Risk analysis support system and risk analysis support method

Abstract

To evaluate an event occurrence risk with high accuracy according to an event occurrence time.SOLUTION: A risk analysis support system includes a processor and a storage device connected to the processor, wherein the storage device stores information indicating a health condition of each of a plurality of persons in each period, and the processor is configured to construct a plurality of risk models for predicting a predetermined risk outcome in accordance with a length of an elapsed period from a reference time point based on the information indicating the health condition of each of the plurality of persons in each period, and predict a risk outcome of a person to be analyzed in accordance with a length of an elapsed period from a predetermined time point by inputting information indicating the health condition of the person to be analyzed into the plurality of risk models.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technology for supporting analysis of risks related to health conditions. [Background technology]

[0002] Disease risk prediction is performed based on the health history information of a subject at a certain point in time, and is based on the occurrence or non-occurrence of future outcomes over a specified period of time, and can be widely used in underwriting assessments for life insurance, health insurers, health guidance by health promotion providers, etc. To accurately predict future risk, it is effective to calculate future risk over time based on past health history information.

[0003] One example of a technology for predicting disease risk is the technology described in JP 2021-189585 (Patent Document 1). Patent Document 1 states, "The prediction device has an input unit, a model generation unit, a standard threshold calculation unit, a time-dependent threshold calculation unit, a risk identification unit, and an output unit. The model generation unit generates a future hospitalization risk model using notification information, and outputs a time-dependent risk score, which is the change in the risk score over time, using known techniques such as statistical methods and machine learning methods with the hospitalization risk score as the objective variable." [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-189585 Summary of the Invention [Problem to be solved by the invention]

[0005] In the method described in Patent Document 1, the accuracy of the prediction model varies depending on the length of time until the time of the target prediction, so a method is disclosed in which a threshold is changed over time in consideration of the prediction error, but a method for constructing a time-varying risk prediction model is not made clear.In underwriting assessments and health guidance settings, input information is acquired as health history for a certain period of time in the past, so a time-varying risk assessment method based on this information is required.

[0006] Therefore, in order to solve the above problems, the present invention aims to provide a method for performing time-expanded conversion of health history information and information on the time of event occurrence, thereby enabling highly accurate evaluation of the risk of event occurrence. [Means for solving the problem]

[0007] In order to solve at least one of the above problems, a representative example of the invention disclosed in the present application is a risk analysis support system comprising a processor and a storage device connected to the processor, wherein the storage device holds information indicating the health status of multiple individuals at each time point, and the processor constructs multiple risk models that predict a predetermined risk outcome depending on the length of time that has elapsed since a reference point based on the information indicating the health status of the multiple individuals at each time point, and predicts the risk outcome depending on the length of time that has elapsed since a predetermined point in time for the individual being analyzed by inputting information indicating the health status of the individual being analyzed into the multiple risk models. [Effects of the Invention]

[0008] According to one aspect of the present invention, it is possible to accurately evaluate the risk of an event occurring according to the time of the event occurrence. Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiments. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a block diagram showing an example of the configuration of a risk analysis support system according to an embodiment of the present invention. [Figure 2]FIG. 3 is an explanatory diagram illustrating an example of basic information managed by a basic information management unit according to the embodiment of the present invention. [Figure 3] FIG. 2 is an explanatory diagram showing an example of health checkup information managed by a health information management unit according to the embodiment of the present invention. [Figure 4] FIG. 2 is an explanatory diagram illustrating an example of medical history information managed by a medical history information management unit according to an embodiment of the present invention. [Figure 5] 10 is an explanatory diagram illustrating an example of time-dependent extension conversion information managed by a data conversion information management unit according to an embodiment of the present invention; FIG. [Figure 6] FIG. 2 is an explanatory diagram showing an example of risk model parameter information managed by a risk model information management unit in an embodiment of the present invention. [Figure 7] FIG. 10 is an explanatory diagram illustrating an example of parameter fitting information managed by a model correction information management unit according to an embodiment of the present invention. [Figure 8] 1 is an explanatory diagram showing an example of risk assessment definition information managed by a risk assessment information management unit in an embodiment of the present invention. [Figure 9] 10 is an explanatory diagram showing an example of a risk value calculation result managed by the risk assessment information management unit of an embodiment of the present invention. FIG. [Figure 10] 10 is a flowchart illustrating an example of processing executed by a data generation unit according to the embodiment of this invention. [Figure 11] 10 is a flowchart illustrating an example of processing executed by a model construction unit according to an embodiment of the present invention. [Figure 12] 10 is a flowchart illustrating an example of processing executed by a risk determination unit and a risk value calculation unit according to the embodiment of the present invention. [Figure 13] 10 is a flowchart illustrating an example of processing executed by a model analysis unit according to the embodiment of this invention. [Figure 14] 10 is a flowchart illustrating an example of processing executed by a model correction unit according to an embodiment of the present invention. [Figure 15] FIG. 10 is an explanatory diagram showing an example of a user interface displayed in the processing of the risk determination unit and the risk value calculation unit according to the embodiment of the present invention. [Figure 16]FIG. 10 is an explanatory diagram showing another example of a user interface displayed during processing by the risk determination unit and the risk value calculation unit in the embodiment of the present invention. [Figure 17] FIG. 10 is an explanatory diagram illustrating an example of a user interface displayed during processing by a model analysis unit according to an embodiment of the present invention. [Figure 18] FIG. 10 is an explanatory diagram illustrating an example of a user interface displayed in the processing of the model correction unit according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0011] FIG. 1 is a block diagram showing an example of the configuration of a risk analysis support system 101 according to an embodiment of the present invention.

[0012] The risk analysis support system 101 is a computer system and includes an input unit 102 such as a keyboard and mouse, an output unit 103 representing a display that outputs display data, a CPU (Central Processing Unit) 104, a memory 105, a communication unit 108, and a storage medium 106.

[0013] The risk analysis support system 101 has a data generation unit 111, a model construction unit 112, a risk determination unit 113, a risk value calculation unit 114, a model analysis unit 115, and a model correction unit 116. The functions of each unit from the data generation unit 111 to the model correction unit 116 are realized by the CPU 104 executing a program stored in the storage medium 106. When these programs are executed by the CPU 104, at least a part of them may be copied to the memory 105 as necessary.

[0014] A database 107 is connected to the risk analysis support system 101. The database 107 has a basic information management unit 121, a health information management unit 122, a medical history information management unit 123, a data conversion information management unit 124, a risk model information management unit 125, a risk assessment information management unit 126, and a model correction information management unit 127.

[0015] As will be described later, the basic information management unit 121 manages basic information 200 (Fig. 2). The health information management unit 122 manages medical checkup information 300 (Fig. 3). The medical history information management unit 123 manages medical history information 400 (Fig. 4). The data conversion information management unit 124 manages time-dependent extension conversion information 500 (Fig. 5). The risk model information management unit 125 manages risk model parameter information 600 (Fig. 6). The risk assessment information management unit 126 manages risk assessment definition information 800 (Fig. 8) and risk value calculation results 900 (Fig. 9). The model correction information management unit 127 manages parameter fitting information 700 (Fig. 7).

[0016] The database 107 may be stored in a storage system connected to the risk analysis support system 101 via a network, for example, or may be built into the risk analysis support system 101 (for example, by being stored in the storage medium 106). When the database 107 is stored in a system external to the risk analysis support system 101, at least a portion of its contents may be copied to the storage medium 106 or memory 105 as necessary. Furthermore, the entire system including the computer having the input unit 102, output unit 103, CPU 104, memory 105, and storage medium 106, and the database 107 may be referred to as the risk analysis support system.

[0017] Furthermore, the risk analysis support system 101 may be realized by one computer having the configuration shown in Fig. 1, for example, or may be realized by multiple computers. For example, the information held in the database 107 described above may be stored in a distributed manner in multiple storage media 106 or memories 105, and the functions of the risk analysis support system 101 described above may be executed in a distributed manner by multiple CPUs 104 of multiple computers.

[0018] FIG. 2 is an explanatory diagram showing an example of basic information 200 managed by the basic information management unit 121 according to the embodiment of the present invention.

[0019] The basic information 200 is basic information about each person. This embodiment can be applied to various purposes, but here, an example will be described in which the risk analysis support system 101 supports information analysis for an insurance company to evaluate the risk of insurance payment for a person who has applied to subscribe to an insurance product. In this example, the basic information 200 is basic information about each person extracted from the notification information submitted by each person to the insurance company in order to apply to subscribe to the insurance product of the insurance company.

[0020] Specifically, the basic information 200 includes a personal ID 201 that identifies each person, a gender 202 that identifies the gender of each person, a date of birth 203 that identifies the date of birth of each person, an observation start date 204 and an observation end date 205 that indicate the start and end dates of observation of the health condition of each person, etc. The observation start date 204 and the observation end date 205 may be, for example, the dates when information indicating the health condition of each person, specifically, medical checkup information 300 and medical history information 400 described below, were first and last acquired. The above basic information 200 is an example, and the basic information 200 may include various information about each person as necessary.

[0021] Although this embodiment describes an example of risk analysis by an insurance company as described above, the present invention is not limited to this. For example, this embodiment can be applied to assess health risks when a local government provides health guidance according to the health risks of residents.

[0022] FIG. 3 is an explanatory diagram showing an example of health checkup information 300 managed by the health information management unit 122 according to the embodiment of the present invention.

[0023] The medical checkup information 300 is information about the results of a medical checkup (health check) that each person has undergone, and may be information that each person has submitted as part of notification information when applying for insurance, for example. The medical checkup information 300 includes a personal ID 301, a test date 302, a BMI 303, a fasting blood glucose 304, an HbA1c 305, a medical interview result 306, and findings 307, etc.

[0024] The personal ID 301 is information that identifies each person and corresponds to the personal ID 201 in the basic information 200. The examination date 302 indicates the date on which the health check was conducted. The BMI 303, fasting blood glucose 304, and HbA1c 305 are examples of test values ​​obtained as a result of the health check. The medical interview result 306 is the result of the medical interview conducted during the health check, and may include information indicating, for example, whether or not the person has a drinking habit, whether or not they have an exercise habit, etc. The findings 307 are information on the findings made as a result of the health check.

[0025] The above information is a typical example of information obtained as a result of a health checkup, and in reality, the health checkup information 300 may not include at least any of these, or may include information on items other than these (e.g., blood pressure, etc.). Furthermore, the health checkup information 300 may also include test results and interview results from previous visits to the hospital or hospitalizations, etc., other than the health checkup.

[0026] FIG. 4 is an explanatory diagram showing an example of medical history information 400 managed by the medical history information management unit 123 according to the embodiment of the present invention.

[0027] The medical history information 400 is information about illnesses or injuries that each person has experienced in the past (so-called pre-existing conditions, etc.) that is extracted from the notification information that each person submitted to the insurance company when applying for insurance, and for example, the information that each person wrote in the item corresponding to the pre-existing condition in the notification information may be retained as is as the medical history information 400. The medical history information 400 includes a personal ID 401, an illness or injury name 402, an illness or injury code 403, a length of hospitalization 404, surgery information 405, and medication 406.

[0028] The personal ID 401 is information that identifies each person, and corresponds to the personal ID 201 in the basic information 200. The injury / illness name 402 and injury / illness code 403 are information that identify the injury or illness of each person. The hospitalization period 404 is information that indicates whether each person was hospitalized or not, and if so, the length of hospitalization. The surgery information 405 is information that identifies whether each person has undergone surgery or not, and if so, the details of the surgery, etc. The medication 406 is information that identifies whether each person has taken medication or not, and if so, the type of medication, etc. Furthermore, for people with no medical history, there are no records in the medical history information 400.

[0029] The above basic information 200, medical examination information 300, and medical history information 400 are merely examples, and the contents and acquisition methods thereof are not limited to those described above. For example, information similar to the above may be acquired from a so-called PHR (Personal Health Record).

[0030] FIG. 5 is an explanatory diagram showing an example of temporal extension conversion information 500 managed by the data conversion information management unit 124 according to the embodiment of the present invention.

[0031] The time-course extended conversion information 500 is information generated as learning data for constructing a risk model in this embodiment based on the basic information 200, the medical checkup information 300, and the medical history information 400. The time-course extended conversion information 500 shown in Fig. 5 includes an individual ID 501, an age 502, a sex 503, a BMI 504, a fasting blood glucose 505, hypertension 506, dyslipidemia 507, diabetes 508, myocardial infarction 509, a drug A 510, a drug B 511, a period of time elapsed 512, whether or not risk outcome A has occurred 513, whether or not risk outcome B has occurred 514, whether or not risk outcome C has occurred 515, and whether or not risk outcome D has occurred 516.

[0032] The personal ID 501 is information that identifies each person, and corresponds to the personal ID 201 in the basic information 200. The age 502 and the gender 503 are information that identify the age and gender of each person, and correspond to the date of birth 203 and the gender 202 in the basic information 200, respectively.

[0033] The BMI 504 and fasting blood glucose 505 are information indicating the health condition of each person, and correspond to the BMI 303 and fasting blood glucose 304, respectively, of the medical checkup information 300. Although omitted in FIG. 5, the time-dependent expansion conversion information 500 may further include other items (e.g., HbA1c) included in the medical checkup information 300.

[0034] Hypertension 506, dyslipidemia 507, diabetes 508, and myocardial infarction 509 are information indicating the medical history of each person, and each item corresponds to an illness or disease recorded in the illness or disease name 402 of the medical history information 400. Although Fig. 5 shows the above four items, the time-dependent extension conversion information 500 may further include items for other illnesses or diseases (for example, lung cancer, acute myocardial infarction, etc. shown in Fig. 4).

[0035] Drug A 510 and drug B 511 are information indicating the medication history of each person, and each item corresponds to a drug recorded in medication 406 of medical history information 400. Although Fig. 5 shows the above two items, time-dependent extension conversion information 500 may further include items for other drugs (for example, drug C, drug D, etc. shown in Fig. 4).

[0036] The elapsed period 512 indicates the period of time (for example, the number of months) that has elapsed since a reference point in time.

[0037] Risk outcome A occurrence or non-occurrence 513, risk outcome B occurrence or non-occurrence 514, risk outcome C occurrence or non-occurrence 515, and risk outcome D occurrence or non-occurrence 516 each indicate whether or not an event set as a risk outcome has occurred. Here, risk outcome indicates the event that is the target of risk analysis (i.e., the risk of the event occurring is the target of prediction by the risk model).

[0038] For example, when the risk analysis support system 101 of this embodiment is used for underwriting assessment of life insurance, an event for which insurance payment is made (e.g., death or hospitalization due to a specified injury or illness) may be set as the risk outcome. In this case, in addition to an event that simply results in the payment of insurance payment, an event that results in the payment of insurance payment of a specified amount or more may be set as the risk outcome, or the amount of insurance payment that occurs may be set as the risk outcome.

[0039] Alternatively, when the risk analysis support system 101 is used to select persons to be subject to insurance guidance by a local government, an event that is a criterion for selecting persons to be subject to the guidance (for example, receiving nursing care certification) may be set as a risk outcome.

[0040] For example, lines 1 to 7 of the time-course extension conversion information 500 shown in Fig. 5 record information about a one-month period from a reference point in time up to 60 months after the reference point in time for a person whose personal ID 501 is "P001." In the following description, a person whose personal ID 501 is "P001" may be referred to as "person "P001," a person whose personal ID 501 is "P002" may be referred to as "person "P002," and so on. Also, the period from the reference point in time up to one month after the reference point in time (i.e., the period when the elapsed period 512 is "1") may be referred to as elapsed period "1," the period from one month after the reference point in time to two months after the reference point in time may be referred to as elapsed period "2," the period from 59 months after the reference point in time to 60 months after the reference point in time may be referred to as elapsed period "60," and so on.

[0041] Here, the information from age 502 to drug B 511 is baseline information obtained up to the reference point. In the example of Fig. 5, in elapsed period "1", person "P001" is a 36-year-old male with a BMI of 22, fasting blood glucose of 90, hypertension, no dyslipidemia, no diabetes, and no myocardial infarction, and taking drug A and drug B. This information may be, for example, the latest information obtained up to the reference point (for example, the latest information about the person included in the basic information 200, health check information 300, and medical history information 400), or may be an average value, minimum value, maximum value, or the like for a predetermined period, such as the past few years.

[0042] Meanwhile, in the fields of Risk Outcome A Occurrence 513 to Risk Outcome D Occurrence 516, information indicating whether an event corresponding to the risk outcome occurred during each period indicated by the elapsed period 512 is stored. In the example of Figure 5, "1" indicates that an event occurred, and "0" indicates that an event did not occur.

[0043] For example, according to the first row of the time-dependent extended conversion information 500 shown in Figure 5, none of the risk outcomes A to D occurred for person "P001" during elapsed period "1." This indicates that, based on the health checkup information 300, medical history information 400, etc., it has been determined that no events corresponding to risk outcomes A to D occurred for person "P001" during elapsed period "1."

[0044] On the other hand, in the seventh line, the occurrence status 513 of risk outcome A for person "P001" in the elapsed period "60" is set to "1." This indicates that, based on the medical checkup information 300, medical history information 400, etc., it has been determined that an event corresponding to risk outcome A occurred in person "P001" in the elapsed period "60."

[0045] For other persons such as person “P002” and person “P003”, similar information is obtained from the basic information 200 , medical examination information 300 and medical history information 400 and stored in the time-dependent extension conversion information 500 .

[0046] In the above example, information about each person from the reference point in time up to 60 months later is collected and stored in the time-over-time extension conversion information 500. However, in reality, depending on the person, the period for which information can be obtained may be shorter than 60 months due to circumstances such as the person starting to collect information late or the person quitting their job and data collection being interrupted. In such cases, information about the period for which information can be obtained is stored in the time-over-time extension conversion information 500 together with the value of the elapsed period 512. Generally, the larger the value of the elapsed period 512, the fewer the number of data samples stored in the time-over-time extension conversion information 500.

[0047] FIG. 6 is an explanatory diagram showing an example of risk model parameter information 600 managed by the risk model information management unit 125 according to the embodiment of the present invention.

[0048] The risk model parameter information 600 includes a risk model ID 601 that identifies the risk model, a target elapsed period 602 that indicates the elapsed period that the risk model targets, a target outcome 603 that indicates the risk outcome that the risk model targets, and model parameters 604 that indicate the structure and parameters of the risk model.

[0049] 6, the first line of the example risk model parameter information 600 stores "A001," "1," "A," and "Age: 1.1, Gender: 2.4, BMI: 1.5, ..." as the risk model ID 601, target elapsed time 602, target outcome 603, and model parameters 604. This indicates that the model with risk model ID 601 "A001" (hereinafter, this will also be simply referred to as risk model "A001." The same applies to other models) is a model for predicting the risk of risk outcome A occurring within one month from a certain point in time, based on the values ​​of a person's age, gender, BMI, etc. at that point in time, and that the coefficients for the age, gender, and BMI are "1.1," "2.4," and "1.5," respectively.

[0050] This risk model "A001" is a model generated by learning the row of the time-dependent extended conversion information 500 where the elapsed period 512 is "1" using age 502, gender 503, BMI 504, etc. as explanatory variables and the occurrence or non-occurrence of risk outcome A 513 as the objective variable.

[0051] The type of model to be generated and the method for generating it are not limited, but in this example, it may be a multiple regression model or a logistic regression model with age, gender, BMI, etc. as explanatory variables and the risk of occurrence of risk outcome A as the objective variable.

[0052] Similarly, the second line of the example risk model parameter information 600 shown in Figure 6 indicates that risk model "A002" is a model for predicting the risk of risk outcome A occurring between one month and two months after a certain point in time based on the values ​​of a person's age, gender, BMI, etc., and that the coefficients for age, gender, and BMI are "1.2," "2.0," and "1.6," respectively.

[0053] This risk model "A002" is a model generated by learning the row of the time-dependent extended conversion information 500 where the elapsed period 512 is "2" using age 502, gender 503, BMI 504, etc. as explanatory variables and the occurrence or non-occurrence of risk outcome A 513 as the objective variable.

[0054] When comparing risk models "A001" and "A002," even if the explanatory variables and objective variables are the same, the elapsed period 512 of the information used for learning is different, and as a result, learning is performed based on different information, and the values ​​of the generated model parameters 604 are also different.

[0055] Similarly, models predicting the risk of occurrence of risk outcome A for each elapsed period other than those mentioned above, and models predicting the risk of occurrence of risk outcomes B, C, D, etc. for each elapsed period are generated and stored in the risk model parameter information 600.

[0056] FIG. 7 is an explanatory diagram showing an example of parameter fitting information 700 managed by the model correction information management unit 127 according to the embodiment of the present invention.

[0057] The parameter fitting information 700 indicates the results of fitting the parameters of the generated risk model. For example, as described below, the parameter values ​​of the risk model are fitted to the target elapsed time period. Specifically, the parameter fitting information 700 shown in FIG. 7 includes a target outcome 701, a risk model ID 702, a fitting method 703, target parameters 704, and a fitting result 705.

[0058] Target outcome 701 indicates the risk outcome to be predicted by a model including the fitted parameters. Risk model ID 702 is information that identifies the risk model including the fitted parameters, and corresponds to the risk model ID 601 in the risk model parameter information 600. Fitting method 703 indicates the method used for fitting. Target parameters 704 indicate the fitted parameters. Fitting result 705 indicates the fitting results.

[0059] For example, the first line of the parameter fitting information 700 shown in Figure 7 indicates that when the coefficient value for age, among the parameters of risk model "A001" that predicts the risk of occurrence of risk outcome A using age, sex, BMI, etc. as explanatory variables, is plotted in a space with the target elapsed time as the horizontal axis, it is fitted by linear regression to a line with a slope of 1.1, with an error of 0.01 and a correlation coefficient of 0.7.

[0060] Similarly, the results of fitting other parameters (for example, gender, BMI) and the results of fitting each parameter of a risk model for predicting the risk of occurrence of other risk outcomes are stored in parameter fitting information 700.

[0061] FIG. 8 is an explanatory diagram showing an example of risk assessment definition information 800 managed by the risk assessment information management unit 126 according to the embodiment of the present invention.

[0062] The risk assessment definition information 800 includes information defining criteria for assessing the level of risk based on a risk value calculated using a risk model. Specifically, the risk assessment definition information 800 includes a definition ID 801, a risk model ID 802, and a assessment threshold 803. The definition ID 801 is information for identifying each definition. The risk model ID 802 is information for identifying the risk model for which the assessment criteria are defined, and corresponds to the risk model ID 601 in the risk model parameter information 600. The assessment threshold 803 is the defined assessment criteria, and in this example indicates the assessment threshold.

[0063] FIG. 9 is an explanatory diagram showing an example of a risk value calculation result 900 managed by the risk assessment information management unit 126 according to the embodiment of the present invention.

[0064] The risk value calculation result 900 indicates a risk value calculated using a risk model. Specifically, the risk value calculation result 900 includes an individual ID 901, a period 902, a risk of outcome A occurring 903, a risk of outcome B occurring 904, a risk of outcome C occurring 905, and a risk of outcome D occurring 906.

[0065] The individual ID 901 is information that identifies each person and corresponds to the individual ID 201 in the basic information 200. The elapsed period indicates the period (e.g., the number of months) that has elapsed since a reference point. The outcome A occurrence risk 903 to outcome D occurrence risk 906 indicate the risk of risk outcome A to risk outcome D occurring, respectively, calculated using the risk model.

[0066] The first row of the risk value calculation result 900 shown in Figure 9 stores "P001," "1," and "0.51" as the individual ID 901, elapsed time period 902, and risk of outcome A occurring 903, respectively. This indicates that by inputting information indicating the health status of person "P001" as an explanatory variable into a risk model (e.g., risk model "A001" shown in Figure 6) in which the target outcome 603 is "A" and the target elapsed time period 602 is "1," the risk of risk outcome A occurring in elapsed time period "1" (i.e., the period from the base point to one month later) was calculated as "0.51." If the explanatory variables input into the risk model here are the current values, then "0.51" is the predicted value for the risk of risk outcome A occurring in the period from the current point to one month later.

[0067] Similarly, the risk of occurrence of outcome A 903 in the second row of the risk value calculation result 900 stores the risk of occurrence of risk outcome A for person P001 in the period from one month to two months after the reference point, calculated using a risk model in which the target outcome 603 is "A" and the target elapsed period 602 is "2." Although some parts are omitted in Figure 9, the risk value calculation result 900 similarly stores the risk of occurrence of risk outcome A for person P001 for each month from the reference point until 60 months later.

[0068] In the example in Figure 9, the risk of risk outcome A initially increases as the elapsed period increases, but then starts to decrease as the elapsed period becomes longer. By constructing a risk model according to the length of the elapsed period and predicting the risk of risk outcome occurrence using that risk model, it is possible to predict risks that show trends other than a monotonically increasing or decreasing trend, as described above.

[0069] Similarly, the risk value calculation result 900 stores the risk of occurrence of risk outcome B to risk outcome D for person "P001" for each elapsed period, calculated using the same method as above, and the calculated results of the risk of occurrence of risk outcome A to risk outcome D for each elapsed period for each person other than person "P001."

[0070] FIG. 10 is a flowchart showing an example of processing executed by the data generation unit 111 according to the embodiment of the present invention.

[0071] The data generation process shown in FIG. 10 is a process for generating learning data for constructing a risk model, and includes a time-dependent extension conversion process for generating time-dependent extension conversion information 500 from basic information 200, medical examination information 300, and medical history information 400.

[0072] When this process starts (step 1001), the data generation unit 111 reads the basic information 200, the medical examination information 300, and the medical history information 400 (steps 1002, 1003, 1004).

[0073] Next, the data generation unit 111 sets the conversion conditions (step 1005). For example, the data generation unit 111 may set the period of data used to construct the risk model (i.e., which period of data to use for learning), the method of generating data during time-extension conversion (e.g., which period of value to use as the baseline value, whether to use the latest value during that period or the average value of values ​​at multiple points in time, etc.), the period length and step size of the data generated by time-extension conversion, and the risk outcome to be predicted. Here, an example will be described in which 60 months and 1 month are specified as the data period and step size, respectively, but other values ​​may also be set.

[0074] Next, the data generation unit 111 collates the acquired data (step 1005). For example, the data generation unit 111 organizes the acquired basic information 200, medical examination information 300, and medical history information 400 for each person based on the personal ID.

[0075] Next, the data generating unit 111 calculates the elapsed period (step 1007). For example, when the data generating unit 111 is set to generate data in one-month increments up to 60 months later as described above, the data generating unit 111 extracts baseline information at the reference time point from the matched data for each person and stores the information in the corresponding item for each one-month period up to 60 months later from the reference time point in the time-course extended conversion information 500.

[0076] Next, the data generation unit 111 determines whether or not an outcome has occurred. For example, if risk outcomes A to D are set as prediction targets, the data generation unit 111 determines whether or not an event corresponding to risk outcomes A to D has occurred during each one-month period based on the collated data for each person (particularly, data related to each person extracted from the health checkup information 300 and medical history information 400), and stores the results in the corresponding item of the time-course extension conversion information 500 (step 1008).

[0077] The elapsed period is calculated based on the conversion condition settings and each person's observation start date 204 and observation end date 205. As an example, we will explain generation from data for a person whose observation start date 204 is April 2015 and whose observation end date 205 is March 2021. Here, the conversion conditions are set to use values ​​from April 2015 to March 2016 as baseline information, and the base point of the elapsed period is April 2016. In this case, the elapsed period is 1 month to 60 months, the confirmation period for the outcome to be predicted is April 2016 to March 2021, and the data generation unit 111 checks the occurrence of the outcome every month during that period. When predicting whether or not an outcome occurs, a "1" is entered in the elapsed period field for the outcome occurrence for the corresponding items, such as "Risk Outcome A Occurrence Presence / Absence" 513 to "Risk Outcome D Occurrence Presence / Absence" 516. In this case, a total of 60 records of time-expanded converted data are created.

[0078] As mentioned above, depending on the person, it may not be possible to acquire data for the entire period up to 60 months in the future. In that case, the acquired data for the period that could be acquired is stored in the time-course extension conversion information 500.

[0079] This completes the data generation process (step 1009).

[0080] FIG. 11 is a flowchart showing an example of processing executed by the model construction unit 112 according to the embodiment of the present invention.

[0081] 11 is a process for constructing a risk model based on generated learning data. When this process starts (step 1101), the model construction unit 112 reads the time-dependent extended conversion information 500 as learning data (step 1101) and sets analysis conditions (step 1103). As analysis conditions, for example, the type of model to be constructed, explanatory variables, objective variables, the length and step size of the period to be predicted, etc. can be set.

[0082] In this embodiment, the type of model to be constructed is not limited. For example, any model, such as a statistical model or a machine learning model, can be constructed. The explanatory variables may include at least one of items of basic information 200, such as age and gender, items of health check information 300, such as BMI, blood glucose level, and blood pressure, and items of medical history information 400, such as a history of hospitalization due to a specified injury or illness. An event corresponding to a risk outcome is set as the objective variable.

[0083] An example will be described in which the length and step size of the period to be predicted are set to 60 months and 1 month, respectively, similar to those set in the time-expansion transformation.

[0084] Next, the model construction unit 112 constructs a risk model according to the conditions set in step S1103 (step 1104). At this time, the model construction unit 112 constructs a risk model that predicts the occurrence of a risk outcome from the reference point up to one month later, using data from rows in which the elapsed period 512 of the time-course extension conversion information 500 is "1." A known method can be used as a specific technique for constructing the risk model, and therefore detailed explanation will be omitted.

[0085] Similarly, the model construction unit 112 constructs a risk model that predicts the occurrence of a risk outcome for each month from one month after the reference time point to 60 months after, using data in rows where the elapsed period 512 is "2" to "60" in the time-dependent extension conversion information 500. Then, when the elapsed period reaches the maximum (60 in the above example) (step 1105: Yes), the model construction unit 112 stores the parameters of the constructed risk model in the risk model parameter information 600 (step 1106).

[0086] As mentioned above, in general, the larger the value of the elapsed time 512, the smaller the number of corresponding data samples. Therefore, in general, the longer the elapsed time, the lower the accuracy of the risk model.

[0087] This completes the model construction process (step 1107).

[0088] FIG. 12 is a flowchart showing an example of processing executed by the risk determining unit 113 and the risk value calculating unit 114 according to the embodiment of the present invention.

[0089] The risk assessment process shown in Figure 12 is a process for assessing the risk of occurrence of a risk outcome based on a constructed risk model. When this process starts (step 1201), the risk assessment unit 113 reads data. Here, data corresponding to the baseline information of the person who is the target of risk assessment is read.

[0090] Next, the risk determination unit 113 reads the constructed risk model (step 1203). At this time, a plurality of risk models are read depending on the risk outcome and period to be determined.

[0091] For example, if you want to determine a person's risk of risk outcome A occurring for each month from the present until 60 months from now, step 1202 loads the person's current baseline information, and step 1202 loads 60 risk models with target outcome 603 of "A" and target elapsed time period 602 ranging from "1" to "60."

[0092] Next, the risk assessment unit 113 reads the risk assessment definition corresponding to the risk model from the risk assessment definition information 800 (step 1204).

[0093] Next, the risk value calculation unit 114 calculates a risk value by inputting the data read in step 1202 into the risk model read in step 1203 (step 1205). When the calculation of risk values ​​using all read risk models (e.g., the above 60 risk models) is completed (step 1206: Yes), the risk determination unit 113 determines the risk of each period by comparing the calculated risk value with the risk determination definition read in step 1204 (e.g., the determination threshold 803 corresponding to each period) (step 1207).

[0094] This completes the risk assessment process (step 1208).

[0095] FIG. 13 is a flowchart showing an example of processing executed by the model analysis unit 115 according to the embodiment of the present invention.

[0096] 13 is a process for analyzing the tendency of a constructed risk model. When this process starts (step 1301), the model analysis unit 115 reads the risk model to be analyzed (step 1302). Specifically, the model analysis unit 115 reads, for example, the parameters 604 of the risk model to be analyzed.

[0097] Next, the model analysis unit 115 sets parameter analysis conditions (step 1303). Specifically, conditions for fitting parameter values ​​may be set. For example, conditions may be set such that parameter values ​​for an elapsed period, such as 1 to 60 months, are fitted using linear regression or other methods. The user may observe parameter values ​​for an elapsed period and manually set conditions that are deemed appropriate.

[0098] Next, the model analysis unit 115 analyzes the tendency of the risk model to be analyzed (step 1304) in accordance with the conditions set in step 1303. For example, it may perform linear or curved fitting in accordance with the set conditions and obtain a regression equation as a result.

[0099] This completes the model analysis process.

[0100] FIG. 14 is a flowchart showing an example of processing executed by the model corrector 116 according to the embodiment of the present invention.

[0101] 14 is a process for correcting a constructed risk model based on the results of its analysis. When this process is started (step 1401), the model correction unit 116 reads the risk model to be corrected (step 1402), sets parameter correction conditions (step 1403), corrects the risk model (step 1404), and stores the resulting correction parameters in the model correction information management unit 127 (step 1405).

[0102] For example, if the risk model that was the subject of the model analysis processing shown in Figure 14 is read in at step 1402 as the correction target, and a correction condition is set at step 1403 that correction is to be made using the regression equation obtained in the model analysis processing, then at step 1404, the parameter values ​​corresponding to each elapsed period are corrected using that regression equation.

[0103] Alternatively, if a correction condition that correction based on prior probability is performed is set in step 1403, the parameter values ​​of the risk model for the elapsed period to be corrected are corrected based on prior probability in step 1404. Correction based on prior probability will be described later with reference to FIG.

[0104] This completes the model correction process (step 1406).

[0105] Next, an example of a user interface provided by the risk analysis support system 101 when executing the processes shown in FIGS. 10 to 14 will be described with reference to FIGS.

[0106] FIG. 15 is an explanatory diagram showing an example of a user interface displayed in the processing of the risk determination unit 113 and the risk value calculation unit 114 according to the embodiment of the present invention.

[0107] A risk analysis execution screen 1500 shown in FIG. 15 is an example of a screen that displays the results of risk analysis performed by the risk analysis support system 101, and includes, for example, an input information display section 1501 and a risk display section 1502.

[0108] The input information display section 1501 includes a plurality of input data fields 1503, a risk ratio display selection field 1504, and a process execution button 1505. The input data field 1503 displays input fields for information identifying the person who is the subject of risk analysis, the prediction period, the prediction target, and the risk judgment definition.

[0109] For example, a subject ID is input as information to identify the person who is the subject of risk analysis, the elapsed period to be predicted is input as the prediction period, and the risk outcome to be predicted is input as the prediction target. In the example of Figure 15, it is specified that the risk of occurrence of risk outcome A for person "P001" from 1 month to 60 months from now is predicted based on risk assessment definition "RA_001."

[0110] The risk ratio display selection field 1504 is a checkbox for specifying whether or not to display the predicted risk value as a ratio. In the example of Fig. 15, the checkbox is not checked, so it is specified that the risk value itself is displayed.

[0111] When the user operates the process execution button 1505, the risk model is read and risk determination processing is performed based on the information entered in the input data field 1503 (see FIG. 12). The results are then displayed in the risk display section 1502 using the method selected in the risk ratio display selection field 1504.

[0112] In the example of Figure 15, a graph is displayed in which the predicted risk of risk outcome A occurring for each period is plotted in space with the horizontal axis representing the prediction period (i.e., the number of months elapsed from 1 month to 60 months) and the vertical axis representing the risk value. In this example, the risk of risk outcome A occurring is higher than the judgment threshold when only a few months have elapsed, then decreases, and falls below the judgment threshold as the number of months elapsed increases. For this reason, the example of Figure 15 shows that the short-term risk is high and the long-term risk is low.

[0113] FIG. 16 is an explanatory diagram showing another example of a user interface displayed in the processing of the risk determination unit 113 and the risk value calculation unit 114 according to the embodiment of the present invention.

[0114] A risk analysis execution screen 1600 shown in FIG. 16 is another example of a screen that displays the results of risk analysis performed by the risk analysis support system 101, and includes, for example, an input information display section 1601 and a risk display section 1602.

[0115] The input information display section 1601 includes a plurality of input data fields 1603, a risk ratio display selection field 1604, and a process execution button 1605. These are similar to the input data field 1503, risk ratio display selection field 1504, and process execution button 1505 shown in Fig. 5, but in the example of Fig. 16, a check mark has been entered in the risk ratio display selection field 1604. In this case, the risk display section 1602 displays the ratio of the predicted risk of occurrence of risk outcome A.

[0116] Here, for example, the ratio in prediction period t represents the risk value of the person being predicted in prediction period t / the risk value of the standard person in prediction period t. In the example of Figure 16, a graph is displayed in which the ratio of the risk of occurrence of risk outcome A for each predicted period is plotted in a space with the horizontal axis representing the prediction period and the vertical axis representing the risk value ratio.

[0117] In this example, the risk of the person being predicted experiencing risk outcome A is approximately twice that of a person in standard health when only a few months have passed, and then decreases to the same as that of a person in standard health when approximately 60 months have passed.

[0118] In insurance appraisal, the risk of the subject is often evaluated in comparison with the risk of a standard person, so when the risk analysis support system 101 is used for insurance underwriting appraisal, user convenience is improved by displaying a screen such as that shown in Figure 16.

[0119] FIG. 17 is an explanatory diagram showing an example of a user interface displayed in the processing of the model analysis unit 115 according to the embodiment of the present invention.

[0120] A model analysis execution screen 1700 shown in FIG. 17 is an example of a screen that displays the results of the risk model analysis performed by the risk analysis support system 101, and includes, for example, an input information display section 1701 and an analysis result display section 1702.

[0121] The input information display section 1701 includes a plurality of input data fields 1703 and a process execution button 1704. The input data field 1703 displays input fields for information identifying the risk model to be analyzed (e.g., risk model ID), information identifying the parameter items to be analyzed, and information identifying the analysis method. In the example of Fig. 17, it is specified that the model parameters of the explanatory variable "BMI" of risk models "A001" to "A060" are to be analyzed by "linear regression".

[0122] Here, risk models "A001" to "A060" are 60 risk models for predicting the risk of occurrence of risk outcome A for each month from 1 month to 60 months later, and the model parameter of the explanatory variable "BMI" is, for example, a coefficient related to the BMI value in the regression equation if the risk models are regression models. This value can be interpreted as the contribution of BMI to the risk of occurrence of the risk outcome.

[0123] When the user operates the process execution button 1704, a model analysis process is performed based on the information entered in the input data field 1703 (see FIG. 13), and the results are displayed in the analysis result display section 1702. In the example of FIG. 17, a graph showing the results of analyzing the parameter values ​​of the risk model to be analyzed for each period using a specified analysis method is displayed in a space with the horizontal axis representing the prediction period and the vertical axis representing the parameter values ​​to be analyzed.

[0124] In this example, a circular shape is plotted showing the parameter values ​​being analyzed. In this example, the contribution of BMI increases as the forecast period increases. The results of linear regression of these parameter values ​​are displayed as a dashed line. The slope is 1.5, the error is 0.01, and the correlation coefficient is 0.8.

[0125] In the above example, the maximum prediction period in the training data is 60 months, so the parameters obtained through training are those up to 60 months into the future. However, by extrapolating the parameters using a regression model, it is possible to estimate model parameters for longer periods.

[0126] For example, if the parameter value is 100 when the prediction period is 60 months, the parameter value when the prediction period is 70 months can be calculated using a slope of 1.5 as 100 + 1.5 × (70 - 60) = 115.

[0127] Alternatively, the parameter values ​​of the constructed risk model can be corrected based on a regression model. Also, a risk model with higher resolution of the elapsed period can be constructed by interpolating parameters using a regression model.

[0128] FIG. 18 is an explanatory diagram showing an example of a user interface displayed during processing by the model corrector 116 according to the embodiment of the present invention.

[0129] A model correction execution screen 1800 shown in FIG. 18 is an example of a screen that displays the results of the risk model analysis performed by the risk analysis support system 101, and includes, for example, an input information display section 1801 and a correction result display section 1802.

[0130] The input information display section 1801 includes a plurality of input data fields 1803 and a process execution button 1804. The input data field 1803 displays input fields for information identifying the risk model to be corrected (e.g., risk model ID), information identifying the parameter item to be corrected, and information identifying the correction method. In the example of FIG. 18, it is specified that the model parameters of the explanatory variable "gender" of risk models "A001" to "A060" are to be corrected based on the "prior probability." Here, the risk models "A001" to "A060" are the same 60 risk models as those shown in FIG. 17, and the model parameters of the explanatory variable "gender" are, for example, coefficients related to gender in the regression equation when these risk models are regression models.

[0131] When the user operates the process execution button 1804, a model correction process is performed based on the information entered in the input data field 1803 (see Fig. 14), and the results are displayed in the correction result display section 1802. In the example of Fig. 18, a graph showing the results of correcting the parameter values ​​of the risk model to be corrected using the specified correction method is displayed in a space with the horizontal axis representing the prediction period and the vertical axis representing the parameter values ​​to be corrected.

[0132] As already explained, the number of samples included in the training data tends to be higher for shorter forecast periods and lower for longer forecast periods. It is known that the fewer the number of samples in a risk model, the more difficult it becomes to estimate parameters (the influence of a particular sample becomes relatively larger).

[0133] In this example, a model is created for each prediction period. When parameter estimation is difficult due to a small number of samples or outcome occurrences, it is effective to make corrections based on the model parameters for the prediction period one step earlier (or one step later, or even several steps earlier or later). For example, in the case of a generalized linear model, parameter values ​​and their error distributions (error bars in the figure) are calculated. The error distribution of the parameters estimated at the previous time point t-1 is used as the prior probability distribution for time point t (the shaded distribution in the figure), thereby setting the parameter estimation range for the risk model at time point t. This has the effect of reducing the problem of parameter estimation results changing significantly with a single step forward or backward. This is particularly effective for outcomes where risk is not expected to increase or decrease sharply over time.

[0134] Note that the correction method shown in Figure 18 is just an example, and other correction methods may also be applied. For example, the parameter values ​​for each prediction period may be corrected using the regression line shown in Figure 17. In this case, the value of the regression line itself may be used, or a value calculated by applying a predetermined correction factor may be used. Alternatively, a method such as smoothing may be used.

[0135] In the above embodiment, a model is constructed that predicts risk outcomes (e.g., events that will be covered by insurance payments) related to a person's health condition according to the length of time that has elapsed since a reference point, based on information indicating the person's health condition (e.g., medical checkup information 300, medical history information 400, etc.), and the model is used to predict risk outcomes for the person being analyzed according to the length of time that has elapsed since a predetermined point in time. This method can also be applied to a machine to predict risk outcomes related to the machine.

[0136] For example, instead of the medical checkup information 300 and medical history information 400, measurements taken at the time of machine inspection and repair, and the findings of the person in charge may be stored together with information indicating when the inspection and repair were performed, and a model may be constructed based on this information to predict a risk outcome related to the state of the machine depending on the length of time that has elapsed since a reference point. Here, a risk outcome related to the state of the machine may be, for example, the occurrence of a malfunction. By inputting measurements and the like related to the state of the machine to be analyzed into this model, it is possible to predict a risk outcome related to the length of time that has elapsed since the reference point for that machine.

[0137] Furthermore, the system according to the embodiment of the present invention may be configured as follows.

[0138] (1) A risk analysis support system (e.g., risk analysis support system 101) comprising a processor (e.g., CPU 104) and a storage device (e.g., memory 104 and storage medium 106) connected to the processor, wherein the storage device holds information indicating the health status of multiple persons at each time period (e.g., at least a portion of basic information 200, medical checkup information 300, and medical history information 400), and the processor constructs multiple risk models (e.g., risk models having parameters stored in risk model parameter information 600) that predict a predetermined risk outcome depending on the length of time elapsed from a reference point in time based on the information indicating the health status of the multiple persons at each time period (e.g., model construction process shown in Figure 11), and predicts the risk outcome depending on the length of time elapsed from a predetermined point in time for the analysis subject by inputting information indicating the health status of the analysis subject into the multiple risk models (e.g., risk judgment process shown in Figure 12).

[0139] This allows for accurate prediction of risk outcomes over time.

[0140] (2) In the risk analysis support system described in (1) above, the processor generates, for each person, learning data (e.g., time-dependent extended conversion information 500) from information indicating the health status of the multiple persons at each time point, which corresponds the values ​​of multiple items indicating the health status at the reference time point (e.g., baseline information) with the value of the risk outcome (e.g., whether or not a risk outcome has occurred) depending on the length of time elapsed since the reference time point, and by learning the learning data, constructs the multiple risk models in which the values ​​of the multiple items indicating the health status at the reference time point are used as explanatory variables and the value of the risk outcome depending on the length of time elapsed since the reference time point is used as a target variable.

[0141] This generates a model that predicts risk outcomes over time.

[0142] (3) In the risk analysis support system described in (2) above, the storage device further stores risk definition information (e.g., risk assessment definition information 800) that defines the criteria for risk assessment, and the processor assesses the risk of the predicted analysis subject based on the risk definition information according to the length of time that has elapsed since a predetermined point in time (e.g., step 1207).

[0143] This makes it possible to determine whether the risk exceeds a predetermined standard for each elapsed period.

[0144] (4) In the risk analysis support system described in (2) above, the processor analyzes the relationship between the length of the period elapsed from the reference point in time and the parameter values ​​of the plurality of risk models, and performs at least one of correction, interpolation, and extrapolation of the parameter values ​​based on the analysis results (for example, the model analysis process shown in Figure 13).

[0145] This makes it possible to predict risk outcomes for periods during which no model has been generated, for example, because no training data was available.

[0146] (5) In the risk analysis support system described in (2) above, the processor corrects the parameter values ​​of the second risk model by treating the distribution of parameter values ​​of a first risk model corresponding to a first elapsed period among the parameter values ​​of the multiple risk models as the distribution of prior probabilities of parameter values ​​of a second risk model corresponding to a second elapsed period different from the first elapsed period, and setting an estimation range of the parameter values ​​of the second risk model based on the distribution of prior probabilities (for example, the model correction process shown in Figure 14).

[0147] This makes it possible to correct a risk model that is inaccurate because, for example, a sufficient amount of training data was not obtained.

[0148] The present invention is not limited to the above-described embodiments and includes various modifications. For example, the above-described embodiments have been described in detail to provide a better understanding of the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0149] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The above-described configurations, functions, etc. may also be implemented in software, with a processor interpreting and executing a program that implements each function. Information such as the programs, tables, and files that implement each function can be stored in storage devices such as nonvolatile semiconductor memory, hard disk drives, and solid-state drives (SSDs), or in computer-readable, non-transitory data storage media such as IC cards, SD cards, and DVDs.

[0150] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are interconnected. [Explanation of symbols]

[0151] 101 Risk Analysis Support System 102 Input section 103 Output section 104 CPU 105 memory 106 Storage medium 107 Database 108 Communications Department 111 Data Generation Unit 112 Model Construction Department 113 Risk Assessment Department 114 Risk Value Calculation Unit 115 Model Analysis Department 116 Model Correction Unit 121 Basic Information Management Department 122 Health Information Management Department 123 Medical History Information Management Department 124 Data Conversion Information Management Department 125 Risk Model Information Management Department 126 Risk Assessment Information Management Department 127 Model Correction Information Management Unit

Claims

1. A risk analysis support system, a processor and a storage device connected to the processor, the storage device holds information indicating the health conditions of a plurality of persons at each time period; The processor: constructing a plurality of risk models that predict predetermined risk outcomes according to the length of time elapsed from a reference point based on the information indicating the health status of the plurality of individuals at each time point; A risk analysis support system characterized by predicting the risk outcome according to the length of time that has elapsed since a specified point in time for the subject of analysis by inputting information indicating the health status of the subject of analysis into the multiple risk models.

2. 2. The risk analysis support system according to claim 1, The processor: generating, for each person, learning data that associates values ​​of a plurality of items indicating the health status at the reference time point with values ​​of the risk outcome according to the length of time elapsed since the reference time point, from information indicating the health status of the plurality of people at each time point; A risk analysis support system characterized by constructing a plurality of risk models by learning the learning data, in which the values ​​of a plurality of items indicating the health status at the reference point in time are used as explanatory variables, and the value of the risk outcome according to the length of time elapsed from the reference point in time is used as a target variable.

3. 3. The risk analysis support system according to claim 2, the storage device further holds risk definition information that defines a criterion for determining a risk; A risk analysis support system characterized in that the processor determines the risk of the predicted analysis subject according to the length of time that has elapsed since a predetermined point in time, based on the risk definition information.

4. 3. The risk analysis support system according to claim 2, The processor analyzes the relationship between the length of the period elapsed since the reference point in time and the parameter values ​​of the plurality of risk models, and performs at least one of correction, interpolation, and extrapolation of the parameter values ​​based on the analysis results.

5. 3. The risk analysis support system according to claim 2, a processor correcting the parameter values ​​of the second risk model by setting a distribution of parameter values ​​of a first risk model corresponding to a first elapsed period among the parameter values ​​of the plurality of risk models as a distribution of prior probabilities of parameter values ​​of a second risk model corresponding to a second elapsed period different from the first elapsed period, and setting an estimation range of the parameter values ​​of the second risk model based on the distribution of prior probabilities.

6. A risk analysis support method executed by a risk analysis support system, the risk analysis support system includes a processor and a storage device connected to the processor; the storage device holds information indicating the health conditions of a plurality of persons at each time period; The risk analysis support method includes: a first step in which the processor constructs a plurality of risk models that predict predetermined risk outcomes according to the length of time elapsed from a reference point in time, based on information indicating the health status of the plurality of persons at each time point; A risk analysis support method characterized by including a second step in which the processor predicts the risk outcome according to the length of time that has elapsed since a predetermined point in time for the subject of analysis by inputting information indicating the health status of the subject of analysis into the multiple risk models.

7. 7. The risk analysis support method according to claim 6, In the first step, the processor generating, for each person, learning data that associates values ​​of a plurality of items indicating the health status at the reference time point with values ​​of the risk outcome according to the length of time elapsed since the reference time point, from information indicating the health status of the plurality of people at each time point; A risk analysis support method characterized by constructing a plurality of risk models by learning the learning data, in which the values ​​of a plurality of items indicating the health status at the reference time point are used as explanatory variables, and the value of the risk outcome according to the length of time elapsed from the reference time point is used as a target variable.

8. 8. The risk analysis support method according to claim 7, the storage device further holds risk definition information that defines a criterion for determining a risk; A risk analysis support method characterized in that in the second step, the processor determines the risk of the predicted analysis subject based on the length of time that has elapsed since a specified point in time, based on the risk definition information.

9. 8. The risk analysis support method according to claim 7, A risk analysis support method characterized by further comprising a third step in which the processor analyzes the relationship between the length of the period elapsed from the reference point in time and the parameter values ​​of the plurality of risk models, and performs at least one of correction, interpolation, and extrapolation of the parameter values ​​based on the analysis results.

10. 8. The risk analysis support method according to claim 7, a fourth step in which the processor corrects the parameter value of the second risk model by setting a distribution of parameter values ​​of a first risk model corresponding to a first elapsed period among the parameter values ​​of the plurality of risk models as a distribution of prior probabilities of parameter values ​​of a second risk model corresponding to a second elapsed period different from the first elapsed period, and setting an estimation range of the parameter values ​​of the second risk model based on the distribution of prior probabilities.

11. A risk analysis support system, a processor and a storage device connected to the processor, the storage device holds information indicating the state of a plurality of machines at each time point; The processor: constructing a plurality of risk models that predict predetermined risk outcomes according to the length of time elapsed from a reference point in time based on information indicating the state of the plurality of machines at each time point; A risk analysis support system characterized by predicting the risk outcome according to the length of time that has elapsed since a specified point in time for the machine being analyzed by inputting information indicating the state of the machine being analyzed into the multiple risk models.

Images