Data analysis apparatus and method for data analysis

Abstract

To provide a data analysis apparatus capable of supporting efficient intervention.SOLUTION: An interruption risk calculation unit 34 calculates, for each individual, the exposure risk value obtained by evaluating the risk of exposure to a risk factor, on the basis of personal information (basic resident information 51 and health examination information 52) and interruption risk analysis information 54. An interruption impact calculation unit 35 calculates, for each individual, an exposure impact value obtained by evaluating the impact that exposure to a risk factor has on the occurrence of an outcome, on the basis of personal information and interruption impact analysis information 55. A clustering unit 36 classifies a plurality of individuals into a plurality of groups according to the exposure risk value and the exposure impact value on the basis of the personal information.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a data analysis device and a data analysis method. [Background technology]

[0002] In order to strengthen the financial foundation of the health insurance system, insurers are being asked to implement measures to rationalize medical expenses, such as encouraging insurers to adopt preventive measures. For example, patients, such as those with diabetes, may experience a worsening of their illness due to discontinuing medical treatment or medication, resulting in unnecessary medical expenses. Therefore, in order to prevent discontinuing medical treatment or medication, medical staff can provide interventions such as medical guidance to patients, thereby rationalizing medical expenses.

[0003] However, since there are limitations on the number of medical staff who can perform the intervention and the costs involved in the intervention, it is difficult to perform the intervention on all candidates, such as patients.

[0004] In response to this, Patent Document 1 discloses a technology for predicting the long-term transition of the effect of an intervention based on the health information of candidates who are candidates for intervention, and selecting a subject from the candidates. Patent Document 2 discloses a technology for constructing a risk model for calculating the risk of changes in health status based on health information on the health of multiple people, and for determining the priority of health guidance, which is a type of intervention, for each person based on the risk model. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-218966 [Patent Document 2] Japanese Patent Application Publication No. 2023-47778 Summary of the Invention [Problem to be solved by the invention]

[0006] However, in the technologies described in Patent Documents 1 and 2, subjects for intervention are selected based on the presence or absence of outcomes such as disease onset or medical expenses, and the risk of exposure is not taken into consideration. Furthermore, the technologies described in Patent Documents 1 and 2 have the problem of low efficiency, as they determine whether or not a subject is a candidate on an individual basis.

[0007] An object of the present disclosure is to provide a data analysis device and a data analysis method that can support efficient intervention. [Means for solving the problem]

[0008] A data analysis device according to one aspect of the present disclosure includes a memory unit that stores personal information for each of a plurality of individuals, exposure information indicating whether or not each individual has been exposed to a risk factor, and measurement target information indicating whether or not each individual has experienced an outcome affected by the exposure; a risk calculation unit that calculates, for each individual, an exposure risk value that evaluates the risk of the exposure occurring based on the personal information and the exposure information; an impact calculation unit that calculates, for each individual, an exposure impact value that evaluates the impact of the exposure on the outcome based on the personal information and the measurement target information; and a classification unit that classifies the plurality of individuals into a plurality of groups according to the exposure risk value and the exposure impact value based on the personal information. [Effects of the Invention]

[0009] The present disclosure makes it possible to support efficient intervention. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 illustrates a data analysis system according to an embodiment of the present disclosure. [Figure 2] FIG. 2 illustrates an example of a functional configuration of an analysis server. [Figure 3] FIG. 2 is a diagram showing an example of basic resident information. [Figure 4] FIG. 10 is a diagram illustrating an example of medical checkup information. [Figure 5] FIG. 10 is a diagram showing an example of medical examination information. [Figure 6] FIG. 10 is a diagram illustrating an example of interruption risk analysis information. [Figure 7] FIG. 10 is a diagram illustrating an example of interruption impact analysis information. [Figure 8] FIG. 10 is a diagram illustrating an example of interruption risk information. [Figure 9] FIG. 10 is a diagram illustrating an example of interruption impact information. [Figure 10] FIG. 10 is a diagram illustrating an example of branch information. [Figure 11] FIG. 10 is a diagram illustrating an example of cluster information. [Figure 12] FIG. 10 is a sequence diagram illustrating an example of a data analysis process. [Figure 13] 10 is a flowchart illustrating an example of a clustering process. [Figure 14] FIG. 10 is a diagram illustrating an example of a basic analysis result screen. [Figure 15] FIG. 10 is a diagram illustrating an example of a clustering result screen. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] Fig. 1 is a diagram showing a data analysis system according to an embodiment of the present disclosure. The data analysis system 100 shown in Fig. 1 is a system for analyzing personal information about each of a plurality of individuals and classifying the individuals into clusters, which are groups according to an exposure risk value that evaluates the risk of exposure to a risk factor and an outcome that is affected by exposure to the risk factor. In this embodiment, a regression tree model is used for classifying the clusters.

[0013] The data analysis system 100 includes a user terminal 1 and an analysis server 2, which are connected to each other via a communication network 101 such as the Internet. Note that there may be multiple user terminals 1.

[0014] The user terminal 1 is a terminal device operated by a user who uses the data analysis system, and includes a main memory device 11, an auxiliary memory device 12, a calculation device 13, an input / output device 14, and a communication I / F 15.

[0015] The main memory device 11 is a device that records data in a writable and readable manner, and stores a program (computer program) that defines the operation of the arithmetic unit 13. The auxiliary memory device 12 stores data used or generated by the program stored in the main memory device 11. The arithmetic unit 13 executes the program stored in the main memory device 11 to realize various functions. The input / output device 14 accepts various information from the user and outputs the various information by displaying it. The communication I / F 15 communicates with the analysis server 2, which is an external device.

[0016] The analysis server 2 is a data analysis device that, in accordance with instructions from the user terminal 1, classifies multiple individuals into multiple groups (clusters) that have similar risks of exposure to risk factors and similar impact values ​​when exposure to the risk factors occurs, and has a main memory device 21, an auxiliary memory device 22, an arithmetic device 23, an input / output device 24, and a communication I / F 25.

[0017] The main memory device 21 is a device that records data in a writable and readable manner, and stores programs that define the operation of the arithmetic device 23. In this embodiment, the main memory device 21 stores the following programs: a consultation interruption identification program 41, an event extraction program 42, a basic analysis program 43, an interruption risk calculation program 44, an interruption impact calculation program 45, and a clustering program 46.

[0018] The auxiliary storage device 22 is a storage unit that stores data used or generated by the programs stored in the main storage device 21. In this embodiment, the auxiliary storage device 22 stores resident basic information 51, health checkup information 52, medical examination information 53, interruption risk analysis information 54, interruption impact analysis information 55, interruption risk information 56, interruption impact information 57, branch information 58, and cluster information 59.

[0019] The resident basic information 51 and the health checkup information 52 constitute personal information about each of the multiple individuals to be analyzed. Specifically, the resident basic information 51 has items of basic attributes about each individual, and the health checkup information 52 has items about the health of each individual obtained through a health checkup (physical examination). Furthermore, the medical examination information 53 indicates the medical examination history of each individual at a medical institution. Note that the medical examination information 53 may be included in the personal information.

[0020] The discontinuation risk analysis information 54 is exposure information indicating whether or not each individual is exposed to a risk factor. In this embodiment, the exposure to a risk factor is discontinuation of medical treatment for a specific disease, but this example is not limited to this. In this embodiment, the specific disease is type 2 diabetes, but this example is not limited to this. The discontinuation impact analysis information 55 is measurement target information indicating whether or not an event occurs, which is an outcome affected by exposure to a risk factor in each individual. Examples of events include a specific disease, a change in test value, an increase or decrease in medical expenses, the performance of a medical procedure, hospitalization, death, nursing care certification, or a change in the level of nursing care required. In this embodiment, the event is chronic kidney disease.

[0021] The interruption risk information 56 is information indicating an interruption risk value, which is an exposure risk value that evaluates the risk of exposure to a risk factor for each individual. The interruption impact information 57 is information indicating an exposure impact value that evaluates the impact that exposure to a risk factor for each individual has on an event.

[0022] Branching information 58 is information indicating branching conditions for reaching each node of the regression tree model for classifying individuals into clusters. Cluster information 59 is information indicating the relationship between each individual and the cluster to which the individual belongs.

[0023] The arithmetic unit 23 executes programs stored in the main memory unit 21 to realize various functions. The input / output unit 24 receives various information from the user and outputs the information by displaying it. The communication I / F 25 communicates with the user terminal 1, which is an external device.

[0024] 2 is a diagram illustrating an example of the functional configuration of the analysis server 2. As illustrated in FIG. 2, the analysis server 2 has, as its functional configuration, a consultation interruption identification unit 31, an event extraction unit 32, a basic analysis unit 33, an interruption risk calculation unit 34, an interruption impact calculation unit 35, a clustering unit 36, an input unit 37, and an output unit 38. The consultation interruption identification unit 31, the event extraction unit 32, the basic analysis unit 33, the interruption risk calculation unit 34, the interruption impact calculation unit 35, and the clustering unit 36 ​​are realized by the calculation device 23 executing a consultation interruption identification program 41, an event extraction program 42, a basic analysis program 43, an interruption risk calculation program 44, an interruption impact calculation program 45, and a clustering program 46, respectively. The input unit 37 and the output unit 38 are realized by the input / output device 24 and the communication I / F 25.

[0025] The consultation interruption identification unit 31 determines whether or not an interruption in consultation, which is exposure to risk factors, has occurred for each individual based on the consultation information 53, and generates interruption risk analysis information 54 based on the determination result.

[0026] The event extraction unit 32 determines whether or not an event that is an outcome has occurred based on the medical examination information 53, and generates interruption impact analysis information 55 based on the determination result.

[0027] Based on the judgment results of the examination interruption identification unit 31 and the event extraction unit 32, the basic analysis unit 33 performs a basic analysis to calculate a basic evaluation value for each item of personal information (resident basic information 51 and health check information 52) by evaluating the degree of association between the item and exposure to risk factors and the degree of association between the item and the occurrence of outcomes.

[0028] The interruption risk calculation unit 34 calculates an exposure risk value that evaluates the risk of exposure to risk factors for each individual based on items corresponding to the basic evaluation value in the personal information and the interruption risk analysis information 54, and generates interruption risk information 56.

[0029] The interruption impact calculation unit 35 calculates an exposure impact value for each individual that evaluates the impact of exposure to risk factors on the occurrence of outcomes based on items corresponding to the basic evaluation values ​​in the personal information and the interruption impact analysis information 55, and generates interruption impact information 57.

[0030] The clustering unit 36 ​​is a classification unit that classifies individuals into multiple clusters according to exposure risk values ​​and exposure impact values ​​based on personal information, interruption risk information 56, and interruption impact information 57, and generates branch information 58 and cluster information 59.

[0031] The input unit 37 receives various information from the user terminal 1 etc. The output unit 38 is a display unit that outputs various information to the user terminal 1 etc. and displays the various information on the user terminal 1.

[0032] 3 is a diagram showing an example of the resident basic information 51. The resident basic information 51 shown in FIG.

[0033] Field 511 stores an individual ID, which is identification information for identifying an individual. Field 512 stores the individual's date of birth. Field 512 stores the individual's gender. Field 513 stores the individual's place of residence.

[0034] 4 is a diagram showing an example of the medical checkup information 52. The medical checkup information 52 shown in FIG.

[0035] Field 521 stores the resident's personal ID. Field 522 stores the medical checkup date, which is the time when the individual underwent the medical checkup, and the year and month (month and year) of the medical checkup. Field 523 stores the test values ​​of a predetermined test performed during the medical checkup. In this embodiment, the test value is the HbA1c value, which is an index for determining the risk of diabetes (blood glucose control state), but is not limited to this example. There may also be multiple types of test values. Fields 524 and 525 store the results of a predetermined medical interview performed during the medical checkup. Specifically, field 524 stores whether or not the individual smokes as a medical interview result, and field 525 stores whether or not the individual participates in society as a medical interview result. However, the medical interview results are not limited to these examples.

[0036] 5 is a diagram showing an example of the medical examination information 53. The medical examination information 53 shown in FIG.

[0037] Field 531 stores the resident's personal ID. Field 532 stores the examination date, which is information indicating the time when the resident received medical treatment, and the year and month (year and month) of the examination. Field 532 stores the name of the illness or injury for which the resident received medical treatment.

[0038] 6 is a diagram showing an example of the interruption risk analysis information 54. The interruption risk analysis information 54 shown in FIG.

[0039] Field 541 stores the resident's personal ID. Field 542 stores the resident's age. Field 543 stores the resident's gender. Field 544 stores the resident's place of residence. Field 545 stores the test results of a specified test for the resident. Field 546 stores whether or not the resident smokes. Field 547 stores whether or not the resident participates in society. Field 547 stores whether or not the diagnosis has been discontinued as a measure of whether or not the resident has been exposed to risk factors.

[0040] In this embodiment, the interruption risk analysis information 54 includes each item of personal information as shown in FIG. 6, but it does not have to include each item of personal information.

[0041] Fig. 7 is a diagram showing an example of interruption impact analysis information 55. The interruption impact analysis information 55 shown in Fig. 7 has a configuration in which a field 551 is added to fields 541 to 547 of the interruption risk analysis information 54. In other words, the interruption impact analysis information 55 includes all of the interruption risk analysis information 54. However, the interruption impact analysis information 55 does not need to include all of the interruption risk analysis information 54.

[0042] A field 551 stores whether or not an event that becomes an outcome has occurred.

[0043] 8 is a diagram showing an example of the interruption risk information 56. The interruption risk information 56 shown in FIG.

[0044] A field 561 stores an individual ID. A field 562 stores a discontinuation risk value, which is an exposure risk value that evaluates the risk of discontinuing medical treatment, which is exposure to a risk factor.

[0045] Fig. 9 is a diagram showing an example of the interruption impact information 57. The interruption impact information 57 shown in Fig. 9 includes fields 571 to 572.

[0046] Field 571 stores the resident's personal ID. Field 572 stores the interruption impact value, which is an exposure impact value that evaluates the impact of interruption of medical visits, which is exposure to each individual's risk factors, on the occurrence of an event (incidence of chronic kidney disease).

[0047] 10 is a diagram showing an example of branch information 58. Branch information 58 shown in FIG.

[0048] Field 581 stores a hierarchical level number indicating the level of the node in the regression tree model for classifying individuals into clusters. Field 582 stores a cluster number that identifies the node in the regression tree model within that level. In other words, a node is uniquely identified by its hierarchical level number and cluster number. Fields 583 and 584 store branching conditions for reaching that node.

[0049] 11 is a diagram showing an example of the cluster information 59. The cluster information 59 shown in FIG.

[0050] Field 591 stores an individual ID. Field 592 stores a hierarchical number indicating the hierarchical level of the cluster to which the individual with the individual ID belongs. Field 593 stores a cluster number indicating the cluster to which the individual with the individual ID belongs.

[0051] The information described above is merely an example, and may include fields other than those described above.

[0052] FIG. 12 is a sequence diagram illustrating an example of data analysis processing by the data analysis system.

[0053] First, the user inputs a basic analysis request to the input / output device 14 of the user terminal 1, requesting the execution of a basic analysis. The calculation device 13 of the user terminal 1 sends the basic analysis request to the analysis server 2 via the communication I / F 15 (step S101). The basic analysis request includes specification information that specifies the analysis conditions for the basic analysis. The analysis conditions include, for example, the aggregation period, which is the period to be analyzed, and the event to be used as the outcome.

[0054] When the input unit 37 of the analysis server 2 receives a basic analysis request, the medical examination interruption identification unit 31 determines whether or not medical examinations related to type 2 diabetes, which is exposure to risk factors, have been interrupted for each individual (each individual ID) based on the medical examination information 53. The medical examination interruption identification unit 31 generates interruption risk analysis information 54 based on the determination result, the resident basic information 51, and the health checkup information 52, and stores it in the auxiliary storage device 22 (step S102). Note that the medical examination interruption identification unit 31 checks, for example, based on the year and month of medical examination in the medical examination information 53, whether or not the interval between medical examinations has ever been a predetermined interval (for example, six months) or longer, and determines that medical examinations have been interrupted if the interval between medical examinations has ever been the predetermined interval or longer.

[0055] Thereafter, the event extraction unit 32 determines whether or not an event specified in the basic analysis request has occurred for each individual based on the medical examination information 53. Based on the determination result and the interruption risk analysis information 54, the event extraction unit 32 generates interruption impact analysis information 55 and stores it in the auxiliary storage device 22 (step S103). The event extraction unit 32 determines whether or not a medical examination related to chronic kidney disease has occurred based on, for example, the illness name in the medical examination information 53, and determines that an event has occurred if a medical examination related to chronic kidney disease has occurred. However, in this embodiment, since the analysis takes into account the causal effect between exposure to risk factors and the occurrence of an event, the event extraction unit 32 determines that an event has not occurred if a medical examination related to chronic kidney disease has occurred when no interruption of a medical examination related to diabetes, which is an exposure to a risk factor, has occurred, or if a medical examination related to chronic kidney disease has occurred before the interruption of a medical examination related to diabetes. In other words, the event extraction unit 32 determines that an event has occurred if a medical examination related to chronic kidney disease has occurred after the interruption of a medical examination related to diabetes.

[0056] Then, based on the interruption impact analysis information 55, the basic analysis unit 33 performs a regression analysis using whether or not the consultation was interrupted and whether or not an event occurred as objective variables, and each item of personal information included in the interruption impact analysis information 55 (gender, age, test results, etc.) as an explanatory variable, to perform a basic analysis to calculate a basic evaluation value that evaluates the influence of each explanatory variable on each objective variable for each explanatory variable (step S104).

[0057] Basic analysis unit 33 transmits a basic analysis result screen (see FIG. 14) showing the results of the basic analysis to user terminal 1 via output unit 38, and displays it on user terminal 1 (step S105).

[0058] The user then checks the basic analysis result screen and specifies the items to be used for clustering the individuals via the input / output device 14 of the user terminal 1. The calculation device 13 of the user terminal 1 receives the specification and sends a clustering request requesting clustering to the analysis server 2 via the communication I / F 15 (step S106). The items to be used for clustering include items to be used for calculating the interruption risk value and items to be used for calculating the interruption impact value, and the user, for example, specifies these separately. The clustering request includes the specified items as specified items.

[0059] When the input unit 37 of the analysis server 2 receives the clustering request, the interruption risk calculation unit 34 constructs an exposure risk model, which is a machine learning model with the risk specification items, which are specified in the clustering request for calculating the interruption risk value, as explanatory variables and the presence or absence of interruption of medical treatment as the objective variable, based on the interruption impact analysis information 55. The interruption risk calculation unit 34 calculates the interruption risk value of each individual by inputting the values ​​of the risk specification items of each individual into the exposure risk model, and stores the calculated value as interruption risk information 56 in the auxiliary storage device 22 (step S107). This machine learning model is, for example, a causal inference model such as Meta-Learners, Causal Tree, or Causal Forest.

[0060] Furthermore, based on the interruption impact analysis information 55, the interruption impact calculation unit 35 constructs an exposure impact model, which is a machine learning model with impact specification items, which are specified items in the clustering request for calculating the interruption impact value, as explanatory variables and whether or not an event has occurred as a response variable. The interruption impact calculation unit 35 calculates the interruption impact value of each individual by inputting the values ​​of the impact specification items of each individual into the exposure impact model, and stores the calculated interruption impact value as interruption impact information 57 in the auxiliary storage device 22 (step S108). This machine learning model is, for example, a prediction model such as Logistic-Regression or GBDT. In this embodiment, since the causal effect is taken into consideration as described above, the interruption impact value is a value obtained by evaluating the causal effect.

[0061] The clustering unit 36 ​​executes a clustering process (see FIG. 13) for clustering each individual into a plurality of clusters based on the interruption impact analysis information 55, the interruption risk information 56, and the interruption impact information 57 (step S109).

[0062] The clustering unit 36 ​​transmits a clustering result screen (see FIG. 15) showing the results of the clustering process to the user terminal 1, and displays it on the user terminal 1 (step S110), and then ends the process.

[0063] In the analysis process described above, the exposure to a risk factor is fixed to the interruption of medical treatment, but it may also be selected by the user.

[0064] FIG. 13 is a flowchart illustrating an example of the clustering process.

[0065] The clustering unit 36 ​​reads the interruption risk information 56 and the interruption impact information 57 from the auxiliary storage device 22 (steps S201, S202), combines the interruption risk information 56 and the interruption impact information 57, and generates cluster determination information indicating the interruption risk value and the interruption impact value for each individual (step S203).

[0066] The clustering unit 36 ​​constructs a regression tree model using both the interruption risk value and the interruption influence value as objective variables based on the cluster determination information (steps S204 and S205).

[0067] Specifically, the clustering unit 36 ​​calculates the amount of change in error, such as the mean square error, before and after branching for each of the interruption risk value and the interruption impact value, and repeats a division process (step S204) of dividing individuals so as to minimize the weighted sum of the amount of change in these errors until the regression tree model satisfies a termination condition (step S205). The termination condition may be, for example, when the hierarchical depth of the regression tree model or the number of samples of the lowest-level node of the regression tree model reaches a predetermined value. The predetermined value may be determined in advance, set by the user, or set according to the number of target individuals (the number of individual IDs included in the cluster determination information).

[0068] Then, the clustering unit 36 ​​generates branch information 58 indicating the determined regression tree model and stores it in the auxiliary storage device 22 (step S206). Furthermore, the clustering unit 36 ​​stores the results of clustering the individuals using the regression tree as cluster information 59 in the auxiliary storage device 22 (step S207), and ends the process.

[0069] 14 and 15 are diagrams showing a diabetes medical consultation discontinuation analysis dashboard, which is an example of a display screen displayed on the user terminal 1. Fig. 14 shows an example of a diabetes medical consultation discontinuation analysis dashboard including a basic analysis result screen, and Fig. 15 shows an example of a diabetes medical consultation discontinuation analysis dashboard including a clustering result screen.

[0070] The diabetes consultation discontinuation analysis dashboard 200a shown in FIG. 14 includes a request input screen 201 for the user to input a basic analysis request, and a basic analysis result screen 202 showing the results of the basic analysis.

[0071] The request input screen 201 includes a designation section 211 for designating analysis conditions for the basic analysis, and a send button 212 for sending a basic analysis request including the analysis conditions designated in the designation section 211. In the illustrated example, the designation section 211 includes a designation section 211a for designating the aggregation period, which is the period to be analyzed, and a designation section 211b for designating an event that will be the outcome.

[0072] The basic analysis result screen 202 includes a result section 221 showing the results of the basic analysis, an item designation section 222 for designating the items to be used in clustering, and a send button 223 for sending a clustering request including the designated items, which are the items designated in the item designation section 222.

[0073] The result section 221 displays, as a result of the basic analysis, odds ratios and p-values ​​for whether or not the consultation was interrupted and whether or not an event occurred, for each item of personal information. The odds ratios and p-values ​​are examples of basic evaluation values. Note that basic evaluation values ​​are not limited to odds ratios and p-values. The item designation section 222 includes check boxes for each item of personal information that indicate whether or not to use the interruption risk value and interruption impact value.

[0074] The diabetes consultation discontinuation analysis dashboard 200b shown in FIG. 15 includes a request input screen 201 and a clustering result screen 203 that displays the results of the clustering process.

[0075] The clustering result screen 203 includes a clustering result section 231 that displays the clustering results, and a patient mapping section 232 that maps the clustering results in a two-dimensional space.

[0076] The clustering result section 231 shows the regression tree model used for clustering. Each node 231a of the regression tree model lists the number of samples N, which is the number of samples (individuals) belonging to that node 231a, and the average values ​​of the interruption risk value and interruption impact value of that sample. Note that each node 231a may list other statistical values, such as medians, instead of the average values ​​of the interruption risk value and interruption impact value. The clustering result section 231 also shows a branching determination condition 231b for each node.

[0077] The patient mapping unit 232 displays mapping information in which information 232a indicating each cluster is mapped into a space defined by the interruption risk value and the interruption impact value based on the number of samples in each leaf node at the end of the regression tree model (i.e., the number of samples in each cluster) and the statistical values ​​of the interruption risk value and the interruption impact value of the samples in each leaf node. In the example of Figure 15, the space defined by the interruption risk value and the interruption impact value is a two-dimensional space with the interruption risk value on the horizontal axis and the exposure impact value on the vertical axis. The information indicating a cluster is displayed as a circle of a size corresponding to the number of samples at the position of the average value of the interruption risk value and the interruption impact value of the samples included in that cluster.

[0078] As described above, according to this embodiment, the interruption risk calculation unit 34 calculates an exposure risk value for each individual that evaluates the risk of exposure to a risk factor based on personal information (resident basic information 51 and health check information 52) and interruption risk analysis information 54. The interruption impact calculation unit 35 calculates an exposure impact value for each individual that evaluates the impact of exposure to a risk factor on the occurrence of an outcome based on the personal information and interruption impact analysis information 55. The clustering unit 36 ​​classifies multiple individuals into multiple groups according to the exposure risk value and exposure impact value based on the personal information. Therefore, it is possible to classify individuals into clusters with similar impacts and risks, taking into account not only the impact of exposure to a risk factor on the occurrence of an outcome but also the risk of exposure to the risk factor. Therefore, it is possible to identify groups that are predicted to be highly effective through intervention and intervene in those groups, thereby supporting efficient intervention.

[0079] Furthermore, in this embodiment, the interruption risk calculation unit 34 constructs an exposure risk model for calculating an exposure risk value based on personal information and interruption risk analysis information 54, and calculates an exposure risk value for each individual using the exposure risk model. This makes it possible to more appropriately evaluate the risk of exposure occurring.

[0080] In addition, in this embodiment, explanatory variables in the exposure risk model are specified according to a basic evaluation value that evaluates the degree of association between personal information items and exposure, making it possible to more appropriately evaluate the risk of exposure occurring.

[0081] Furthermore, in this embodiment, the interruption effect calculation unit 35 constructs an exposure effect model for calculating an exposure effect value based on personal information and interruption effect analysis information 55, and calculates an exposure effect value for each individual using the exposure effect model. This makes it possible to more appropriately evaluate the effects of exposure.

[0082] In addition, in this embodiment, explanatory variables in the exposure impact model are specified according to the basic evaluation value that evaluates the degree of association between personal information items and the occurrence of outcomes, making it possible to more appropriately evaluate the impact of exposure.

[0083] Furthermore, in this embodiment, clustering is performed using a regression tree model in which personal information is used as an explanatory variable and exposure risk values ​​and exposure impact values ​​are used as objective variables, thereby enabling appropriate clustering.

[0084] Furthermore, in this embodiment, a clustering result section 231 showing the results of clustering and a patient mapping section 232 that maps the results of clustering in two-dimensional space are displayed, making it possible to appropriately predict groups that will be highly effective through intervention, thereby supporting efficient intervention.

[0085] In this embodiment, the personal information also includes a place of residence, which is information about the area of ​​residence. The place of residence is closely related to the ease (or difficulty) of accessing medical institutions, and therefore it becomes possible to appropriately evaluate the risk of exposure to risk factors (such as interruption of diagnosis).

[0086] The above-described embodiments of the present disclosure are merely illustrative examples of the present disclosure, and are not intended to limit the scope of the present disclosure to these embodiments alone. Those skilled in the art may implement the present disclosure in various other forms without departing from the scope of the present disclosure. [Explanation of symbols]

[0087] 1: User terminal 2: Analysis server 11: Main memory device 12: Auxiliary memory device 13: Computing unit 14: Input / output device 15: Communication I / F 21: Main memory device 22: Auxiliary memory device 23: Computing unit 24: Input / output device 25: Communication I / F 31: Consultation interruption identification unit 32: Event extraction unit 33: Basic analysis unit 34: Interruption risk calculation unit 35: Interruption impact calculation unit 36: Clustering unit 37: Input unit 38: Output unit

Claims

1. a storage unit that stores personal information about each of a plurality of individuals, exposure information indicating whether or not each individual is exposed to a risk factor, and measurement target information indicating whether or not each individual has developed an outcome affected by the exposure; a risk calculation unit that calculates an exposure risk value for each individual by evaluating the risk of the exposure occurring based on the personal information and the exposure information; an influence calculation unit that calculates an exposure influence value for each individual, the exposure having an influence on the occurrence of the outcome, based on the personal information and the measurement target information; a classification unit that classifies the plurality of individuals into a plurality of groups according to the exposure risk value and the exposure effect value based on the personal information.

2. 2. The data analysis device of claim 1, wherein the risk calculation unit constructs an exposure risk model for calculating the exposure risk value from the personal information based on the personal information and the exposure information, and calculates the exposure risk value for each individual using the exposure risk model.

3. a basic analysis unit that calculates a basic evaluation value for each item by evaluating the degree of association between the item of personal information and the exposure based on the personal information and the exposure information; an output unit that displays a basic analysis result indicating the basic evaluation value for each of the items; and an input unit that receives designation information that designates the items to be used as explanatory variables of the exposure risk model, The data analysis device according to claim 2 , wherein the risk calculation unit constructs the exposure risk model using the items specified in the specification information as explanatory variables.

4. 2. The data analysis device according to claim 1, wherein the influence calculation unit constructs an exposure influence model that calculates the exposure influence value from the personal information based on the personal information and the measurement target information, and calculates the exposure influence value for each individual using the exposure influence model.

5. a basic analysis unit that calculates a basic evaluation value for each item by evaluating the degree of association between the item of personal information and the occurrence of the outcome based on the personal information and the measurement target information; an output unit that displays a basic analysis result indicating the basic evaluation value for each of the items; an input unit that receives designation information that designates the items to be used as explanatory variables of the exposure-effect model; The data analysis device according to claim 4 , wherein the risk calculation unit constructs the exposure-effect model using the items specified in the specification information as explanatory variables.

6. 2. The data analysis device according to claim 1, wherein the classification unit constructs a regression tree model with the personal information as an explanatory variable and the exposure risk value and the exposure impact value as objective variables, and classifies the individuals into the groups using the regression tree model.

7. 7. The data analysis device according to claim 6, further comprising an output unit that displays information indicating a branching decision condition at each node of the regression tree model and statistical values ​​of the exposure risk value and the exposure impact value of the sample at each node.

8. 7. The data analysis device according to claim 6, further comprising an output unit that displays mapping information in which information indicating each of the plurality of populations is mapped onto a space defined by the exposure risk values ​​and the exposure impact values, based on the number of samples in each leaf node at the end of the regression tree model and the statistical values ​​of the exposure risk values ​​and the exposure impact values ​​of the samples in each leaf node.

9. The data analysis device of claim 1 , wherein the exposure is a discontinuation of medical treatment for a specific disease.

10. The data analysis device according to claim 1 , wherein the outcome is a specific disease, a change in test value, an increase or decrease in medical expenses, the performance of a medical procedure, hospitalization, death, certification of need for care, or a change in the level of need for care.

11. The data analysis device according to claim 1 , wherein the personal information includes information about the individual's residential area.

12. A data analysis method using a data analysis device, storing personal information about each of a plurality of individuals, exposure information indicating whether or not each individual is exposed to a risk factor, and measurement target information indicating whether or not each individual has developed an outcome affected by the exposure; calculating an exposure risk value for each individual that evaluates the risk of the exposure occurring based on the personal information and the exposure information; calculating an exposure effect value for each individual that evaluates the effect of the exposure on the outcome based on the personal information and the measurement target information; A data analysis method that classifies the plurality of individuals into a plurality of groups according to the exposure risk value and the exposure effect value based on the personal information.

Images