CN-115331765-B - Quality inspection method, device and equipment for consistency of discharge diagnosis and treatment process

CN115331765BCN 115331765 BCN115331765 BCN 115331765BCN-115331765-B

Abstract

The application discloses a quality inspection method, a device and equipment for consistency of discharge diagnosis and treatment processes, the method comprises the steps of firstly determining a target medical record to be inspected, acquiring a target discharge diagnosis and a target diagnosis and treatment process from the target medical record, then acquiring a typical treatment medical record corresponding to the target discharge diagnosis from a pre-constructed typical treatment medical record library, and acquiring a typical diagnosis and treatment process from the typical treatment medical record; and then calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process, and judging the consistency of the target discharge diagnosis and the target diagnosis and treatment process according to the calculation result of the similarity. Therefore, the similarity calculation can be carried out on the target diagnosis and treatment process of the target medical record to be inspected and the typical diagnosis and treatment process of the typical treatment medical record corresponding to the target discharge diagnosis in the medical record library by utilizing the pre-constructed typical treatment medical record library, and the quality inspection result with higher accuracy can be obtained according to the similarity calculation result.

Inventors

YOU XINXIN
LIU XIEN
ZHOU KAIYIN
WU JI

Assignees

北京惠及智医科技有限公司

Dates

Publication Date: 20260508
Application Date: 20220810

Claims (8)

1. A method for quality testing of consistency of discharge diagnosis and treatment processes, comprising: Determining a target medical record to be inspected, and acquiring a target discharge diagnosis and a target diagnosis and treatment process from the target medical record; Acquiring a typical treatment medical record corresponding to the target discharge diagnosis from a pre-constructed typical treatment medical record library, and acquiring a typical diagnosis and treatment process from the typical treatment medical record; calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process; Judging the consistency of the target discharge diagnosis and the target diagnosis and treatment process according to the calculation result of the similarity; the calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process comprises the following steps: Performing word segmentation processing on the typical diagnosis and treatment process to obtain each word segmentation word contained in the typical diagnosis and treatment process; traversing each word-segmentation word contained in the typical diagnosis and treatment process through a sliding window, and calculating the dependency relationship among each word-segmentation word by utilizing a mutual information PMI formula; Generating a context graph corresponding to the typical diagnosis and treatment process according to the dependency relationship among the word segmentation words; Word segmentation processing is carried out on the target diagnosis and treatment process and the typical diagnosis and treatment process, so that word segmentation words contained in the target diagnosis and treatment process and the typical diagnosis and treatment process are respectively obtained; calculating word vectors corresponding to each word segmentation word contained in the target diagnosis and treatment process and the typical diagnosis and treatment process respectively; Performing pairwise similarity calculation between word vectors corresponding to each word segmentation word contained in the target diagnosis and treatment process and word vectors corresponding to each word segmentation word contained in the typical diagnosis and treatment process to obtain an adjacency matrix of a heterogeneous graph corresponding to the target diagnosis and treatment process; The context diagram corresponding to the typical diagnosis and treatment process and the heterogeneous diagram corresponding to the target diagnosis and treatment process are input to a pre-constructed diagram network quality inspection model together, the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process is obtained through identification, and the diagram network quality inspection model is obtained through comparison training by using a loss function according to positive sample medical record data and negative sample medical record data.
2. The method of claim 1, wherein the graph network quality inspection model is constructed as follows: Constructing training sample medical record data, wherein the training sample medical record data comprises positive sample medical record data and negative sample medical record data; Constructing a training context diagram and a training heterogram corresponding to the diagnosis and treatment process in the training sample medical record data; Inputting a training context graph and a training heterogeneous graph corresponding to the diagnosis and treatment process in the training sample medical record data into an initial graph network quality inspection model, and training to obtain the graph network quality inspection model by adjusting a loss function; the initial graph network quality inspection model comprises a graph convolution layer, a new graph construction layer and a characteristic output layer.
3. The method of claim 1 or 2, wherein the loss function is a hinge loss function, the loss function being used to shorten the distance between positive sample medical record data representations and to lengthen the distance between negative sample medical record data representations.
4. The method according to claim 2, wherein the method further comprises: acquiring a test medical record, and acquiring a test discharge diagnosis and a test diagnosis and treatment process from the test medical record; acquiring a test typical treatment medical record corresponding to the test discharge diagnosis from a pre-constructed typical treatment medical record library, and acquiring a test typical diagnosis and treatment process from the test typical treatment medical record; building a context diagram corresponding to the test typical diagnosis and treatment process and building a heterogeneous diagram corresponding to the test typical diagnosis and treatment process; the context graph corresponding to the test typical diagnosis and treatment process and the heterogeneous graph corresponding to the test typical diagnosis and treatment process are input to the graph network quality inspection model together, and the test similarity between the test diagnosis and treatment process and the test typical diagnosis and treatment process is obtained through recognition; And when the test similarity does not exceed the preset threshold, the test medical record is re-used as a training sample medical record, and the graph network quality inspection model is updated.
5. The method of claim 1, wherein the graph network quality inspection model comprises a first layer of graph roll lamination, a first layer of new graph construction, a second layer of graph convolution, a second layer of new graph construction and a feature readout layer, wherein the step of inputting the context graph corresponding to the typical diagnosis and treatment process and the heterogeneous graph corresponding to the target diagnosis and treatment process to a pre-constructed graph network quality inspection model together, and the step of identifying the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process comprises the steps of: The context graph corresponding to the typical diagnosis and treatment process and the heterogeneous graph corresponding to the target diagnosis and treatment process are input to the first layer graph convolution layer of the graph network quality inspection model together for convolution processing, and a first initial feature vector after convolution processing is obtained; inputting the first initial feature vector into the first new graph construction layer, and extracting the features of the core words to obtain a first feature vector; Inputting the first feature vector into the second layer graph convolution layer to carry out convolution processing to obtain a second initial feature vector after convolution processing; inputting the second initial feature vector into the second new graph construction layer, and extracting the features of the core words again to obtain a second feature vector; Splicing the feature vector, the first feature vector and the second feature vector corresponding to the heterogeneous graph corresponding to the typical diagnosis and treatment process to obtain a spliced feature vector, and inputting the spliced feature vector into the feature reading layer to perform dimension reduction processing to obtain a dimension reduced feature vector; and determining the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process according to the feature vector after dimension reduction.
6. A discharge diagnosis and treatment process consistency quality inspection device, comprising: The first acquisition unit is used for determining a target medical record to be inspected and acquiring a target discharge diagnosis and a target diagnosis and treatment process from the target medical record; the second acquisition unit is used for acquiring typical treatment medical records corresponding to the target discharge diagnosis from a pre-constructed typical treatment medical record library and acquiring a typical diagnosis and treatment process from the typical treatment medical records; The calculating unit is used for calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process; the quality inspection unit is used for judging the consistency of the target discharge diagnosis and the target diagnosis and treatment process according to the calculation result of the similarity; The calculation unit includes: The first word segmentation subunit is used for carrying out word segmentation processing on the typical diagnosis and treatment process to obtain each word segmentation word contained in the typical diagnosis and treatment process; the first calculation subunit is used for traversing each word-segmentation word contained in the typical diagnosis and treatment process through a sliding window and calculating the dependency relationship among the word-segmentation words by utilizing a mutual information PMI formula; The generation subunit is used for generating a context graph corresponding to the typical diagnosis and treatment process according to the dependency relationship among the word segmentation words; The second word segmentation subunit is used for carrying out word segmentation processing on the target diagnosis and treatment process and the typical diagnosis and treatment process to respectively obtain word segmentation words contained in the target diagnosis and treatment process and the typical diagnosis and treatment process; The second calculating subunit is used for calculating word vectors corresponding to each word segmentation word contained in each of the target diagnosis and treatment process and the typical diagnosis and treatment process; A third calculation subunit, configured to perform similarity calculation between word vectors corresponding to each word segmentation word included in the target diagnosis and treatment process and word vectors corresponding to each word segmentation word included in the typical diagnosis and treatment process, so as to obtain an adjacency matrix of a heterogeneous graph corresponding to the target diagnosis and treatment process and the typical diagnosis and treatment process; The identification subunit is used for inputting the context graph corresponding to the typical diagnosis and treatment process and the heterogeneous graph corresponding to the target diagnosis and treatment process to a pre-constructed graph network quality inspection model together, and identifying and obtaining the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process, wherein the graph network quality inspection model is obtained by performing contrast training by using a loss function according to positive sample medical record data and negative sample medical record data.
7. The quality inspection device for discharge diagnosis and diagnosis process consistency is characterized by comprising a processor, a memory and a system bus; The processor and the memory are connected through the system bus; The memory is for storing one or more programs, the one or more programs comprising instructions, which when executed by the processor, cause the processor to perform the method of any of claims 1-5.
8. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein instructions, which when run on a terminal device, cause the terminal device to perform the method of any of claims 1-5.

Description

Quality inspection method, device and equipment for consistency of discharge diagnosis and treatment process Technical Field The application relates to the technical field of intelligent medical treatment, in particular to a quality inspection method, device and equipment for consistency of discharge diagnosis and treatment processes. Background "Discharge diagnosis" in medical records generally refers to the first data in the discharge diagnosis list, i.e., discharge first diagnosis, which is the disease that the patient is primarily treating for the primary medical purpose during the present hospital stay. Meanwhile, the hospitalization data of the patient in the medical record also comprises an "discharge record", wherein the diagnosis and treatment process field completely records the diagnosis and treatment process of the patient in this time. Whether the treatment process in the discharge diagnosis and diagnosis process can be accurately and consistently checked, the purpose of checking the discharge diagnosis rationality is achieved, and the method is particularly important for effectively developing and implementing the payment of the disease Diagnosis Related Group (DRG). At present, two quality inspection methods for consistency of discharge diagnosis and treatment processes in medical records are usually adopted, namely, manual quality inspection is adopted, the method can ensure very high accuracy, but has the defects of high labor cost, low quality inspection speed, low efficiency and the like, and the other common quality inspection method is adopted for quality inspection based on rules of a knowledge base, and the method solves the problem of partial manual quality inspection, but the traditional method only has a shallow knowledge base and has less coverage because of the research and the less attention of the layer, so that the quality inspection is basically carried out according to the knowledge base, only a few part of the related treatment methods can be covered, and the final quality inspection result is not accurate enough. Disclosure of Invention The embodiment of the application mainly aims to provide a quality inspection method, device and equipment for consistency of discharge diagnosis and treatment processes, which can improve accuracy of quality inspection results, thereby providing technical support for effective development and implementation of DRG payment. The embodiment of the application provides a quality inspection method for consistency of discharge diagnosis and treatment processes, which comprises the following steps: Determining a target medical record to be inspected, and acquiring a target discharge diagnosis and a target diagnosis and treatment process from the target medical record; Acquiring a typical treatment medical record corresponding to the target discharge diagnosis from a pre-constructed typical treatment medical record library, and acquiring a typical diagnosis and treatment process from the typical treatment medical record; calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process; And judging the consistency of the target discharge diagnosis and the target diagnosis and treatment process according to the calculation result of the similarity. In a possible implementation manner, the calculating the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process includes: Constructing a context diagram corresponding to the typical diagnosis and treatment process, and constructing a heterogeneous diagram corresponding to the target diagnosis and treatment process and the typical diagnosis and treatment process; The context diagram corresponding to the typical diagnosis and treatment process and the heterogeneous diagram corresponding to the target diagnosis and treatment process are input to a pre-constructed diagram network quality inspection model together, the similarity between the target diagnosis and treatment process and the typical diagnosis and treatment process is obtained through identification, and the diagram network quality inspection model is obtained through comparison training by using a loss function according to positive sample medical record data and negative sample medical record data. In a possible implementation manner, the graph network quality inspection model is constructed as follows: Constructing training sample medical record data, wherein the training sample medical record data comprises positive sample medical record data and negative sample medical record data; Constructing a training context diagram and a training heterogram corresponding to the diagnosis and treatment process in the training sample medical record data; Inputting a training context graph and a training heterogeneous graph corresponding to the diagnosis and treatment process in the training sample medical record data into an initial graph net