A prediction model for the risk of venous thromboembolism in hospitalized children
By constructing a VTE risk prediction model based on the decision tree algorithm, seven risk factors affecting the occurrence of VTE in hospitalized children were screened out, which solved the problems of high missed VTE rate and insufficient preventive measures in hospitalized children, and improved the accuracy of VTE risk assessment and clinical application value.
Patent Information
- Application Number
- CN202411864737.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-13
- Publication Date
- 2025-04-04
AI Technical Summary
The missed diagnosis rate of venous thromboembolic disease in hospitalized children is high, existing preventive measures are insufficient in terms of economic and bleeding risks, and there is a lack of effective VTE risk assessment tools.
A prediction model based on decision tree algorithm was developed to construct a risk prediction model for clinical VTE occurrence by screening seven risk factors that affect the occurrence of VTE in hospitalized children.
It improves the accuracy of VTE risk assessment, can efficiently identify children with high risk of VTE, provide clinicians with early intervention basis, and optimizes VTE prevention management of hospitalized children.
Smart Images
Figure CN119763843A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of disease risk prediction, and in particular to a prediction model for the risk of venous thromboembolism in hospitalized children. Background Art
[0002] Venous thromboembolism (VTE) refers to abnormal blood coagulation in the veins, which leads to complete or incomplete vascular blockage and venous reflux disorders, including deep vein thrombosis and pulmonary thromboembolism. VTE can lead to recurrent thrombosis, venous access termination, post-thrombotic syndrome, pulmonary embolism and even death.
[0003] In recent years, with the promotion and application of extracorporeal membrane oxygenation, blood purification therapy, medium-length venous catheters and central venous infusion devices in pediatrics, the incidence of VTE in hospitalized children has been increasing year by year. From the early 1990s to 2005, the incidence of VTE in children has increased from 0.053% to 0.580%. Failure to provide preventive measures can lead to a significant increase in the incidence of VTE in hospitalized children. According to statistics, the incidence of VTE in children who do not receive thrombus prevention is as high as 20% to 58%.
[0004] The impact of VTE on children is not only in terms of economic burden, but also in terms of disease progression. In the United States, the occurrence of deep vein thrombosis causes a total economic loss of more than $27.2 billion each year. One case of VTE in a child can cost nearly $30,000, increase the average length of hospital stay by 8 days, and increase the number of outpatient follow-up visits. Acute complications of VTE include organ and limb function decline and even individual death, of which acute pulmonary embolism accounts for 15% to 20%, with a mortality rate of nearly 10%; 50% of children with VTE have unablated thrombi, 12% will develop postphlebitic syndrome, 20% will develop recurrent venous thrombosis, and up to 30% will develop thrombotic syndrome, manifested as chronic pulmonary hypertension or painful limb swelling secondary to venous insufficiency.
[0005] It is very important to improve the recognition of VTE. Early and appropriate prevention and treatment can reduce the mortality and thrombotic syndrome of hospitalized children. Because the symptoms of VTE in children are non-specific and often masked by potential complications, relevant examinations are often performed only when the children show clinical symptoms in clinical work. At the same time, due to the lack of experience of clinical medical staff, they cannot make timely judgments and provide treatment and care even after the children show symptoms, resulting in a very high rate of missed diagnosis of VTE in hospitalized children. Although the use of preventive anticoagulant therapy can reduce the risk of VTE formation, it will increase the risk of bleeding in children. There is little benefit from preventive anticoagulant therapy for children with a low risk of VTE. Therefore, before preventive anticoagulant therapy, VTE risk assessment should be performed on hospitalized children to identify children with high susceptibility to VTE. Studies have shown that the use of risk assessment tools before imaging examinations can identify and better manage high-risk adult VTE patients. Therefore, it is reasonable, feasible and necessary to develop high-quality, low-cost, pediatric-specific VTE risk assessment tools.
[0006] To this end, the present invention aims to provide a method and system for predicting the risk of venous thromboembolism in hospitalized children, so as to solve the above-mentioned problems. Summary of the invention
[0007] The purpose of the present invention is to solve the above problems and provide a prediction model for the risk of venous thromboembolism in hospitalized children. The scheme of the present invention screens out the risk factors that affect the occurrence of VTE in hospitalized children, and uses a decision tree algorithm model for training, thereby establishing a VTE risk prediction model for hospitalized children that can be applied clinically. The decision tree risk prediction model can efficiently identify children with high risk of VTE, provide a basis for early intervention for clinicians, so as to take preventive measures in time and optimize the VTE prevention and management of hospitalized children. The establishment of this model not only improves the accuracy of VTE risk assessment, but also provides strong data support for clinical decision-making, and has important clinical application value.
[0008] In order to achieve the above object, the technical solution of the present invention is as follows: The present invention provides a prediction model for the risk of venous thromboembolism in hospitalized children, which is characterized by: The following steps are involved: S1. Data collection: Data of hospitalized children were collected as training sets for building prediction models; S2. Patient grouping: Children who developed VTE were classified as the VTE group, and children who did not develop VTE during the same period were classified as the non-VTE group; S3. Screening of risk factors: Compare the clinical data of children in the VTE group and non-VTE group to screen out the risk factors that affect the occurrence of VTE in hospitalized children; S4. Construct a decision tree prediction model: Based on the risk factors for VTE in hospitalized children, a decision tree prediction model was constructed using R software; S5. Evaluation of the predictive efficacy of the model: Hospitalized children with and without VTE were collected as a validation set to verify the decision tree prediction model.
[0009] Furthermore, in step S1, a total of 16 items of clinical data of children were included, including 3 items of general data, 2 items of laboratory tests, 2 items of catheter data, and 9 items of treatment conditions.
[0010] Furthermore, in step S2, children who were hospitalized in the pediatric department and had VTE were selected as the VTE group, and children who were hospitalized in the pediatric department during the same period and did not have VTE were selected as the non-VTE group at a ratio of 1:4. Inclusion criteria for children in the VTE group: ① VTE was confirmed by vascular ultrasound, magnetic resonance imaging or angiography during hospitalization; ② Age < 18 years old. Exclusion criteria: Discharged due to interruption of treatment for personal reasons. In the end, there were 71 cases in the VTE group and 284 cases in the non-VTE group.
[0011] Furthermore, in step S3, chi-square test and logistic regression analysis were used to screen out 7 risk factors that may affect the occurrence of VTE in hospitalized children, including D-dimer ≥ 0.55 mg / L, PICC indwelling time ≥ 3 days, infection, chemotherapy, ICU hospitalization time ≥ 4 days, use of ventilator, and immobilization time > 72 hours.
[0012] Furthermore, in step S4, the seven risk factors were used as predictive indicators, and the “rpart” package in R software was used to draw a decision tree prediction model. The decision tree prediction model included a total of four features (PICC indwelling duration, chemotherapy, ICU hospitalization duration, and D-dimer level), and seven judgment rules were determined, including four for judging that VTE would not occur and three for judging that VTE would occur.
[0013] Furthermore, in step S5, 31 hospitalized children with VTE and 124 hospitalized children without VTE were collected as validation sets, and the constructed decision tree prediction model was used to verify these 155 children, and the prediction accuracy, sensitivity, and specificity were calculated to evaluate the prediction effect of the model.
[0014] The present invention also provides an application of the above-mentioned model for predicting the risk of VTE in hospitalized children, and the decision tree model for predicting the risk of venous thromboembolism in hospitalized children constructed by the method is used for assessing the risk of VTE in hospitalized children.
[0015] The difficulty and significance of the present invention in solving the technical problem are: the symptoms of VTE in children are non-specific and often masked by potential complications. In clinical work, relevant examinations are often performed only when the child has clinical symptoms. At the same time, due to the lack of experience of clinical medical staff, even if the child has symptoms, they cannot make timely judgments and provide treatment and care, resulting in a very high rate of missed diagnosis of VTE in hospitalized children. Although preventive anticoagulant therapy can reduce the risk of VTE formation, it will increase the risk of bleeding in children, and there is little benefit from preventive anticoagulant therapy for children with a low risk of VTE. Therefore, before preventive anticoagulant therapy, VTE risk assessment should be performed on hospitalized children to identify children with high susceptibility to VTE. At present, there are studies on the construction of VTE risk prediction models for middle-aged and elderly patients, but there are few studies on the risk of VTE in hospitalized children, and they mainly focus on special populations such as children after congenital heart disease surgery, lacking universality. Some scholars have also constructed a VTE probability prediction tool for hospitalized children, but the study did not include possible influencing factors such as biomarkers, and its prediction effect needs to be further verified.
[0016] The scheme of the present invention studies a method for model prediction based on screening risk factors for VTE, which is not only expected to provide more targeted insights for the prevention, timely diagnosis and treatment of VTE in children, but also to optimize the comprehensive management of hospitalized children, reduce the occurrence of serious complications, and improve the treatment effect and quality of life of children.
[0017] Compared with the prior art, the present invention has the following beneficial effects: the present invention provides a multi-dimensional prediction model by combining the general information of hospitalized children, laboratory test indicators, catheter information and treatment conditions, which can conduct a more comprehensive and in-depth risk assessment of VTE. The decision tree risk prediction model constructed by the present invention can efficiently identify children with high risk of VTE, provide a basis for early intervention for clinicians, so as to take preventive measures in time and optimize the VTE prevention and management of hospitalized children. The establishment of this model not only improves the accuracy of VTE risk assessment, but also provides strong data support for clinical decision-making, and has important clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a flow chart of constructing a decision tree prediction model for the risk of VTE occurrence in hospitalized children in an embodiment of the present invention.
[0019] Figure 2 It is a decision tree model for predicting the risk of VTE in hospitalized children in an embodiment of the present invention.
[0020] Figure 3 This is a receiver operating characteristic curve diagram of the present invention for evaluating the prediction effect using data of 355 hospitalized children as a training set in an embodiment of the present invention.
[0021] Figure 4 This is a receiver operating characteristic curve diagram of the present invention for evaluating the prediction effect using data of 155 hospitalized children as a validation set in an embodiment of the present invention. DETAILED DESCRIPTION
[0022] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution of the present invention will be further described in detail below in conjunction with the embodiments of the present invention and the accompanying drawings. 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 should fall within the scope of protection of the present invention.
[0023] Example: The solution provided by the embodiment of the present invention is: a prediction model for the risk of venous thromboembolism in hospitalized children, the prediction model construction method is as follows Figure 1 As shown, the following steps are included.
[0024] Step 1: Data collection: In this embodiment, data of 355 children admitted to the pediatric department of a hospital in Nanning were collected as a training set for building a prediction model. A total of 16 clinical data of the included patients were collected, including 3 general data: age (≤2 years, 3-10 years, >10 years), gender (female, male), blood type (O type, non-O type); 2 laboratory tests: fibrinogen (<2g / L, 2-4g / L, >4g / L), D-dimer (<0.55mg / L, ≥0.55mg / L); 2 catheter data: central venous catheter (CVC) indwelling time (<3d, ≥3d), peripherally inserted intravenous catheter (PICC) indwelling time (<3d, ≥3d); 9 treatment conditions: blood purification treatment (no, yes), infection (no, yes), severe trauma such as fracture and spinal cord injury (no, yes), surgical treatment (no, yes), chemotherapy (no, yes), duration of hormone treatment (7d, >7d), duration of ICU stay (<4d, ≥4d), use of ventilator (no, yes), and duration of immobilization (≤72h, >72h).
[0025] Step 2: Patient grouping: In this example, 71 children who were hospitalized in the pediatric department of a hospital in Nanning and had VTE were selected as the VTE group, marked as 1, and 284 children who were hospitalized in the pediatric department during the same period and did not have VTE were selected as the non-VTE group at a ratio of 1:4, marked as 0. The inclusion criteria for children in the VTE group: ① VTE was confirmed by vascular color Doppler ultrasound, magnetic resonance imaging or angiography during hospitalization; ② Age < 18 years old. Exclusion criteria: Discharged due to interruption of treatment for personal reasons.
[0026] Step 3: Screening of risk factors: In this example, the general data, laboratory tests, catheter data, and treatment conditions of the VTE group and the non-VTE group were compared, and the chi-square test and logistic regression were used to screen out 7 indicators that may affect the occurrence of VTE in hospitalized children, including D-dimer ≥ 0.55 mg / L, PICC indwelling time ≥ 3 days, infection, chemotherapy, ICU hospitalization time ≥ 4 days, use of ventilator, and immobilization time > 72 hours.
[0027] Step 4: Build a decision tree prediction model: In this embodiment, the above 7 risk factors were used as prediction indicators, and the "rpart" package in R software was used to draw a decision tree prediction model. Figure 2 As shown, the constructed decision tree prediction model contains 4 important features, 13 nodes, the tree depth is 4, and 7 judgment rules are determined.
[0028] The four important features are PICC retention time (divided into "<3d" and "≥3d"), chemotherapy (divided into "chemotherapy" and "no chemotherapy"), and ICU hospitalization time (divided into "<4d" and "≥4d"). The PICC retention time is the first-level influencing factor affecting the risk of VTE in hospitalized children. The depth of the decision tree refers to the number of nodes on the longest path from the root node to the farthest leaf node. A depth of 4 means that it takes 4 nodes from the root node to the farthest leaf node, that is, it takes at most 4 steps to predict the risk of VTE in hospitalized children.
[0029] Among the 7 judgment rules, 4 judge that VTE will not occur and 3 judge that VTE will occur. The first rule is that if the hospitalized child has a PICC indwelling time of less than 3 days, has received chemotherapy, and has been in the ICU for less than 4 days, the probability that the child will not have VTE is 9%; the second rule is that if the hospitalized child has a PICC indwelling time of less than 3 days, has received chemotherapy, has been in the ICU for ≥4 days, and has a D-dimer <0.55mg / L, the probability that the child will not have VTE is 16%; the third rule is that if the hospitalized child has a PICC indwelling time of less than 3 days, has received chemotherapy, has been in the ICU for ≥4 days, and has a D-dimer ≥0.55mg / L, the probability that the child will have VTE is 64%; the fourth rule is that if the hospitalized child has a P If the ICC is in place for less than 3 days, the child has not received chemotherapy, and the ICU stay is less than 4 days, the probability of the child not developing VTE is 32%; Article 5, if the hospitalized child has a PICC in place for less than 3 days, has not received chemotherapy, and the ICU stay is ≥4 days, the probability of the child developing VTE is 80%; Article 6, if the hospitalized child has a PICC in place for ≥3 days and the D-dimer is <0.55mg / L, the probability of the child not developing VTE is 32%; Article 7, if the hospitalized child has a PICC in place for ≥3 days and the D-dimer is ≥0.55mg / L, the probability of the child developing VTE is 68%.
[0030] The predicted probability was used as the test variable, whether the hospitalized child had VTE was used as the state variable, and the state variable was assigned a value of "1" to obtain the receiver operating characteristic curve of the model training set (such as Figure 3 As shown in the figure, the training set AUC is 0.769 (95%CI: 0.706-0.831), the model has good discrimination, and the best cutoff value of the model calculated by the receiver operating characteristic curve is 0.237, the sensitivity corresponding to the best cutoff value is 0.671, and the specificity is 0.801. The accuracy of the model is 0.841, the recall rate is 0.958, the precision rate is 0.858, and the F1 value is 0.905.
[0031] Step 5: Evaluate the model prediction performance: In this embodiment, in order to verify the predictive effect of the above decision tree model on the risk of VTE in hospitalized children, data of 155 children admitted to the pediatric department of a hospital in Nanning were collected as a validation set. 31 hospitalized children with VTE were selected as the VTE group, marked as 1, and 124 children who were hospitalized in the pediatric department during the same period and did not develop VTE were selected as the non-VTE group at a ratio of 1:4 and marked as 0.
[0032] The receiver operating characteristic curve of the validation set is shown (eg Figure 4As shown in the figure, the AUC value is 0.718 (95%CI: 0.617-0.819), the optimal cutoff value of the model is 0.237, the sensitivity corresponding to the optimal cutoff value is 0.576, and the specificity is 0.803. The accuracy of the model is 0.832, the recall rate is 0.950, the precision rate is 0.852, and the F1 value is 0.899. The decision tree model performs well in the validation data set and has certain practical value.
[0033] It is worth noting that, through the scheme of the above-mentioned embodiment of the present invention, the present invention successfully screens out the risk factors that affect the occurrence of VTE in hospitalized children, and establishes a VTE occurrence risk prediction model for hospitalized children that can be applied clinically. The decision tree risk prediction model constructed by the present invention can efficiently identify children with high risk of VTE, provide a basis for early intervention for clinicians, so as to take preventive measures in time and optimize the VTE prevention management of hospitalized children. The establishment of this model not only improves the accuracy of VTE risk assessment, but also provides strong data support for clinical decision-making, and has important clinical application value.
[0034] The above specific embodiments are merely explanations of the present invention and are not limitations of the present invention. After reading this specification, those skilled in the art may make modifications to the embodiments without any creative contribution as needed. However, such modifications are protected by the patent law as long as they are within the scope of the claims of the present invention.
Claims
1. A prediction model for the risk of venous thromboembolism in hospitalized children, characterized by: The following steps are involved: S1. Data collection: Data of hospitalized children were collected as training sets for building prediction models; S2. Patient grouping: Children with VTE were divided into the VTE group, and children without VTE during the same period were divided into the non-VTE group; S3. Screening of risk factors: Compare the clinical data of children in the VTE group and non-VTE group to screen out the risk factors that affect the occurrence of VTE in hospitalized children; S4. Construct a decision tree prediction model: Based on the risk factors for VTE in hospitalized children, a decision tree prediction model was constructed using R software; S5. Evaluate the predictive efficacy of the model: Hospitalized children with and without VTE were collected as a validation set to verify the decision tree model.
2. A prediction model for the risk of venous thromboembolism in hospitalized children according to claim 1, characterized in that: In step S1, a total of 16 items of clinical data of children were included, including 3 items of general data, 2 items of laboratory tests, 2 items of catheter data, and 9 items of treatment conditions.
3. A prediction model for the risk of venous thromboembolism in hospitalized children according to claim 1, characterized in that: In step S2, children who were hospitalized in the pediatric department and had VTE were selected as the VTE group, and children who were hospitalized in the pediatric department during the same period and did not have VTE were selected as the non-VTE group at a ratio of 1:
4. The inclusion criteria for children in the VTE group were: ① VTE was confirmed by vascular ultrasound, magnetic resonance imaging or angiography during hospitalization; ② Age < 18 years old. Exclusion criteria: Discharged due to interruption of treatment for personal reasons. In the end, there were 71 cases in the VTE group and 284 cases in the non-VTE group.
4. A prediction model for the risk of venous thromboembolism in hospitalized children according to claim 1, characterized in that: In step S3, chi-square test and logistic regression analysis were used to screen out seven risk factors that may affect the occurrence of VTE in hospitalized children, including D-dimer ≥ 0.55 mg / L, PICC indwelling time ≥ 3 days, infection, chemotherapy, ICU stay ≥ 4 days, use of ventilator, and immobilization time > 72 hours.
5. A prediction model for the risk of venous thromboembolism in hospitalized children according to claim 1, characterized in that: In step S4, the seven risk factors were used as predictive indicators, and the "rpart" package in R software was used to draw a decision tree prediction model. The decision tree prediction model included a total of four features (PICC indwelling duration, chemotherapy, ICU hospitalization duration, and D-dimer level), and seven judgment rules were determined, including four for judging the absence of VTE and three for judging the occurrence of VTE.
6. A prediction model for the risk of venous thromboembolism in hospitalized children according to claim 1, characterized in that: In step S5, 31 hospitalized children with VTE and 124 hospitalized children without VTE were collected as validation sets. The constructed decision tree prediction model was used to verify these 155 children, and the prediction accuracy, sensitivity, and specificity were calculated to evaluate the prediction effect of the model.
7. A prediction model for the risk of venous thromboembolism in hospitalized children according to any one of claims 1 to 6, characterized in that: The decision tree model for predicting the risk of venous thromboembolism in hospitalized children constructed by the method is used to assess the risk of VTE in hospitalized children.