Method and device for screening risk factors for venous thromboembolism in orthopedic trauma patients

Abstract

The present application relates to the medical-related technical field, and specifically to a method and device for screening risk factors for venous thromboembolism in orthopedic trauma patients. The method includes: obtaining original data; based on the original data, combining multiple data-driven feature screening methods to screen a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients; based on literature review, using a meta-analysis method to determine a second group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients; integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and preset algorithms, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph; deleting the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients, and obtaining the target risk factors.

Description

Technical Field

[0001] The present application relates to the field of medical-related technology, and specifically to a method and device for screening risk factors for venous thromboembolism in orthopedic trauma patients. Background Art

[0002] The incidence of venous thromboembolism (VTE) in orthopedic trauma patients is quite high, which seriously threatens the health and life of patients and has become the focus of global medical attention. Effective identification of VTE risk factors, accurate screening and implementation of intervention are of great significance to improving patient prognosis and reducing mortality.

[0003] However, the method of determining risk factors is mainly manual analysis, which is subject to greater subjective influence and the determined risk factors are unreasonable. Summary of the invention

[0004] In view of this, the embodiments of the present application are directed to providing a method and device for screening risk factors for venous thromboembolism in orthopedic trauma patients, so as to determine the risk factors for venous thromboembolism in orthopedic trauma patients.

[0005] The present application provides a method for screening risk factors for venous thromboembolism in orthopedic trauma patients, comprising:

[0006] Get the original data;

[0007] Based on the raw data, combining multiple data-driven feature screening methods, screening a first set of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients;

[0008] Based on the literature review, a meta-analysis approach was used to identify a second set of candidate risk factors associated with VTE in orthopedic trauma patients;

[0009] Integrate the first group of candidate risk factors and the second group of candidate risk factors, combine expert experience, based on causal theory and preset algorithms, identify the causal relationship of the candidate risk factors, and draw a directed acyclic graph;

[0010] Delete the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients to obtain target risk factors.

[0011] In some embodiments, the raw data includes: general information, injury characteristics, and post-injury characteristics;

[0012] The general information included: gender, age, height, weight, body mass index, admission time, smoking history, drinking history, and comorbidities;

[0013] The injury characteristics include: fracture site, injury mechanism, injury severity score, time from injury to hospital admission, whether it is multiple injury, whether it is accompanied by shock, and vital signs at admission;

[0014] The post-injury characteristics include: drug prevention, mechanical prevention, number of surgeries, number of red suspension infusions, and whether or not the patient was admitted to the ICU.

[0015] In some embodiments, the multiple data-driven feature screening methods include: Pearson coefficient, Spearman coefficient, distance correlation coefficient, mutual information, maximum information coefficient, average accuracy reduction and penalized regression method.

[0016] In some embodiments, the first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients are screened based on the raw data in combination with multiple data-driven feature screening methods, including:

[0017] Based on the preset multiple data-driven feature screening methods, screening is performed respectively to obtain a feature set corresponding to each data-driven feature screening method;

[0018] Each feature set is counted and screened using the voting method to obtain the first set of candidate risk factors.

[0019] In some embodiments, the second group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients is determined based on literature review using a meta-analysis method, including:

[0020] The databases were searched for literature related to orthopedic trauma, fractures, deep vein thrombosis, venous thromboembolism, and risk factors;

[0021] The adjusted odds ratios and confidence intervals of each risk factor were extracted based on the data provided in the literature;

[0022] A random effects model was used to merge and screen out risk factors that had a definite impact on venous thromboembolism to obtain the second group of candidate factors.

[0023] In some embodiments, the step of integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and a preset algorithm, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph includes:

[0024] Construct an implicit graph containing all candidate risk factors and perform simple links on all candidate risk factors;

[0025] Identify the causal relationship of candidate risk factors based on causal theory and preset algorithms;

[0026] Based on the causal relationship, the implicit graph is modified, and the directions of the links are added to obtain a directed acyclic graph.

[0027] The present application also provides a device for screening risk factors of venous thromboembolism in orthopedic trauma patients, comprising:

[0028] An acquisition module is used to obtain raw data;

[0029] A screening module, for screening a first group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients based on the raw data and in combination with a plurality of data-driven feature screening methods;

[0030] A module was developed to identify a second set of candidate risk factors associated with VTE in orthopedic trauma patients based on a literature review using a meta-analysis approach;

[0031] An integration module is used to integrate the first group of candidate risk factors and the second group of candidate risk factors, combine expert experience, based on causal theory and preset algorithms, identify the causal relationship of the candidate risk factors, and draw a directed acyclic graph; delete the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients to obtain target risk factors.

[0032] The present application also provides an electronic device, including:

[0033] A processor, and a memory for storing a program executable by the processor;

[0034] The processor is used to implement the above-mentioned method for screening risk factors of venous thromboembolism in orthopedic trauma patients by running the program in the memory.

[0035] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes the above-mentioned method for screening risk factors for venous thromboembolism in orthopedic trauma patients.

[0036] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned method for screening risk factors for venous thromboembolism in orthopedic trauma patients.

[0037] The present application provides a method for screening risk factors for venous thromboembolism in orthopedic trauma patients, firstly obtaining original data; based on the original data, combining multiple data-driven feature screening methods, screening a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients; based on literature review, using a meta-analysis method, determining a second group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients; integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and preset algorithms, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph; deleting the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients, and obtaining the target risk factors. In this way, in the data-driven method, multiple methods are used to screen variables, and finally the voting method is used to select the optimal risk factor, which helps to ensure the robustness of the results. Combining the data-driven method with literature review, expert experience, and drawing a causal graph to achieve the screening of influencing factors can ensure the simplicity and accuracy of the final screening of risk factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0039] Figure 1 It is a flowchart of a method for screening risk factors for venous thromboembolism in orthopedic trauma patients provided in one embodiment of the present application.

[0040] Figure 2 It is a partial flow chart of a method provided by an embodiment of the present application.

[0041] Figure 3 It is a directed acyclic graph of a method provided by an embodiment of the present application.

[0042] Figure 4 It is a schematic diagram of the structure of a device for screening risk factors for venous thromboembolism in orthopedic trauma patients provided by one embodiment of the present application.

[0043] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Venous thromboembolism is a common complication in patients with orthopedic trauma. Effective prevention of venous thromboembolism is an integral part of the national medical quality improvement goals. Identifying risk factors associated with venous thromboembolism is an important prerequisite for assessing the risk of thrombosis in patients with orthopedic trauma and providing targeted prevention. Thrombosis risk factors in patients with orthopedic trauma can be divided into three main categories: one is the patient's basic conditions such as gender, age, and comorbidities; one is the characteristic factors of trauma, such as injury site, injury energy, and severity; and one is post-injury factors, which are mainly corresponding treatment interventions after injury. The above factors are numerous and have complex interactions. For example, age may be associated with chronic comorbidities, and injury site and injury severity will also affect post-injury treatment and intervention. How to screen out the most important risk factors that have a definite causal relationship with thrombosis from many risk factors is an important and unresolved issue.

[0046] At present, there are three main types of risk factor screening methods in the risk assessment of venous thromboembolism in patients with orthopedic trauma: one is based on medical data, combined with or without a prediction model, to determine which thrombosis-related risk factors to select according to the degree of influence of risk factors on thrombosis; one is based on previous literature research, integrating existing risk factors in the literature to select risk factors; one is based on expert opinions, brainstorming or expert meetings based on the clinical experience of experts, and selecting risk factors that most experts agree on. The above methods may be used alone or in combination with two or more methods. However, when the above methods are used alone or in combination, there is a lack of identification and verification of the causal relationship. Using these methods to screen risk factors may result in the screened risk factors being closely related to the occurrence of venous thromboembolism, but there is still mutual influence, which may lead to inaccurate evaluation results in risk assessment.

[0047] In order to solve the above problems, the present application provides a solution, which is based on identifying the causal relationship between each factor, combined with other methods to assist in the screening of risk factors, which can make the screening results more concise and accurate, and help to build a later risk assessment model or find accurate intervention factors. In this way, in the data-driven method, multiple methods are used to screen variables, and finally the voting method is used to select the optimal risk factor, which helps to ensure the robustness of the results. Combining data-driven methods with literature review, expert experience, and drawing causal diagrams to achieve the screening of influencing factors can ensure the simplicity and accuracy of the final screening of risk factors.

[0048] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0049] Figure 1 FIG. 1 is a flow chart of a method for screening risk factors for venous thromboembolism in orthopedic trauma patients provided in one embodiment of the present application. Figure 1 As shown, the method includes the following contents.

[0050] Step S110, obtaining original data;

[0051] In this step, we need to collect medical data related to orthopedic trauma patients, which will serve as the basis for subsequent analysis. The raw data include: general information, injury characteristics, and post-injury characteristics;

[0052] The general information included: gender, age, height, weight, body mass index, admission time, smoking history, drinking history, and comorbidities;

[0053] The injury characteristics include: fracture site, injury mechanism, injury severity score, time from injury to hospital admission, whether it is multiple injury, whether it is accompanied by shock, and vital signs at admission;

[0054] The post-injury characteristics include: drug prevention, mechanical prevention, number of surgeries, number of red suspension infusions, and whether or not the patient was admitted to the ICU.

[0055] This data can be extracted from a hospital's electronic health record (EHR) system or collected through clinical research. Ensuring the quality and integrity of the data is critical for subsequent analysis.

[0056] Step S120, based on the original data, combining multiple data-driven feature screening methods, screening a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients;

[0057] After obtaining the raw data, we used a variety of data-driven feature screening methods to identify factors that are closely associated with VTE. These methods include:

[0058] Statistical methods: such as Pearson correlation coefficient, Spearman correlation coefficient, etc., are used to evaluate the linear or rank correlation between variables. Information theory methods: such as mutual information, are used to measure the interdependence between variables. Model-based methods: such as random forest, Lasso regression, etc., are used to evaluate the importance of features. Through these methods, we can screen out the first group of candidate risk factors that are highly correlated with VTE risk.

[0059] Step S130, based on the literature review, using a meta-analysis method, determining a second group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients;

[0060] In this step, we systematically reviewed the existing medical literature and used meta-analysis methods to combine risk factors from different studies. This included:

[0061] Literature search: Medical databases were searched for studies related to orthopedic trauma and VTE.

[0062] Data extraction: Adjusted odds ratios (ORs) and their confidence intervals for risk factors were extracted from eligible studies.

[0063] Meta-analysis: These risk factors were combined using a random-effects model to identify a second set of candidate risk factors that have a definite impact on VTE.

[0064] Step S140, integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and a preset algorithm, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph;

[0065] The first group and the second group of candidate risk factors obtained in steps S120 and S130 are integrated, and combined with expert experience, causal theory is used to identify the causal relationship between these factors. This step involves:

[0066] Causal Diagram Drawing: Use software tools such as DAGitty to draw a directed acyclic graph (DAG) to visualize the relationship between risk factors.

[0067] Causal relationship identification: Identify and confirm direct and indirect causal relationships between risk factors based on preset algorithms and causal theories.

[0068] Step S150, deleting the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients, to obtain target risk factors.

[0069] In the directed acyclic graph, we further reviewed and verified the causal relationship between each risk factor and VTE. This step includes: Verification of causal relationship: Verify the causal relationship of each risk factor through expert review and additional statistical tests. Removal of non-causal factors: Remove those risk factors that have no direct causal relationship with VTE from the causal graph to refine the target risk factor set.

[0070] Through this series of steps, we finally obtained a set of target risk factors directly related to the risk of VTE in orthopedic trauma patients, which can be used for further risk assessment and formulation of intervention strategies.

[0071] Specifically, the multiple data-driven feature screening methods include: Pearson coefficient, Spearman coefficient, distance correlation coefficient, mutual information, maximum information coefficient, average accuracy reduction and penalized regression method.

[0072] In some embodiments, the first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients are screened based on the raw data in combination with multiple data-driven feature screening methods, including:

[0073] Based on the preset multiple data-driven feature screening methods, screening is performed respectively to obtain the feature set corresponding to each data-driven feature screening method; each feature set is counted, and the voting method is used to screen to obtain the first group of candidate risk factors.

[0074] Specifically, the process of screening based on the preset multiple data-driven feature screening methods and using the voting method to obtain the first group of candidate risk factors can be described in detail as follows:

[0075] 1. Choose a data-driven feature screening method

[0076] First, determine which data-driven feature selection methods will be used. These methods may include, but are not limited to:

[0077] Pearson Correlation Coefficient: Used to evaluate the linear relationship between two continuous variables.

[0078] Spearman Correlation Coefficient: It is used to evaluate the rank correlation between two variables and is applicable to data with non-normal distribution.

[0079] Mutual Information: Evaluates the interdependence between variables and is applicable to any type of variables.

[0080] Maximum Information Coefficient (MIC): Used to quantify the strength and form of the relationship between two variables.

[0081] Model-based feature importance: such as the Mean Decrease Accuracy (MDA) of random forest and the Lasso and Ridge methods based on penalized regression.

[0082] 2. Application feature screening method

[0083] Apply each of the above feature screening methods to the original data set to obtain the feature set corresponding to each method. For example:

[0084] Pearson correlation coefficient was used to screen out features that were highly correlated with VTE.

[0085] The Spearman correlation coefficient was used to screen out the features associated with VTE grade.

[0086] The features with the most information about VTE were identified by mutual information method.

[0087] Identify important features using MDA with random forests.

[0088] Lasso and Ridge regression were applied to identify important features, and cross-validation was used to determine the optimal penalty coefficient.

[0089] 3. Statistical feature set

[0090] For each method, record the number of times each feature is selected. This can be done by creating a counter or using a data frame, where each row represents a feature, each column represents a feature selection method, and the value in the cell represents the number of times the feature is selected in the corresponding method.

[0091] 4. Use voting method for screening

[0092] Set a threshold to determine how many methods a feature needs to be selected by in order to be considered important. This threshold can be set based on experience, for example, a feature needs to be selected by at least 3 out of 6 methods. Then:

[0093] Count the total number of times each feature is selected.

[0094] Select those features whose number of selections exceeds or equals the threshold.

[0095] 5. Get the first set of candidate risk factors

[0096] After the voting method was used to screen, the remaining feature set was the first set of candidate risk factors. These factors were considered to be highly correlated with the risk of VTE in orthopedic trauma patients and could be used for further analysis and model building.

[0097] Through this approach, we are able to combine multiple data-driven feature screening methods, reduce the bias that may be introduced by a single method, and improve the robustness and accuracy of feature selection.

[0098] Furthermore, based on the literature review, a meta-analysis method was used to identify a second group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients, including:

[0099] Literature related to orthopedic trauma, fracture, deep vein thrombosis, venous thromboembolism, and risk factors were retrieved from the database; the adjusted odds ratio and confidence interval of each risk factor were extracted based on the data provided by the literature; the random effects model was used to merge and screen out the risk factors that had a definite impact on venous thromboembolism to obtain the second group of candidate factors.

[0100] This process involves systematic literature search, data analysis and statistical synthesis. The following are the detailed steps:

[0101] Step 1: Search the database

[0102] Identify search terms: Keywords and terms related to orthopedic trauma, fractures, deep vein thrombosis (DVT), venous thromboembolism (VTE), and risk factors were selected.

[0103] Select database: Decide which medical and scientific databases you want to search, such as PubMed, EMBASE, Cochrane Library, etc.

[0104] Perform a search: Combine search terms using Boolean logic (AND, OR, NOT) to find relevant articles.

[0105] Literature screening: irrelevant literature was excluded based on the title and abstract, and then the full text was read to determine the literature that was finally included in the analysis.

[0106] Step 2: Extract Data

[0107] Extracted information: Data related to risk factors were extracted from each eligible literature, including but not limited to study design, sample size, risk factors, adjusted odds ratio (aOR) and its confidence interval (CI).

[0108] Record data: Organize the extracted data into tables for subsequent analysis.

[0109] Step 3: Pooling using a random effects model

[0110] Choose appropriate software: Use statistical software such as R, Stata, RevMan, etc. for meta-analysis.

[0111] Enter data: The extracted data were entered into the software in preparation for meta-analysis.

[0112] Model selection: Due to heterogeneity among studies, a random effects model was chosen to combine the results of individual studies.

[0113] Heterogeneity test: Heterogeneity tests (such as I² statistic and Cochran's Q test) were performed to assess the magnitude of heterogeneity among studies.

[0114] Sensitivity analysis: Sensitivity analysis was performed to assess the impact of individual studies on the pooled results.

[0115] Pooled results: A random-effects model was used to combine the aOR values ​​of each study to obtain an overall effect estimate and confidence interval for each risk factor.

[0116] Step 4: Screen risk factors for definite impact

[0117] Evaluation of confidence intervals: The confidence interval of the combined aOR for each risk factor was checked to see whether it contained 1. If it did not contain 1, the risk factor was considered to be statistically significantly associated with VTE.

[0118] Consider publication bias: Assess publication bias using methods such as funnel plots and Egger's test and correct for it as needed.

[0119] Determine the second group of candidate factors: Based on the above analysis, determine the risk factors that have a definite impact on VTE to form the second group of candidate risk factors.

[0120] Through this series of steps, we were able to systematically identify and confirm factors associated with VTE risk in orthopedic trauma patients from the existing literature, providing a scientific basis for clinical practice and further research.

[0121] In some embodiments, the step of integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and a preset algorithm, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph includes:

[0122] Construct an implicit graph containing all candidate risk factors and simply link all candidate risk factors; identify the causal relationship of candidate risk factors based on causal theory and preset algorithms; based on the causal relationship, modify the implicit graph, add the direction of the link, and obtain a directed acyclic graph.

[0123] Constructing a directed acyclic graph (DAG) to identify causal relationships between candidate risk factors is a complex process involving multiple steps. Here are the detailed steps:

[0124] Step 1: Construct the implicit graph

[0125] List all candidate risk factors: List all the first and second group candidate risk factors screened through literature review and data-driven methods.

[0126] Create nodes: Create a node for each candidate risk factor in a DAG software tool such as DAGitty.

[0127] Simple links: All nodes are simply connected without considering the causal direction, forming an implicit graph containing all candidate risk factors. These links represent possible associations between risk factors.

[0128] Step 2: Identify causal relationships

[0129] Application of causal theory: Based on known medical knowledge and causal theory, determine which risk factors may be causes and which may be effects.

[0130] Preset Algorithms: Use preset algorithms, such as those based on conditional independence tests, to identify and infer causal relationships between variables.

[0131] Expert knowledge: Combine the knowledge of domain experts to verify and adjust the algorithm results to ensure the rationality of causal relationships.

[0132] Step 3: Modify the implicit graph

[0133] Add direction to links: Based on the identified causal relationships, add direction to the links in the implicit graph to indicate the causal direction. For example, if risk factor A causes risk factor B, then the direction is from A to B.

[0134] Remove undirected links: If some links cannot be directed or if they cannot possibly be causally related according to causal theory, remove those links from the graph.

[0135] Form a directed acyclic graph: Make sure there are no cycles in the graph, that is, there is no path from any node to the node. Such a graph is called a directed acyclic graph (DAG).

[0136] Step 4: Verify and adjust the DAG

[0137] Verification: Verify the causal relationships in the DAG through additional statistical tests, such as Granger causality tests or structural equation modeling (SEM).

[0138] Adjustment: Make necessary adjustments to the DAG based on the verification results to ensure that the links in the graph correctly reflect the causal relationship between risk factors.

[0139] Iterative process: Identifying causal relationships and building DAGs may be an iterative process that requires constant adjustment and verification until a stable and accurate DAG is obtained.

[0140] Step 5: Get the final DAG

[0141] Final Review: After all adjustments are completed, a final review of the DAG is performed to ensure that all links are supported by sufficient evidence.

[0142] Explain the DAG: Explain the DAG to clarify which risk factors are causes of other factors and which are results, as well as the direct and indirect relationships between them.

[0143] Through this process, we can obtain a directed acyclic graph that clearly shows the causal relationship between VTE risk factors in orthopedic trauma patients, which is of great significance for understanding how risk factors affect the occurrence of VTE and formulating preventive measures.

[0144] Reference Figure 2 , the scheme provided by the present application is explained below in conjunction with specific embodiments: the main problem solved by the present invention is: to provide a method for screening risk factors for venous thromboembolism in patients with orthopedic trauma, which is mainly used in the medical field, based on the screening of patient medical-related data and literature review, combined with the idea of ​​causal graph, and drawing a relationship diagram between variables to screen venous thromboembolism risk factors with causal relationships, so that the obtained risk factors are more accurate and comprehensive, which is conducive to risk assessment and the implementation of targeted intervention measures.

[0145] The technical solution of the present invention comprises the following steps:

[0146] Step 1. Obtain relevant medical data of orthopedic trauma patients based on the database, including general patient information, trauma characteristics, and post-traumatic treatment information.

[0147] (1) General information: gender, age, height, weight, body mass index, admission time, smoking history, drinking history, and comorbidities.

[0148] (2) Injury characteristics: fracture site, injury mechanism, injury severity score, time from injury to hospital admission, whether it is multiple injury, whether there is shock, and vital signs on admission.

[0149] (3) Post-injury characteristics: drug prevention, mechanical prevention, number of surgeries, number of hemoglobin infusions, and whether or not the patient was admitted to the ICU.

[0150] After obtaining the relevant data, batch cleaning of the data is performed. The data is not binned, and all measurement data enter the database using the original values; then, based on the random forest method, other variables other than the missing values ​​are used as predictive variables, and a random forest model is established to fill in the missing data.

[0151] Filling missing data based on random forest method is a method that uses machine learning technology to estimate and fill missing values ​​in a data set. The following is a detailed description of the process:

[0152] 1. Understanding the Random Forest Algorithm

[0153] Random forest is an ensemble learning method that builds multiple decision trees and combines their results to improve the accuracy and robustness of predictions. Each decision tree is trained by randomly selecting samples from the dataset (with replacement sampling), and only a randomly selected subset of features is considered at each split.

[0154] 2. Data Preprocessing

[0155] Before applying random forest to fill missing values, the following preprocessing steps are required:

[0156] Data cleaning: Make sure there are no outliers or erroneous data in the dataset.

[0157] Identify missing values: Determine which variables contain missing values ​​and record the location of those variables.

[0158] 3. Build a random forest model to fill missing values

[0159] Step 3.1: Split the data

[0160] The dataset is divided into two parts: one part contains missing values ​​(data to be filled), and the other part does not contain missing values ​​(for training the model).

[0161] Step 3.2: Select predictor variables

[0162] Identify variables that are used to predict missing values. These should be other variables that are correlated with the missing value variable.

[0163] Step 3.3: Build a Random Forest Model

[0164] For each variable with missing values, a random forest model is built using the other variables as features (predictors).

[0165] When building the model, you can set the parameters of the random forest, such as the number of trees (n_estimators), maximum depth (max_depth), minimum number of sample splits (min_samples_split), etc.

[0166] Step 3.4: Train the model

[0167] The random forest model is trained using data that does not contain missing values. The model learns how to predict the missing value variable based on the predictor variables.

[0168] Step 3.5: Predict missing values

[0169] For each record with missing values, the trained random forest model is used to make predictions and obtain the estimated value of the missing value.

[0170] 4. Filling missing values

[0171] The predicted missing value estimates are filled into the corresponding missing positions in the original data set.

[0172] You can choose to use the mean, median, or mode of the model's predictions as filler values, or use randomly sampled values ​​from the predicted distribution to increase the diversity of the dataset.

[0173] 5. Verification and Adjustment

[0174] Verify the filling effect: Check whether the filled data set maintains the original data distribution characteristics through statistical analysis and visualization methods.

[0175] Adjust model parameters: If the imputation effect is not ideal, you can adjust the parameters of the random forest model or choose a different imputation strategy and re-implant.

[0176] 6. Complete the filling

[0177] After filling in all missing values, a complete data set is obtained, which can be used for subsequent data analysis and modeling.

[0178] The advantage of using the random forest method to fill missing values ​​is that it can handle nonlinear relationships and interaction effects, and is robust to outliers and noise. In addition, the random forest model can provide an assessment of feature importance, which helps to understand which variables are most important for predicting missing values.

[0179] Step 2. Through a variety of data-driven feature screening methods, including Pearson correlation coefficient (PearsonCorr), Spearman correlation coefficient (SpearmanCorr), distance correlation coefficient (DC), mutual information (MI), maximum information coefficient (MIC), mean decrease accuracy (MDA) and penalized regression (lasso, ridge) methods, the importance of the above features was scored using the voting method to screen out factors closely associated with venous thromboembolism in patients with orthopedic trauma.

[0180] For the five univariate analysis methods (filtering methods), the correlation coefficient between each feature and the target VTE was first calculated, and the bootstrap method was used to obtain the background value distribution of the corresponding irrelevant data through 1000 rounds of random sampling. When the obtained coefficient value is significantly greater than the background value, that is, the coefficient mean coeff minus two times the standard deviation coeff_std is greater than 0 (approximately), that is, it satisfies the 95% confidence level one-sided test of the normal distribution, it can be considered that there is a significant correlation or association between the feature and the target, and it is selected into the feature set. Among the three feature screening methods based on the parcel method, the MDA based on random forest can obtain the importance of each feature for target prediction through permutation test at the same time, and the result form is the same as that of the univariate analysis. Before the penalized regression method performs feature selection, the optimal penalty coefficient α for the prediction problem is first determined by ten-fold cross validation, and then it is substituted into the lasso or ridge equation, and the feature subset finally selected is determined by combining recursive feature elimination. Finally, the feature selection results of each method are summarized, and the importance of each feature is scored by voting. The features selected by at least 6 methods are selected as the screened risk factors to obtain the first group of candidate factors;

[0181] Among them, the random forest method divides the samples into training set and test set in a ratio of 7:3. The model is first trained using the training set, and then a permutation test is performed on the test set to obtain feature importance and significance information. The parameter settings are as follows:

[0182] Table 1 Parameter setting table

[0183] Parameter name Parameter meaning Parameter Value n_estimators Number of base learners 100 criterion Sample splitting criterion "gini" max_depth Maximum tree depth 4 min_samples_split Minimum number of split samples 6 min_samples_leaf Minimum number of leaf samples 2

[0184] Lasso feature selection: Lasso optimal penalty coefficient is selected using three paradigms based on cross-validation CV, AIC and BIC criteria. A search is performed, where CV corresponds to the highest point of the curve (the model has the highest generalization prediction accuracy score), while AIC and BIC correspond to the lowest information criterion values ​​(i.e., achieving bias-variance balance). Finally, the result obtained by CV is selected for feature selection based on Lasso and RFE.

[0185] Summarizing all feature selection methods, the following comprehensive evaluation results are obtained as shown in the following table

[0186] Table 2 Comprehensive evaluation results

[0187] feature_name pearson spearman dist mi mic lasso ridge md count AGE 1 1 1 0 1 1 1 1 7 ISS 1 1 1 1 1 1 1 0 7 RED_CELL_SUSPENSION 1 1 1 0 1 1 1 1 7 NUM_SURGERIES 1 1 1 0 1 1 1 0 6 PULMONARY_DISEASE 1 1 1 0 1 1 1 0 6 AROUND_KNEE 1 1 1 0 1 1 1 0 6 CHF 1 1 1 0 1 1 1 0 6 PELVIC_ACETABULAR 1 1 1 0 1 1 1 0 6 INJURY_MECHANISM_highfall 1 1 1 0 1 1 1 0 6 ICU 1 1 1 0 1 1 1 0 6 INJURY_TO_ADMISSION 1 1 1 0 1 1 1 0 6 TIBIAL_FIBULAR 1 1 1 0 1 1 1 0 6 MULTIPLE_TRAUMA 1 1 1 0 1 1 1 0 6 FEMUR_SHAFT 1 1 1 0 1 1 1 0 6 PROXIMAL_FEMUR 1 1 1 0 1 1 1 0 6 T 1 1 1 0 1 1 1 0 6 PVD 1 1 1 0 1 1 1 0 6 R 1 1 1 0 1 0 1 0 5 INJ_PROPHYLAXIS 0 0 1 0 1 1 1 1 5 NUM_FRACTURES 1 1 1 0 1 0 1 0 5 HBP 1 1 1 0 1 0 0 0 4 ULNA_RADIUS 0 0 1 0 1 1 1 0 4 BMI 0 0 0 1 1 1 1 0 4 PEPTIC_ULCER 1 1 1 0 0 0 1 0 4 SHOCK_IN_ADMISSION 1 1 1 0 1 0 0 0 4 INJURY_MECHANISM_fall 0 0 1 0 1 1 1 0 4 INJURY_MECHANISM_other 0 0 1 0 1 1 1 0 4 KIDNEY_DISEASE 1 1 1 0 1 0 0 0 4 MEC_PROPHYLAXIS 0 0 1 0 1 1 1 0 4 MI 1 1 1 0 0 0 1 0 4 PROPHYLAXIS 0 0 1 0 1 0 1 1 4 ORAL_PROPHYLAXIS 0 0 1 0 1 1 1 0 4 INJURY_MECHANISM_traffic accident 1 1 1 0 1 0 0 0 4 AROUND_SHOULDER 0 0 1 0 1 1 1 0 4 CVD 1 1 1 0 0 0 1 0 4 DBP 0 0 1 0 1 1 1 0 4 P 1 1 1 0 1 0 0 0 4 DIABETES 1 1 0 0 0 0 1 0 3 SBP 0 0 0 0 1 1 1 0 3 WEIGHT 0 0 0 0 1 1 1 0 3 SMOKING_abandoned 0 0 0 0 0 1 1 0 2 LIVER_DISEASE 0 0 0 0 0 1 1 0 2 HEIGHT 0 0 1 0 1 0 0 0 2 SOLID_TUMOR 0 0 0 0 0 0 1 1 2 ANKLE 0 0 0 0 0 1 1 0 2 FOOT 0 0 0 0 0 0 1 1 2 MALIGANT_LYMPHOMA 0 0 0 0 1 0 1 0 2 INJURY_MECHANISM_sharpinjury 0 0 0 0 0 0 1 0 1 DRINKING_no 0 0 0 0 0 0 1 0 1 HUMERAL 0 0 1 0 0 0 0 0 1 LEUKEMIA 0 0 0 0 0 0 1 0 1 INJURY_MECHANISM_machine 0 0 1 0 0 0 0 0 1 INJURY_MECHANISM_firearm 0 0 0 0 0 0 1 0 1 CTD 0 0 0 0 0 0 1 0 1 WHOLE_BLOOD 0 0 0 0 0 0 0 0 0 AIDS 0 0 0 0 0 0 0 0 0 SEX 0 0 0 0 0 0 0 0 0 INJURY_MECHANISM_crush 0 0 0 0 0 0 0 0 0 METASTATIC_TUMOR 0 0 0 0 0 0 0 0 0 DRINKING_yes 0 0 0 0 0 0 0 0 0 SMOKING_yes 0 0 0 0 0 0 0 0 0 DRINKING_abandoned 0 0 0 0 0 0 0 0 0 DEMENTIA 0 0 0 0 0 0 0 0 0 SMOKING_no 0 0 0 0 0 0 0 0 0 total 28 28 39 2 38 31 46 6 -

[0188] The English representation and Chinese name of each feature in the table are as follows:

[0189] AGE: age; ISS: Injury Severity Score; RED_CELL_SUSPENSION: red blood cell suspension; NUM_SURGERIES: number of surgeries; PULMONARY_DISEASE: pulmonary disease; AROUND_KNEE: knee injury; CHF: congestive heart failure; PELVIC_ACETABULAR: pelvic acetabular injury; INJURY_MECHANISM_high fall: mechanism of high fall injury; ICU: intensive care unit; INJURY_TO_ADMISSION: time from injury to hospital admission; TIBIAL_FIBULAR: tibia and fibula injury; MULTIPLE_TRAUMA: multiple trauma; FEMUR_SHAFT: femoral shaft fracture; PROXIMAL_FEMUR: proximal femur fracture; T: body temperature; PVD: peripheral vascular disease =VascularDisease); R: Respiratory rate; INJ_PROPHYLAXIS: Post-injury prophylaxis; NUM_FRACTURES: Number of fractures; HBP: Hypertension; ULNA_RADIUS: Radius and ulna injury; BMI: Body Mass Index; PEPTIC_ULCER: Peptic ulcer; SHOCK_IN_ADMISSION: Shock on admission; INJURY_MECHANISM_fall: Fall injury mechanism; INJURY_MECHANISM_other: Other injury mechanism; KIDNEY_DISEASE: Kidney disease; MEC_PROPHYLAXIS: Mechanical prophylaxis; MI: Myocardial Infarction; PROPHYLAXIS: Prophylaxis; ORAL_PROPHYLAXIS: Oral prophylaxis; INJURY_MECHANISM_traffic accident: Traffic accident injury mechanism; AROUND_SHOULDER: Shoulder injury; CVD: Cardiovascular disease Disease); DBP: Diastolic Blood Pressure; P: Blood Pressure; DIABETES: Diabetes; SBP: Systolic Blood Pressure; WEIGHT: Weight; SMOKING_abandoned: Smoking cessation; LIVER_DISEASE: Liver disease; HEIGHT: Height;SOLID_TUMOR: solid tumor; ANKLE: ankle injury; FOOT: foot injury; MALIGANT_LYMPHOMA: malignant lymphoma; INJURY_MECHANISM_sharp injury: sharp injury mechanism; DRINKING_no: no drinking; HUMERAL: humeral injury; LEUKEMIA: leukemia; INJURY_MECHANISM_machine: mechanical injury mechanism; INJURY_MECHANISM_firearm: gunshot injury mechanism; CTD: chronic thromboembolic disease (ChronicThromboembolic Disease); WHOLE_BLOOD: whole blood; AIDS: Acquired Immune Deficiency Syndrome); SEX: gender; INJURY_MECHANISM_crush: crush injury mechanism; METASTATIC_TUMOR: metastatic tumor; DRINKING_yes: drinking; SMOKING_yes: smoking; DRINKING_abandoned: quit drinking; DEMENTIA: dementia; SMOKING_no: no smoking. ;

[0190] Step 3. Based on the literature review, meta-analysis was used to combine the existing VTE-related risk factors in the literature;

[0191] Retrieve the literature related to orthopedic trauma, fracture, deep vein thrombosis, venous thromboembolism, and risk factors in the database, extract the adjusted odds ratio (ajusted OR) and its confidence interval of each risk factor based on the data provided by the study, merge them using a random effects model, and screen out the risk factors that have a definite impact on venous thromboembolism (the confidence interval of the ajusted OR does not include 1) to obtain the second group of candidate factors;

[0192] Step 4. Integrate the first and second groups of candidate risk factors, combine expert experience, use DAGitty to draw a directed acyclic graph, sort out and clarify the association between predictors, and on this basis merge duplicate risk factors and delete risk factors that are not directly causal. Constructing a causal graph follows the following steps:

[0193] (1) Preliminary screening of more certain factors based on literature knowledge and data-driven;

[0194] (2) Construct an implied graph containing all factors and simply link all factors;

[0195] (3) Based on the causal theory, the relationship between factors is identified, and the basic principles to be followed include four aspects:

[0196] Temporality: Does the hypothesized cause precede the effect? ​​If so, it is a directed link; if it is later, there is an inverse relationship;

[0197] Face validity: Is the hypothesized relationship reasonable? If yes, keep it; if no, evaluate the reverse relationship.

[0198] Regression theory: Is the hypothesized relationship supported by the theory?

[0199] Counterfactual thought experiments: Whether the hypothesized relationship is supported by systematic thought experiments informed by the counterfactual framework.

[0200] The connection relationship is established according to the above principles, and after deleting the factors that are finally considered irrelevant, a new directed acyclic graph is formed.

[0201] (4) Based on the similarity between the parent node and the child node, merge similar factors into one.

[0202] Specifically, the drawing form of the directed acyclic graph is as follows Figure 3 Finally, the risk factors affecting the occurrence of VTE were screened out by combining multiple considerations, including age, gender, hypertension, diabetes, coronary heart disease, chronic heart failure, cerebrovascular disease, history of VTE, multiple injuries, vascular and nerve injuries, high-energy injuries, comminuted fractures, open fractures, fracture sites, amount of red blood cell transfusion, time from injury to hospital admission, anesthesia method, surgical method, duration of operation, whether drug prevention was used, whether admission to the ICU, and number of days in bed after surgery.

[0203] In summary, the scheme provided in this application combines data-driven, meta-analysis methods and causal inference ideas into the risk factor screening process to comprehensively obtain the risk factors for venous thromboembolism in patients with orthopedic trauma. The risk factors obtained are more accurate, closely related to thrombosis, and more practical, which is helpful for the accurate construction of risk assessment scales or models. At the same time, it can effectively identify the intervenible risk factors related to thrombosis, which is helpful for the implementation of targeted intervention strategies.

[0204] The device embodiments of the present application can be used to execute the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0205] Figure 4 The block diagram of a device for screening risk factors of venous thromboembolism in orthopedic trauma patients provided by one embodiment of the present application is shown. Figure 4 As shown, the device comprises:

[0206] An acquisition module 41 is used to acquire original data;

[0207] A screening module 42, for screening a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients based on the raw data and in combination with a plurality of data-driven feature screening methods;

[0208] Determine Module 43 for identifying a second set of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients based on literature review using a meta-analysis approach;

[0209] The integration module 44 is used to integrate the first group of candidate risk factors and the second group of candidate risk factors, combine expert experience, based on causal theory and preset algorithms, identify the causal relationship of the candidate risk factors, and draw a directed acyclic graph; delete the risk factors in the directed acyclic graph that are not directly causally related to venous thromboembolism in orthopedic trauma patients to obtain the target risk factors.

[0210] Below, reference Figure 5 To describe an electronic device according to an embodiment of the present application. Figure 5 A block diagram of an electronic device according to an embodiment of the present application is illustrated.

[0211] like Figure 5 As shown, electronic device 500 includes one or more processors 510 and memory 520 .

[0212] The processor 510 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 500 to perform desired functions.

[0213] The memory 520 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 510 may run the program instructions to implement the above-described methods for screening risk factors for venous thromboembolism in orthopedic trauma patients of various embodiments of the present application and / or other desired functions. Various contents such as category correspondences may also be stored in the computer-readable storage medium.

[0214] In one example, the electronic device 500 may further include: an input device 530 and an output device 540 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0215] In addition, the input device 530 may also include, for example, a keyboard, a mouse, an interface, etc. The output device 540 may output various information to the outside, including analysis results, etc. The output device 540 may include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, etc.

[0216] Of course, to simplify, Figure 5 Only some of the components in the electronic device related to the present application are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, the electronic device may further include any other appropriate components according to specific application conditions.

[0217] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0218] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0219] In addition, an embodiment of the present application may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the steps of the method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to various embodiments of the present application described in the above “Exemplary Method” section of this specification.

[0220] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0221] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for screening risk factors for venous thromboembolism in orthopedic trauma patients, characterized in that: include: Get the original data; Based on the raw data, combining multiple data-driven feature screening methods, screening a first set of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients; Based on the literature review, a meta-analysis approach was used to identify a second set of candidate risk factors associated with VTE in orthopedic trauma patients; Integrate the first group of candidate risk factors and the second group of candidate risk factors, combine expert experience, based on causal theory and preset algorithms, identify the causal relationship of the candidate risk factors, and draw a directed acyclic graph; Deleting, in the directed acyclic graph, risk factors that are not directly causally related to venous thromboembolism in patients with orthopedic trauma to obtain target risk factors; Among them, based on the literature review, a meta-analysis method was used to identify a second group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients, including: The databases were searched for literature related to orthopedic trauma, fractures, deep vein thrombosis, venous thromboembolism, and risk factors; The adjusted odds ratios and confidence intervals of each risk factor were extracted based on the data provided in the literature; A random effects model was used to merge and screen out the risk factors that had a definite impact on venous thromboembolism to obtain the second group of candidate factors; Based on the raw data, a plurality of data-driven feature screening methods are combined to screen a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients, including: Based on the preset multiple data-driven feature screening methods, screening is performed respectively to obtain a feature set corresponding to each data-driven feature screening method; Count each feature set and use voting method to screen and obtain the first set of candidate risk factors; This also includes: identifying the causal relationship of candidate risk factors based on known medical knowledge and causal theory; using an algorithm based on conditional independence testing to identify the causal relationship of candidate risk factors; and verifying and adjusting the algorithm results in combination with the knowledge of domain experts to determine the causal relationship of candidate risk factors.

2. The method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to claim 1, characterized in that: The raw data include: general information, injury characteristics and post-injury characteristics; The general information included: gender, age, height, weight, body mass index, admission time, smoking history, drinking history, and comorbidities; The injury characteristics include: fracture site, injury mechanism, injury severity score, time from injury to hospital admission, whether it is multiple injury, whether it is accompanied by shock, and vital signs at admission; The post-injury characteristics include: drug prevention, mechanical prevention, number of surgeries, number of red suspension infusions, and whether or not the patient was admitted to the ICU.

3. The method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to claim 1 or 2, characterized in that: The multiple data-driven feature screening methods include: Pearson coefficient, Spearman coefficient, distance correlation coefficient, mutual information, maximum information coefficient, average accuracy reduction and penalized regression method.

4. The method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to claim 1, characterized in that: The step of integrating the first group of candidate risk factors and the second group of candidate risk factors, combining expert experience, based on causal theory and a preset algorithm, identifying the causal relationship of the candidate risk factors, and drawing a directed acyclic graph includes: Construct an implicit graph containing all candidate risk factors and perform simple links on all candidate risk factors; Identify the causal relationship of candidate risk factors based on causal theory and preset algorithms; Based on the causal relationship, the implicit graph is modified, and the directions of the links are added to obtain a directed acyclic graph.

5. A device for screening risk factors for venous thromboembolism in orthopedic trauma patients, characterized in that: include: An acquisition module is used to obtain raw data; A screening module, for screening a first group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients based on the raw data and in combination with a plurality of data-driven feature screening methods; A module was developed to identify a second set of candidate risk factors associated with VTE in orthopedic trauma patients based on a literature review using a meta-analysis approach; An integration module is used to integrate the first group of candidate risk factors and the second group of candidate risk factors, identify the causal relationship of the candidate risk factors based on causal theory and preset algorithms in combination with expert experience, and draw a directed acyclic graph; Deleting, in the directed acyclic graph, risk factors that are not directly causally related to venous thromboembolism in patients with orthopedic trauma to obtain target risk factors; Among them, based on the literature review, a meta-analysis method was used to identify a second group of candidate risk factors associated with venous thromboembolism in orthopedic trauma patients, including: The databases were searched for literature related to orthopedic trauma, fractures, deep vein thrombosis, venous thromboembolism, and risk factors; The adjusted odds ratios and confidence intervals of each risk factor were extracted based on the data provided in the literature; A random effects model was used to merge and screen out the risk factors that had a definite impact on venous thromboembolism to obtain the second group of candidate factors; Based on the raw data, a plurality of data-driven feature screening methods are combined to screen a first group of candidate risk factors related to venous thromboembolism in orthopedic trauma patients, including: Based on the preset multiple data-driven feature screening methods, screening is performed respectively to obtain a feature set corresponding to each data-driven feature screening method; Count each feature set and use voting method to screen and obtain the first set of candidate risk factors; This also includes: identifying the causal relationship of candidate risk factors based on known medical knowledge and causal theory; using an algorithm based on conditional independence testing to identify the causal relationship of candidate risk factors; and verifying and adjusting the algorithm results in combination with the knowledge of domain experts to determine the causal relationship of candidate risk factors.

6. An electronic device, characterized in that: include: A processor, and a memory for storing a program executable by the processor; The processor is used to implement the method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to any one of claims 1 to 4 by running the program in the memory.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to any one of claims 1 to 4.

8. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for screening risk factors for venous thromboembolism in orthopedic trauma patients according to any one of claims 1 to 4 is implemented.

Images