A method and device for constructing a clinical pathway for a single disease

Abstract

The present invention discloses a method and device for constructing a clinical pathway for a single disease, and belongs to the technical field of medical systems. The present invention first obtains the medical record information of all patients in the hospital, and screens out medical records that meet the requirements; classifies the medical records that meet the requirements according to the name of the disease, and if the number of patients with a certain disease is higher than a preset threshold, the disease is used as a single disease for which a clinical pathway is to be constructed; obtains the historical medical text data and historical diagnosis and treatment process of the patient corresponding to the single disease, and extracts text features and time series data; constructs a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; then inputs the text features into the deep learning model to obtain the key features of the single disease; finally, based on the key features, generates a clinical pathway for a single disease, and stores it in a designated database. The present invention can gradually increase the number of single diseases in the hospital itself, reduce the waste of medical resources, improve the quality level of hospital diagnosis and treatment, and reduce the medical expenses of patients; since there is no human involvement in the process, it can also reduce management costs, so that it can be applied on a large scale.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical systems, and specifically relates to a method and device for constructing a clinical pathway for a single disease. Background Art

[0002] Single disease refers to a specific disease or disease type, which usually has clear diagnostic criteria and treatment plans. Long-term practice has proved that single disease quality control is an effective means and entry point for improving medical quality.

[0003] However, in the actual diagnosis and treatment process, in addition to the 51 single diseases specified, there are other diseases that meet the single disease conditions that have not been included in the standardized management. As a result, the diagnosis and treatment plans for the same disease are different, which also causes unstable medical quality and high and low charges. Therefore, gradually increasing the number of single diseases and establishing corresponding clinical management pathways play an important role in reducing the waste of medical resources, improving the level of hospital management, and reducing patients' medical expenses. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for constructing a clinical pathway for a single disease, for diseases that meet the conditions of a single disease but have not been included in standardized management.

[0005] The technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for constructing a clinical pathway for a single disease, comprising the following steps:

[0007] S1. Obtain the medical records of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria; the medical records include personal information, disease name, symptoms and historical diagnosis and treatment process;

[0008] S2. Classify the medical records that meet the requirements obtained in S1 according to the name of the disease; if the number of patients with a certain disease is higher than the preset threshold, the disease will be used as a single disease for the construction of the clinical pathway;

[0009] S3, obtaining historical medical text data of patients corresponding to the single disease, and extracting text features in the historical medical text data;

[0010] S4, obtaining the historical diagnosis and treatment process of the patient corresponding to the single disease, and extracting time series data from the historical diagnosis and treatment process;

[0011] S5. Construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; input the text features into the deep learning model to obtain the key features of the single disease;

[0012] S6. Generate a clinical pathway for the single disease based on the key features and store it in a designated database.

[0013] Furthermore, generating a clinical pathway for a single disease based on the key features also includes:

[0014] A1. Determine the frequency of occurrence of the key features according to the historical diagnosis and treatment process and the historical medical text data; sort the frequency of occurrence of the key features, and determine the weight of each key feature according to the sorting result;

[0015] A2. Retrieve the actual diagnosis and treatment process of the patient corresponding to the single disease at the current moment, and determine whether the actual diagnosis and treatment process contains the key feature according to the weight of each key feature;

[0016] A3. If any key feature is missing in the actual diagnosis and treatment process, the weight of the missing key feature is retrieved, and based on the weight, it is determined whether the missing key feature is a necessary feature; if the missing key feature is a necessary feature, a clinical pathway corresponding to the actual diagnosis and treatment process is generated and stored in the designated database.

[0017] Furthermore, in step S1, the exclusion criteria include:

[0018] B1. Eliminate medical records of 51 types of single diseases stipulated by the state;

[0019] B2. Eliminate suspected diagnosis medical records;

[0020] B3. Eliminate medical records of patients who were treated for two or more diseases at the same time;

[0021] B4. Eliminate medical records of patients who have been transferred to other departments during hospitalization;

[0022] B5. Eliminate medical records of deaths within 48 hours of admission;

[0023] B6. Eliminate medical records of patients who were diagnosed but discharged without completing normal treatment.

[0024] Furthermore, in step S3, extracting text features from the historical medical text data includes:

[0025] C1. Establishing a medical data map based on the historical medical text data;

[0026] C2. Determine the semantic structure of the historical medical text data according to the medical data graph;

[0027] C3. Classifying the historical medical text data according to preset categories based on the semantic structure;

[0028] C4: Build a medical topic model, extract features from all historical medical text data in each category obtained in C3, and obtain text features corresponding to all historical medical text data in each category.

[0029] Furthermore, in step S4, extracting time series data from the historical diagnosis and treatment process includes:

[0030] Analyze the historical diagnosis and treatment process, determine the diagnosis and treatment stage, and determine the corresponding time window according to the diagnosis and treatment stage;

[0031] Analyze the time nodes in the time window to determine whether there is any medical treatment behavior in each time node; if there is any medical treatment behavior in the corresponding time node, extract the corresponding time node as time series data; if there is no medical treatment behavior in any time node, eliminate the time node without medical treatment behavior.

[0032] In a second aspect, the present invention provides a clinical pathway construction device for a single disease, comprising: a medical record screening module, a single disease screening module, a text feature extraction module, a time series data extraction module, a key feature acquisition module, and a single disease generation module;

[0033] The medical record screening module is used to obtain the medical record information of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria;

[0034] The single disease screening module is used to classify the medical records screened by the medical record screening module according to the disease name; if the number of patients with a certain disease is higher than a preset threshold, the disease is screened out as a single disease for which a clinical pathway is to be constructed;

[0035] The text feature extraction module is used to obtain the historical medical text data corresponding to the single disease type and extract text features in the historical medical text data;

[0036] The time series data extraction module is used to obtain the historical diagnosis and treatment process corresponding to the single disease, and extract time series data from the historical diagnosis and treatment process;

[0037] The key feature acquisition module is used to construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data, and input the text features to obtain the key features of the single disease;

[0038] The single disease generation module is used to generate a clinical pathway for a single disease based on the key features and store it in a designated database.

[0039] The solution provided by the present invention first obtains the medical record information of all patients in the hospital, and screens the medical record information that meets the requirements according to the exclusion criteria; classifies all medical records that meet the requirements according to the name of the disease, and if the number of patients with a certain disease is higher than a preset threshold, the disease is used as a single disease for which a clinical pathway is to be constructed; obtains the historical medical text data and historical diagnosis and treatment process of the patient corresponding to the single disease, extracts text features in the historical medical text data, and extracts time series data in the historical diagnosis and treatment process; constructs a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; then inputs the text features into the deep learning model to obtain the key features of the single disease; finally, generates a clinical pathway for the single disease based on the key features, and stores it in a designated database.

[0040] The present invention first screens out the single diseases that meet the requirements, and then constructs the clinical pathways of the single diseases through the patient's historical diagnosis and treatment process and historical medical text data, which can gradually increase the number of single diseases in the hospital itself, reduce the waste of medical resources, improve the quality of hospital diagnosis and treatment, and reduce the medical expenses of patients; compared with processing all data, the present invention reduces the amount of data processing, making the construction of clinical pathways more responsive and efficient. In addition, since there is no human involvement in the process, it can also reduce management costs, making it applicable on a large scale. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of a method for constructing a clinical pathway for a single disease provided in an embodiment;

[0042] Figure 2 A schematic diagram of the structure of a clinical pathway construction device for a single disease provided in an embodiment. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are 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.

[0044] This embodiment provides a method for constructing a clinical pathway for a single disease, such as Figure 1 As shown, the following steps are included:

[0045] S1. Obtain the medical records of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria; the medical records include personal information, disease name, symptoms and historical diagnosis and treatment process.

[0046] Specifically, the exclusion criteria include:

[0047] B1. Eliminate medical records of 51 types of single diseases stipulated by the state;

[0048] B2. Eliminate suspected diagnosis medical records;

[0049] B3. Eliminate medical records of patients who were treated for two or more diseases at the same time;

[0050] B4. Eliminate medical records of patients who have been transferred to other departments during hospitalization;

[0051] B5. Eliminate medical records of deaths within 48 hours of admission;

[0052] B6. Eliminate medical records of patients who were diagnosed but discharged without completing normal treatment.

[0053] S2. Classify the medical records that meet the requirements obtained in S1 according to the name of the disease; if the number of patients with a certain disease is higher than the preset threshold, the disease will be taken as a single disease for which the clinical pathway is to be constructed.

[0054] The preset threshold can be set according to actual conditions. For example, when the number of patients with the disease exceeds 500, the disease will be considered as a single disease for which a clinical pathway is to be constructed.

[0055] S3, obtaining historical medical text data of patients corresponding to the single disease, and extracting text features in the historical medical text data;

[0056] Specifically, extracting text features from the historical medical text data includes:

[0057] C1. Establishing a medical data map based on the historical medical text data;

[0058] C2. Determine the semantic structure of the historical medical text data according to the medical data graph;

[0059] C3. Classifying the historical medical text data according to preset categories based on the semantic structure;

[0060] C4: Build a medical topic model, extract features from all historical medical text data in each category obtained in C3, and obtain text features corresponding to all historical medical text data in each category.

[0061] Historical medical text data refers to text data such as medical records and medical reports left by patients who have completed their medical treatment.

[0062] Since the information and data involved in the hospital are diverse, in order to take all the information and data into consideration more comprehensively, a medical data map is established to enrich the semantic expression of certain words. By using the upper and lower relationships and upper and lower information formed in the process of establishing the medical data map, a hierarchical structure is established to obtain the semantic structure of historical medical text data.

[0063] The preset category refers to the category corresponding to all the categories involved in the hospital. In order to make the classification more accurate, the category can be set from different angles. Based on the semantic structure of the historical medical text data determined above, the historical medical text data is mapped to extract the information in the historical medical text data, and it is matched one by one to the corresponding preset category to achieve classification.

[0064] In addition, a medical topic model based on deep learning is constructed. By inputting a large amount of medical text data of different categories and corresponding text features into this medical topic model, the medical topic model is trained so that this medical topic model can recognize medical text data of different categories and output corresponding text features. After the historical medical text data is classified according to the preset categories in the above manner, it is input into this trained medical topic model to directly obtain the corresponding text features.

[0065] S4, obtaining the historical diagnosis and treatment process of the patient corresponding to the single disease, and extracting time series data from the historical diagnosis and treatment process;

[0066] Specifically, extracting time series data from the historical diagnosis and treatment process includes:

[0067] Analyze the historical diagnosis and treatment process, determine the diagnosis and treatment stage, and determine the corresponding time window based on the diagnosis and treatment stage; analyze the time nodes in the time window to determine whether there is diagnosis and treatment behavior in each time node; if there is diagnosis and treatment behavior in the corresponding time node, extract the corresponding time node as time series data; if there is no diagnosis and treatment behavior in any time node, eliminate the time node where there is no diagnosis and treatment behavior.

[0068] Since each patient may have different stages of treatment in the hospital, such as observation period, surgery period, recovery period, etc., these different stages can be understood as the diagnosis and treatment stages of the patient.

[0069] Diagnostic and treatment behavior refers to the specific diagnostic and treatment actions performed by the patient's attending physician on the patient. If the patient is only comforted or inspected to understand the situation, it may not be considered a diagnostic and treatment behavior.

[0070] Analyze the diagnosis and treatment content in the historical diagnosis and treatment process to determine several diagnosis and treatment stages. Determine the time window corresponding to each diagnosis and treatment stage based on the start time and the corresponding end time of the diagnosis and treatment stage. Analyze the corresponding time nodes one by one in each time window to determine whether there is any diagnosis and treatment behavior in each recorded time node. If there is any diagnosis and treatment behavior in the corresponding time node, extract the corresponding time node as time series data. If there is no diagnosis and treatment behavior in the corresponding time node, this time node can be ignored. In order to facilitate subsequent applications, this time node can be removed to avoid data confusion caused by confusion.

[0071] S5. Construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; input the text features into the deep learning model to obtain the key features of the single disease;

[0072] S6. Generate a clinical pathway for the single disease based on the key features and store it in a designated database.

[0073] Specifically, based on the key features, a clinical pathway for a single disease is generated, which also includes:

[0074] A1. Determine the frequency of occurrence of the key features according to the historical diagnosis and treatment process and the historical medical text data; sort the frequency of occurrence of the key features, and determine the weight of each key feature according to the sorting result;

[0075] A2. Retrieve the actual diagnosis and treatment process of the patient corresponding to the single disease at the current moment, and determine whether the actual diagnosis and treatment process contains the key feature according to the weight of each key feature;

[0076] A3. If any key feature is missing in the actual diagnosis and treatment process, the weight of the missing key feature is retrieved, and based on the weight, it is determined whether the missing key feature is a necessary feature; if the missing key feature is a necessary feature, a clinical pathway corresponding to the actual diagnosis and treatment process is generated and stored in the designated database.

[0077] The more frequently a key feature appears, the more likely it is that there will be corresponding features for different patients. In this case, this key feature can be considered an indispensable feature. According to this situation, all key features are sorted according to the frequency of their occurrence, and the weights of the key features are determined in turn. According to the obtained weights, the key features are divided into necessary features and general features. Among them, necessary features can be considered as features that are indispensable during the diagnosis and treatment process; general features can be considered as features that are more common during the diagnosis and treatment process and do not play a key role in the patient's recovery.

[0078] In the actual diagnosis and treatment process, there may be a situation where the actual situation does not match the planning of the clinical pathway. In this case, the actual diagnosis and treatment process corresponding to the patient of the clinical pathway can be retrieved, and then the actual diagnosis and treatment process is analyzed to see whether it lacks key features. If so, it is determined whether the missing feature is a necessary feature. If so, it may mean that the patient does not involve this necessary feature in the actual diagnosis and treatment process. In this case, the clinical pathway can be adjusted according to the actual diagnosis and treatment process, and a clinical pathway corresponding to the actual diagnosis and treatment process can be generated and stored in the designated database for subsequent application to the same type of patients.

[0079] This embodiment also provides a clinical pathway construction device for a single disease, such as Figure 2 As shown, it includes: medical record screening module, single disease screening module, text feature extraction module, time series data extraction module, key feature acquisition module, single disease generation module;

[0080] The medical record screening module is used to obtain the medical record information of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria;

[0081] The single disease screening module is used to classify the medical records screened by the medical record screening module according to the disease name; if the number of patients with a certain disease is higher than a preset threshold, the disease is screened out as a single disease for which a clinical pathway is to be constructed;

[0082] The text feature extraction module is used to obtain the historical medical text data corresponding to the single disease type and extract text features in the historical medical text data;

[0083] The time series data extraction module is used to obtain the historical diagnosis and treatment process corresponding to the single disease, and extract time series data from the historical diagnosis and treatment process;

[0084] The key feature acquisition module is used to construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data, and input the text features to obtain the key features of the single disease;

[0085] The single disease generation module is used to generate a clinical pathway for a single disease based on the key features and store it in a designated database.

[0086] The solution provided in this embodiment first obtains the medical records of all patients in the hospital, and screens the medical records that meet the requirements according to the exclusion criteria; classifies all medical records that meet the requirements according to the disease name, and if the number of patients with a certain disease is higher than the preset threshold, the disease is used as a single disease for which a clinical pathway is to be constructed; obtains the historical medical text data and historical diagnosis and treatment process of the patient corresponding to the single disease, and extracts the text features in the historical medical text data, and extracts the time series data in the historical diagnosis and treatment process; constructs a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; then inputs the text features into the deep learning model to obtain the key features of the single disease; finally, generates a clinical pathway for the single disease based on the key features, and stores it in a designated database. Through the technical solution of the present invention, the number of single diseases in the hospital itself can be gradually increased, the waste of medical resources can be reduced, the quality of hospital diagnosis and treatment can be improved, and the medical expenses of patients can be reduced.

Claims

1. A method for constructing a clinical pathway for a single disease, characterized in that: The following steps are involved: S1. Obtain the medical records of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria; the medical records include personal information, disease name, symptoms and historical diagnosis and treatment process; S2. Classify the medical records that meet the requirements obtained in S1 according to the name of the disease; if the number of patients with a certain disease is higher than the preset threshold, the disease will be used as a single disease for the construction of the clinical pathway; S3, obtaining historical medical text data of patients corresponding to the single disease, and extracting text features in the historical medical text data; S4, obtaining the historical diagnosis and treatment process of the patient corresponding to the single disease, and extracting time series data from the historical diagnosis and treatment process; S5. Construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data; input the text features into the deep learning model to obtain the key features of the single disease; S6. Generate a clinical pathway for the single disease based on the key features and store it in a designated database; In step S6, a clinical pathway for a single disease is generated according to the key features, which also includes: A1. Determine the frequency of occurrence of the key features according to the historical diagnosis and treatment process and the historical medical text data; sort the frequency of occurrence of the key features, and determine the weight of each key feature according to the sorting result; A2. Retrieve the actual diagnosis and treatment process of the patient corresponding to the single disease at the current moment, and determine whether the actual diagnosis and treatment process contains the key feature according to the weight of each key feature; A3. If any key feature is missing in the actual diagnosis and treatment process, the weight of the missing key feature is retrieved, and based on the weight, it is determined whether the missing key feature is a necessary feature; if the missing key feature is a necessary feature, a clinical pathway corresponding to the actual diagnosis and treatment process is generated and stored in a designated database; In step S1, the exclusion criteria include: B1. Eliminate medical records of 51 types of single diseases stipulated by the state; B2. Eliminate suspected diagnosis medical records; B3. Eliminate medical records of patients who were treated for two or more diseases at the same time; B4. Eliminate medical records of patients who have been transferred to other departments during hospitalization; B5. Eliminate medical records of deaths within 48 hours of admission; B6. Eliminate medical records of patients who were diagnosed but discharged without completing normal treatment.

2. The method for constructing a clinical pathway for a single disease according to claim 1, characterized in that: In step S3, extracting text features from the historical medical text data includes: C1. Establishing a medical data map based on the historical medical text data; C2. Determine the semantic structure of the historical medical text data according to the medical data graph; C3. Classifying the historical medical text data according to preset categories based on the semantic structure; C4: Build a medical topic model, extract features from all historical medical text data in each category obtained in C3, and obtain text features corresponding to all historical medical text data in each category.

3. The method for constructing a clinical pathway for a single disease as claimed in claim 2, characterized in that: In step S4, extracting time series data from the historical diagnosis and treatment process includes: Analyze the historical diagnosis and treatment process, determine the diagnosis and treatment stage, and determine the corresponding time window according to the diagnosis and treatment stage; Analyze the time nodes in the time window to determine whether there is any medical treatment behavior in each time node; if there is any medical treatment behavior in the corresponding time node, extract the corresponding time node as time series data; if there is no medical treatment behavior in any time node, eliminate the time node without medical treatment behavior.

4. The method for constructing a clinical pathway for a single disease as claimed in claim 3, characterized in that: The device for implementing the clinical pathway construction method for a single disease includes: a medical record screening module, a single disease screening module, a text feature extraction module, a time series data extraction module, a key feature acquisition module, and a single disease generation module; The medical record screening module is used to obtain the medical record information of all patients in the hospital and screen the medical records that meet the requirements according to the exclusion criteria; The single disease screening module is used to classify the medical records screened by the medical record screening module according to the disease name; if the number of patients with a certain disease is higher than a preset threshold, the disease is screened out as a single disease for which a clinical pathway is to be constructed; The text feature extraction module is used to obtain the historical medical text data corresponding to the single disease type and extract text features in the historical medical text data; The time series data extraction module is used to obtain the historical diagnosis and treatment process corresponding to the single disease, and extract time series data from the historical diagnosis and treatment process; The key feature acquisition module is used to construct a deep learning model based on the time series data and the diagnosis and treatment behavior corresponding to the time series data, and input the text features to obtain the key features of the single disease; The single disease generation module is used to generate a clinical pathway for a single disease based on the key features and store it in a designated database.

Images