Cardiovascular disease risk assessment system and method based on heart and kidney metabolism indexes

Abstract

The invention relates to a cardiovascular disease risk assessment system and method. On the basis of fully considering the influence of dimension index information of kidney-related risk factors on the occurrence risk of cardiovascular diseases, heart and kidney metabolic indexes are integrated. And a mathematical model Fine-Gray is used for screening predictive factors by a step-by-step method. On the premise that traditional cardiovascular disease risk factors are reserved, in order to give consideration to the accuracy and easy generalization of risk assessment, prediction factors are simplified through strict calculation and derivation; the method is constructed based on a small amount of clinically ubiquitous detection index information such as age, sex, smoking or not, systolic pressure (SBP), fasting blood glucose (FBG), non-high density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) as predictive factors. The method has the advantages of simplicity and convenience in operation, less required information and low professional requirement. Meanwhile, the risk of the outcome of various clinical cardiovascular diseases (including CVD, ASCVD and HF) can be evaluated at the same time, and the method can be suitable for various scenes such as primary medical care and clinical diagnosis and treatment.

Description

Technical Field

[0001] The present invention belongs to the field of public health and clinical medicine, specifically to the field of cardiovascular disease risk assessment, and in particular to a cardiovascular disease risk assessment system and method. Background Art

[0002] With the aging of the population and changes in lifestyle, the burden of chronic diseases such as cardiovascular disease (CVD), obesity, diabetes, and chronic kidney disease (CKD) continues to increase. These cardiorenal metabolic diseases often cluster in the same individual, acting synergistically and causing huge health hazards.

[0003] In 2023, the American Heart Association first proposed the concept of cardiorenal metabolic syndrome (CKMS), which specifically refers to abnormal health conditions caused by obesity, diabetes, CKD and CVD, aiming to more systematically explore the interaction between metabolic diseases and chronic kidney disease and cardiovascular disease and its impact on the incidence and mortality of cardiovascular disease. With the emergence of some new metabolic drugs with cardiorenal protective effects, comprehensive intervention of cardiorenal metabolic abnormalities has become possible (NDUMELE CE, RANGASWAMI J, CHOW SL, et al. Cardiovascular-Kidney-Metabolic Health: AP residential Advisory From the American Heart Association [J]. Circulation, 2023). However, due to potential side effects and the influence of health economic factors, it is still necessary to conduct a comprehensive risk assessment of relevant patients and screen out high-risk patients for treatment. Based on this concept, there is an urgent need to develop new cardiovascular disease risk assessment systems and methods, so as to fully integrate kidney disease-related risk factors on the basis of traditional risk factors, provide simple and easy-to-use new tools for precise intervention of cardiorenal metabolic diseases, identify suitable populations for different intervention measures, so as to maximize the cost-effectiveness of intervention measures and avoid the potential hazards of excessive intervention.

[0004] The PREVENT model (AHA Predicting Risk of CVD EVENTs) released by the American Heart Association is the first CVD prediction model that integrates CKM health indicators and can accurately predict the 10-year and 30-year CVD risk of adults aged 30-79 in the United States (KHAN SS, CORESH J, PENCINAM J, et al. Novel Prediction Equations for Absolute Risk Assessment of Total Cardiovascular Disease Incorporating Cardiovascular-Kidney-Metabolic Health: A Scientific Statement From the American Heart Association [J]. Circulation, 2023.). However, there are huge differences in metabolic risk factors and CKD epidemic characteristics in different regions, as well as the impact of social determinants of health, which limits the application and promotion of models such as PREVENT in other regions, such as the Chinese population.

[0005] CN 114783606 A discloses a method for predicting the risk of cardiovascular disease that is easy to promote and apply, which is based on a small number of traditional risk factors other than blood lipids. However, the patent does not consider the dimensional indicators of kidney-related risk factors. In addition, the outcome definition does not include heart failure events. When this method is applied in areas with a high incidence of heart failure, it will potentially underestimate the risk of individual cardiovascular disease.

[0006] CN 115458172A discloses a cardiac risk assessment system, device and medium. The cardiac risk assessment requires the collection of multi-dimensional data of patients, such as personal information, medical history information, symptom information, family information and daily routine information, and has a high usage premise. Moreover, the impact of dimensional indicator information of kidney-related risk factors on the risk of cardiovascular disease is not considered.

[0007] CN 117116490A discloses a cardiovascular disease assessment model construction method, dietary therapy and health management system, which combines risk prediction with clinical decision-making of dietary therapy and health management. Based on the data of cardiovascular disease patients, the assessment model is constructed by proportional risk regression method, which includes many predictive factors and has a high usage premise, making it difficult to be widely used.

[0008] Moreover, current risk assessment tools mainly evaluate the risk of occurrence of composite CVD endpoints, and lack risk assessment of major CVD subtypes such as atherosclerotic cardiovascular disease [ASCVD] and heart failure [HF]. Summary of the invention

[0009] In order to solve the above technical problems, the present invention provides a cardiovascular disease risk assessment system and method. On the basis of fully considering the impact of dimensional indicator information of kidney-related risk factors on the risk of cardiovascular disease, cardiorenal metabolic indicators are integrated. On the premise of retaining traditional cardiovascular disease risk factors, in order to take into account the accuracy and easy promotion of risk assessment, the prediction factors are simplified through rigorous calculation and derivation, and constructed based on a small amount of detection indicator information commonly found in clinical practice. It has the advantages of simple operation, less information required, and low professional requirements. At the same time, the present invention can simultaneously evaluate the risks of multiple (including CVD, ASCVD and HF) clinical cardiovascular disease outcomes, and can be applied to various scenarios such as primary health care and clinical diagnosis and treatment.

[0010] In one aspect, the present invention provides a cardiovascular disease risk assessment system, comprising:

[0011] The data collection module collects data from two completely independent cohorts, deletes samples with CVD and renal failure at baseline, and samples with missing related variables, and finally forms a data set containing baseline information and outcome indicators. One cohort data is used as a training set for modeling, and the other cohort data is used as an external validation set.

[0012] Feature screening module, in the training set, based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, the range of candidate predictors is determined, including demographic data, lifestyle, medical history and medication status, metabolic and renal health indicators, subclinical atherosclerosis and subclinical heart failure indicators, inflammatory indicators, and social determinants of health. Then, the Fine-Gray model is used to screen predictors stepwise, while retaining traditional cardiovascular disease risk factors, deleting indicators that are not easy to obtain in primary health care and clinical diagnosis and treatment scenarios, and age, gender, smoking, systolic blood pressure (SBP), fasting blood glucose (FBG), non-high-density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) are screened as predictors according to the statistical significance level;

[0013] In the model building module, the Fine-Gray model was used to build a CVD risk prediction model. The training set was fitted to build a competing risk model. The independent variables were the seven predicted factors screened out, the dependent variables were the total CVD event occurrence and time, and the competing events were non-CVD deaths and time. Based on the above method, it was extended to the prediction of the risk of CVD subtypes (ASCVD and HF). The calculation formula is as follows:

[0014]

[0015] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ CVD (t), λ ASCVD (t), λ HF (t) are the risk functions of CVD, ASCVD and HF at time t, respectively, CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t), P ASCVD (t), P HF (t) are the incidence risks of CVD, ASCVD and HF at time t, respectively.

[0016] In a preferred embodiment, the cardiovascular disease risk assessment system of the present invention further comprises: a model validation module, which uses a 10-fold cross validation method to perform internal validation on the model to prevent overfitting of the model.

[0017] Furthermore, in a preferred embodiment of the present invention, the model verification module randomly divides the training set into 10 equal parts, selects any 9 equal parts for training the model, and uses the remaining 1 equal part for testing.

[0018] Furthermore, in a preferred embodiment of the present invention, the model verification module uses the C statistic to reflect the discrimination, denoted as C trajn and C test , select when C train +C test The model constructed when the maximum

[0019] In a preferred embodiment, the cardiovascular disease risk assessment system of the present invention further includes: a performance evaluation module for performing performance evaluation in a training set and an external validation set, wherein the model evaluation indicators include C statistics and calibration slope.

[0020] In a preferred embodiment, the cardiovascular disease risk assessment system of the present invention further includes: an assessment result stratification module, which stratifies an individual's CVD, ASCVD and HF risk <5%, 5%-9.9% and ≥10% into low, medium and high risks based on commonly used clinical thresholds.

[0021] In a preferred embodiment, the cardiovascular disease risk assessment system of the present invention further includes: an assessment result display module, which uses a cross-classification chart to intuitively display the proportion of individuals who are correctly stratified in the three risk strata, reflecting the risk stratification effect of the present invention.

[0022] In another aspect, the present invention provides a method for assessing cardiovascular disease risk, comprising the following steps:

[0023] S1) Dataset preparation: Data from two completely independent cohorts were collected, and samples with CVD and renal failure at baseline and missing related variables were deleted to eventually form a dataset containing baseline information and outcome indicators. One cohort data was used as a training set for modeling, and the other cohort data was used as an external validation set.

[0024] S2) Feature screening: In the training set, based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, the range of candidate predictors was determined, including demographic data, lifestyle, medical history and medication status, metabolic and renal health indicators, subclinical atherosclerosis and subclinical heart failure indicators, inflammatory indicators, and social determinants of health. The Fine-Gray model was then used to perform a stepwise screening of predictors. The indicators that were not easily obtained in primary health care and clinical diagnosis and treatment scenarios were deleted, with the traditional cardiovascular disease risk factors as the premise. Age, gender, smoking, systolic blood pressure (SBP), fasting blood glucose (FBG), non-high-density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) were screened as predictors based on the statistical significance level;

[0025] S3) Model construction, the Fine-Gray model was used to build a CVD risk prediction model, and the training set was fitted to build a competing risk model. The independent variables were the seven predicted factors screened out, the dependent variables were the total CVD event occurrence and time, and the competing events were non-CVD deaths and time. Based on the above method, it was extended to the prediction of the risk of CVD subtypes (ASCVD and HF). The calculation formula is as follows:

[0026]

[0027] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ CVD (t), λ ASVCD (t), λ HF (t) are the risk functions of CVD, ASCVD and HF at time t, respectively, CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD(t), P ASCVD (t), P HF (t) are the incidence risks of CVD, ASCVD and HF at time t, respectively.

[0028] In a preferred embodiment, the cardiovascular disease risk assessment method of the present invention further comprises: step S4) model validation, using a 10-fold cross-validation method to perform internal validation on the model to prevent overfitting of the model.

[0029] Furthermore, in a preferred embodiment of the present invention, in step S4) model verification, the training set is randomly divided into 10 equal parts, and any 9 equal parts are selected for training the model, and the remaining 1 equal part is used for testing.

[0030] Furthermore, in a preferred embodiment of the present invention, in step S4) model verification, the C statistic is used to reflect the discrimination, denoted as C train and C test , select when C train +C test The model constructed when the maximum

[0031] In a preferred embodiment, the cardiovascular disease risk assessment method of the present invention further comprises: step S5) performance evaluation, performing performance evaluation in a training set and an external validation set, and the model evaluation indicators include C statistics and calibration slope.

[0032] In a preferred embodiment, the cardiovascular disease risk assessment method of the present invention further comprises: step S6) stratifying the assessment results, stratifying the individual's CVD, ASCVD and HF risk <5%, 5%-9.9% and ≥10% into low, medium and high risk according to commonly used clinical thresholds.

[0033] The present invention has the following beneficial effects:

[0034] 1) Higher accuracy: Based on the traditional cardiovascular disease risk factors, the present invention fully considers the impact of renal system disease factors on the risk of cardiovascular disease, and thus has a higher accuracy in CVD risk prediction compared with existing cardiovascular disease risk scoring tools.

[0035] 2) Improve the efficiency of risk assessment: The present invention can simultaneously assess the risk of occurrence of three cardiovascular disease clinical outcome events, including CVD, ASCVD and HF risks. Only one risk prediction is required to complete the risk assessment of the three clinical outcomes, thereby providing a reference for the formulation of individualized intervention plans for patients in different scenarios.

[0036] 3) Conducive to promotion and application: The present invention is based on easily available data in primary health care and clinical diagnosis and treatment scenarios. It ensures the accuracy of risk assessment while taking into account the ease of promotion. After strict calculation and derivation of simplified prediction factors, it is constructed based on a small amount of clinically common detection index information. It has the advantages of simple operation, less information required, and low professional requirements. It can be applied to a variety of scenarios such as primary health care and clinical diagnosis and treatment.

[0037] 4) It is conducive to the rational allocation of medical resources: The present invention can help clinical medical staff to formulate individualized intervention measures for patients with different risk stratifications according to the risk stratification status of different cardiovascular outcomes of patients, which can improve the cost-effectiveness of intervention for suitable populations and avoid the harm that may be caused by excessive intervention for populations that are not suitable for intervention, and contribute to the rational allocation of health resources in medical institutions.

[0038] 5) Helps to enhance public health awareness: The present invention not only provides an effective risk assessment tool for medical staff, but also can encourage the public to pay more attention to their own health and adopt a more active lifestyle, such as healthy diet and regular exercise, by accurately assessing the risk of individual cardiovascular disease, thereby effectively preventing and reducing the occurrence of cardiovascular disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Receiver operating characteristic curve of CVD in the validation set of the risk assessment system of the present invention.

[0040] Figure 2 Receiver operating characteristic curve of ASCVD in the validation set of the risk assessment system of the present invention.

[0041] Figure 3 This is the receiver operating characteristic curve of HF in the validation set of the risk assessment system of the present invention.

[0042] Figure 4 The cross-classification diagram of the risk assessment system of the present invention in the modeling training set and the validation set. A is the cross-classification diagram of the modeling training set; B is the cross-classification diagram of the validation set DETAILED DESCRIPTION

[0043] The technical solution of the present invention is clearly and completely described below in conjunction with specific embodiments. The embodiments given are only for better illustrating the present invention rather than for limiting the scope of the present invention.

[0044] The present invention provides a cardiovascular disease risk assessment system, comprising:

[0045] The data collection module collects data from two completely independent cohorts, deletes samples with CVD and renal failure at baseline, and samples with missing related variables, and finally forms a data set containing baseline information and outcome indicators. One cohort data is used as a training set for modeling, and the other cohort data is used as an external validation set.

[0046] The feature screening module determines the range of candidate predictors in the training set based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, including demographic data, lifestyle, medical history and medication status, metabolic and renal health indicators, subclinical atherosclerosis and subclinical heart failure indicators, inflammatory indicators, and social determinants of health. The Fine-Gray model is then used to perform a stepwise screening of predictors. While retaining traditional cardiovascular disease risk factors, age, gender, smoking, systolic blood pressure (SBP), fasting blood glucose (FBG), non-high-density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) are screened as predictors based on the statistical significance level;

[0047] In the model building module, the Fine-Gray model was used to build a CVD risk prediction model. The training set was fitted to build a competing risk model. The independent variables were the seven predicted factors screened out, the dependent variables were the total CVD event occurrence and time, and the competing events were non-CVD deaths and time. Based on the above method, it was extended to the prediction of the risk of CVD subtypes (ASCVD and HF). The calculation formula is as follows:

[0048]

[0049] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ CVD (t), λ ASVCD (t), λ HF (t) are the risk functions of CVD, ASCVD and HF at time t, respectively, CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to y years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t), P ASCVD (t), P HF (t) are the incidence risks of CVD, ASCVD and HF at time t, respectively.

[0050] In this invention, the main outcome total CVD was defined as: coronary heart disease (CHD), stroke and HF. Among them, acute CHD events were defined as: acute myocardial infarction, CHD death, revascularization; stroke events were defined as ischemic stroke, hemorrhagic stroke and unclassified stroke; HF events were defined according to hospitalization diagnosis.

[0051] The present invention fully considers the influence of cardiovascular competing events. When total CVD is the outcome, non-CVD death is a competing event; when ASCVD is the outcome, non-ASCVD death and non-ischemic stroke events are competing events; when HF is the outcome, non-HF death is a competing event.

[0052] Example 1

[0053] This technical solution provides a prediction model that integrates the dimensions of cardiorenal metabolic health indicators and can simultaneously predict the risk of total CVD and two CVD subtypes (ASCVD and HF) in an individual in the next 10 years. The steps for building the model are as follows:

[0054] 1. Dataset preparation

[0055] Based on the large prospective cohort study database established by our team, the second review data in 2007 were selected as baseline characteristics, including demographic information, lifestyle, physical examination, questionnaire survey, and laboratory test index information. Individuals with CVD at baseline were excluded, and samples with missing blood pressure, blood sugar, blood lipids, metabolic and kidney health (serum creatinine) related factor indicators were deleted.

[0056] In this example, the primary outcome of total CVD was defined as: CHD, stroke, and HF. Acute CHD events were defined as: acute myocardial infarction, CHD death, and revascularization; stroke events were defined as ischemic stroke, hemorrhagic stroke, and unclassified stroke; and HF events were defined according to hospitalization diagnosis. CHD and stroke events were actively followed up and recorded and supplemented through external data linkage, and HF events were linked through the hospital electronic health record hospitalization system. Secondary outcomes included ASCVD (defined as acute CHD and ischemic stroke events) and HF events.

[0057] In this example, the impact of competing cardiovascular events was fully considered. When total CVD was the outcome, non-CVD death was a competing event; when ASCVD was the outcome, non-ASCVD death and non-ischemic stroke were competing events; when HF was the outcome, non-HF death was a competing event.

[0058] The above dataset or cohort is called a training set or modeling cohort.

[0059] To further improve the extrapolation of the model, based on another large prospective cohort study established by our team earlier, the clinical characteristics of more than 1,000 people were collected during the baseline survey in 2017 (specifically the same as the training set or modeling cohort above). The definition of outcomes and competing events, as well as the follow-up methods are the same as the training set or modeling cohort. The data set is called an external data set or external validation cohort. The relevant information of the modeling cohort population and the validation cohort population is shown in Tables 1 and 2.

[0060] Table 1 Relevant information of the modeling cohort population

[0061]

[0062] Table 2 Relevant information of the validation cohort population

[0063]

[0064] 2. Feature screening

[0065] Based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, the range of candidate predictors was determined. These included demographic data (age, gender), lifestyle (smoking, eating red meat, eating fruit, physical activity, snoring), metabolic risk factors (BMI and waist circumference), blood pressure, laboratory test indicators (fasting blood glucose FBG, cholesterol, triglycerides TG), medication status (antihypertensive drugs, hypoglycemic drugs, lipid-lowering drugs), kidney health indicators (estimated glomerular filtration rate eGFR, calculated using serum creatinine levels according to the CKD-EPI 2009 formula), subclinical atherosclerosis indicators (high-sensitivity troponin T, hs-cTnT) and subclinical heart failure indicators (N-terminal B-type natriuretic peptide pro, NT-proBNP), inflammatory indicators (high-sensitivity C-reactive protein, hc-CRP) and social determinants of health (education level, occupation, marital status, household per capita monthly income, medical insurance type). These features constitute the training data set.

[0066] To avoid the influence of extreme values, the above characteristics were preprocessed before model construction, and the continuous variables were transformed by natural logarithm. Continuous variables included age, BMI, waist circumference, SBP, FBG, HDL-C, TG, non-HDL-C, eGFR, NT-proBNP, and hs-cTnT. Categorical characteristics included gender, snoring, moderate-intensity physical activity, smoking, eating red meat, eating fruit, taking antihypertensive drugs, taking hypoglycemic drugs, and taking lipid-lowering drugs. The characteristic value was set as 0 or 1 variable.

[0067] Subsequently, the Fine-Gray model was used to stepwise screen important predictors associated with total CVD in the above-mentioned dataset, and the screening principle was Akaike Information Criterion AIC. Eleven predictors were screened out, including age, gender, smoking, daily fruit intake, BMI, SBP, FBG, non-HDL-C, eGFR, NT-proBNP, and hs-cTnT. Under the premise of forcibly retaining traditional cardiovascular disease risk factors, considering the clinical significance and data accessibility of medical environments such as primary care and clinical diagnosis and treatment, screening was performed among statistically significant predictors, and finally a total of 7 important predictors were screened out, including age, gender, smoking, SBP, FBG, non-HDL-C, and eGFR levels.

[0068] 3. Model construction

[0069] Before model construction, continuous variables were preprocessed (every 10-unit change), including age, SBP, FBG, non-HDL-C, and eGFR; categorical variables were set to 0 or 1, including sex and smoking.

[0070] The Fine-Gray model was used to fit the training set to construct a competing risk prediction model, with the 7 predictors mentioned above as independent variables; the dependent variables were the total CVD occurrence status and occurrence time, and non-CVD death and occurrence time were competing events.

[0071] The above steps can obtain a simplified CVD model, which is used to calculate the risk of CVD in subjects within 10 years. The calculation formula is as follows:

[0072]

[0073] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ CVD (t) is the risk function of CVD at time t, λ CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t) is the risk of CVD at time t.

[0074] The regression coefficients and hazard ratios of the obtained simplified CVD model are shown in Table 3.

[0075] Table 3 Regression coefficients and hazard ratios of simplified models

[0076]

[0077] Based on the simplified CVD model, the 10-year CVD risk function λ in the general population is calculated by CVD (t) is replaced by 10-year ASCVD (i.e., λ ASCVD (t)) and HF(λ HF (t)) hazard function (where the exponential part is The model is extended to predict the risk of CVD subtypes (ASCVD and HF). The calculation formula is as follows:

[0078]

[0079]

[0080] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ ASCVD (t), λ HF (t) are the risk functions of ASCVD and HF at time t, S(t) is the overall survival rate of the study population at time t, represents the integral of the product of the ASCVD risk function and the overall survival function for time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, and P ASCVD (t), P HF (t) are the incidence risks of ASCVD and HF at time t, respectively.

[0081] 4. Model validation and performance evaluation

[0082] The 10-fold cross-validation method was used to internally validate the simplified CVD model to prevent overfitting of the model. The training set was randomly divided into 10 equal parts (or folds), and any 9 equal parts (or folds) were selected as the "training set" to train the model, and the remaining 1 equal part (or fold) was used as the "test set". The predictive performance of the model was then evaluated in both the "training set" and the "test set". In this example, the discrimination (C statistic) was used to reflect this, denoted as C train and C test , select when C train +C test The model constructed when the maximum

[0083] The performance of the simplified CVD model constructed above was evaluated in the training set, which is an internal evaluation. The model evaluation indicators include discrimination (C statistic) and calibration (calibration slope). At the same time, in order to demonstrate the extrapolation of the model, the simplified CVD model was applied to the external validation set for performance evaluation, which is an external evaluation.

[0084] By substituting the 10-year CVD risk function λ for the general population CVD (t) is replaced by 10-year ASCVD (i.e., λ ASCVD (t)) and HF(λ HF The risk function of ASCVD (t) was used to extend the model to predict the risk of ASCVD and HF. The evaluation of ASCVD and HF outcomes included discrimination and calibration, and the evaluation indicators were the same as above.

[0085] The predictive performance of the simplified model for the internal evaluation of total CVD, ASCVD, and HF is shown in Table 4 .

[0086] Table 4 Prediction performance of the simplified model for internal evaluation of total CVD, ASCVD and HF

[0087]

[0088] At the same time, in order to demonstrate the extrapolation of the model, the simplified CVD model was applied to an external data set for performance evaluation. The prediction performance in the external validation set is shown in the attached Figure 1-3 The areas under the time-dependent curves (AUCs) were 0.722, 0.716, and 0.763, respectively. The above results indicate that the simplified CVD model has a good predictive performance for the three cardiovascular disease events.

[0089] Comparative Example 1

[0090] The same data set and model building steps as in the embodiment were used, and 11 predictors were screened out, including age, gender, smoking, daily fruit consumption, BMI, SBP, FBG, non-HDL-C, eGFR, NT-proBNP and hs-cTnT. The Fine-Gray model was used to fit the training set to construct a competing risk model, with the 11 predictors as independent variables, the dependent variables being the occurrence and time of total CVD events, and the competing events being non-CVD deaths and time of occurrence.

[0091] According to the steps described above, a full factor model can be obtained to calculate the 10-year CVD risk of the subject, as follows:

[0092]

[0093] Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor of the training set population, λ CVD (t) is the risk function of CVD at time t, λ CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t) is the risk of CVD at time t.

[0094] The full factor model was validated and evaluated for performance using the same method as in Example 1, and was extended to the prediction of the risk of CVD subtypes (ASCVD and HF), and the prediction performance was evaluated. The internal evaluation prediction performance of the full factor model for total CVD, ASCVD, and HF is shown in Table 5.

[0095] Table 5 Prediction performance of the full factorial model for internal evaluation of total CVD, ASCVD, and HF

[0096]

[0097] It can be seen from the evaluation results of the embodiments of the present invention and the comparative examples that the present invention, on the basis of traditional cardiovascular disease risk factors, fully considers the impact of renal system disease factors on the risk of cardiovascular disease, and thus has a higher accuracy in risk prediction of CVD compared with existing cardiovascular disease risk scoring tools. In addition, the present invention has been fully verified in external data, further illustrating the predictive performance of the present invention, and the advantage of being able to be extrapolated to other risk assessment areas. The present invention is based on easily accessible information in primary health care and clinical diagnosis and treatment scenarios, while ensuring the accuracy of risk assessment and taking into account easy promotion. After strict calculation and derivation, the simplified predictive factors are constructed based on a small amount of detection index information commonly found in clinical practice. It has the advantages of simple operation, less information required, and low professional requirements. It can be applied to a variety of scenarios such as primary health care and clinical diagnosis and treatment.

Claims

1. A cardiovascular disease risk assessment system, characterized in that: include: The data collection module collects data from two completely independent cohorts, deletes samples with CVD and renal failure at baseline, and samples with missing related variables, and finally forms a data set containing baseline information and outcome indicators. One cohort data is used as a training set for modeling, and the other cohort data is used as an external validation set. Feature screening module, in the training set, based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, the range of candidate predictors is determined, including demographic data, lifestyle, medical history and medication status, metabolic and renal health indicators, subclinical atherosclerosis and subclinical heart failure indicators, inflammatory indicators, and social determinants of health. Then, the Fine-Gray model is used to screen predictors stepwise, while retaining traditional cardiovascular disease risk factors, deleting indicators that are not easy to obtain in primary health care and clinical diagnosis and treatment scenarios, and age, gender, smoking, systolic blood pressure (SBP), fasting blood glucose (FBG), non-high-density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) are screened as predictors according to the statistical significance level; In the model building module, the Fine-Gray model was used to build a CVD risk prediction model, and the training set was fitted to build a competing risk model. The independent variables were the seven predicted factors screened out, the dependent variables were the total CVD event occurrence and time, and the competing events were non-CVD deaths and time of occurrence. Based on the above method, it was extended to the prediction of the risk of CVD subtypes ASCVD and HF. The calculation formula is as follows: Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor in the training set, λ CVD (t), λ ASCVD (t), λ HF (t) are the risk functions of CVD, ASCVD and HF at time t, respectively, CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t), P ASCVD (t), P HF (t) are the risks of CVD, ASCVD and HF at time t, respectively.

2. The cardiovascular disease risk assessment system according to claim 1, characterized in that: Also includes: The model validation module uses the 10-fold cross validation method to perform internal validation on the model to prevent overfitting of the model.

3. The cardiovascular disease risk assessment system according to claim 2, characterized in that The model validation module randomly divides the training set into 10 equal parts, selects any 9 equal parts for training the model, and uses the remaining 1 equal part for testing.

4. The cardiovascular disease risk assessment system according to claim 2, characterized in that Model validation module, using C statistic to reflect the degree of discrimination, denoted as C train and C test , select when C train +C test The model constructed when the maximum 5. The cardiovascular disease risk assessment system according to claim 1, characterized in that It also includes: a performance evaluation module, which performs performance evaluation in the training set and the external validation set. The model evaluation indicators include C statistics and calibration slope.

6. The cardiovascular disease risk assessment system according to claim 1, characterized in that It also includes: an assessment result stratification module, which stratifies individual CVD, ASCVD and HF risks <5%, 5%-9.9% and ≥10% into low, medium and high risks based on commonly used clinical thresholds.

7. The cardiovascular disease risk assessment system according to claim 1, characterized in that It also includes: an assessment result display module, which uses a cross-classification chart to visually display the proportion of individuals who are correctly stratified in the three risk strata.

8. A method for assessing cardiovascular disease risk, characterized in that: The following steps are involved: S1) Dataset preparation: Data from two completely independent cohorts were collected, and samples with CVD and renal failure at baseline and missing related variables were deleted to eventually form a dataset containing baseline information and outcome indicators. One cohort data was used as a training set for modeling, and the other cohort data was used as an external validation set. S2) Feature screening: In the training set, based on the predictors of the current mainstream cardiovascular disease risk prediction model and the CKMS guidelines issued by the American Heart Association, the range of candidate predictors was determined, including demographic data, lifestyle, medical history and medication status, metabolic and renal health indicators, subclinical atherosclerosis and subclinical heart failure indicators, inflammatory indicators, and social determinants of health. The Fine-Gray model was then used to screen predictors in a stepwise manner, with the traditional cardiovascular disease risk factors as the premise, deleting indicators that were not easily obtained in primary health care and clinical diagnosis and treatment scenarios. Age, gender, smoking, systolic blood pressure (SBP), fasting blood glucose (FBG), non-high-density lipoprotein cholesterol (non-HDL-C) and estimated glomerular filtration rate (eGFR) were screened as predictors based on the statistical significance level; S3) Model construction, using the Fine-Gray model to build a CVD risk prediction model, fitting the training set to build a competing risk model, the independent variables are the seven predicted factors screened out, the dependent variables are the total CVD event occurrence and time, the competing events are non-CVD deaths and time, and based on the above method, it is extended to the prediction of CVD subtype ASCVD and HF incidence risk, the calculation formula is as follows: Where β refers to the regression coefficient of the predictor, X refers to the level of the individual predictor, refers to the average level of the predictor in the training set, λ CVD (t), λ ASCVD (t), λ HF (t) are the risk functions of CVD, ASCVD and HF at time t, respectively, CR1 (t) is the risk function of CVD corresponding to the competing risk event at time t, The risk function for CVD and its corresponding competing events is the integral of time u from the start of follow-up to t years of follow-up, where du indicates that u is a dummy variable for follow-up time, S(t) is the overall survival rate of the study population at time t, and P CVD (t), P ASCVD (t), P HF (t) are the risks of CVD, ASCVD and HF at time t, respectively.

9. The cardiovascular disease risk assessment method according to claim 8, characterized in that The method also includes: step S4) model validation, using a 10-fold cross validation method to perform internal validation on the model to prevent overfitting of the model.

10. The cardiovascular disease risk assessment method according to claim 9, characterized in that Step S4) In model verification, the training set is randomly divided into 10 equal parts, 9 equal parts are selected for model training, and the remaining 1 equal part is used for testing. The C statistic is used to reflect the discrimination, which is recorded as C train and C test , select when C train +C test The model constructed when the maximum 11. The cardiovascular disease risk assessment method according to claim 8, characterized in that The method also includes: step S5) performance evaluation, performing performance evaluation in a training set and an external validation set, and the model evaluation indicators include C statistics and calibration slope.

12. The cardiovascular disease risk assessment method according to claim 8, characterized in that The method also includes: step S6) stratifying the evaluation results, stratifying the individual's CVD, ASCVD and HF risk <5%, 5%-9.9% and ≥10% into low, medium and high risk according to commonly used clinical thresholds.

Images