Judgment method for dynamic adjustment of hypertension diagnosis boundary value based on network structure learning

Abstract

The invention relates to a judgment method for dynamic adjustment of a hypertension diagnosis boundary value based on network structure learning. The method comprises the following steps: firstly, carrying out data-related preprocessing of repeated values, missing values, abnormal values and the like on crowd health-related data; secondly, selecting a logically reasonable index combination through a network structure learning method, constructing a Bayesian network by using three algorithms of Tabu Search, HC and MMHC, and determining direct related factors of hypertension through network node and structure analysis; and finally, analyzing and comparing the difference of each factor in different years by utilizing a statistical method, and judging whether the diagnosis threshold of hypertension needs to be dynamically adjusted or not according to a statistical index result.

Description

Technical Field

[0001] The present invention relates to an evaluation method for monitoring chronic disease diagnostic thresholds, and in particular to a determination method for dynamic adjustment of hypertension diagnostic thresholds based on network structure learning. Background Art

[0002] The diagnostic threshold for hypertension has been verified through long-term epidemiological investigations and clinical evidence-based medicine. The diagnostic criteria have been adjusted many times over time, but it has been a long time since the last revision (2018). With the rapid changes in people's daily behavior patterns such as living habits and eating habits, this standard may no longer be able to reflect the actual blood pressure health status of the current population. Therefore, it is necessary to re-evaluate and adjust the diagnostic threshold for hypertension. At present, there is no method to determine whether the diagnostic threshold for hypertension needs to be dynamically adjusted.

[0003] 1. Technical issues to be solved

[0004] In view of the deficiencies of existing research, the present invention provides a determination method for dynamic adjustment of hypertension diagnostic thresholds based on network structure learning, which solves the problems raised in the background technology.

[0005] (II) Technical solution

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a determination method for dynamic adjustment of hypertension diagnostic threshold based on network structure learning, the steps comprising the following:

[0007] S1: Collect relevant data on population health in a certain area for at least 10 consecutive years;

[0008] S2: Preprocess the collected data, use network structure learning to build a Bayesian network and select a logically reasonable combination of indicator arrangements;

[0009] S3: Construct different Bayesian networks and perform data analysis, and select the direct related factors of hypertension based on the data analysis results;

[0010] S4: Use statistical methods to analyze and compare the differences of various factors in different years;

[0011] S5: Output the results and determine whether the hypertension diagnostic threshold needs to be adjusted.

[0012] Preferably, the population health-related data in S1 mainly includes the following contents:

[0013] (1) Baseline data of the population: gender, age, total family population, annual family income, height (cm), weight (kg), BMI, race / ethnicity, education level, marital status, occupation, etc.;

[0014] (2) Lifestyle data: whether smoking, drinking, exercise intensity, etc.;

[0015] (3) Health data: systolic blood pressure, diastolic blood pressure, diabetes, kidney weakness, kidney stones, coronary heart disease, angina pectoris, stroke, sleep disorders, congestive heart failure, high cholesterol level, heart rate, pulse, whether overweight, etc.

[0016] Preferably, in step S2, the specific content of data processing is to preprocess the data, such as eliminating duplicate, missing, and abnormal data, filling in missing values ​​according to the data type and actual situation, and using the bnlearn package in the R software to draw the network structure of the indicators, and screen out logically reasonable combinations for the next step of analysis.

[0017] Preferably, in step S3, a Bayesian network is constructed, and network node analysis and network structure analysis are performed using three algorithms, Tabu Search, HC, and MMHC, and a network topology is drawn. And factors directly related to hypertension are determined based on the analysis results. In the network topology, the parent node is the cause, i.e., the factor affecting hypertension; the child node is the result, i.e., the result that hypertension will cause. If two or more algorithms show that a certain factor is a directly related factor under any logical combination, the factor is defined as a cause (parent node) or a result (child node).

[0018] Preferably, the statistical analysis method in step S4 is: let the actual frequency of illness in a certain year be P 11 , the theoretical frequency is D 11 , the actual frequency of people without disease is P 12 , the theoretical frequency is D 12 , the total number for that year is n R1 The actual frequency of illness in the next year of the counting year is P 21 , the theoretical frequency is D 21 , the actual frequency of people without disease is P 22 , the theoretical frequency is D 22 , the total number for that year is n R2 The total number of illnesses in two years is n C1 , the total number of non-diseased people is n C2 , the total number is n. Then use the actual frequency and theoretical frequency to perform statistical tests:

[0019]

[0020] The test was a two-sided test, with α set at 0.05, and the χ 2 The distribution table shows that the boundary value is 3.96. 2 When the value is greater than 3.96, it means that the difference is statistically significant, indicating that there are differences in this factor between different years.

[0021] Preferably, in step S5, the judgment criterion according to the output result is: a significant change is represented as "1", and no significant change is represented as "0", then the difference results of significant changes in causal factors are (1, 0), (0, 1), (1, 1), (0, 0), respectively, the difference results of blood pressure are "1" and "0", respectively, the situations that need adjustment are (1, 1, 1), (0, 0, 0), respectively, and the situations that do not need adjustment are (1, 0, 0), (1, 0, 1), "(0, 1, 0), (0, 1, 1), (1, 1, 0), (0, 0, 1), respectively.

[0022] (III) Beneficial effects

[0023] The present invention provides a method for determining the dynamic adjustment of the hypertension diagnosis threshold based on network structure learning. It has the following beneficial effects:

[0024] The present invention mainly uses Bayesian network structure learning as the main method, determines the direct related factors of hypertension through network node analysis and structure analysis of three algorithms, Tabu Search, HC and MMHC, and determines the direct cause and direct result of hypertension according to the position of the parent node and the child node. Compared with ordinary correlation analysis, the present invention can not only determine the related factors, but also clarify the causal factors. The causal factors of hypertension and whether there is a significant change between different years of hypertension can be determined by statistical analysis methods, and the dynamic monitoring of the diagnostic threshold of hypertension can be carried out according to the data of different years to judge whether the diagnostic threshold of hypertension needs to be adjusted, which is more suitable for the actual health status of the population. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A step diagram of a determination method for dynamic adjustment of hypertension diagnosis threshold based on network structure learning according to the present invention;

[0026] Figure 2 This is the preprocessing result of population health-related data from 2005 to 2018;

[0027] Figure 3 This is a logically reasonable network structure diagram selected under the three algorithms;

[0028] Figure 4 The direct causes and direct consequences of hypertension obtained by three different algorithms under six structures;

[0029] Figure 5 The direct causes and direct consequences of high blood pressure are calculated;

[0030] Figure 6 These are the hypertension data results from 2005 to 2006 and 2007 to 2008;

[0031] Figure 7 The results are for diabetes data from 2005 to 2006 and 2007 to 2008;

[0032] Figure 8 These are the stroke data results from 2005 to 2006 and 2007 to 2008. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0034] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0035] See also Figure 1 A method for dynamically adjusting the diagnostic threshold of hypertension based on network structure learning comprises the following steps:

[0036] S1: Collect relevant data on population health in a certain area for at least 10 consecutive years;

[0037] S2: Preprocess the collected data, use network structure learning to build a Bayesian network and select a logically reasonable combination of indicator arrangements;

[0038] S3: Construct different Bayesian networks and perform data analysis, and select the direct related factors of hypertension based on the data analysis results;

[0039] S4: Use statistical methods to analyze and compare the differences of various factors in different years;

[0040] S5: Output the results and determine whether the hypertension diagnostic threshold needs to be adjusted.

[0041] Preferably, the population health-related data in S1 mainly includes the following contents:

[0042] (1) Baseline data of the population: gender, age, total family population, annual family income, height (cm), weight (kg), BMI, race / ethnicity, education level, marital status, occupation, etc.;

[0043] (2) Lifestyle data: whether smoking, drinking, exercise intensity, etc.;

[0044] (3) Health data: systolic blood pressure, diastolic blood pressure, diabetes, kidney weakness, kidney stones, coronary heart disease, angina pectoris, stroke, sleep disorders, congestive heart failure, high cholesterol level, heart rate, pulse, whether overweight, etc.

[0045] Preferably, in step S2, the specific content of data processing is to preprocess the data, such as eliminating duplicate, missing, and abnormal data, filling in missing values ​​according to the data type and actual situation, and using the bnlearn package in the R software to draw the network structure of the indicators, and screen out logically reasonable combinations for the next step of analysis.

[0046] Preferably, in step S3, a Bayesian network is constructed, and network node analysis and network structure analysis are performed using three algorithms, Tabu Search, HC, and MMHC, and a network topology is drawn. And factors directly related to hypertension are determined based on the analysis results. In the network topology, the parent node is the cause, i.e., the factor affecting hypertension; the child node is the result, i.e., the result that hypertension will cause. If two or more algorithms show that a certain factor is a directly related factor under any logical combination, the factor is defined as a cause (parent node) or a result (child node).

[0047] Preferably, the statistical analysis method in step S4 is: let the actual frequency of illness in a certain year be P 11 , the theoretical frequency is D 11 , the actual frequency of people without disease is P 12 , the theoretical frequency is D 12 , the total number for that year is n R1 The actual frequency of illness in the next year of the counting year is P 21 , the theoretical frequency is D 21 , the actual frequency of people without disease is P 22 , the theoretical frequency is D 22 , the total number for that year is n R2 The total number of illnesses in two years is n C1 , the total number of non-diseased people is n C2 , the total number is n. Then use the actual frequency and theoretical frequency to perform statistical tests:

[0048]

[0049] The test was a two-sided test, with α set at 0.05, and the χ 2 The distribution table shows that the boundary value is 3.96. 2 When the value is greater than 3.96, it means that the difference is statistically significant, indicating that there are differences in this factor between different years.

[0050] Preferably, in step S5, the judgment criterion according to the output result is: a significant change is represented as "1", and no significant change is represented as "0", then the difference results of significant changes in causal factors are (1, 0), (0, 1), (1, 1), (0, 0), respectively, the difference results of blood pressure are "1" and "0", respectively, the situations that need adjustment are (1, 1, 1), (0, 0, 0), respectively, and the situations that do not need adjustment are (1, 0, 0), (1, 0, 1), "(0, 1, 0), (0, 1, 1), (1, 1, 0), (0, 0, 1), respectively.

[0051] Implementation examples:

[0052] Taking the population health data from 2005 to 2018 in a database as an example, the specific implementation steps are as follows:

[0053] S1: Import population health data related to hypertension, including:

[0054] (1) Baseline data of the population: sex (male, female), age (≥18 years), BMI, race (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other race), education level (less than 9th grade, 9-11th grade, high school graduate / GED or equivalent, college or AA degree, college graduate or above), marital status (married / living with partner, widowed / divorced / separated, never married);

[0055] (2) Lifestyle data: whether smoking or drinking;

[0056] (3) Health data: Surveys on residents’ health status, including whether they suffer from high blood pressure, diabetes, kidney weakness, coronary heart disease, angina pectoris, stroke, sleep disorders, etc.

[0057] S2: Preprocess the data, such as deleting duplicates, filling missing values ​​according to data type and actual situation, and converting data into categorical variables. The results of preprocessing some population data are as follows: Figure 2 shown.

[0058] Analyze the data set, use the bnlearn package in R software to draw the network topology structure of the indicators, and screen out logically reasonable combinations. Logical combinations are as follows: Figure 3 shown.

[0059] S3: Using Tabu Search, HC, and MMHC algorithms to analyze network nodes and network structure and draw network topology, we can get the direct causes and direct results of hypertension obtained by three different algorithms under six structures, such as Figure 4 shown.

[0060] If two or more algorithms show that a factor is a directly related factor under any logical combination, then the factor is defined as a cause (parent node) or a result (child node). Therefore, the direct causes of high blood pressure are race, marital status, and diabetes; the direct results of high blood pressure are kidney deficiency and stroke. Figure 5 shown.

[0061] S4: Use statistical analysis methods to compare whether the changes of various factors between different years are significantly different:

[0062] Taking the data of blood pressure, diabetes and stroke from 2005 to 2008 as an example, we conducted a χ 2 The test was used to compare whether the factors were significantly different between 2005-2006 and 2007-2008.

[0063] Hypertension data results such as Figure 6 Compared with the data from 2005 to 2006, the number of patients with hypertension increased in 2007 to 2008. 2 The value is 2.721<3.96, which means that the difference is not statistically significant, indicating that there is no significant difference in blood pressure between 2005-2006 and 2007-2008.

[0064] Diabetes data results such as Figure 7 As shown in Figure 2, the number of people with diabetes increased compared with the data from 2005 to 2006. 2 The value is 9.005>3.96, which means that the difference is statistically significant, indicating that there is a significant difference in diabetes between 2005-2006 and 2007-2008.

[0065] Stroke data results Figure 8 As shown in Figure 2, the number of stroke cases increased in 2007-2008 compared with 2005-2006. 2 The value is 3.961>3.96, which means that the difference is statistically significant, indicating that there is a significant difference in stroke between 2005-2006 and 2007-2008.

[0066] S5: Determine whether the hypertension diagnostic threshold needs to be adjusted based on the output results.

[0067] The judgment process is explained by taking the changes in cause (diabetes), effect (stroke) and hypertension from 2005 to 2008 as an example. According to the statistical output results of the three, there are significant differences in diabetes and stroke between 2005-2006 and 2007-2008, but there is no significant difference in hypertension between 2005-2006 and 2007-2008. Therefore, the diagnostic threshold of hypertension needs to be adjusted.

[0068] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0069] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for dynamically adjusting the threshold of hypertension diagnosis based on network structure learning, characterized in that: The steps include: S1: Collect relevant data on population health in a certain area for at least 10 consecutive years; S2: Preprocess the collected data, use network structure learning to build a Bayesian network and select a logically reasonable combination of indicator arrangements; S3: Construct different Bayesian networks and perform data analysis, and select the direct related factors of hypertension based on the data analysis results; S4: Use statistical methods to analyze and compare the differences of various factors in different years; S5: Output the results and determine whether the hypertension diagnostic threshold needs to be adjusted.

2. The method for dynamically adjusting the hypertension diagnosis threshold based on network structure learning according to claim 1, characterized in that: The population health-related data in S1 mainly include the following: (1) Baseline data of the population: gender, age, total family population, annual family income, height (cm), weight (kg), BMI, race / ethnicity, education level, marital status, occupation, etc.; (2) Lifestyle data: whether smoking, drinking, exercise intensity, etc.; (3) Health data: systolic blood pressure, diastolic blood pressure, diabetes, kidney weakness, kidney stones, coronary heart disease, angina pectoris, stroke, sleep disorders, congestive heart failure, high cholesterol level, heart rate, pulse, whether overweight, etc.

3. The method for dynamically adjusting the threshold of hypertension diagnosis based on network structure learning according to claim 1, characterized in that: In step S2, the specific content of data processing is to pre-process the data, such as eliminating duplicate, missing, and abnormal data, filling missing values ​​according to the data type and actual situation, and using the bnlearn package in the R software to draw the network structure of the indicators, and screen out logically reasonable combinations for the next step of analysis.

4. The method for dynamically adjusting the hypertension diagnosis threshold based on network structure learning according to claim 1, characterized in that: In the step S3, a Bayesian network is constructed, and network node analysis and network structure analysis are performed using three algorithms: Tabu Search, HC, and MMHC, and a network topology is drawn. And factors directly related to hypertension are determined based on the analysis results. In the network topology, the parent node is the cause, i.e., the factor affecting hypertension; the child node is the result, i.e., the result that hypertension will cause. If two or more algorithms show that a certain factor is a directly related factor under any logical combination, the factor is defined as a cause (parent node) or a result (child node).

5. The method for dynamically adjusting the threshold of hypertension diagnosis based on network structure learning according to claim 1, characterized in that: The statistical analysis method in step S4 is: let the actual frequency of illness in a certain year be P 11 , the theoretical frequency is D 11 , the actual frequency of people without disease is P 12 , the theoretical frequency is D 12 , the total number for that year is n R1 The actual frequency of illness in the next year of the counting year is P 21 , the theoretical frequency is D 21 , the actual frequency of people without disease is P 22 , the theoretical frequency is D 22 , the total number for that year is n R2 The total number of illnesses in two years is n C1 , the total number of non-diseased people is n C2 , the total number is n. Then use the actual frequency and theoretical frequency to perform statistical tests: The test was a two-sided test, with α set at 0.05, and the χ 2 The distribution table shows that the boundary value is 3.

96. 2 When the value is greater than 3.96, it means that the difference is statistically significant, indicating that there are differences in this factor between different years.

6. The method for dynamically adjusting the hypertension diagnosis threshold based on network structure learning according to claim 1, characterized in that: In step S5, the judgment criteria of the output results are: significant changes are represented as "1", and no significant changes are represented as "0". The difference results of significant changes in causal factors are (1, 0), (0, 1), (1, 1), (0, 0), and the difference results of blood pressure are "1" and "0", respectively. The situations that need adjustment are (1, 1, 1), (0, 0, 0), respectively. The situations that do not need adjustment are (1, 0, 0), (1, 0, 1), "(0, 1, 0), (0, 1, 1), (1, 1, 0), (0, 0, 1).

Images