A method for predicting lifetime risk of cardiovascular disease in a type 2 diabetes patient

Abstract

The present application relates to the technical field of risk prediction of type 2 diabetes patients, and discloses a method for predicting the life-long cardiovascular disease risk of type 2 diabetes adult patients, comprising: including research objects meeting the criteria and integrating population characteristics and disease information; using a Fine & Gray proportional risk function adjusted by competition risk to construct two models with the same prediction variables, wherein model a takes non-vascular disease death as a competition event to predict the cardiovascular disease risk of an individual per life year, and model b takes a cardiovascular disease event as a competition event to predict the non-vascular disease death risk of an individual per life year; the cumulative calculation of the two models is up to 85 years old, and the cardiovascular disease life-long risk of an individual is obtained in combination. The method can evaluate the life-long cardiovascular disease risk of type 2 diabetes patients, so that medical workers, patients, policy makers and insurance practitioners can more clearly understand the long-term risk situation of type 2 diabetes patients, screen out high-risk patients, intervene as early as possible, and achieve the effects of disease prevention and health promotion.

Description

Technical Field

[0001] This invention relates to the field of risk prediction technology for patients with type 2 diabetes, specifically a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes. Background Technology

[0002] Electronic health records (EHRs) are standardized and scientific records of the public's health management process (disease prevention, health protection, health promotion, etc.). Centered on individual health, they encompass various health-related factors, dynamically collect information from multiple channels, and serve as information resources to meet residents' needs for self-care, health management, and health decision-making. EHR data covers a large population and provides a relatively accurate and complete reflection of an individual's multidimensional health-related information, leading to its increasing application in research on chronic disease risk prediction and prevention management.

[0003] Current cardiovascular risk prediction models for patients with type 2 diabetes mostly predict short-term risk over 5 or 10 years. Baseline age is a significant factor in short-term risk prediction, leading to lower short-term risk in younger patients with risk factors and resulting in insufficient intervention. Furthermore, these models do not consider the impact of competing events, potentially overestimating the risk in middle-aged and older patients. Therefore, assessing only short-term risk is detrimental to disease prevention and health promotion.

[0004] Domestic and international guidelines recommend using lifetime risk prediction models for cardiovascular disease to guide individualized disease prevention and intervention for patients. For example, the European Society of Cardiology (ESC) guidelines recommend using the diabetes lifetime-perspective prediction model (DIAL) to assess the lifetime risk of patients with type 2 diabetes. However, this model is based on a Swedish population of type 2 diabetes patients, whose population differs significantly from that of the Chinese population in terms of race, socioeconomic development, disease and risk factor prevalence, and baseline survival rates. This could lead to inaccurate risk assessment and inapplicability of the model in the Chinese population. Furthermore, the availability of predictive variables included in the model in the Chinese patient population is poor, limiting the model's application. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] In view of the shortcomings of the existing technology, the present invention provides a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes, wherein the competing risk-adjusted Fine & Gray proportional hazards function includes two risk prediction models with the same predictor variables. Model a predicts the individual's cardiovascular disease risk by using non-vascular disease death as a competing event; Model b predicts the individual's non-vascular disease death risk by using cardiovascular disease events as competing events. The combination of the two models can predict the lifetime risk of cardiovascular disease in patients with type 2 diabetes.

[0009] Preferably, the risk prediction method specifically includes the following steps:

[0010] S1: Clean up EHR data and determine the target population based on the type 2 diabetes patient inclusion flowchart;

[0011] S2: Based on a health big data platform, determine the baseline date, cardiovascular outcome date, competing event date, and study termination date for the target population. The follow-up termination date is the earliest recorded date among the first occurrence of a cardiovascular outcome, competing event, or study termination date for the study participants.

[0012] S3: When compiling patient baseline data, multidimensional characteristic levels, follow-up time, cardiovascular outcomes, and occurrence of competing events;

[0013] S4: Based on the variables included in the previously constructed cardiovascular disease prediction model for patients with type 2 diabetes, and considering the predictive value and accessibility of each variable in actual application scenarios, determine the predictive variables in the lifetime risk prediction model.

[0014] S5: For missing predictor variables at baseline, multiple imputation is performed using chain equations;

[0015] S6: Organize the target population data to obtain the occurrence of cardiovascular events and competing events for each individual at each age level from baseline age to the age at which follow-up was terminated;

[0016] S7: Using a competing risk-adjusted Fine & Gray prediction model in each age group of the target population, without adjusting any variables, the baseline risk of cardiovascular disease and competing events in each age group of the target population is obtained. "1-baseline risk" is the baseline survival probability of no cardiovascular disease and no competing events in each age group of the target population.

[0017] S8: In the target population, set individual birth dates, baseline dates, and follow-up termination dates. The study outcome is cardiovascular disease events, and the competing event is non-vascular disease death. Incorporate predefined predictive variables and construct a Fine&Gray competing risk prediction model to obtain the coefficients of each predictive variable in the cardiovascular disease model. With other settings unchanged, set the study outcome as non-vascular disease death events and the competing event as cardiovascular disease outcomes. Incorporate the same predictive variables and construct a Fine&Gray competing risk prediction model to obtain the coefficients of each predictive variable in the competing event model.

[0018] S9: Substitute the baseline survival probabilities for each age group obtained in S7 and the coefficients of each variable obtained in S8 into Equation 1.

[0019] P c(1-yearrisk) =1-(age-specific1-yrbaselinesurvival c )^exp(β ic X i (Equation 1);

[0020] S10: The baseline survival probability at the beginning of the year for the target population is set to 1. The individual's survival probability at the beginning of the next life year is obtained by using (survival probability at the beginning of the year - cardiovascular disease risk in the next life year interval - risk of competing events in the next life year interval); and so on, to obtain the individual's survival probability at the beginning of each life year.

[0021] S11: Substitute the individual's risk of cardiovascular events and competing events per lifespan year obtained from S9, and the initial survival probability per lifespan year obtained from S10, into Formula 2.

[0022] S c(eachlifeyear) =S c(beginningofeachlifeyear) *[1-(P c(1-yearrisk) (Equation 2);

[0023] S12: Calculate the individual's adjusted cardiovascular disease risk per life-year for competing events = S c(eachlifeyear) ×P c(eachlifeyear) ;

[0024] S13: The lifetime cardiovascular risk for patients with type 2 diabetes is the cumulative risk of cardiovascular disease per lifespan, adjusted for competing event risks, starting from baseline age and ending at age 85.

[0025] Preferably, in S6, the age groups are spaced one year apart.

[0026] Preferably, in S7, the baseline survival probability of the target population at each age level without cardiovascular disease and without competing events is age-specific 1-yrbaselinesurvival.

[0027] Preferably, in S8, the Fine&Gray competition risk function is expressed as Equation 3;

[0028]

[0029] Preferably, in the Fine&Gray competing hazard function, c represents the outcome event type, c=1 indicates a CVD event, c=2 indicates a non-cardiovascular death, x is a covariate matrix, and β... ic These are the partial regression coefficients for different covariates.

[0030] Preferably, in S9, P c(1-yearrisk) To calculate the risk of cardiovascular events and competing events per lifespan for an individual.

[0031] Preferably, in S11, S c(eachlifeyear) To calculate the cumulative survival probability of an individual per lifespan year.

[0032] (III) Beneficial Effects

[0033] Compared with existing technologies, this invention provides a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes, which has the following beneficial effects:

[0034] 1. This invention discloses a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes. Based on EHR patient data, the method involves cleaning the database to determine the baseline date, cardiovascular outcome date, competing event date, and study termination date for the target population; compiling the patient's baseline multidimensional characteristic levels, follow-up time, outcomes, and competing event occurrences; determining the predictive variables in the lifetime risk prediction model based on variables included in previously constructed cardiovascular disease prediction models for type 2 diabetes patients, considering the predictive value and accessibility of these variables in practical application scenarios; filling in missing variables at baseline; calculating the baseline survival probability of type 2 diabetes patients at different age groups in the region based on the above parameters; constructing a Fine&Gray competing risk prediction model with age as a time function based on the above patient information to obtain the coefficients of each predictive variable; and calculating the individual's cumulative cardiovascular disease risk at age 85 using the formula. This invention solves the problems of inapplicable parameters and inaccurate model evaluation caused by relying on existing foreign models. It considers the individual's competing event risk during the calculation process, making the individual cardiovascular disease risk assessment more accurate. Meanwhile, by considering multiple dimensions when including predictor variables, the model's application in patient populations is limited, making it more convenient to use in clinical and healthcare settings.

[0035] 2. This method predicts the lifetime risk of cardiovascular disease in patients with type 2 diabetes. By assessing the lifetime risk of cardiovascular disease in patients with type 2 diabetes, it enables medical professionals, patients, and policymakers to better understand the long-term risk status of patients with type 2 diabetes, screen out high-risk patients, and intervene as early as possible to achieve the effects of disease prevention and health promotion. Attached Figure Description

[0036] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0037] Figure 1 This is a flowchart for including patients with type 2 diabetes.

[0038] Figure 2 This is an intentional representation of the follow-up status of patients with type 2 diabetes in EHR data;

[0039] Figure 3 This is a diagram illustrating the calculation of cardiovascular disease risk per lifespan for patients with type 2 diabetes.

[0040] In the figure: BMI, Body Mass Index; SBP, Systolic Blood Pressure; eGFR, Estimated Glomerular Filtration Rate; TC / HDL-C, Ratio of Total Cholesterol to High-Density Lipoprotein Cholesterol. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0042] like Figure 1-3 As shown, this invention provides a method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes. The competing risk-adjusted Fine & Gray proportional hazards function includes two risk prediction models with the same predictor variables. Model a predicts the individual's cardiovascular disease risk by using non-vascular disease death as a competing event; Model b predicts the individual's non-vascular disease death risk by using cardiovascular disease events as competing events. The combination of the two models can predict the lifetime risk of cardiovascular disease in patients with type 2 diabetes.

[0043] Preferably, the risk prediction method specifically includes the following steps:

[0044] S1: Extract and clean EHR data from a regional health big data platform based on individual identification numbers, determine the target population based on the type 2 diabetes patient inclusion flowchart, select type 2 diabetes patients aged 25-85 years from January 1, 2010, and exclude patients with baseline kidney transplantation history, kidney dialysis or other renal insufficiency, cancer history or cardiovascular disease history.

[0045] S2: Based on the data collection characteristics of each database, determine the baseline date, cardiovascular outcome date, competing event date, and study termination date for the target population. Define the baseline date for newly diagnosed patients as the date of diabetes diagnosis, and define the study termination date as December 31, 2020.

[0046] S3: Compile the patient's baseline multidimensional characteristic levels, follow-up time, cardiovascular outcomes, and occurrence of competing events. The follow-up time is from the baseline time to the earliest date among the occurrence of cardiovascular outcomes, occurrence of competing events, or study termination date.

[0047] S4: Based on the variables included in the previously constructed cardiovascular disease prediction model for patients with type 2 diabetes, and considering the predictive value and accessibility of each variable in practical application scenarios, determine the predictive variables in the lifetime risk prediction model, including gender, duration of diabetes, body mass index (BMI), systolic blood pressure (SBP), glycated hemoglobin (HbA1c), total cholesterol to high-density lipoprotein cholesterol ratio (TC / HDLC), estimated glomerular filtration rate (eGFR), education level, current smoking status, smoking cessation status, insulin therapy status, antihypertensive drug therapy status, and aspirin therapy status.

[0048] S5: For missing predictor variables at baseline, multiple imputation is performed using chain equations;

[0049] S6: Organize the target population data to obtain the occurrence of cardiovascular events and competing events for each individual from baseline age to age at the end of follow-up (each age is separated by 1 year);

[0050] S7: In each age group of the target population, the Fine & Gray prediction model with competing risk adjustment is used without adjusting any variables to obtain the baseline risk of cardiovascular disease and competing events in each age group of the target population. "1-baseline risk" is the baseline survival probability of no cardiovascular disease and no competing events in each age group of the target population (the baseline survival probability of no cardiovascular disease and no competing events in each age group of the target population is age-specific 1-yr baseline survival).

[0051] S8: In the target population, set individual birth dates, baseline dates, and follow-up termination dates. The study outcome is cardiovascular disease events, and the competing event is non-vascular disease death. Incorporate predefined predictive variables and construct a Fine&Gray competing risk prediction model to obtain the coefficients of each predictive variable in the cardiovascular disease model. With other settings unchanged, set the study outcome as non-vascular disease death events and the competing event as cardiovascular disease outcomes. Incorporate the same predictive variables and construct a Fine&Gray competing risk prediction model to obtain the coefficients of each predictive variable in the competing event model.

[0052] The Fine&Gray competitive hazard function is expressed as Equation 1, where c represents the outcome event type, c=1 indicates a CVD event, c=2 indicates a non-cardiovascular death, x is the covariate matrix, and β... ic These are the partial regression coefficients for different covariates;

[0053]

[0054] S9: Substitute the baseline survival probabilities for each age group obtained in S7 and the coefficients of each variable obtained in S8 into Equation 2, P c(1-year risk) To calculate the risk of cardiovascular events and competing events per life-year interval for an individual;

[0055] P c(1-yearrisk) =1-(age-specific 1-yrbaseline survival c )^exp(β ic X i (Equation 2);

[0056] S10: The baseline survival probability at the beginning of the year for the target population is set to 1. The individual's survival probability at the beginning of the next life year is obtained by using (survival probability at the beginning of the year - cardiovascular disease risk in the next life year interval - risk of competing events in the next life year interval); and so on, to obtain the individual's survival probability at the beginning of each life year.

[0057] S11: Substitute the individual's risk of cardiovascular events and competing events per life-year interval obtained from S9, and the initial survival probability per life-year obtained from S10, into Equation 3. In S11, S... c(each life year) To calculate the cumulative survival probability of an individual per lifespan year interval;

[0058] S c(each life year) =S c(beginning ofeach life year) *[1-(P c(1-year risk) (Equation 3);

[0059] S12: Calculate the individual's adjusted cardiovascular disease risk per life-year for competing events = S c(eachlifeyear) ×Pc(each life year) ;

[0060] S13: The lifetime cardiovascular risk for patients with type 2 diabetes is the cumulative risk of cardiovascular disease per lifespan, adjusted for competing event risks, starting from baseline age and ending at age 85.

[0061] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes, characterized in that: The competing risk-adjusted Fine & Gray proportional hazards function comprises two risk prediction models with the same predictor variables. Model a uses non-vascular death as a competing event to predict an individual's cardiovascular disease risk; Model b uses cardiovascular events as competing events to predict an individual's non-vascular death risk. Combining the two models can predict the lifetime cardiovascular disease risk in patients with type 2 diabetes. The risk prediction method specifically includes the following steps: S1: Clean up EHR data and determine the target population based on the type 2 diabetes patient inclusion flowchart; S2: Based on the health big data platform, determine the baseline date, cardiovascular disease outcome date, competing event date, and study termination date for the target population. The follow-up termination date is the earliest date among the first occurrence of cardiovascular disease outcome, competing event, or study termination date of the study subjects. S3: When compiling patient baseline data, multidimensional characteristic levels, follow-up time, cardiovascular outcomes, and occurrence of competing events; S4: Based on the variables included in the previously constructed cardiovascular disease prediction model for patients with type 2 diabetes, and considering the predictive value and accessibility of each variable in actual application scenarios, determine the predictive variables in the lifetime risk prediction model. S5: For missing predictor variables at baseline, multiple imputation is performed using chain equations; S6: Organize the target population data to obtain the occurrence of cardiovascular events and competing events for each individual from baseline age to age at the end of follow-up; S7: Using a competing risk-adjusted Fine & Gray prediction model in each age group of the target population, without adjusting any variables, the baseline risk of cardiovascular disease and competing events in each age group of the target population is obtained. "1-baseline risk" is the baseline survival probability of no cardiovascular disease and no competing events in each age group of the target population. S8: In the target population, set individual birth dates, baseline dates, and follow-up termination dates. The study outcome is cardiovascular disease events, and the competing event is non-vascular disease death. Incorporate predefined predictive variables and construct a Fine & Gray competing risk prediction model to obtain the coefficients of each predictive variable in the cardiovascular disease model. With other settings unchanged, set the study outcome as non-vascular disease death events and the competing event as cardiovascular disease outcomes. Incorporate the same predictive variables and construct a Fine & Gray competing risk prediction model to obtain the coefficients of each predictive variable in the competing event model. S9: Substitute the baseline survival probabilities for each age group obtained in S7 and the coefficients of each variable obtained in S8 into Equation 1. P c(1-year risk) =1-(age-specific 1-year baseline survival c )^exp(β ic X i ) Equation 1; S10: The baseline survival probability at the beginning of the year for the target population is set to 1. The individual's survival probability at the beginning of the next life year is obtained by using "the survival probability at the beginning of the year - the cardiovascular disease risk of the individual in the next life year interval - the risk of competing events in the individual in the next life year interval". This process is repeated to obtain the individual's survival probability at the beginning of each life year. S11: Substitute the individual's risk of cardiovascular events and competing events per lifespan year obtained from S9, and the initial survival probability per lifespan year obtained from S10, into Equation 2. S c(each life year) =S c(beginning of each life year) *[1- (P c(1-year risk) Formula 2; S12: Calculate the individual's adjusted cardiovascular disease risk per life-year for competing events = S c(each life year) ×P c(each life year) ; S13: The lifetime cardiovascular risk for patients with type 2 diabetes is the cumulative risk of cardiovascular disease per lifespan, adjusted for competing event risks, starting from baseline age and ending at age 85.

2. The method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes according to claim 1, characterized in that: In S6, each age group is spaced one year apart.

3. The method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes as described in claim 1, characterized in that: In S7, the baseline survival probability of the target population at each age level without cardiovascular disease and without competing events is age-specific 1-yr baseline survival.

4. The method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes as described in claim 1, characterized in that: In S8, the Fine & Gray competing hazard function is expressed as Equation 3; in the Fine & Gray competing hazard function, c is the outcome event type, c=1 is the CVD event, c=2 is the non-cardiovascular death, x is the covariate matrix, and β is the covariate matrix. ic These are the partial regression coefficients for different covariates; Formula 3.

5. The method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes according to claim 1, characterized in that: In S9, P c(1-year risk) To calculate the risk of cardiovascular events and competing events per lifespan for an individual.

6. The method for predicting the lifetime risk of cardiovascular disease in patients with type 2 diabetes according to claim 1, characterized in that: In S11, S c(each life year) To calculate the cumulative survival probability of an individual per lifespan year.

Images