adverse event sample to obtain a medical adverse event model.
5. The system according to claim 4, wherein in the aspect of training the medical model according to the historical medical adverse event samples to obtain a medical adverse event model, the training module is specifically configured to: carrying out coding processing on the description of the sample medical adverse event to obtain a word vector corresponding to the sample medical adverse event; inputting the word vector into a feature calculation layer of the medical model to extract features of the word vector through a self-attention mechanism, so as to obtain a feature vector for representing the sample medical adverse event; Inputting the feature vector into a classification full-connection layer of the medical model, and outputting the category prediction probability of the sample medical adverse event belonging to each medical adverse event category; inputting the feature vector into a grading full-connection layer of the medical model, and outputting the grade prediction probability of the sample medical adverse event belonging to each medical adverse event grade; And training the medical model according to the category prediction probability, the level prediction probability, the true grade and the true category of the sample medical adverse event to obtain the medical adverse event model.
6. The system of claim 5, wherein the loss function when training the medical model based on the class prediction probability and the true class of sample medical adverse events is: ; wherein N represents the number of samples, C represents the total number of classification categories of the medical adverse events, A category prediction probability indicating that the ith sample healthcare adverse event belongs to the c-th healthcare adverse event category, Representing the true probability that the true category of the ith sample healthcare adverse event belongs to the c-th healthcare adverse event category, when the true category of the ith sample healthcare adverse event is the c-th healthcare adverse event category, =1, Otherwise =0; The loss function when training the medical model according to the level prediction probability and the actual level of the sample medical adverse event is as follows: ; wherein N represents the number of samples, M represents the total number of classes of adverse medical events, A level prediction probability indicating that the ith sample medical adverse event belongs to the mth medical adverse event level, Representing a true probability that the true level of the ith sample healthcare adverse event belongs to the mth healthcare adverse event level, when the true level of the ith sample healthcare adverse event belongs to the mth healthcare adverse event level, =1, Otherwise =0。
7. The system according to claim 6, wherein, in terms of the classification fully-connected layer of the input feature vector into the medical model to make the output sample medical adverse event belong to the category prediction probability of each medical adverse event category, the training module is specifically configured to: Inputting the feature vector into a classification full-connection layer of the medical model, and obtaining a category logistic regression vector, wherein the category logistic regression vector consists of initial category prediction probabilities of sample medical adverse events belonging to various medical adverse event categories; Applying a softmax function to the category logistic regression vector to obtain category prediction probabilities of the sample medical adverse events belonging to the categories of the medical adverse events; Wherein, the , The c-th initial class prediction probability in the class logistic regression vector representing the i-th sample medical adverse event.
8. The system according to claim 7, wherein, in terms of the hierarchical fully-connected layer of the input feature vector into the medical model such that the output sample medical adverse event belongs to the level prediction probability of the respective medical adverse event level, the training module is specifically configured to: Inputting the feature vector into a hierarchical full-connection layer of the medical model, and obtaining a hierarchical logistic regression vector, wherein the hierarchical logistic regression vector consists of initial level prediction probabilities of sample medical adverse events belonging to various medical adverse event classes; applying a softmax function to the grade logistic regression vector to obtain the grade prediction probability that the sample medical adverse event belongs to each medical adverse event grade; Wherein, the , The m-th initial level prediction probability in the level logistic regression vector representing the i-th sample medical adverse event.
9. The system of claim 8, wherein the medical model is trained from an original large language model using a medical knowledge base comprising medical expertise, care and health education knowledge, and clinical paths and policy files.
10. The system according to claim 9, wherein, in the aspect of inputting the word vector into the feature computation layer of the medical model to perform feature extraction on the word vector by a self-attention mechanism to obtain a feature vector characterizing the sample medical adverse event, the training module is specifically configured to: the feature vector of the sample medical adverse event is obtained through the following formula: ; Wherein, the Is three mapping matrices that can be learned and, Is a key vector Is a feature vector of length d, d being a hyper-parameter of the medical model, e representing a word vector.

Description

Intelligent monitoring and early warning system for medical adverse events Technical Field The application relates to the technical field of medical treatment, in particular to an intelligent monitoring and early warning system for medical adverse events. Background Medical adverse events (including: medical quality adverse events and medical safety adverse events) refer to medical related events that occur during a medical procedure that are unexpected and may cause injury to a patient, and are characterized by three general characteristics, unexpected (inconsistent with expected medical outcomes), damaging (causing physical or psychological damage to the patient), and preventable (avoidable by modification of the system or procedure). Therefore, establishing a reporting system of the adverse medical events has double values, namely, experience sharing and system improvement are realized through active reporting, and basis is provided for continuous improvement of medical quality through data analysis. Currently, the traditional adverse medical event reporting system commonly adopted by medical institutions in China mainly depends on an active reporting mechanism of clinical medical staff. However, due to the fact that the clinical first-line work load is heavy, time resources are tense, and part of medical staff is unfamiliar with or has worry about the reporting flow, the problems of low reporting rate, untimely information reporting and the like exist in the actual operation of the system. Disclosure of Invention The application aims to provide an intelligent monitoring and early warning system for medical adverse events, which can improve reporting efficiency of the medical adverse events. In order to achieve the above object, the present application provides the following solutions: The application provides an intelligent monitoring and early warning system for medical adverse events, which comprises: The medical adverse event prediction system comprises an ETL module, a hospital clinical information system, a medical adverse event prediction module and a display module, wherein the medical adverse event prediction module comprises a medical adverse event model obtained through training, the medical adverse event model is obtained through training the medical model according to a historical medical adverse event sample, and the medical adverse event model can output corresponding medical adverse event types and grades according to input medical text data; The ETL module is used for extracting medical text data from the hospital clinical information system; the medical adverse event prediction module is used for predicting the medical text data and acquiring the class and grade of the medical adverse event corresponding to the medical text data; the display module is used for outputting prompt information, wherein the prompt information comprises medical text data and medical adverse event categories and grades corresponding to the medical text data. According to the specific embodiment provided by the application, the application discloses the following technical effects: The application provides an intelligent monitoring and early warning system for medical adverse events, which automatically extracts medical text data from a clinical information system through an ETL (extract-transform-load) module, replaces manual collection and arrangement, omits tedious processes, reduces labor and time cost, relies on a trained medical adverse event model, can rapidly analyze the extracted data, outputs the type and grade of the medical adverse event, avoids the inefficiency problem of manual judgment, and simultaneously, can directly output prompt information containing complete information, simplifies reporting operation of medical staff, does not need manual form filling and arrangement, avoids reporting delay caused by information lag, and remarkably improves reporting efficiency and timeliness. Drawings In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. FIG. 1 is a flowchart illustrating a medical adverse event model training method, according to an exemplary embodiment; FIG. 2 is a diagram illustrating partial data content in a database according to an exemplary embodiment; FIG. 3 is a schematic diagram illustrating a configuration of a medical adverse event reporting system according to an exemplary embodiment; FIG. 4 is a push interface schematic diagram of a medical adverse event reporting system, according to an example embodiment; FIG. 5 is a schematic diagram of a reporting interface of a medical adverse event rep