Acute myocardial infarction risk prediction method and system

Abstract

The invention discloses an acute myocardial infarction risk prediction method and system, and the method comprises the following steps: S1, sample collection: collecting sample data, and updating a sample in a database in real time; s2, creating a risk prediction model according to the current database; an unconditional Logistic regression method is adopted to construct a diabetes attack risk prediction model; s3, screening out a plurality of variables of which the association strength with the attack risk of acute myocardial infarction meets a predetermined requirement from the factors incorporated into the model; screening prediction factors by adopting Lasso regression, and screening final variables used in the risk prediction model; s4, inputting information of a to-be-evaluated user into the risk prediction fitting model, and calculating a comprehensive score according to the input information of the to-be-evaluated user; and S5, predicting the risk level of the to-be-evaluated user according to the comprehensive score, and displaying the risk level through an interactive interface. The acute myocardial infarction risk prediction model disclosed by the invention is high in sensitivity and specificity, and the accuracy of identifying acute myocardial infarction high-risk people can be effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data analysis and prediction, and specifically relates to an acute myocardial infarction risk prediction method and system for evaluating and grading the risk of acute myocardial infarction in a population. Background Art

[0002] Acute myocardial infarction is a disease caused by multiple causes. The acute myocardial infarction risk assessment model is a good tool for comprehensive assessment of multiple influencing factors. It establishes a quantitative relationship between multiple risk factors and health status through a statistical model, and predicts the probability of an individual suffering from acute myocardial infarction within a certain period of time. It plays an important role in effectively identifying high-risk groups for acute myocardial infarction and conducting early intervention.

[0003] At present, the samples based on the more influential cardiovascular disease risk prediction models abroad are all from Western populations, whose races and lifestyles are different from those of the Chinese population. The cohorts based on the existing models in China are all established a long time ago and are applicable to people over 35 years old. There is currently no risk prediction model for AMI alone. It is necessary to establish an AMI risk prediction model suitable for the Chinese population to provide a suitable tool for early diagnosis and treatment of AMI in my country. The accuracy and applicability of my country's acute myocardial infarction risk prediction model need to be improved.

[0004] Regarding influencing factors such as drinking, waist circumference, prediabetes, hypertension subtypes, and blood lipid levels, some factors have been less studied, and some factors are controversial between different studies. Therefore, it is necessary to conduct relevant cohort studies on the Chinese population to analyze the relationship between the above-mentioned influencing factors and the incidence of AMI. Summary of the invention

[0005] In response to the problems existing in the prior art, the purpose of the present invention is to use nationally representative cohort study data to establish a high-sensitivity and high-specificity acute myocardial infarction risk prediction model suitable for adults in my country, so as to provide a basis for effectively identifying people at high risk of acute myocardial infarction and scientifically formulating targeted intervention measures and strategies.

[0006] To achieve the above object, the present invention provides a method for predicting the risk of acute myocardial infarction, the method comprising the following steps:

[0007] S1. Sample collection, collect sample data and update samples in the database in real time;

[0008] S2. Create a risk prediction model based on the current database; use the unconditional Logistic regression method to build a diabetes risk prediction model. The basic form of the model is:

[0009] Logit(p)=β0+β1X1+β2X2+β3X3+…++βm X m

[0010] S3. Select several variables whose correlation strength with the risk of acute myocardial infarction meets the predetermined requirements from the factors included in the model; use Lasso regression to select predictive factors, select variables by adjusting the penalty coefficient λ, fit the model multiple times and perform collinearity test on each variable, finally determine the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction, and select the final variables used in the risk prediction model;

[0011] S4. Input the information of the user to be evaluated into the risk prediction fitting model, and calculate a comprehensive score based on the input information of the user to be evaluated;

[0012] S5. predict the risk level of the user to be assessed based on the comprehensive score and display it through the interactive interface;

[0013] Among them, the risk prediction fitting model is: Logit (P) = -11.175073 + 0.049014 × age + 0.365518 × annual family income + 0.008830 × systolic blood pressure + 0.076421 × body mass index + 0.007213 × waist circumference + 0.322536 × insufficient physical activity.

[0014] Furthermore, in step S1, the database includes: a baseline database and a follow-up database; the baseline database and the follow-up database are merged, and after the database is merged, the variables in the data are cleaned.

[0015] Furthermore, in step S2, the factors included in the model include: gender, age, urban and rural areas, regions, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

[0016] Further, step S2 specifically includes:

[0017] S2.1 Sampling weight: The sampling weight of the sample individual Ww s as follows:

[0018] W S =W S1 ×W S2 ×W S3 ×W S4

[0019] W S1 is the sampling weight of the sample county / district; W S2 is the sampling weight of the sample township / street; W S3is the sampling weight of the sample village / neighborhood committee; W S4 is the sampling weight of the sample individuals;

[0020] S2.2 Post-stratification weights: The stratification factors are gender, age, urban and rural areas, and regions. The monitoring samples weighted by the sampling weights and the national samples are divided into 156 layers according to the stratification factors, and the weight value of each layer is calculated;

[0021] S2.3 Describe and analyze the baseline characteristics of the survey subjects; describe the basic situation of the baseline data;

[0022] S2.4 Establish an acute myocardial infarction risk prediction model using logistic regression.

[0023] Further, step S3 specifically includes:

[0024] S3.1 Lasso regression was used to screen predictive factors so that the final variables included in the model were all related to the dependent variable while considering the problem of variable collinearity;

[0025] S3.2Using Cox proportional hazards regression model to conduct univariate analysis and estimate the strength of association between single risk factors and acute myocardial infarction events;

[0026] S3.3 Proportional hazard regression multivariate analysis; variables with P < 0.10 in the univariate Cox proportional hazard regression analysis were included in the multivariate Cox proportional hazard regression model analysis;

[0027] S3.4 Variable screening of acute myocardial infarction risk prediction model; screening variables by adjusting the penalty coefficient λ to obtain the final model variables;

[0028] S3.5 Fitting of acute myocardial infarction risk prediction model; after using Lasso regression to screen out several final model variables, the prediction model was fitted using the Logistic regression model.

[0029] Furthermore, according to the calculation results of the risk prediction fitting model, the risk of disease is divided into four levels, namely: low risk: <0.23%, general risk: 0.23%~1.52%, medium risk: 1.52%~3.90%, and high risk: ≥3.90%.

[0030] Further, the cleaning method comprises:

[0031] Variable validity cleaning: including checking whether there are logical errors and correcting them if there are any; checking whether there are outliers and correcting or eliminating observations with outliers; and

[0032] Variable integrity cleaning: For variables with missing values ​​less than 5%, interpolation is performed using the mean.

[0033] On the other hand, a system for predicting the risk of acute myocardial infarction is provided, the system comprising a sample collection module, a model building module, a variable determination module, a prediction module and a display output module.

[0034] Furthermore, the model building module is used to create a risk prediction model based on the current basic database; Logistic regression is used to build the model, and the basic form of the model is:

[0035] Logit(p)=β0+β1X1+β2X2+β3X3+…++β m X m

[0036] Factors included in the model include: gender, age, urban or rural area, region, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

[0037] Furthermore, the variable screening module is used to screen out several variables whose correlation strength with the risk of acute myocardial infarction meets the predetermined requirements from the factors included in the model; according to the literature review and combined with the cohort data variables, the variable selection and the form of variable entry into the model are reasonably determined, the Lasso regression is used to screen the predictive factors, the variables are screened by adjusting the penalty coefficient λ, the model is fitted multiple times and the collinearity test is performed on each variable, and finally the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction are determined, and the final variables used in the risk prediction model are screened.

[0038] Beneficial effects:

[0039] More than 7 million people worldwide are affected by acute myocardial infarction every year. More than 1 / 3 of people in developed countries die from acute myocardial infarction each year. In recent years, the incidence and mortality of acute myocardial infarction in my country have both been on the rise. Effective intervention strategies and measures are urgently needed to prevent the increase in the number of people suffering from acute myocardial infarction. The acute myocardial infarction risk prediction method and system of the present invention can improve the prediction accuracy, effectively identify the high-risk group of acute myocardial infarction, take intervention measures early, control the occurrence of acute myocardial infarction, improve the health level of residents, and reduce the disease burden. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of a method for predicting the risk of acute myocardial infarction according to the present invention;

[0041] Figure 2 It is a schematic diagram of database information of the present invention;

[0042] Figure 3 It is a schematic diagram of the relationship between the penalty coefficient λ and the variables of the present invention;

[0043] Figure 4 A nomogram used in the AMI incidence prediction model of the present invention;

[0044] Figure 5 It is a schematic diagram of the optimal situation of the penalty coefficient λ of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention will be described clearly and completely below in conjunction with 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 are within the scope of protection of the present invention.

[0046] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0047] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0049] like Figure 1-Figure 5 As shown, according to an acute myocardial infarction risk prediction method and system of the present invention, the system includes a sample collection module, a model building module, a variable determination module, a prediction module and a display output module.

[0050] Among them, the sample collection module is used to collect sample data and update the basic database in real time;

[0051] The model building module is used to create a risk prediction model based on the current basic database; Logistic regression is used to build the model, and the basic form of the model is:

[0052] Logit(p)=β0+β1X1+β2X2+β3X3+…++β m X m

[0053] Factors included in the model include: gender, age, urban or rural area, region, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

[0054] The variable screening module is used to screen out several variables whose correlation strength with the risk of acute myocardial infarction reaches the predetermined requirements from the factors included in the model. The variables are the factors included in the model as described above. According to the cohort data variables, the variable selection and the form of the variable entering the model are reasonably determined, the prediction factors are screened by Lasso regression, the variables are screened by adjusting the penalty coefficient λ, the model is fitted multiple times and each variable is tested for collinearity, and finally the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction are determined, and the final variables used in the risk prediction model are screened. Variables include risk factors, and variables can enter the model in different forms. For example, the variable of body mass index can first enter the model in the form of a numerical variable, and then enter the model in the form of underweight, normal, overweight, and obesity, and select the variable form of the model with better performance. In a specific embodiment, the final variables include age, annual family income, systolic blood pressure, body mass index, waist circumference, and physical inactivity.

[0055] The prediction module is used to input the information of the user to be evaluated into the risk prediction fitting model, and calculate the comprehensive score based on the input information of the user to be evaluated. The information at least includes the final variables used in the risk prediction model, for example, in this embodiment, the information at least includes age, annual family income, systolic blood pressure, body mass index, waist circumference, and physical inactivity.

[0056] The display output module is used to predict the risk level of the user to be evaluated based on the comprehensive score and display it through the interactive interface, where: low risk: <0.23%; general risk: 0.23%~1.52%; medium risk: 1.52%~3.90%; high risk: ≥3.90%.

[0057] Correspondingly, a method for predicting the risk of acute myocardial infarction according to the present invention comprises the following steps:

[0058] S1. Sample collection, collect sample data and update samples in the database in real time;

[0059] S2. Create a risk prediction model based on the current database; use the unconditional Logistic regression method to build a diabetes risk prediction model. The basic form of the model is:

[0060] Logit(p)=β0+β1X1+β2X2+β3X3+…++β m X m

[0061] Factors included in the model include: gender, age, urban or rural area, region, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

[0062] S3. Select several variables whose correlation strength with the risk of acute myocardial infarction meets the predetermined requirements from the factors included in the model. According to the literature review and the variables of the cohort data, the variables are reasonably selected and the form of variables entering the model is determined. The predictive factors are screened by Lasso regression, and the variables are screened by adjusting the penalty coefficient λ. The model is fitted multiple times and the collinearity test is performed on each variable. Finally, the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction are determined, and the final variables used in the risk prediction model are screened. For example, in a specific embodiment, the final variables include age, annual family income, systolic blood pressure, body mass index, waist circumference, and lack of physical activity. The specific calculation method is: using the Logistic regression model to fit the prediction model: Logit (P) = -11.175073 + 0.049014 × age (years) + 0.365518 × annual family income (1 = less than 20,000 yuan, 0 = 20,000 yuan and above) + 0.008830 × systolic blood pressure (mmHg) + 0.076421 × body mass index (kg / m2) + 0.007213 × waist circumference (cm) + 0.322536 × lack of physical activity (1 = yes, 0 = no).

[0063] S4. Input the information of the user to be evaluated into the risk prediction fitting model, and calculate the comprehensive score based on the input information of the user to be evaluated. The information at least includes the final variables used in the risk prediction model, for example, in this embodiment, the information at least includes age, annual family income, systolic blood pressure, body mass index, waist circumference, and physical inactivity.

[0064] S5. Predict the risk level of the user to be assessed based on the comprehensive score and display it through the interactive interface, where: low risk: <0.23%; general risk: 0.23%~1.52%; medium risk: 1.52%~3.90%; high risk: ≥3.90%.

[0065] Specifically, in step S1,

[0066] The database in the present invention includes: a baseline database and a follow-up database (including a family database and a cardiovascular and cerebrovascular database). The baseline database and the follow-up database are merged to form a master database. During the database merging process, the baseline database and the follow-up database are first accurately matched, and the ID number, name and gender are all consistent. For the missing baseline or follow-up ID number, the date of birth can be used for fuzzy matching, and the matching principle is that the date of birth, name and gender are all consistent. After the date of birth and age and gender are matched, if it is still not possible to match the same person, the object is eliminated. After the database merging is completed, the variables in the data are cleaned.

[0067] The cleaning method is as follows:

[0068] Each variable in the data is cleaned. This includes: (1) Variable validity cleaning: including checking whether there are logical errors, and correcting them if there are; checking whether there are outliers, and correcting or eliminating observations with outliers. (2) Variable integrity cleaning: handling missing values ​​of variables. For some variables, such as missing age and gender, the ID card number information can be used to fill in the missing values. If the age is missing, the ID card number is used to query the birth date of the research subject, and the age of the research subject is obtained by subtracting the birth date from the survey date. If the gender is missing, the second to last digit of the ID card number is used, with odd numbers representing males and even numbers representing females. At the same time, for some variables with fewer missing values, such as those with missing values ​​less than 5%, the mean is used for interpolation.

[0069] The present invention uses a special calculation formula for the sample size of a prospective cohort study and uses estimated parameters to estimate the sample size. The formula is as follows:

[0070]

[0071] In the formula, p0 and p1 represent the expected incidence rates of the control group and the exposed group, respectively, p is the mean of the two incidence rates, q = 1-p, z1-α / 2 and z β The area of ​​the curve under the standard normal distribution can be obtained by looking up the table.

[0072] Under the conditions of α = 0.05 (bilateral) and β = 0.10, the incidence of AMI in the control group was P0 = 1.82%, and RR = 1.43

[102] . Then z1-α / 2 = 1.96, z β=1.282, p0=0.0182, q0=0.9818, p1=RR×p0=0.02603, q1=0.97397. Substituting into the above formula, n=15,732 is calculated. Taking into account the possible 30% loss of follow-up, the required sample size is calculated to be 20,452 people. In a specific embodiment, the research population included in the analysis of the present invention is 26,794 people, which meets the required sample size. In a specific embodiment, the present invention selects all the survey subjects of 60 monitoring points in 11 provinces as the baseline data of this study. Following the inclusion and exclusion criteria of the research subjects (the age of the survey subjects is ≥18 years old, not suffering from AMI disease, non-pregnant or lactating women), a total of 36,426 people were finally included as survey subjects, including 16,650 males (45.71%) and 19,776 females (54.29%). A follow-up survey was conducted from 2016 to 2017, and 8,929 people were lost to follow-up. A total of 27,497 people were actually followed up (follow-up rate 75.49%), including 12,428 males (45.20%) and 15,069 females (54.80%). After excluding 541 people who died of other diseases and 162 people with missing baseline BMI and waist circumference information, a total of 26,794 people were included in this analysis. Figure 2 .

[0073] In step S2, since the monitoring adopts a multi-stage complex sampling design, it is necessary to perform sampling weighting on the samples; since the sampling causes deviations between the sample and the overall distribution of some important indicators (mainly deviations in age and gender), post-stratification adjustment is performed. Therefore, step S2 specifically includes:

[0074] S2.1 Sampling weight: The sampling weight of the sample individual Ww s as follows:

[0075] W S =W S1 ×W S2 ×W S3 ×W S4

[0076] W S1 is the sampling weight of the sample county / district, and its value is the inverse of the sampling probability of the sample county / district under stratified simple random sampling. The urban areas of the country are divided into 27 layers according to the east, middle and west, the proportion of non-agricultural population, and the number of population (among which, the proportion of non-agricultural population and the number of population are divided according to three tertiles respectively), and the rural areas are divided into 27 layers according to the east, middle and west, GDP, and population (among which, GDP and population are divided according to three tertiles respectively), with a total of 54 layers. Its calculation formula is as follows:

[0077]

[0078] W S2It is the sampling weight of the sample township / street, which is calculated by SAS during the population proportional cluster sampling (PPS sampling) process. Its value is the inverse of the sampling probability of the sample township / street under PPS sampling proportional to the population size.

[0079] W S3 is the sampling weight of the sample village / neighborhood committee, which is calculated by SAS during the PPS sampling process. Its value is the inverse of the sampling probability of the sample village / neighborhood committee under the PPS sampling proportional to the population.

[0080] W S4 is the sampling weight of the sample individual, and its value is the inverse of the sampling probability of the sample individual. The calculation formula is as follows:

[0081]

[0082] S2.2 Post-stratification weights: stratification factors are 2 layers of gender (male, female), 13 layers of age (18-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80-), 2 layers of urban and rural areas (urban, rural), and 3 layers of regions (east, central, and western). The monitoring samples weighted by the sampling weights and the national samples are divided into 156 layers according to the stratification factors. The formula for calculating the weight value of each layer is as follows:

[0083]

[0084] The heaviest weight of the sample individual: W = W S ×W PS .

[0085] S2.3 Descriptive analysis of baseline characteristics of the respondents.

[0086] Describe the basic situation of baseline data (such as age, gender, occupation, marital status, annual household income, body mass index and other factors included in the model). For continuous data, if they conform to normal distribution or approximate normal distribution, the mean ± standard deviation is used. The data were presented in Table 1. The t-test or analysis of variance was used for inter-group comparison. The median and interquartile range were used for skewed distribution, and the Wilcoxon rank sum test or Kruskal-Wallis rank sum test was used for inter-group comparison. The categorical data were presented in the number of cases and percentages, and the chi-square test was used for inter-group comparison. The rank data were presented in the number of cases and percentages, and the rank sum test was used according to the purpose of the study. All statistical tests were performed using two-sided tests, with an α level of 0.05, and P < 0.05 was considered statistically significant.

[0087] S2.4 Establishment of an acute myocardial infarction risk prediction model.

[0088] The establishment and verification of the AMI risk prediction model of the present invention was carried out in R software, and the model was established using Logistic regression. The basic form of the model is:

[0089] Logit(p)=β0+β1X1+β2X2+β3X3+…++β m X m

[0090] In the embodiment of the present invention, the follow-up period is 6 years, so a 6-year AMI risk prediction model is established.

[0091] Step S3 includes:

[0092] S3.1 Lasso regression was used to screen predictive factors in combination with literature and professional knowledge. The core idea of ​​the Lasso regression algorithm is to find an optimal balance between the estimation accuracy of the model and the absolute value of the coefficient. The principle is to add a penalty term (the sum of the absolute values ​​of the coefficients) on the basis of minimizing the residual sum, and to screen variables by adjusting the penalty coefficient (λ). As λ increases, the regression coefficient β approaches 0. The degree of Lasso regression complexity adjustment is controlled by λ. As λ increases, the penalty for model complexity increases. Lasso regression can make the independent variables included in the model correlated with the dependent variable (P < 0.05) while considering the problem of variable collinearity. The methods for model screening predictive factors include lambda.min and lambda.1se. Lambda.min refers to the lambda value that minimizes the model estimation error; while lambda.1se refers to the lambda value that keeps the model estimation error within an acceptable range (the simplest model). After λ increases to a certain value, continuing to add independent variables does not significantly improve the performance of the model. Therefore, lambda.1se can provide a model with excellent performance and the least number of independent variables.

[0093] In the present invention, the assignment of variables is shown in Table 3.1.

[0094] Table 3.1 Assignment of research variables

[0095]

[0096] Among them: BMI: body mass index; ISH: isolated systolic hypertension; IDH: isolated diastolic hypertension; SDH: combined systolic and diastolic hypertension; TG: triglycerides; TC: total cholesterol; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol.

[0097] S3.2 Cox proportional hazards regression model was used for univariate analysis.

[0098] The present invention uses the new AMI event as the dependent variable and each possible influencing factor as the independent variable, and uses the Cox proportional hazard regression model for univariate analysis (test level α=0.10) to estimate the strength of association between a single risk factor and AMI events. The factors included in the model include: gender, age, urban and rural areas, regions, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking. In this embodiment, the results show that: age (45-59 years old: HR=2.64, 95%CI: 1.79-3.89; ≥60 years old: HR=6.32, 95%CI: 4.35-9.19), overweight (HR=1.37, 95%CI: 1.01-1.84), obesity (HR=2.56, 95%CI: 1.86-3.54), abdominal obesity (HR=2.22, 95%CI: 1.71-2.89), hypertension (HR=3.18, 95%CI: 2.42-4.19), prediabetes (HR=2.12, 95%CI: 1.57-2.87), diabetes (HR=2.50, 95%CI: 1.78-3.50), abnormal LDL-C (HR=2.53, 95%CI: 1.38-4.64), insufficient physical activity (HR=1.59, 95%CI: 1.19-2.11) and insufficient intake of fruits and vegetables (HR=1.25, 95%CI: 0.96-1.61) can increase the risk of AMI. Alcohol consumption (HR=0.71, 95%CI: 0.54-0.93), excessive red meat intake (HR=0.54, 95%CI: 0.34-0.87), high education level (junior high school: HR=0.53, 95%CI: 0.33-0.83; high school and above: HR=0.45, 95%CI: 0.22-0.92) and high income (10,000-20,000 yuan: HR=0.61, 95%CI: 0.44-0.84; 20,000-30,000 yuan: HR=0.43, 95%CI: 0.28-0.65; ≥30,000 yuan: HR=0.43, 95%CI: 0.30-0.60) were protective factors for the incidence of AMI.

[0099] S3.3 Proportional hazards regression multivariate analysis.

[0100] The variables with P < 0.10 in the univariate Cox proportional hazard regression analysis were included in the multivariate Cox proportional hazard regression model analysis (all the variables with statistical significance in the univariate analysis met the equal proportional hazard assumption, and the collinearity diagnosis results showed that there was no collinearity problem between the variables, the tolerance of each variable was > 0.1, and the variance expansion factor was < 10). After adjusting for age, gender, education level, income, drinking, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, BMI, abdominal obesity, hypertension, abnormal TC, abnormal TG, abnormal HDL-C, abnormal LDL-C and blood glucose grouping, the results showed that age (45-59 years: HR = 2.08, 95% CI: 1.38-3.12; ≥ 60 years: HR = 4.23, 95% CI: 2.80-6.41), hypertension (HR = 1.80, 95% CI: 1. 34~2.43), prediabetes (HR=1.39, 95%CI: 1.02~1.91), diabetes (HR=1.42, 95%CI: 1.00~2.02), abdominal obesity (HR=1.65, 95%CI: 1.15~2.35), abnormal LDL-C (HR=3.19, 95%CI: 1.42~7.19) and lack of physical activity (HR=1.34, 95%CI: 1.00~1.79) were the risk factors for AMI. High income (10,000-20,000 yuan: HR=0.81, 95%CI: 0.58-1.13; 20,000-30,000 yuan: HR=0.56, 95%CI: 0.36-0.87; ≥30,000 yuan: HR=0.66, 95%CI: 0.45-0.95) is a protective factor for AMI.

[0101] S3.4 Variable screening for acute myocardial infarction risk prediction model.

[0102] In this embodiment, based on the literature, professional knowledge and the results of this study, age, gender, annual family income, smoking, drinking, insufficient intake of fruits and vegetables, excessive intake of red meat, insufficient physical activity, systolic blood pressure, diastolic blood pressure, diabetes, blood lipids, body mass index and waist circumference are included in the model, and Lasso regression is used to screen predictive factors. The assignment principle of the variables included in the model is shown in Table 4.1, and the relationship between the penalty coefficient λ and the variables is shown in Figure 3 , Figure 5, the coefficients of each variable are gradually compressed to 0 as the parameter λ increases. By adjusting the penalty coefficient λ to select variables, lambda.min selects 12 predictors, namely age, annual family income, body mass index, waist circumference, abnormal TC, abnormal HDL-C, abnormal LDL-C, systolic blood pressure, diabetes, drinking, insufficient fruit and vegetable intake, and insufficient physical activity. Combining professional knowledge and multiple fitting models to further select predictive factors, λ was finally set to 0.001069583, that is, log(λ)=-6.840487, and 6 predictive factors were selected: age, annual family income, systolic blood pressure, body mass index, waist circumference, and insufficient physical activity.

[0103] Table 4.1 Principles of variable assignment

[0104]

[0105] Among them: BMI: body mass index; TG: triglycerides; TC: total cholesterol; HDL-C: high-density lipoprotein cholesterol; LDL-C: low-density lipoprotein cholesterol.

[0106] S3.5 Fitting of the acute myocardial infarction risk prediction model.

[0107] After screening out 6 predictive factors (age, annual household income, systolic blood pressure, body mass index, waist circumference, and insufficient physical activity) using Lasso regression, the prediction model was fitted using the Logistic regression model.

[0108] The fitted model in this embodiment is: Logit (P) = -11.175073 + 0.049014 × age (years) + 0.365518 × annual family income (1 = less than 20,000, 0 = 20,000 and above) + 0.008830 × systolic blood pressure (mmHg) + 0.076421 × body mass index (kg / m 2 )+0.007213×waist circumference (cm)+0.322536×lack of physical activity (1=yes, 0=no).

[0109] The numbers in the model are calculated by using the corresponding code of R software to screen the model and obtain the corresponding values ​​of each variable. The model numbers are automatically obtained according to the corresponding code of R software based on the sample data, and adjusted according to the increase and change of sample data.

[0110] The risk level is divided according to the predicted probability of AMI in healthy people (those who have not suffered from AMI). Because the predicted probability of occurrence is skewed, the predicted probability of occurrence is normalized (i.e., natural logarithm) when dividing the risk level. The mean and standard deviation after log conversion to normal are calculated. The mean and standard deviation are -5.129667 and 0.94257, respectively. Then, the risk of occurrence is divided into 4 levels according to mean-S, mean+S, and mean+2S. The corresponding predicted probability of occurrence is obtained by taking the inverse natural logarithm of these three boundary points. The risk levels are: low risk: <0.23%, general risk: 0.23%~1.52%, medium risk: 1.52%~3.90%, and high risk: ≥3.90%.

[0111] The variance inflation factor (VIF) was used to judge the collinearity of the fitted model. The results showed that the VIFs of age, annual household income, systolic blood pressure, BMI, waist circumference and physical inactivity were 1.206946, 2.513450, 1.040361, 1.216051, 2.492717 and 1.018596, respectively, all less than 5, indicating that there was no collinearity between the variables. The fitted model was evaluated using discrimination and calibration. The AUC of the fitted model in this study was 0.751, and the Brier score of the calibration was 0.009, indicating that the discrimination and calibration of the fitted model in this study were good.

[0112] In step S4, the AMI incidence prediction model of the present invention is displayed using a nomogram (see Figure 4 ). The implementation principle of the nomogram of the present invention is to integrate multiple prediction factors on the basis of multi-factor regression analysis, and use line segments with scales to draw them on the same plane according to a set ratio, so as to express the relationship between the various factors in the prediction model. The specific implementation process of the nomogram is as follows:

[0113] S4.1 Construct a multi-factor regression model, and assign points to each value level of each predictive factor according to the contribution of each influencing factor in the model to the outcome variable, so as to obtain a score for each predictive factor. The contribution degree is the value of the regression coefficient. The larger the regression coefficient, the greater the contribution degree and the higher the score.

[0114] S4.2 Add the scores of each predictor to get the total score;

[0115] S4.3 Calculate the predicted value of the individual outcome event through the functional conversion relationship between the total score and the probability of occurrence of the individual outcome event;

[0116] S4.4 uses line segments with scales to draw predicted values ​​according to the set ratio to express the relationship between the various factors in the prediction model. The nomogram is drawn using the nomogram function in the rms package of the R software.

[0117] The nomogram transforms complex regression equations into simple and visual graphs, making the results of the prediction model more readable. At the same time, it is economically feasible, can be well promoted and widely used at the grassroots level, and has higher use value. Figure 4 is the score scale corresponding to different variables, age (years), BMI (kg / m 2 ), annual family income (yuan), systolic blood pressure (mmHg), waist circumference (cm) and insufficient physical activity, total score and predicted probability of AMI risk.

[0118] In step S4, according to Figure 4 The AMI incidence prediction model nomogram shows that the age is 80 years old, the first line score is 70 points, and BMI = 30kg / m 2 , the first score is 23 points, and so on. After the total score is calculated, the score corresponding to the second-to-last total score line is used to calculate the corresponding percentage value. For example: an 80-year-old (score 70 points) resident with a BMI of 30kg / m 2 (Score: 23 points) Annual family income is less than 20,000 yuan (score: 7.5 points), systolic blood pressure is 180 mmHg (score: 22.5 points), waist circumference is 110 cm (score: 12.5 points), and physical activity is insufficient (score: 7 points). The total score of this resident is 70+23+7.5+22.5+5+7.5=135.5 points, corresponding to an AMI risk of 7%.

[0119] In step S5, the risk level of the user to be evaluated is predicted to be high risk based on the comprehensive score and displayed through the interactive interface, where the specific risk levels are divided as follows: low risk: <0.23%; general risk: 0.23%~1.52%; medium risk: 1.52%~3.90%; high risk: ≥3.90%.

[0120] In addition, the present invention also includes the validation of the model. The validation of the model is the process of examining the performance indicators of the model, and the indicators include model discrimination and model calibration. Discrimination is also called the discriminative ability or ranking ability of the model. The consistency statistic is also called the C statistic, which is the most commonly used indicator to measure discrimination. The endpoint event of the present invention is a binary variable. When the endpoint event is a binary variable, the C statistic is the same as the area under the receiver operating characteristic curve (ROC) (AUC). The calibration degree is used to evaluate the consistency between the probability of occurrence of the model-predicted event and the observed probability of occurrence of the outcome event. The Brier score is used for quantitative measurement, and the Brier score ranges from 0 to 0.25. The closer the Brier score is to 0, the better the model calibration is. The closer the Brier score is to 0.25, the worse the model calibration is. When the Brier score is equal to 0.25, it means that the model has no predictive ability. According to the data acquisition situation, the ten-fold crossover method was used for internal validation. The samples were divided into 10 equal parts, 9 of which were training sets and the other was a validation set. The calculation was repeated 10 times, and the mean of the 10 results was finally regarded as the final value. Finally, the nomogram function in R software was used to draw a nomogram to display the model.

[0121] Any process or method description in the flowchart of the present invention or described in other ways herein can be understood as a module, segment or part of a code including one or more executable instructions for implementing the steps of a specific logical function or process, which can be implemented in any computer-readable medium for use by an instruction execution system, device or equipment, and the computer-readable medium can be any medium containing storage, communication, propagation or transmission programs for use by execution systems, devices or equipment, including read-only memories, magnetic disks or optical disks, etc.

[0122] In the description of this specification, the description with reference to the terms "embodiment", "example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, those skilled in the art can combine or combine different embodiments or examples described in this specification and the features therein without causing any contradiction.

[0123] Although the above content has shown and described the embodiments of the present invention, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in the field can change, modify, replace, modify and other update operations on the above embodiments within the scope of the present invention.

Claims

1. A method for predicting the risk of acute myocardial infarction, characterized in that: The method comprises the following steps: S1. Sample collection, collect sample data and update samples in the database in real time; S2. Create a risk prediction model based on the current database; use the unconditional Logistic regression method to build a diabetes risk prediction model. The basic form of the model is: Logit(p)=β0+ϊ1X1+β2X2+β3X3+...++β m X m S3. Select several variables whose correlation strength with the risk of acute myocardial infarction meets the predetermined requirements from the factors included in the model; use Lasso regression to select predictive factors, select variables by adjusting the penalty coefficient λ, fit the model multiple times and perform collinearity test on each variable, finally determine the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction, and select the final variables used in the risk prediction model; S4. Input the information of the user to be evaluated into the risk prediction fitting model, and calculate a comprehensive score based on the input information of the user to be evaluated; S5. predict the risk level of the user to be assessed based on the comprehensive score and display it through the interactive interface; Among them, the risk prediction fitting model is: Logit (P) = -11.175073 + 0.049014 × age + 0.365518 × annual family income + 0.008830 × systolic blood pressure + 0.076421 × body mass index + 0.007213 × waist circumference + 0.322536 × insufficient physical activity.

2. The method for predicting the risk of acute myocardial infarction according to claim 1, characterized in that: In step S1, the database includes: a baseline database and a follow-up database; the baseline database and the follow-up database are merged, and after the database is merged, the variables in the data are cleaned.

3. The method for predicting the risk of acute myocardial infarction according to claim 1, characterized in that: In step S2, the factors included in the model include: gender, age, urban and rural areas, region, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

4. The method for predicting the risk of acute myocardial infarction according to claim 1, characterized in that: Step S2 specifically includes: S2.1 Sampling weight: The sampling weight of the sample individual Ww s as follows: IN S =In S1 ×W S2 ×W S3 ×W S4 W S1 is the sampling weight of the sample county / district; W S2 is the sampling weight of the sample township / street; W S3 is the sampling weight of the sample village / neighborhood committee; W S4 is the sampling weight of the sample individuals; S2.2 Post-stratification weights: The stratification factors are gender, age, urban and rural areas, and regions. The monitoring samples weighted by the sampling weights and the national samples are divided into 156 layers according to the stratification factors, and the weight value of each layer is calculated; S2.3 Describe and analyze the baseline characteristics of the survey subjects; describe the basic situation of the baseline data; S2.4 Establish an acute myocardial infarction risk prediction model using logistic regression.

5. The method for predicting the risk of acute myocardial infarction according to claim 1, characterized in that: Step S3 specifically includes: S3.1 Lasso regression was used to screen predictive factors so that the final variables included in the model were all related to the dependent variable while considering the problem of variable collinearity; S3.2Using Cox proportional hazards regression model to conduct univariate analysis and estimate the strength of association between single risk factors and acute myocardial infarction events; S3.3 Proportional hazard regression multivariate analysis; variables with P < 0.10 in the univariate Cox proportional hazard regression analysis were included in the multivariate Cox proportional hazard regression model analysis; S3.4 Variable screening of acute myocardial infarction risk prediction model; screening variables by adjusting the penalty coefficient λ to obtain the final model variables; S3.5 Fitting of acute myocardial infarction risk prediction model; after using Lasso regression to screen out several final model variables, the prediction model was fitted using the Logistic regression model.

6. The method for predicting the risk of acute myocardial infarction according to claim 5, characterized in that: According to the calculation results of the risk prediction fitting model, the risk of disease is divided into four levels, namely: low risk: <0.23%, general risk: 0.23%~1.52%, medium risk: 1.52%~3.90%, high risk: ≥3.90%.

7. The method for predicting the risk of acute myocardial infarction according to claim 2, characterized in that: The cleaning method comprises: Variable validity cleaning: including checking whether there are logical errors and correcting them if there are any; checking whether there are outliers and correcting or eliminating observations with outliers; and Variable integrity cleaning: For variables with missing values ​​less than 5%, interpolation is performed using the mean.

8. An acute myocardial infarction risk prediction system, characterized in that: The system is used to implement the method for predicting the risk of acute myocardial infarction according to any one of claims 1 to 7, and the system includes a sample collection module, a model building module, a variable determination module, a prediction module and a display output module.

9. The acute myocardial infarction risk prediction system according to claim 8, characterized in that: The model building module is used to create a risk prediction model based on the current basic database; Logistic regression is used to build the model, and the basic form of the model is: Logit(p)=β0+β1X1+β2X2+β3X3+…++β m X m Factors included in the model include: gender, age, urban or rural area, region, education level, annual family income, hypertension, diabetes, abdominal obesity, BMI, TG abnormalities, TC abnormalities, HDL-C abnormalities, LDL-C abnormalities, insufficient physical activity, insufficient intake of fruits and vegetables, excessive intake of red meat, smoking and drinking.

10. The acute myocardial infarction risk prediction system according to claim 8, characterized in that: The variable screening module is used to screen out several variables whose correlation strength with the risk of acute myocardial infarction meets the predetermined requirements from the factors included in the model; according to the literature review and combined with the cohort data variables, the variable selection and the form of variable entry into the model are reasonably determined, the Lasso regression is used to screen the predictive factors, the variables are screened by adjusting the penalty coefficient λ, the model is fitted multiple times and the collinearity test is performed on each variable, and finally the variables of the proposed model and the correlation strength between each risk factor and the risk of acute myocardial infarction are determined, and the final variables used in the risk prediction model are screened.

Images