A device for predicting the effects of lifestyle interventions, a system for predicting the effects of lifestyle interventions, a method for predicting the effects of lifestyle interventions, and a program.

Abstract

Based on the data entered by the medical examinee, the system encourages the examinee to proactively change their lifestyle habits. [Solution] The lifestyle intervention effect prediction device 1 includes: a patient ID identification unit 11 that uses a medical examination data database 17 storing medical examination data for multiple patient IDs to identify each patient ID related to data whose distance from the medical examination data for multiple medical examination items entered by the user is closer than a predetermined relative standard to the data of the user's improved lifestyle; an intervention effect prediction unit 12 that refers to each evaluation value evaluated by the intervention effect prediction model 16 based on each patient ID identified by the patient ID identification unit 11 and the medical examination data database 17 as each predicted result of the user's improved lifestyle; and an external communication unit 13 that outputs each predicted result referenced by the intervention effect prediction unit 12.

Description

Technical Field

[0001] The present invention relates to a lifestyle intervention effect prediction device, a lifestyle intervention effect prediction system, a lifestyle intervention effect prediction method, and a program.

Background Art

[0002] Although a large amount of medical examination data has been accumulated over many years in many hospitals, the effect of lifestyle guidance by doctors cannot be predicted. As a result, the realization of preventive medicine has been delayed. Here, the applicants have developed an AI (Artificial Intelligence) that predicts the effect of intervention on lifestyle using the medical examination data of hospitals. This can be a basic technology for realizing preventive medicine.

[0003] In Patent Document 1 of the AI developed by the applicants for predicting the effect of intervention on lifestyle, an invention for predicting the effects of a plurality of interventions on a person is described. In the abstract of this Patent Document 1, it is described that "a computer system manages a first model that generates feature quantities by mapping a vector composed of values of a plurality of factors representing a person's state to a feature quantity space by machine learning, and a second model that outputs predicted values of the effects of a plurality of interventions on a person from the feature quantities. The first model maps the plurality of learning data to the feature quantity space so that the difference in the distribution of the plurality of learning data in the feature quantity space used in the machine learning is small. The computer system receives input data including values of a plurality of factors, generates the feature quantity of the input data by inputting the input data into the first model, and calculates the predicted values of the effects of a plurality of interventions by inputting the feature quantity of the input data into the second model."

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] AI for predicting the intervention effect on lifestyle has been developed in many studies using open data, but there is no established scheme for providing feedback to the examinee on the predicted intervention effect on lifestyle and prompting the examinee to change their lifestyle themselves. The disease prediction conducted by insurance companies is a long-term probability prediction of whether a person will develop a disease in the future, and does not involve prediction, feedback, or intervention on lifestyle.

[0006] Therefore, an object of the present invention is to prompt an examinee to change their lifestyle based on the input data of the examinee.

Means for Solving the Problems

[0007] To solve the above problems, a lifestyle intervention effect prediction device of the present invention uses a medical examination data database storing medical examination data and lifestyle data of a plurality of items related to a plurality of examinee IDs, and from the data related to a plurality of items input by a user, identifies each examinee ID related to data whose distance from the data in a state where each lifestyle of the user is improved is closer than a predetermined relative standard by an examinee ID identification unit, and uses each evaluation value evaluated by an intervention effect prediction model based on each examinee ID identified by the examinee ID identification unit and the medical examination data database as each prediction result when the user improves each lifestyle, and is characterized by including an intervention effect prediction unit for referring to the prediction results and an output unit for outputting each prediction result referred to by the intervention effect prediction unit.

[0008] The lifestyle intervention effect prediction system of the present invention is characterized by comprising: a patient ID identification unit that uses a medical examination data database storing medical examination data and lifestyle data for multiple items related to multiple patient IDs to identify each patient ID related to data whose distance from data for multiple items entered by the user into a terminal is closer than a predetermined relative standard to data for the user's improved lifestyle; an intervention effect prediction unit that refers to each evaluation value evaluated by an intervention effect prediction model based on each patient ID identified by the patient ID identification unit and the medical examination data database as each predicted result of the user's improved lifestyle; and an output unit that outputs each of the predicted results referred by the intervention effect prediction unit to the display unit of the terminal.

[0009] The present invention provides a method for predicting the effectiveness of lifestyle interventions, comprising the steps of: using a medical examination data database containing medical examination data and lifestyle data for multiple items related to multiple medical examination IDs, a medical examination ID identification unit identifies each medical examination ID related to data whose distance from data relating to the user's improved lifestyle is closer than a predetermined relative standard, based on data relating to multiple items entered by the user; an intervention effect prediction unit references each evaluation value evaluated by an intervention effect prediction model based on each medical examination ID identified by the medical examination ID identification unit and the medical examination data database as each predicted result of the user's improved lifestyle; and an output unit displays each of the predicted results referenced by the intervention effect prediction unit on a display unit.

[0010] The present invention provides a program for causing a computer to perform the following steps: use a medical examination data database containing medical examination data and lifestyle data for multiple items related to multiple medical examination IDs to identify each medical examination ID related to data that is closer to a predetermined relative standard than the data of the user's improved lifestyle from data for multiple items entered by the user; refer to each evaluation value evaluated by an intervention effect prediction model based on each identified medical examination ID and the medical examination data database as each predicted result of the user's improved lifestyle; and display each of the referenced predicted results on a display unit. Other means will be described within the descriptions of embodiments for carrying out the invention. [Effects of the Invention]

[0011] According to the present invention, it is possible to encourage examinees to make changes to their lifestyles based on the data they input. [Brief explanation of the drawing]

[0012] [Figure 1] This figure shows the configuration and operation of the lifestyle intervention effect prediction system according to this embodiment. [Figure 2] This is a physical configuration diagram of a device for predicting the effects of lifestyle interventions. [Figure 3] This is a flowchart of the application process for inter-item networks. [Figure 4A] This figure (Part 1) shows the correlations between each factor and the factors themselves. [Figure 4B] This figure (part 2) shows the correlations between each factor and the factors themselves. [Figure 5] This figure (part 3) shows the correlations between each factor and the factors themselves. [Figure 6A] This figure (Part 1) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 6B] This figure (part 2) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 6C]This figure (part 3) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 6D] This figure (part 4) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 7A] This figure (part 5) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 7B] This figure (part 6) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 7C] This figure (part 7) shows the process of narrowing down covariates using an inter-item network that excludes confounding factors. [Figure 8] This figure shows an example of an inter-item network. [Figure 9A] This is a diagram (part 1) illustrating correction rule #1 for the inter-item transition matrix. [Figure 9B] This is a diagram (part 2) illustrating correction rule #1 for the inter-item transition matrix. [Figure 10A] This is a diagram (part 1) illustrating correction rule #2 for the inter-item transition matrix. [Figure 10B] This is a diagram (part 2) illustrating correction rule #2 for the inter-item transition matrix. [Figure 11A] This is a diagram (part 1) illustrating the rules that determine opposing correlations. [Figure 11B] This is a diagram (part 2) illustrating the rules that determine opposing correlations. [Figure 12A] This is a diagram (part 1) illustrating the rules that determine opposing correlations. [Figure 12B] This is a diagram (part 2) illustrating the rules that determine opposing correlations. [Figure 12C] This is a diagram (part 3) that explains the rules for determining opposing correlations. [Figure 13] This is a flowchart of the process for creating inter-item networks. [Figure 14] This is a flowchart for the covariate filtering process using an inter-item transition matrix. [Figure 15] This is a flowchart for predicting the intervention effect for first-time patients undergoing screening. [Figure 16] This diagram shows an example of an input screen displayed on a terminal. [Figure 17] This figure shows an example of a prediction results screen displayed on a device. [Figure 18] This figure shows an example of a feedback screen displayed on a device. [Figure 19] This is a flowchart for predicting the intervention effect. [Figure 20] This diagram shows an example of an input screen displayed on a terminal. [Figure 21] This figure shows an example of a prediction results screen displayed on a device. [Modes for carrying out the invention]

[0013] Hereafter, embodiments for carrying out the present invention will be described in detail with reference to the figures.

[0014] The lifestyle intervention effect prediction system of the present invention uses a smartphone application program to provide feedback to the person undergoing health checkups on the predicted effects of lifestyle interventions predicted by an intervention effect prediction model. The person undergoing health checkups inputs methods for changing their lifestyle into the smartphone app. In response to the person's input, the intervention effect prediction model, which is an AI that predicts the effects of interventions, predicts the effects and provides feedback to the person's smartphone. This input and output of prediction results can be easily repeated many times, encouraging the person undergoing health checkups to make changes to their lifestyle themselves. In this way, the lifestyle intervention effect prediction system can recommend lifestyle improvements and provide feedback that is easy for the user to understand.

[0015] Figure 1 shows the configuration and operation of the lifestyle intervention effect prediction system 100 according to this embodiment.

[0016] The lifestyle intervention effect prediction system 100 is configured such that the user's terminal 2 and the lifestyle intervention effect prediction device 1 are connected via a network or the like. Based on the data input by the person undergoing the health checkup, the lifestyle intervention effect prediction system 100 displays information on the terminal 2 that encourages the person undergoing the health checkup to make changes to their lifestyle.

[0017] Terminal 2 is, for example, a smartphone or computer owned by the user, and is composed of an input unit 21 and a display unit 22. The input unit 21 is, for example, the touch panel portion of a touch panel display, and the display unit 22 is, for example, the display portion of a touch panel display. The display unit 22 functions as an output unit that outputs the predicted results of improvements to each lifestyle habit in order of their effectiveness. Terminal 2 is a terminal on which the user, who is undergoing a medical examination, inputs their own information. The user who inputs medical examination data into Terminal 2 is not limited to the medical examination data stored in the medical examination data database 17 described later, but may also be a medical examination for the first time.

[0018] The lifestyle intervention effect prediction device 1 is a server installed, for example, in a data center, and consists of a patient ID identification unit 11, an intervention effect prediction unit 12, an external communication unit 13, a comprehensive PCIT unit 18, an inter-item network creation unit 14, a covariate narrowing unit 15, a model learning unit 19, an intervention effect prediction model 16, and a health checkup data database 17. Each functional unit is realized by executing a program installed on this server.

[0019] The information request unit 10 requests health checkup data for each health checkup item and lifestyle data for each lifestyle habit from the user's terminal 2, based on the covariates narrowed down by the health checkup item network. The information request unit 10 outputs the health checkup data and / or lifestyle data for multiple items obtained from terminal 2 to the health checkup ID identification unit 11.

[0020] The examiner ID identification unit 11 uses the examination data database 17 to identify each examiner ID corresponding to the data that is the shortest distance from the data showing the user's improved lifestyle, based on the examination data and / or lifestyle data for multiple items entered by the user.

[0021] The examiner ID identification unit 11 normalizes the numerical values ​​of each item in the examination data database 17 and the data showing the user's improved lifestyle habits, and calculates an evaluation value for the distance between the data showing the user's improved lifestyle habits and the data for each examiner ID. Based on the evaluation value of the distance, the examiner ID identification unit 11 identifies the examiner IDs associated with the data that have the smallest distance to the data showing the user's improved lifestyle habits. As the evaluation value of distance in the data space, a function with Euclidean distance or norm as a variable can be used, but a function with Manhattan distance or Chebyshev distance as a variable may also be used and is not limited to this.

[0022] If the distance evaluation value increases as the distance increases, the examiner ID identification unit 11 identifies each examiner ID related to the data with the minimum distance evaluation value for data showing the user's improved lifestyle habits. If the distance evaluation value decreases as the distance increases, the examiner ID identification unit 11 identifies each examiner ID related to the data with the minimum distance evaluation value for data showing the user's improved lifestyle habits.

[0023] Furthermore, the examiner ID identification unit 11 may identify each examiner ID related to data whose proximity to the data showing the user's improved lifestyle habits is higher than a predetermined order. For example, when the predetermined order is defined as 5, the examiner ID may identify the examiner ID related to any of the data from 1st to 4th in terms of proximity. In other words, the examiner ID identification unit 11 may identify each examiner ID related to data whose distance to the data showing the user's improved lifestyle habits is closer than a predetermined relative standard. Moreover, the examiner ID identification unit 11 may identify each examiner ID related to data whose distance to the data showing the user's improved lifestyle habits is less than or equal to a predetermined value.

[0024] The intervention effect prediction unit 12 refers to each evaluation value evaluated by the intervention effect prediction model 16 based on each examinee ID identified by the examinee ID identification unit 11 and the examination data database 17 as each predicted result of the user's improvement in lifestyle habits. The intervention effect prediction unit 12 causes the intervention effect prediction model 16 to evaluate each evaluation value based on each examinee ID identified by the examinee ID identification unit 11 and the normalized values ​​of each item in the examination data database 17.

[0025] The external communication unit 13 transmits the intervention effect predicted by the intervention effect prediction unit 12 to the terminal 2, causing the predicted intervention effect to be displayed on the display unit 22 of the terminal 2. The external communication unit 13 also functions as an output unit that outputs each prediction result referenced by the intervention effect prediction unit 12.

[0026] The health checkup data database 17 is a database that stores health checkup data from multiple individuals and lifestyle data of these individuals. The comprehensive PCIT unit 18 comprehensively performs PCIT (Partial Correlation Coefficient) determination. The comprehensive PCIT unit 18 functions as a mathematical determination processing unit that determines whether or not to exclude confounding factors. The present invention is not limited to the comprehensive PCIT unit 18 that comprehensively performs PCIT determination, but may also be a functional unit that comprehensively performs ARACNE determination as described in Figure 5 below, and is not limited thereto.

[0027] The inter-item network creation unit 14 creates an inter-item network that excludes confounding factors between screening items, based on the Partial Correlation Coefficient (PCIT) judgment results performed by the comprehensive PCIT unit 18. The inter-item network creation unit 14 may, but is not limited to, generate the inter-item network by excluding confounding factors between screening items using mutual information based on a normalized frequency distribution.

[0028] The covariate filtering unit 15 uses the inter-item network created by the inter-item network creation unit 14 to filter the covariates and calculate the inter-item transition matrix. This allows the information request unit 10 and the intervention effect prediction unit 12 to narrow down the examination items to be used for evaluation. The model learning unit 19 retrains the intervention effect prediction model 16, which will be described later, using the covariates narrowed down by the covariate narrowing unit 15. The intervention effect prediction model 16 accepts input items that are limited to screening items narrowed down based on the inter-item transition matrix, and then predicts the intervention effect by referring to the screening data database 17.

[0029] The user inputs their latest health checkup results and lifestyle improvement items into the input unit 21 of terminal 2. The intervention effect prediction model 16 then predicts the effect of lifestyle improvements. The display unit 22 of terminal 2 provides feedback to the user by displaying the predicted health status for 1 to 5 years in the future. This makes it possible to encourage the health checkup participant to make changes to their lifestyle based on the data they input. The number of years that can be predicted depends on how many years of data are available in the database; if there is more data than the number of years to predict, predictions of 6 years or more are possible.

[0030] The health checkup data database 17 stores data from standard health checkups and cancer screening checkups collected over several years. The health checkup data database 17 contains multiple health checkup items K i And, Lifestyle item S j It is composed of the following. Missing values ​​in the health checkup data database 17 are imputed using the median values ​​of the preceding and succeeding values ​​of the health checkup items.

[0031] The intervention effect prediction unit 12 uses the screening item K. i Each time, this examination item K i The overall median or mean value μ i And, examination item K i Overall variance σ i Normalization is performed on all examination items using this method. The intervention effect prediction unit 12 uses the lifestyle item S j The same normalization process is applied to all items.

[0032] The intervention effect prediction model 16 is a machine learning model that predicts multiple interventions simultaneously. The intervention effect prediction model 16 can learn from time-series data, can handle multiple intervention items, and can represent intervention effects as continuous values. Physicians have confirmed that the prediction results of the intervention effect prediction model 16 do not deviate significantly from common medical knowledge.

[0033] This invention allows for the refinement of covariates in the intervention effect prediction model 16 using the intervention effect prediction model 16 and an item network. Therefore, in this embodiment, when providing a smartphone application program that presents the intervention prediction results to the user using the intervention effect prediction model 16, it is possible to reduce user input using the item network. Furthermore, in this embodiment, it is possible to reduce the amount of training required for the intervention effect prediction model 16.

[0034] Furthermore, currently, the items collected from health checkups and lifestyle habits may not be consistent across hospitals. Therefore, in this embodiment, by narrowing down the target items using an item network, the focus is narrowed down to the main items, making it easier to apply to multiple hospitals.

[0035] Figure 2 is a physical diagram of the lifestyle intervention effect prediction system 100. The lifestyle intervention effect prediction system 100 is configured such that the user's terminal 2 and the lifestyle intervention effect prediction device 1 are connected via a network or the like to enable communication. The lifestyle intervention effect prediction device 1 is comprised of a calculation unit 101, a memory 103, an external communication unit 13, and a storage unit 104.

[0036] The arithmetic unit 101 is a central processing unit that executes programs using the memory 103 as a working area. The memory 103 is, for example, RAM (Random Access Memory) and ROM (Read Only Memory), and is accessed by the arithmetic unit 101.

[0037] The external communication unit 13 is, for example, a network interface card, which sends and receives information to and from terminal 2 via network 3.

[0038] The memory unit 104 is a large-capacity storage device such as a hard disk or an SSD (Solid State Drive). The memory unit 104 is composed of an intervention effect prediction program 1041, a comprehensive PCIT (Partial Correlation Coefficient with Information Theory) program 1042, an inter-item network creation program 1043, a covariate narrowing program 1044, the latest health checkup data 1045, past health checkup data 1046, inter-item network data 1047, and inter-item transition matrix data 1048.

[0039] The calculation unit 101 executes the intervention effect prediction program 1041, thereby realizing the information request unit 10, the examinee ID identification unit 11, and the intervention effect prediction unit 12 shown in Figure 1.

[0040] The calculation unit 101 executes the comprehensive PCIT program 1042, thereby realizing the comprehensive PCIT unit 18 in Figure 1. The calculation unit 101 executes the inter-item network creation program 1043, thereby realizing the inter-item network creation unit 14 in Figure 1. The calculation unit 101 executes the inter-item network creation program 1043, thereby creating inter-item network data 1047 and inter-item transition matrix data 1048.

[0041] The calculation unit 101 executes the covariate filtering program 1044, thereby realizing the covariate filtering unit 15 shown in Figure 1.

[0042] The latest screening data 1045 consists of data collected this fiscal year for standard health checkups and cancer screening checkups. The past screening data 1046 consists of data collected last year and earlier for standard health checkups and cancer screening checkups. The latest screening data 1045 and the past screening data 1046 are stored in the screening data database 17.

[0043] Figure 3 is a flowchart of the application process for the inter-item network. First, the calculation unit 101 determines whether or not to use an inter-item network that excludes confounding factors (step S10). If an inter-item network that excludes confounding factors is to be used (Yes), the process proceeds to step S11. If an inter-item network that excludes confounding factors is not to be used (No), the process proceeds to step S15. In some diseases, accuracy may be better without using an inter-item network, so the user is given the option to choose.

[0044] In step S11, the calculation unit 101 performs comprehensive PCIT processing. Then, the calculation unit 101 creates an inter-item network with confounding factors removed (step S12). Next, the calculation unit 101 narrows down the covariates using the inter-item transition matrix (step S13). Finally, the calculation unit 101 retrains the intervention effect prediction model 16 with the narrowed-down covariates (step S14).

[0045] In step S15, the calculation unit 101 calls the intervention effect prediction model 16 and performs a prediction using machine learning, after which it terminates the process shown in Figure 3.

[0046] In this embodiment, the calculation unit 101 improves the accuracy of predicting health improvement through lifestyle interventions and increases the variety of predictions by combining an inter-item network with confounding factors excluded and an intervention effect prediction model 16.

[0047] The calculation unit 101 uses an item - to - item network to perform operations such as narrowing down co - variables (step S13). For example, the calculation unit 101 can narrow down to "only examination items directly connected to the prediction item" or "do not use examination items connected via age or gender". It is also possible to search for the best narrowing down through prediction accuracy evaluation.

[0048] Figure 4A is a diagram showing the correlation between each factor and factors (part 1). Here, factors X, Y, and Z are shown. The correlation between X and Y is r XY . The correlation between X and Y excluding the influence of factor Z is r XY,Z and is calculated by the following formula (1).

Number

[0049] The correlation between Y and Z is r YZ . The correlation between Y and Z excluding the influence of factor X is r YZ,X . The correlation between X and Z is r XZ . The correlation between X and Z excluding the influence of factor Y is r XZ,Y . In the PCIT determination, if the correlation between X and Y satisfies the inequality of the following formula (2), the calculation unit 101 determines that the correlation between X and Y is indirect due to the influence of Z. Here, ε is calculated by the following formula (3).

Number

Number

[0050] The part of the coefficient 1 / 3 of ε used in the exhaustive PCIT determination can be parameterized. For example, changing the coefficient 1 / 3 of ε to 1 / 2 or 1 / 4, etc.

Number

[0051] Making k smaller than 1 / 3 tightens the criteria and reduces the number of remaining edges. Conversely, making k larger than 1 / 3 loosens the criteria and increases the number of remaining edges. This also includes cases where confounding factors are not excluded.

[0052] Figure 4B shows the correlations between each factor and the factors (part 2). In Figure 4B, the factors X, Y, Z1, ..., Z n This is illustrated in the diagram. In the PCIT assessment, the correlation between X and Y is such that Z is Z in equation (2). P If this can be replaced and the result is true, then the calculation unit 101 determines that the correlation between X and Y is Z P It is determined to be an indirect effect caused by the influence of [the relevant factor].

[0053] For the correlation between X and Y, in equation (2), Z is Z P If the condition substituted by does not hold true for all values ​​of P from 1 to n, the calculation unit 101 determines that there is a direct correlation between X and Y.

[0054] Figure 5 shows a mathematical method for removing confounding factors using ARACNE. In addition to the comprehensive PCIT judgment process, another mathematical method for excluding confounding factors is ARACNE, which uses a normalized frequency distribution and mutual information I(X,Y).

[0055] In ARACNE, if the mutual information I(X,Y) between X and Y satisfies the inequality (5) below, then the correlation between X and Y is determined to be indirect due to the influence of Z.

number

[0056] In this way, it is also possible to replace the comprehensive PCIT assessment with ARACNE, perform a comprehensive ARACNE, and create an inter-item network that excludes confounding factors.

[0057] Parameter μ is used to determine confounding factors. Increasing parameter μ loosens the criteria for determining confounding factors, and may include cases where confounding factors are not excluded.

[0058] Figures 6A to 6D show the process of narrowing down covariates using an inter-item network that excludes confounding factors (parts 1 to 4). The links represent the correlations that remained after assessments such as comprehensive PCIT and comprehensive ARACNE.

[0059] Figure 6A illustrates prediction item 41, measurement items 42-46, and the links between them (part 1). Prediction item 41 is linked to measurement items 42 and 43. Measurement item 43 is further linked to measurement item 44. Measurement item 44 is linked to measurement item 45. Measurement item 46 is an isolated item that is not linked to any other items. The inter-item network after processing with comprehensive PCIT judgment will be explained with reference to Figures 6B to 6D below.

[0060] Figure 6B illustrates prediction item 41, measurement items 42-43, and the links between them (part 2). Prediction item 41 is linked to measurement items 42 and 43. Here, only measurement items that can be reached in one step starting from prediction item 41 are shown.

[0061] Figure 6C illustrates prediction item 41, measurement items 42-44, and the links between them (part 3). Prediction item 41 is linked to measurement items 42 and 43. Measurement item 43 is further linked to measurement item 44. Here, only measurement items that can be reached in two steps starting from prediction item 41 are shown.

[0062] Figure 6D illustrates prediction item 41, measurement items 42-45, and the links between them (part 4). Prediction item 41 is linked to measurement items 42 and 43. Measurement item 43 is further linked to measurement item 44. Measurement item 44 is linked to measurement item 45. Here, the measurement items that can be reached in three steps starting from prediction item 41 are shown.

[0063] Figures 7A to 7C show the process of narrowing down covariates using an inter-item network that excludes confounding factors (parts 5 to 7).

[0064] Figure 7A shows the network of predicted and measured items after comprehensive PCIT (Part 5). Prediction item 61 is linked to measurement items 62, 63, and 65. Measurement item 63 is linked to measurement item 64. Measurement item 65 is linked to measurement item 66.

[0065] Figure 7B shows a network of measurement items limited to only items up to two steps from the prediction item (part 6). Prediction item 61 is linked to measurement items 62, 63, and 65. Measurement item 63 is linked to measurement item 64. Measurement item 65 is linked to measurement item 66. Here, a network similar to that after comprehensive PCIT processing shown in Figure 7A is formed.

[0066] Figure 7C shows a network that is limited to items up to two steps, but with age-related items removed (Part 7). Age is measurement item 65. The link to measurement item 65 via this item has been removed. Age is an item that has correlations with many measurement items. By removing other items that are linked via age, the covariates using the inter-item network can be narrowed down.

[0067] Figure 8 shows an example of an inter-item network. This inter-item network consists of screening items a-d and links that show the correlations between these items. There are two types of correlations between screening items: positive correlation and negative correlation. If we represent positive correlation as 1 and negative correlation as -1, we can create an inter-item transition matrix from the inter-item network excluding confounding factors. The inter-item transition matrix is ​​a symmetric matrix.

[0068] Specifically, examination items a, b, and d are connected by links that show a positive correlation. Examination items b, c, and d are also connected by links that show a positive correlation. The inter-item transition matrix illustrating this is shown in equation (6).

number

[0069] To distinguish between positive and negative correlations, we can obtain a matrix X obtained by multiplying the inter-item transition matrix X by itself n times. n It is necessary to separate the positive and negative correlations. The matrix obtained by extracting only the positive correlations from the inter-item transition matrix X is X + Let's assume that. X + An example is shown in equation (7).

number

[0070] The matrix obtained by extracting only negative correlations from the inter-item transition matrix X is X. - Let's assume that. X - An example is shown in equation (8).

number

[0071] X + and X - X is a symmetric matrix. + and X - Let all diagonal elements of X be zero. + and X - The absolute value of the off-diagonal elements is assumed to be 1.

[0072] The following explains an example of calculating the negative correlation in the second step. First, X, which shows from positive to negative... - ×X + This is shown in equation (9).

number

[0073] X that goes from negative to positive + ×X - This is shown in equation (10).

number

[0074] X representing positive to negative - ×X + And X, which shows from negative to positive. + ×X - The sum of these is the negative correlation of the second step, which is shown in equation (11).

number

[0075] Let's explain an example of calculating the positive correlation in the second step. X indicates positive from positive to positive. + ×X + This is shown in equation (12).

number

[0076] X that shows negative from negative - ×X - This is shown in equation (13).

number

[0077] X that shows positive from positive + ×X + And X that shows negative from negative - ×X -The sum of these is the positive correlation X in the second step. + 2 This is shown in equation (14).

number

[0078] The inter-item transition matrix X of the second step 2 This is the negative correlation X in the second step. - 2 and positive correlation X + 2 This is the sum of the two. This is shown in equation (15).

number

[0079] The inter-item transition matrix (X+X) is the result of combining the first and second steps. 2 The following illustrates this. In the example in Figure 8, all items can be connected within two steps: X+X 2 This can be seen from the facts.

number

[0080] By repeating this process, the matrix X of the activity at step n is obtained. + n and the matrix X of inhibitory effects - n We can separate them. From this, we can use equations (17), (18), and (19) below to determine X n+1 It is possible to find this.

number

[0081]

number

[0082]

number

[0083] For any n, matrix X + n ,X - n ,X n This is corrected according to the following rules. Correction Rule #1: All diagonal elements are zero. Correction Rule #2: The absolute value of off-diagonal components is 1

[0084] Figures 9A and 9B illustrate correction rule #1 for the inter-item transition matrix. Figure 9A shows that screening item a and screening item b are positively correlated. Screening item b and screening item c are negatively correlated (Part 1). As shown in Figure 9A, the transition from a ⇒ b ⇒ c is acceptable because it does not return to itself and conforms to correction rule #1.

[0085] Figure 9B shows a correlation that deviates from correction rule #1 (part 2). As shown in Figure 9B, the transition from a⇒b, then back to a, and then to a⇒b⇒c deviates from correction rule #1 because it returns to itself, and is therefore not acceptable.

[0086] Figures 10A and 10B illustrate correction rule #2 for the inter-item transition matrix. Correction rule #2, "the absolute value of off-diagonal elements is 1," means that even if there are multiple paths for correlation, they are treated as one.

[0087] Figure 10A shows the positively correlated path a⇒b and the negatively correlated path b⇒c (Part 1). Figure 10B shows the positively correlated path a⇒d and the negatively correlated path d⇒c (Part 2).

[0088] If the correlation from a to c is a two-step process, and there are two paths, a⇒b⇒c and a⇒d⇒c, and both are negatively correlated (-1), then X 2 The ac component is -2. However, since the correlation between a and c is considered to be one, X 2The AC component is corrected to -1.

[0089] Figures 11A and 11B illustrate the rules for determining conflicting correlations. Conflicting correlations may appear as the number of steps increases. However, the correlation is automatically determined by the principle of majority rule.

[0090] Figure 11A shows the positively correlated path a⇒b and the negatively correlated path b⇒c (Part 1). Figure 11B shows the positively correlated path a⇒d and the positively correlated path d⇒c (Part 2).

[0091] The two-step correlation from a to c is represented by the path a⇒b⇒c shown in Figure 11A and the path a⇒d⇒c shown in Figure 11B. However, the path a⇒b⇒c is negatively correlated (-1), and the path a⇒d⇒c is positively correlated (+1). In this case, X 2 The ac component in this case becomes 0. If there are an equal number of opposing correlations, the correlation between a and c is considered nonexistent and can remain at zero.

[0092] Figures 12A to 12C illustrate the rules that determine opposing correlations. The two-step correlation from a to c can be represented by the paths a⇒b⇒c (1) shown in Figure 12A, a⇒d⇒c (2) shown in Figure 12B, and a⇒e⇒c (3) shown in Figure 12C. However, the a⇒b⇒c path is negatively correlated (-1), the a⇒d⇒c path is positively correlated (+1), and the a⇒e⇒c path is positively correlated (+1). In this case, X 2 The ac component in this case becomes 1. If the opposing correlations are not equal in number, the correlation between a and c is automatically determined by majority vote.

[0093] Figure 13 is a flowchart of the process for creating inter-item networks. First, the inter-item network creation unit 14 reads the health checkup data (step S20). In step S21, the comprehensive PCIT unit 18 selects any two items as items A and B, and performs a comprehensive PCIT check on items A, item B, and the remaining items Z. The inter-item network creation unit 14 determines whether a PCIT check is successful for any of the items Z (step S22). In step S22, if a PCIT check is successful for any of the items Z (Yes), the process proceeds to step S24. If a PCIT check is not successful for any of the items Z (No), the process proceeds to step S23.

[0094] In step S23, the inter-item network creation unit 14 connects item A and item B at an edge. Then, in step S24, the inter-item network creation unit 14 determines whether or not a comprehensive PCIT check has been performed between all pairs of items. If a comprehensive PCIT check has not been performed between any of the pairs of items (No), the process returns to step S21. If a comprehensive PCIT check has been performed between all pairs of items (Yes), the process proceeds to step S25.

[0095] In step S25, the inter-item network creation unit 14 outputs the inter-item network with confounding factors removed, and then terminates the process shown in Figure 13.

[0096] Figure 14 is a flowchart of the covariate filtering process using an inter-item transition matrix. Initially, the covariate filtering unit 15 receives input of the inter-item network with confounding factors removed (step S30) and calculates the inter-item transition matrix X (step S31).

[0097] In step S32, the covariate filtering unit 15 determines whether the path has passed through a predetermined item such as age or gender. If the path has passed through a predetermined item (Yes), the process proceeds to step S34. If the path has not passed through a predetermined item (No), the process proceeds to step S33.

[0098] In step S33, the covariate filtering unit 15 sets all the column and row values ​​of predetermined items such as age or gender to zero in the inter-item transition matrix X.

[0099] In step S34, the covariate filtering unit 15 determines whether the number of passes N is 2 or greater. If the number of passes N is 2 or greater (Yes), the process proceeds to step S35. Then, the covariate filtering unit 15 substitutes n+1 for n (step S35) and, according to the rules for creating the inter-item transition matrix, X n The calculation is performed (step S36). If the number of passes N is less than 2 (No), the process proceeds to step S38.

[0100] In step S37, the covariate filtering unit 15 determines whether n is equal to N. If n is not equal to N (No), the process returns to step S35. If n is equal to N (Yes), the process proceeds to step S38.

[0101] In step S38, the covariate filtering unit 15 is X 1 +…+X N Selecting only the non-zero items in the column corresponding to the prediction item will terminate the process shown in Figure 14.

[0102] Figure 15 is a flowchart of the process for predicting the intervention effect on first-time patients undergoing screening. Initially, the examinee ID identification unit 11 executes the processes in steps S40 and S41 and steps S42 and S43 in parallel. The examinee ID identification unit 11 receives the input of the examination data and / or lifestyle data of the first-time examinee (step S40) and normalizes the data of that first-time examinee across all items (step S41). In parallel, the examinee ID identification unit 11 reads the examination data database 17 (step S42) and normalizes the data for each examinee ID (step S43).

[0103] Next, the examiner ID identification unit 11 calculates an evaluation value of the distance in the data space between the data of the first examiner and the data of all examiner IDs (step S44). In step S44, the examiner ID identification unit 11 calculates an evaluation value L(k) of the distance between the normalized first-year examination data of ID=k and the normalized data of the first examiner.

[0104] The examiner ID identification unit 11 then identifies the examiner ID corresponding to the data where the distance between the normalized first-year examination data and the normalized data of the first-time examiner is minimized, based on the distance evaluation value (step S45). The examiner ID identification unit 11 identifies the ID=k that minimizes the distance in the data space, based on the distance evaluation value L(k).

[0105] The examiner ID identification unit 11 may calculate the Euclidean distance as the distance evaluation value L(k) and determine the ID=k that minimizes the distance evaluation value L(k), or it may calculate the reciprocal of the Euclidean distance as the distance evaluation value L(k) and determine the ID=k that maximizes the distance evaluation value L(k), and is not limited to this. In either case, the examiner ID identification unit 11 will identify the ID=k that minimizes the distance in the data space between the normalized first-year examination data of ID=k and the normalized data of the first-time examinee.

[0106] The intervention effect prediction unit 12 predicts the intervention effect on this examinee with the support of the intervention effect prediction model 16 (step S46). This intervention effect prediction process will be explained in detail in Figure 19 below. Subsequently, when the external communication unit 13 displays the predicted intervention effect for first-time examinees on the display unit 22 of the terminal 2 (step S47), the process shown in Figure 15 is completed. This enables the prediction of first-time examinees for whom there is no data in the examination data database 17. Furthermore, the individuals included in the prediction are not limited to first-time patients; they may also include patients who have previously received examinations and have obtained a patient ID.

[0107] Figure 16 shows an example of an input screen 51 displayed on terminal 2. Information regarding the glucose metabolism model is entered on this input screen 51. In this embodiment, the improvement effect of items related to the glucose metabolism model is evaluated based on the information regarding the glucose metabolism model, but the present invention is not limited to this. Combo box 511 is for entering whether or not you smoke. Combo box 512 is for entering whether or not you exercise for 30 minutes or more. From here on, the respondent scrolls down the list of questions and enters their answers.

[0108] Combo box 513 is a combo box that specifies how many years into the future to predict. The response submission button 514 is a button to tap after completing the input. After tapping the response submission button 514, the response results are sent to the lifestyle intervention effect prediction device 1.

[0109] After transmitting data to the lifestyle intervention effect prediction device 1, and receiving results from the lifestyle intervention effect prediction device 1, the display unit 22 of the terminal 2 switches to the prediction result screen 52.

[0110] Figure 17 shows an example of the prediction results screen 52 displayed on terminal 2. The prediction results screen 52 displays the prediction results table 521 of the glucose metabolism model based on the intervention effect prediction model 16, and input screen buttons 522. Tapping any row in the prediction results table 521 will take the user to a feedback screen 53 for improving the item indicated in that row.

[0111] The prediction results table 521 displays the prediction results of the glucose metabolism model in tabular format. The input screen button 522 is a button that returns to the input screen 51.

[0112] Figure 18 shows an example of the feedback screen 53 displayed on terminal 2. The feedback screen 53 displays the prediction results table 531 and the input screen button 532.

[0113] The prediction results table 531 on the feedback screen 53 recommends that increasing exercise by 30 minutes is the best way to lower blood sugar levels. It also recommends that eating breakfast with every meal and eating dinner earlier are also effective to some extent. The intervention effect prediction model 16 can use the prediction results 531 to automatically recommend the optimal lifestyle improvement method for each individual undergoing the examination. Therefore, the optimal lifestyle improvement method recommendation is output to the feedback screen 53 on terminal 2. Here, multiple recommendations are displayed in order of effectiveness.

[0114] Figure 19 is a flowchart of the intervention effect prediction process. This flowchart details the process in step S46 and will be explained with reference to Figure 1. Initially, the intervention effect prediction unit 12 predicts the examinee's future data when the items are not changed, based on the examinee ID associated with the data that is the least far from the examinee's data (step S50). The intervention effect prediction unit 12 predicts the examinee's future data by referring to the examinee ID identified in step S45 of Figure 15 and each evaluation value evaluated by the intervention effect prediction model 16 based on the examination data database 17 as the user's future prediction result. The examinee ID identification unit 11 and the intervention effect prediction unit 12 repeat the processing from steps S51 to S57 for all items related to the examinee's lifestyle data.

[0115] First, the examiner ID identification unit 11 changes the value of this item by ± a predetermined value (step S52), and then normalizes the examiner data (step S53). Then, the examiner ID identification unit 11 calculates an evaluation value of the distance between the data of the target examiner and all the data related to examiner IDs in the data space, and based on this evaluation value of distance, identifies the examiner ID related to the data that has the smallest distance to the examiner's data (step S54).

[0116] The intervention effect prediction unit 12 refers to the examinee ID corresponding to the data with the minimum distance to the examinee's data and each evaluation value evaluated by the intervention effect prediction model 16 based on the examination data database 17 as each predicted result of the user improving each of the lifestyle habits, and predicts the examinee's future data when the item is changed (step S55). The intervention effect prediction unit 12 uses the examinee's future data when the item is changed as the prediction of the intervention effect.

[0117] The intervention effect prediction unit 12 records the degree of improvement in the examinee's future evaluation items due to the change in items (step S56). Here, the intervention effect prediction unit 12 uses the difference between the examinee's future data when the items are changed and the examinee's future data when the items are not changed as the degree of improvement in the evaluation items due to the change in items.

[0118] In step S57, the examinee ID identification unit 11 and the intervention effect prediction unit 12 determine whether the processing has been repeated for all items related to lifestyle habits. If there are any unprocessed items, the process returns to step S51. If the processing has been repeated for all items, the process proceeds to step S58.

[0119] In step S58, the intervention effect prediction unit 12 rearranges the items and their changes in order of the effect that improves the health check data. Then the process in Figure 19 is completed, and the process returns to the flowchart in Figure 15, and the process in step S47 is executed.

[0120] Figure 20 shows an example of the input screen 54 displayed on terminal 2. Information regarding the glucose metabolism model is entered on this input screen 54. Radio button 541 is for entering whether or not to exclude confounding factors. Combo box 542 is for entering the number of steps to make a prediction. Combo box 543 is for entering the items from which you want to exclude confounding factors for other items via this item.

[0121] From this point onward, respondents scroll down the question content and enter their answers. This input screen 54 allows them to specify how many steps of the health checkup item network should be used as data for prediction. Furthermore, they can specify that paths via specific items such as age should be removed.

[0122] Combo box 544 is a combo box that specifies how many years into the future to predict. The response submission button 545 is a button to tap after completing the input. After tapping the response submission button 545, the response results are sent to the lifestyle intervention effect prediction device 1.

[0123] After transmitting data to the lifestyle intervention effect prediction device 1, and receiving results from the lifestyle intervention effect prediction device 1, the display unit 22 of the terminal 2 switches to the prediction result screen 55 shown in Figure 21.

[0124] Figure 21 shows an example of the prediction results screen 55 displayed on terminal 2. The prediction results screen 55 displays the prediction results table 551 of the glucose metabolism model by the intervention effect prediction model 16, a covariate graph 552, and input screen buttons 553. Tapping any row in the prediction results table 521 will take the user to a feedback screen 53 for improving the item indicated in that row.

[0125] The prediction results table 521 displays the prediction results of the glucose metabolism model in tabular format. The input screen button 522 is a button that returns to the input screen 51.

[0126] The configuration and effects of the present invention are described below.

[0127] [1] A medical examination ID identification unit (11) uses a medical examination data database (17) that stores medical examination data and lifestyle data for multiple items related to multiple medical examination IDs to identify each medical examination ID related to data that is closer to the data of the user's improved lifestyle than a predetermined relative standard, from the data for multiple items entered by the user. An intervention effect prediction unit (12) refers to each evaluation value evaluated by the intervention effect prediction model (16) based on each of the examinee IDs identified by the examinee ID identification unit (11) and the examination data database (17) as each predicted result of the user improving each of the lifestyle habits, The output unit (external communication unit 13) outputs each of the prediction results referenced by the intervention effect prediction unit (12), A device for predicting the effects of lifestyle interventions (1), characterized by being equipped with the following:

[0128] This makes it possible to encourage examinees to make changes to their lifestyles based on the data they input.

[0129] [2] The examiner ID identification unit (11) normalizes the numerical values ​​of each item in the examination data database (17) and identifies each examiner ID related to data whose distance from the data showing the improved state of each of the user's lifestyle habits is closer than a predetermined relative standard. A lifestyle intervention effect prediction device (1) according to claim 1.

[0130] This allows us to identify each patient ID that best reflects the user's improved lifestyle based on their health checkup data and database, even for first-time users, enabling us to output accurate prediction results.

[0131] [3] The intervention effect prediction unit (12) causes the intervention effect prediction model (16) to evaluate each evaluation value based on the normalized values ​​of each of the examinee IDs identified by the examinee ID identification unit (11) and the numerical values ​​of each item in the examination data database (17). A lifestyle intervention effect prediction device (1) according to feature 2.

[0132] This allows the influence of each parameter to be equalized, making it possible to identify the examinee ID that is closest to the appropriate state.

[0133] [4] An inter-item network creation unit (14) creates an inter-item network that excludes confounding factors between items, The covariate filtering unit (15) uses the inter-item network created by the inter-item network creation unit (14) to narrow down the covariates and select the items to be used for evaluation. The covariate narrowing unit (16) includes a model learning unit (19) that trains the intervention effect prediction model, It also has, A lifestyle intervention effect prediction device (1) according to claim 1.

[0134] This allows us to narrow down the number of screening items used for evaluation, enabling the intervention effect prediction model to more accurately predict improvement outcomes.

[0135] [5] The covariate filtering unit (15) calculates an inter-item transition matrix to narrow down the items to be used for evaluation. A lifestyle intervention effect prediction device (1) according to claim 1.

[0136] This makes it easy to narrow down the number of examination items used for evaluation.

[0137] [6] The system further includes an information request unit (10) that requests from the user's terminal health checkup data for each item and lifestyle data for each lifestyle habit based on covariates narrowed down by the health checkup item network. A device for predicting the effect of lifestyle interventions according to feature 4.

[0138] This allows for a reduction in the number of information fields that users need to input, making it easier for them to enter the information.

[0139] [7] The system further includes a mathematical determination processing unit (comprehensive PCIT unit 18) that determines whether or not to exclude confounding factors. The item-to-item network creation unit (14) generates the item-to-item network by excluding confounding factors between items based on the results of the judgment performed by the mathematical judgment processing unit (comprehensive PCIT unit 18). A device for predicting the effect of lifestyle interventions according to feature 4.

[0140] This allows for the generation of an inter-item network that excludes confounding factors.

[0141] [8] The system further includes a comprehensive PCIT unit (18) that performs PCIT (Partial Correlation Coefficient) determination, The item-to-item network creation unit (14) generates the item-to-item network by excluding confounding factors between items based on the results of the PCIT (Partial Correlation Coefficient) judgment performed by the comprehensive PCIT unit (18). A lifestyle intervention effect prediction device (1) according to feature 4.

[0142] This allows for the generation of inter-item networks based on PCIT judgments.

[0143] [9] The inter-item network creation unit (14) generates the inter-item network by excluding confounding factors between items using mutual information based on a normalized frequency distribution. A lifestyle intervention effect prediction device (1) according to feature 4.

[0144] This allows for the generation of inter-item networks based on mutual information using normalized frequency distributions.

[0145]

[10] The aforementioned user is either a patient who has previously received a medical examination and has obtained a medical examination ID, or a patient receiving a medical examination for the first time. A lifestyle intervention effect prediction device (1) according to claim 1.

[0146] This allows for accurate prediction results to be output for both first-time users and users with previous consultation experience.

[0147]

[11] The output unit (external communication unit 13) outputs the predicted results for each lifestyle improvement in order of their effectiveness. A lifestyle intervention effect prediction device (1) according to claim 1.

[0148] This allows you to learn about lifestyle improvement methods that are more effective in bringing about positive changes.

[0149]

[12] A medical examination ID identification unit (11) uses a medical examination data database (17) that stores medical examination data and lifestyle data for multiple items related to multiple medical examination IDs to identify each medical examination ID related to data that is closer to the data of the user's improved lifestyle than a predetermined relative standard, from the data for multiple items entered by the user into the terminal. An intervention effect prediction unit (12) refers to each evaluation value evaluated by the intervention effect prediction model (16) based on each of the examinee IDs identified by the examinee ID identification unit (11) and the examination data database (17) as each predicted result of the user improving each of the lifestyle habits, An output unit (external communication unit 13) outputs each of the prediction results referenced by the intervention effect prediction unit (12) to the display unit (22) of the terminal, A lifestyle intervention effect prediction system (100) characterized by comprising the following:

[0150] This makes it possible to encourage examinees to make changes to their lifestyles based on the data they input.

[0151]

[13] Using a medical examination data database (17) that stores medical examination data and lifestyle data for multiple items related to multiple medical examination IDs, the medical examination ID identification unit (11) identifies each medical examination ID related to data whose distance from the data of the user's improved lifestyle is closer than a predetermined relative standard, based on the data for multiple items entered by the user. The intervention effect prediction unit (12) refers to each evaluation value evaluated by the intervention effect prediction model (16) based on each of the examinee IDs identified by the examinee ID identification unit (11) and the examination data database, as each predicted result of the user improving each of the lifestyle habits. The steps include: the output unit (external communication unit 13) displaying each of the prediction results referenced by the intervention effect prediction unit (12) on the display unit (22); A method for predicting the effects of lifestyle interventions, characterized by comprising the following features.

[0152] This makes it possible to encourage examinees to make changes to their lifestyles based on the data they input.

[0153]

[14] A procedure for identifying each examiner ID whose data is closer to the data showing the improved state of each of the user's lifestyle habits than a predetermined relative standard, using an examination data database that stores examination data and lifestyle data for multiple items related to multiple examiner IDs. A procedure for the user to refer to each evaluation value evaluated by the intervention effect prediction model based on each identified examinee ID and the examination data database as each predicted result of improvement in each of the lifestyle habits, A procedure for displaying each of the referenced prediction results on the display unit. A program for predicting the effects of lifestyle interventions, which is designed to be executed by a computer.

[0154] This makes it possible to encourage examinees to make changes to their lifestyles based on the data they input.

[0155] Variant form The present invention is not limited to the embodiments described above, and includes various modifications. For example, the embodiments described above are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. It is possible to replace parts of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add configurations from other embodiments to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations.

[0156] Each of the above configurations, functions, processing units, and processing means may be implemented in part or in whole by hardware, such as an integrated circuit. Each of the above configurations and functions may also be implemented in software by a processor interpreting and executing a program that implements each function. Information such as programs, tables, and files that implement each function can be stored in a recording device such as memory, a hard disk, or an SSD (Solid State Drive), or on a recording medium such as a flash memory card or a DVD (Digital Versatile Disk).

[0157] In each embodiment, the control lines and information lines shown are those deemed necessary for explanation and do not necessarily represent all control lines and information lines in the actual product. In practice, it can be assumed that almost all components are interconnected.

[0158] Examples of modifications of the present invention include the following (a) to (e). (a) Variations in the presentation of prediction results to the user are not limited to the embodiments. (b) As users continue to use the app, the app may request additional information from them regarding the input items narrowed down by the intervention effect prediction model and the screening item network. (c) The method of displaying the improvement status to the user on the terminal display is not limited to a table format. (d) The mathematical methods for excluding confounding factors are not limited to PCIT and ARACNE. (e) The diseases to which this applies are not limited to diseases related to glucose metabolism, such as diabetes. [Explanation of Symbols]

[0159] 100 Lifestyle Intervention Effectiveness Prediction System 2 terminals 1. Lifestyle intervention effect prediction device 21 Input section 22 Display section 10 Information request section 11. Examination Participant ID Identification Department 12. Intervention Effect Prediction Section 13. External Communication Unit (Output Unit) 14. Inter-item network creation section 15. Covariate selection section 16. Intervention Effect Prediction Model 17. Health checkup data database 18 Comprehensive PCIT Department 101 Arithmetic section 103 memory 104 Storage section 3 Network 1041 Intervention Effect Prediction Program 1042 Comprehensive PCIT Program 1043 Inter-item network creation program 1044 Covariate Refinement Program 1045 Latest medical examination data 1046 Past medical examination data 1047 Inter-item network data 1048 item-to-item transition matrix data 41 Prediction Items 42~46 Measurement items 61 Prediction Items 62~66 Measurement items 51 Input screen 511 Combo Box 512 Combo Box 513 Combo Box 514 Submit response button 52 Prediction Results Screen 522 Input screen button 53 Feedback screen 532 Input screen button

Claims

1. A patient ID identification unit uses a medical examination data database containing medical examination data and lifestyle data for multiple items related to multiple patient IDs to identify each patient ID whose data is closer to the data showing the improved state of each of the user's lifestyle habits than a predetermined relative standard, based on the data for multiple items entered by the user. An intervention effect prediction unit refers to each evaluation value evaluated by the intervention effect prediction model based on each of the examinee IDs identified by the examinee ID identification unit and the examination data database, as each predicted result of the user improving each of the lifestyle habits. An output unit that outputs each of the prediction results referenced by the intervention effect prediction unit, A device for predicting the effects of lifestyle interventions, characterized by being equipped with the following features.

2. The aforementioned examiner ID identification unit normalizes the numerical values ​​of each item in the examination data database and identifies each examiner ID related to data whose distance from the data showing the improved state of each of the user's lifestyle habits is closer than a predetermined relative standard. A device for predicting the effect of lifestyle intervention according to claim 1.

3. The intervention effect prediction unit causes the intervention effect prediction model to evaluate each evaluation value based on the normalized values ​​of each examinee ID identified by the examinee ID identification unit and each item in the examination data database. The lifestyle intervention effect prediction device according to claim 2.

4. An inter-item network creation unit creates an inter-item network that excludes confounding factors between items, A covariate narrowing unit uses the inter-item network created by the inter-item network creation unit to narrow down the covariates and narrow down the items to be used for evaluation, The system further comprises a model learning unit that trains the intervention effect prediction model in the covariate narrowing unit. A device for predicting the effect of lifestyle intervention according to claim 1.

5. The aforementioned covariate filtering unit calculates an inter-item transition matrix to narrow down the items to be used for evaluation. The lifestyle intervention effect prediction device according to feature 4.

6. The system further includes an information request unit that requests health checkup data and lifestyle data for each item from the user's terminal based on covariates narrowed down by an inter-item network. The lifestyle intervention effect prediction device according to feature 4.

7. It further includes a mathematical determination processing unit that determines whether or not to exclude confounding factors. The item-inter-item network creation unit generates the item-inter-item network by excluding confounding factors between items based on the results of the determination performed by the mathematical determination processing unit. The lifestyle intervention effect prediction device according to feature 4.

8. It further includes a comprehensive PCIT unit that performs PCIT (Partial Correlation Coefficient) judgment, The item-to-item network creation unit generates the item-to-item network by excluding confounding factors between items based on the results of the PCIT (Partial Correlation Coefficient) judgment performed by the comprehensive PCIT unit. The lifestyle intervention effect prediction device according to feature 4.

9. The item-to-item network creation unit generates the item-to-item network by excluding confounding factors between items using mutual information based on a normalized frequency distribution. The lifestyle intervention effect prediction device according to feature 4.

10. The aforementioned user is either a patient who has previously received a medical examination and has obtained a medical examination ID, or a patient receiving a medical examination for the first time. A device for predicting the effect of lifestyle intervention according to claim 1.

11. The output unit outputs the predicted results for each lifestyle improvement in order of their effectiveness. A device for predicting the effect of lifestyle intervention according to claim 1.

12. A patient ID identification unit uses a medical examination data database containing medical examination data and lifestyle data for multiple items related to multiple patient IDs to identify each patient ID whose data is closer to the data showing the improved state of each of the user's lifestyle habits than a predetermined relative standard, based on the data for multiple items entered by the user into the terminal. An intervention effect prediction unit refers to each evaluation value evaluated by the intervention effect prediction model based on each of the examinee IDs identified by the examinee ID identification unit and the examination data database, as each predicted result of the user improving each of the lifestyle habits. An output unit that outputs each of the prediction results referenced by the intervention effect prediction unit to the display unit of the terminal, A lifestyle intervention effect prediction system characterized by having the following features.

13. Using a medical examination data database containing medical examination data and lifestyle data for multiple items related to multiple medical examination IDs, the medical examination ID identification unit identifies each medical examination ID whose data is closer to the data showing the improved state of each of the user's lifestyle habits than a predetermined relative standard, based on the data for multiple items entered by the user. The intervention effect prediction unit refers to each evaluation value evaluated by the intervention effect prediction model based on each of the examinee IDs identified by the examinee ID identification unit and the examination data database as each predicted result of the user improving each of the lifestyle habits. The step of having the output unit display each of the prediction results referenced by the intervention effect prediction unit on the display unit, A method for predicting the effects of lifestyle interventions, characterized by comprising the following features.

14. A procedure for identifying each examiner ID whose data is closer to the data showing the improved state of each of the user's lifestyle habits than a predetermined relative standard, using an examination data database that stores examination data and lifestyle data for multiple items related to multiple examiner IDs. A procedure for the user to refer to each evaluation value evaluated by the intervention effect prediction model based on each identified examinee ID and the examination data database as each predicted result of improving each of the lifestyle habits, A procedure for displaying each of the aforementioned prediction results on the display unit. A program that causes a computer to execute something.

Images