Information processing device, information processing method and program

Abstract

To provide a technique capable of facilitating prediction of a tendency of future changes in estimated glomerular filtration rate (eGFR).SOLUTION: An information processing device includes: acquisition means for acquiring time-series medical insurance information of a specific person; and prediction means for predicting a tendency of future changes in estimated glomerular filtration rate (eGFR) of the specific person based on time-series medical insurance information by using a trained machine learning model. The machine learning model predicts the tendency of the future changes in eGFR of the specific person using time-series eGFR and time-series measurements other than eGFR contained in the time-series medical insurance information.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] If diabetic nephropathy becomes severe, artificial dialysis becomes necessary, which is a major burden on medical expenses and daily life. For this reason, a program is provided to prevent diabetic nephropathy from becoming severe (Non-Patent Document 1). In such a program, the criteria for determining whether a person has diabetic nephropathy are diabetes and impaired renal function. In addition, an index is provided that uses the measured value of estimated glomerular filtration rate (eGFR) as an index of impaired renal function, and for people with a relatively high measured eGFR, for example, when the eGFR is less than 60 mL / min / 1.73 m2, a decrease of 5 mL / min / 1.73 m2 or more per year is indicated as a reference.

[0003] On the other hand, it is known that eGFR measurements contain large measurement errors due to the influence of the amount of water in the blood, etc. Even if eGFR is measured frequently (e.g., seven times) in a year, it is often impossible to obtain a correct downward trend, so by measuring and analyzing (e.g., regression analysis) frequently for five years or more, it is finally possible to identify a certain level of downward trend in eGFR. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] "Diabetic Nephropathy Severity Prevention Program," [online], [searched August 9, 2023],<https: / / www.mhlw.go.jp / content / 12401000 / program.pdf> Summary of the Invention [Problem to be solved by the invention]

[0005] The measured value of eGFR decreases with age, but the rate of decline (amount of change per year) differs from person to person. For people with eGFR of around 60mL / min / 1.73m2, whose renal function is not significantly deteriorating, there are overwhelmingly more people whose eGFR declines slowly than those whose eGFR declines rapidly. For this reason, for the majority of people, there are high barriers from the standpoint of time and cost to frequently measure eGFR over several years to determine whether their own rate of decline in eGFR exceeds the standard rate (for example, whether it is 5mL / min / 1.73m2 or more per year).

[0006] The present invention has been made in consideration of the above problems, and an object of the present invention is to realize a technology that makes it possible to easily predict the trend of future changes in estimated glomerular filtration rate (eGFR). [Means for solving the problem]

[0007] In order to solve this problem, for example, the information processing device of the present invention has an acquisition means for acquiring time-series medical insurance information of a specific person, and a prediction means for predicting the trend of future changes in the estimated glomerular filtration rate (eGFR) of the specific person based on the time-series medical insurance information using a trained machine learning model, wherein the machine learning model predicts the trend of future changes in the eGFR of the specific person using the time-series eGFR and time-series measurement values ​​other than eGFR from health checkups contained in the time-series medical insurance information. Effect of the Invention

[0008] According to the present invention, it is possible to easily predict the future trend of changes in eGFR. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing an example of a prediction system according to an embodiment of the present invention. [Diagram 2] FIG. 1 is a diagram for explaining an outline of the operation of a prediction system according to an embodiment of the present invention; [Diagram 3] FIG. 1 is a block diagram showing an example of a functional configuration of an information processing device according to an embodiment of the present invention; [Figure 4] FIG. 2 is a block diagram showing an example of the functional configuration of a communication device of an intervention institution according to the present embodiment; [Diagram 5] FIG. 13 is a diagram showing an example of specific medical checkup information that can be acquired from a database that stores a user's medical insurance information. [Figure 6] A diagram explaining examples of various data correlated with changes in eGFR [Figure 7] A flowchart showing the operation of the eGFR prediction process according to this embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the invention according to the claims, and not all combinations of features described in the embodiments are essential to the invention. Two or more features among the multiple features described in the embodiments may be arbitrarily combined. In addition, the same reference numbers are used for the same or similar configurations, and duplicated descriptions are omitted.

[0011] (Configuration of prediction system) First, referring to FIG. 1, the configuration of a prediction system 10 according to an embodiment of the present invention will be described. The prediction system 10 includes, for example, an information processing device 100, a communication device 110 used by an interventionist 20 of an intervention institution, and a medical insurance information database 130. The information processing device 100 is, for example, a server managed by a service company that provides predictions regarding estimated glomerular filtration rate (eGFR). The interventionist 20 is, for example, an employee of an intervention institution such as a local government who provides (intervenes in) guidance to improve the health condition of a specific individual according to the health condition of the individual. For example, the interventionist provides guidance to improve the health condition of an individual whose future change tendency of eGFR meets certain conditions. The interventionist 20 is not limited to this example, and may include, for example, a doctor of a medical institution who provides advice or judgment according to the future change tendency of eGFR.

[0012] The information processing device 100 accesses the medical insurance information database 130 to acquire medical insurance information including medical checkup results of each person. The medical insurance information database 130 may be, for example, a database of a MynaPortal that stores personal medical insurance information collected and managed by a public mechanism and can be accessed via a predetermined API. Information on daily medical checkups and medical treatments of individuals is collected and accumulated in the medical insurance information database 130 from medical institutions and the like. In this embodiment, a case where the medical insurance information database 130 is a database of a MynaPortal that collects and manages medical insurance information will be described as an example. However, the medical insurance information database 130 may be a database that a service company that manages the information processing device 100 collects medical insurance information from health insurance associations, medical institutions, and the like and builds it. Alternatively, the medical insurance information database 130 may be a database that a company other than the above-mentioned service company collects medical insurance information from health insurance associations, medical institutions, and the like and builds it. That is, the medical insurance information according to this embodiment may include information on medical care acquired from health insurance associations, the Japan Health Insurance Association, mutual aid associations, national health insurance, and the medical care system for the elderly. In addition, medical insurance information may include, for example, medical receipt data, medical treatment, procedures, prescriptions, tests, test results, and other vital data, medications, medical history, and disability information that can be obtained from PHRs, etc., contained in electronic medical record data.

[0013] The information processing device 100 can communicate with the communication device 110 and the medical insurance information database 130 via the network 120. In this embodiment, the information processing device 100 is a server on the cloud, but the information processing device 100 is not limited to this and may be realized by a virtual machine, an edge server, a computer placed on a small network, or the like. In addition, in this embodiment, the communication device 110 is, for example, a personal computer. However, the communication device 110 is not limited to this and may be another terminal that can use a service related to eGFR prediction using a browser or a dedicated application, such as a smartphone or a tablet terminal.

[0014] (Summary of the Invention) Next, an overview of the present invention will be described with reference to FIG. 2. As described above, it is not realistic to measure eGFR frequently for several years or more in order to determine the rate of decline in eGFR for people with a relatively high eGFR value (eGFR less than 60 mL / min / 1.73 m2). However, if the future trend of change in eGFR for each person can be easily and accurately predicted, individuals whose eGFR value is currently relatively high but is rapidly declining can be discovered and extracted early, and early intervention can be performed. If early intervention can be realized, the rate of decline in renal function can be slowed early, and for example, the time when diabetic nephropathy becomes severe can be delayed. Delaying the time when the condition becomes severe leads to reducing the burden of living and medical expenses on individuals and governments. The prediction system can easily and accurately predict the future trend of change in eGFR at any eGFR value. This is particularly advantageous in that it can easily and accurately provide future trends in eGFR changes for people with relatively high eGFR values ​​(e.g., eGFR of 45 mL / min / 1.73 m2 or more and less than 60 mL / min / 1.73 m2).

[0015] The input 201 to the machine learning model 202 includes time-series data including measurements measured at multiple medical checkups (several years' worth of medical checkups). The measurements measured at the medical checkups include, for example, measurements at each medical checkup in the most recent three consecutive years. In this embodiment, by using measurements obtained at the medical checkups, it is easy to obtain input data, and it is easy to predict the future trend of changes in eGFR. In addition, since a prediction result can be obtained by using measurements from at most three years' worth of measurements, it is also easy to predict the future trend of changes in eGFR in terms of the number of samples being reduced. For example, in the first, second, and third consecutive medical checkups for three consecutive years, about two years have passed between the first and third medical checkups. In this embodiment, measurements from three consecutive years' worth of medical checkups may be used, or measurements from medical checkups for a longer period of time (for example, measurements from any three years' worth of medical checkups out of five years, or measurements from five years' worth of medical checkups out of five years, etc.) may be used. The time-series data according to this embodiment may be any data over time, and may include discrete, discontinuous data.

[0016] In this embodiment, the time series data of past eGFR measurements obtained in a medical checkup and the time series data of measurements other than eGFR obtained in the same medical checkup are input into a trained machine learning model to obtain the tendency of future changes in eGFR. That is, the tendency of future changes in eGFR is predicted by considering the behavior of measurements other than eGFR in addition to the behavior of the eGFR measurements. In addition, the time series data of eGFR measurements obtained in a medical checkup before the first time point and the time series data of measurements other than eGFR obtained in the same medical checkup are input, and the annual change in eGFR calculated from the time series data of eGFR measurements in the past and future from the first time point is used as correct answer data to train the machine learning model to predict the tendency of future changes in eGFR. The correct answer data may be in various formats, and the correct answer data may be, for example, time series data of eGFR measurements in the future from the first time point.

[0017] Measurement values ​​in health checkups other than eGFR include at least one of renal function-related values ​​and risk factors. Although renal function-related values ​​and risk factors are shown in FIG. 2, either one may be used. Renal function-related values ​​are measurements related to a decline in renal function, and include, for example, hematocrit value (anemia index). Risk factors are risk factors related to a decline in renal function, and include, for example, HbA1c value, hypertension (systolic blood pressure), dyslipidemia (HDL cholesterol), and increased BMI. Note that renal function-related values ​​and risk factors are examples of measurements in health checkups other than eGFR, and measurements other than these may be used.

[0018] In addition to the above, health checkup measurements other than eGFR may include at least one of, for example, BMI, waist circumference, diastolic blood pressure, triglycerides, LDL cholesterol, GOT, GPT, γ-GTP, urinary sugar, urinary protein, hemoglobin, and red blood cell count.

[0019] Furthermore, the input to the machine learning model 202 may include further data in addition to the measurements at the medical checkup. For example, the input 201 may include the results of a medical interview in a medical checkup. The results of the interview may include at least any of the following: whether or not the patient is taking blood sugar-lowering medication or insulin injections, whether or not the patient has kidney disease, whether or not the patient is smoking, exercise habits, walking habits, walking speed, eating early, eating before bedtime, snacking, eating habits, drinking frequency, amount of alcohol, and sleep time.

[0020] The input 201 may also include the amount of change in eGFR on an annual basis obtained by statistical processing of past eGFR (e.g., the slope of the regression line for eGFR for the past three years). Furthermore, the input 201 may further use fluctuations in measurements from health checkups other than eGFR over a predetermined period of time in the past (e.g., fluctuations in hematocrit and HbA1c values ​​over the past three years). In this way, by inputting the amount of change in eGFR and fluctuations in measurements over a predetermined period of time in the past into the machine learning model, not only time-series data but also the rate of change and fluctuations can be explicitly given to the machine learning model. The input 201 may include other data such as sex, age, and medication information. In addition, the input 201 may use any data of the above-mentioned medical insurance information.

[0021] The output 203 by the machine learning model 202 is a trend of future eGFR changes. In this embodiment, the case where the trend of future eGFR changes is a time series predicted value of future eGFR will be described as an example. That is, the case where the machine learning model is a model that inputs time series data and outputs a time series predicted value will be described as an example. However, the machine learning model may output data in other formats as long as the output 203 represents a trend of future eGFR changes. For example, the trend of future eGFR changes may be the magnitude of eGFR changes in a predetermined future period (for example, the magnitude of change in the future three years). Alternatively, the trend of future eGFR changes may be the annual change amount of future eGFR (i.e., the slope of eGFR). Note that the trend of future eGFR changes may be data obtained using, for example, the predicted future eGFR predicted value and past eGFR, as long as the predicted future eGFR predicted value is used. Such data may be, for example, the annual change amount calculated from the predicted future eGFR predicted value and past eGFR.

[0022] The machine learning model 202 may be, for example, a gradient boosting decision tree model such as LightGBM or XGBoost, but may also be other machine learning models such as a deep neural network.

[0023] (Configuration of information processing device) Next, a functional configuration example of the information processing device 100 will be described with reference to Fig. 3. Each of the functional blocks described may be integrated or separated, and the functions described may be realized by another block. Also, what is described as hardware may be realized by software, and vice versa.

[0024] The communication unit 301 includes a communication circuit that communicates with various devices via a network. The communication unit 301 receives information processed by the control unit 304 from a communication partner device (e.g., the medical insurance information database 130, etc.) and transmits information processed by the control unit 304 to a communication partner device (e.g., the communication device 110). The power supply unit 302 is a power supply that provides power required for the operation of the information processing device 100.

[0025] The storage unit 303 includes a non-volatile storage medium such as a hard disk or a semiconductor memory, and includes various programs executed by the control unit 304 of the information processing device 100 and various data and DBs used by the control unit 304. The various programs include an operating system, a framework, a library, and the like, in addition to a program for executing the eGFR prediction process according to this embodiment. The various data include, for example, setting values ​​of the information processing device 100 and parameter values ​​after learning of the machine learning model. The storage unit 303 may also store learning data for learning the machine learning model.

[0026] The control unit 304 includes a CPU 310, which is a central processing unit, and a RAM 311. The control unit 304 expands a program stored in the storage unit 303 into the RAM 311 and executes it, thereby controlling the operation of each unit inside the control unit 304 and controlling the operation of each unit of the information processing device 100. In addition, the control unit 304 uses each unit inside the control unit 304 to execute an eGFR prediction process, which will be described later.

[0027] The RAM 311 includes a volatile storage medium such as a DRAM, and temporarily stores parameters and processing results for the control unit 304 to execute a program.

[0028] The medical insurance information acquisition unit 320 acquires medical insurance information from the medical insurance information database 130 via the communication unit 301. The medical insurance information includes, for example, specific medical checkup information 501 shown in Fig. 5. Note that in this example, specific medical checkup information is used as an example of medical checkup information, but the medical checkup information in this embodiment is not limited to information obtained from specific medical checkups and includes information obtained from any medical checkup.

[0029] The specific health check information shown in Fig. 5 includes, for example, information on the specific health check taken at a health check institution (measurements and answers in the health check). The specific health check information includes, for example, information in tables such as qualification information, specific health check reception information, health check information basic information, health check information doctor's judgment, health check information questionnaire (specific health check), and specific health check questionnaire (health check for the elderly). The health check information basic items 502 in this specific health check information include information such as BMI, eGFR, HbA1c, and hematocrit value. Note that the information in the table of the specific health check information shown in Fig. 5 shows information included in the medical insurance information of the known My Number Portal as an example of this embodiment.

[0030] When performing inference stage processing (i.e., when predicting the future trend of changes in eGFR for a specific person), the medical insurance information acquisition unit 320 acquires time-series medical insurance information for a specific person. On the other hand, when training a machine learning model or generating training data used for training a machine learning model, the medical insurance information acquisition unit 320 acquires time-series medical insurance information for the number of people required for training the machine learning model.

[0031] The trend prediction unit 321 predicts the trend of future changes in eGFR of a specific person based on the time-series medical insurance information by using a trained machine learning model. As described above, the machine learning model uses the time-series eGFR included in the time-series medical insurance information and the measurement values ​​in the medical checkup other than eGFR to output, for example, a time-series prediction value of future eGFR corresponding to three years. As described above, the input of the machine learning model may include, in addition to the measurement values ​​in the medical checkup, the interview result for the questionnaire, or the annual change in eGFR obtained by statistical processing of the subject's past eGFR (for example, the slope of the regression line for the eGFR for the past three years). By using the annual change in eGFR based on the subject's past eGFR, the machine learning model can add a rough trend to the prediction. Furthermore, the input can further use the fluctuation in the measurement values ​​in the medical checkup over a predetermined period in the past (for example, the fluctuation in the hematocrit value or HbA1c value over the past three years). By using fluctuations over a given period of time in the past, the machine learning model can take into account trends of change in measurements other than eGFR.

[0032] The intervention candidate extraction unit 322 determines whether the trend of future changes in eGFR output by the trend prediction unit 321 satisfies the intervention condition, and extracts the subject as an intervention candidate if the intervention condition is satisfied. The intervention condition includes a condition for determining that eGFR is rapidly declining, such as, for example, a rate of decline in eGFR of 5 mL / min / 1.73 m2 or more per year. If the determination is positive, the subject is extracted as an intervention candidate. If the intervention candidate extraction unit 322 determines that the trend of future changes in eGFR satisfies the intervention condition, it can transmit a notification (candidate extraction notification) to the communication device 110 that there is a trend of changes that satisfies the intervention condition. For example, the candidate extraction notification may include that the subject of the prediction is a candidate for intervention and the rate of decline in eGFR.

[0033] The model generation unit 323 trains a machine learning model that predicts the tendency of future changes in eGFR of a specific person. For example, the model generation unit 323 inputs time series data of eGFR measurements obtained in a medical checkup prior to the first time point and time series data of measurements other than eGFR obtained in the same medical checkup, and trains the machine learning model to predict the tendency of future changes in eGFR using the annual change in eGFR calculated from the time series data of eGFR measurements prior to and in the future from the first time point as correct answer data. As described above, for example, LightGBM, XGBoost, etc. can be used as the machine learning model, and the learning of the machine learning model can be performed using a known method. When a decision tree model such as LightGBM or XGBoost is used as the machine learning model, the influence of each input of the machine learning model on the output can be grasped. Therefore, the influence of each input on the output can be used as the basis for the prediction result. For example, when the tendency of future changes in eGFR output by the machine learning model satisfies the intervention condition, the above-mentioned candidate extraction notice may indicate the basis for the prediction. The candidate extraction notice may include evidence associated with the prediction, such as, for example, "The reasons why the rate of decline in renal function is determined to be high are: (1) a trend of decline in renal function in the past, (2) high HbA1c, and (3) high blood pressure."

[0034] (Configuration of communication device) Furthermore, a configuration example of the communication device 110 will be described with reference to Fig. 4. Fig. 4 shows a functional configuration example of the communication device 110 of this embodiment. Each of the functional blocks described may be integrated or separated, and the functions described may be realized by another block. Also, what is described as hardware may be realized by software, and vice versa.

[0035] The communication unit 401 includes, for example, a communication circuit, and communicates with the information processing device 100 via mobile communication such as LTE, or via wireless communication such as WiFi, to transmit and receive necessary data.

[0036] The operation unit 403 includes a keyboard and a touch panel, and accepts operations on various operation screens displayed on the display unit 405. The display unit 405 includes a display panel such as an LCD or an OLED, and displays a GUI for various operations. For example, the display unit 405 displays a candidate extraction notification sent by the intervention candidate extraction unit 322 of the information processing device 100 on the display unit 405 via, for example, a web browser or an email application.

[0037] The recording unit 406 includes a non-volatile memory such as a semiconductor memory, and stores set user information, programs executed by the control unit 402, and the like.

[0038] The control unit 402 includes a CPU 410 and a RAM 411, and controls the operation of each functional block in the control unit 402 and each unit in the communication device 110 by the CPU 410 executing a program recorded in the recording unit 406, for example.

[0039] (A series of operations for eGFR prediction processing in an information processing device) Next, a series of operations of eGFR prediction processing in the information processing device will be described with reference to Fig. 7. Each operation described in this series of operations is realized by the control unit 304 expanding a program stored in the storage unit 303 into the RAM 311 and executing it to cause each unit of the control unit 304 described above to function. This series of operations is started, for example, when a request for predicting eGFR for a specific person (subject) is received from the communication device 110.

[0040] In S701, the medical insurance information acquisition unit 320 acquires time-series medical insurance information of a specific person from the medical insurance information database 130. The medical insurance information acquisition unit 320 acquires only a predetermined number of years (for example, the most recent three consecutive years) of the medical insurance information of the specific person using information that identifies the specific person. In addition to the specific medical checkup information, the medical insurance information acquisition unit 320 may further acquire other information used by the machine learning model 202 from the storage unit 303 or an external device.

[0041] In S702, the trend prediction unit 321 acquires a time series of eGFR and other time series information from the time series medical insurance information acquired in S701. In S703, the trend prediction unit 321 predicts the trend of future changes in eGFR of a specific person using, for example, the time series of eGFR and the time series of measured values ​​in a health check other than eGFR, using a trained machine learning model as described above. The machine learning model outputs, for example, a time series predicted value of future eGFR as a prediction result.

[0042] The trend prediction unit 321 may refer to the most recent eGFR of a particular person, and if the value falls below eGFR 45 mL / min / 1.73 m2, exclude the person from the processing target. In other words, people whose most recent eGFR is equal to or greater than eGFR 45 mL / min / 1.73 m2 and less than 60 mL / min / 1.73 m2 may be subject to this processing. In this way, when this eGFR prediction processing is repeatedly applied to many people, the processing speed can be increased by excluding people whose condition is somewhat severe.

[0043] As described above, the measurement value in the medical checkup other than eGFR may include at least one of a renal function-related value and a risk factor. FIG. 6 shows an example of a correlation between a renal function-related value or a risk factor and a change in eGFR. As will be clear to those skilled in the art, the state of the correlation may vary depending on the data used. It should be noted that the example shown in FIG. 6 is used for the purpose of showing that there is a certain degree of correlation between a renal function-related value or a risk factor and a change in eGFR. 602 indicates that the more the HbA1c value increases, the greater the future eGFR will decrease. 604 indicates that the more the systolic blood pressure (sbp) value increases, the greater the future eGFR will decrease. 605 indicates that the more the HDL cholesterol value decreases, the greater the future eGFR will decrease. Also, 606 indicates that the greater the increase in BMI over the past two years, the greater the future eGFR will decrease. One or more of 602, 604, 605, and 606 shown in FIG. 6 can be used as risk factors in the measurement value in the medical checkup. That is, by using risk factors as inputs to the machine learning model, predictions can be made that take into account the impact of these factors on future changes in eGFR. Although the gender shown in 601 is not a measurement value from a medical checkup, it shows a correlation between gender and changes in eGFR, and indicates that being male can be used as a risk factor for prediction. That is, by inputting gender into the machine learning model as an input other than the measurement value from a medical checkup, predictions that take into account the risk factor become possible. The measurements shown in FIG. 6 are examples for explanation, and other measurements may be used as renal function-related values ​​or risk factors.

[0044] Also, 603 indicates that the lower the hematocrit (ht), the greater the drop in future eGFR. 603 in FIG. 6 indicates that it can be used for prediction as a renal function-related value in the measurement value in the health check. In other words, by using the renal function-related value as an input for the machine learning model, it is possible to make a prediction that takes into account the effect of the related value on future changes in eGFR.

[0045] In S704, the intervention candidate extraction unit 322 determines whether the trend of future changes in eGFR output by the trend prediction unit 321 satisfies a predetermined intervention condition. If the intervention candidate extraction unit 322 determines that the predetermined intervention condition is satisfied, the process proceeds to S705, and if not, the process ends. The intervention condition includes a condition that determines that eGFR is rapidly declining, such as a rate of decline in eGFR of 5 mL / min / 1.73 m2 per year or more, as described above. In order to identify the trend of future changes in eGFR, the intervention candidate extraction unit 322 may further apply statistical processing (e.g., regression analysis, etc.) to the time-series predicted value output by the machine learning model.

[0046] In S705, the intervention candidate extraction unit 322 can transmit the above-mentioned candidate extraction notification to the communication device 110.

[0047] In the above embodiment, the case where a person whose future change tendency of eGFR is rapidly decreased by eGFR prediction processing is described as an example. However, this embodiment is not limited to the case where the change of eGFR indicates a decrease. This embodiment may also be used to judge the treatment effect when a drug is used. That is, the measured value of eGFR of a patient who is continuously receiving treatment and other measured values ​​are obtained, and if the future change tendency of eGFR is more gradual than a predetermined standard by the above machine learning model, it may be judged that the treatment is effective. Alternatively, a similar judgment may be used as a KPI for verifying the effectiveness of a health project.

[0048] As described above, in this embodiment, time-series medical insurance information of a specific person is acquired, and a trained machine learning model is used to predict the future trend of eGFR change in the specific person using the time-series eGFR and a measurement value other than the time-series eGFR (for example, a measurement value at a medical checkup). In this way, it is possible to easily predict the future trend of eGFR change. In this embodiment, since the trend of eGFR change can be predicted with high accuracy using the measurement value at a medical checkup, the time cost and monetary cost for predicting eGFR can be significantly reduced. Prediction results can be easily obtained even for people with a relatively high current eGFR value, and more subjects who actually need intervention can have the opportunity to receive intervention.

[0049] The invention is not limited to the above-described embodiment, and various modifications and variations are possible within the scope of the gist of the invention. [Explanation of symbols]

[0050] 100: information processing device, 110: communication device, 320: medical insurance information acquisition unit, 321: trend prediction unit

Claims

1. A means for obtaining time series of health insurance information of a particular person; A prediction means for predicting a trend of change in the future estimated glomerular filtration rate (eGFR) of the specific person based on the time-series medical insurance information by a trained machine learning model, The information processing device, characterized in that the machine learning model predicts the future trend of eGFR changes for the specific person using a time series of eGFR and a time series of measurement values ​​other than eGFR contained in the time series of medical insurance information.

2. The information processing device according to claim 1 , wherein the future trend of eGFR change includes a time-series predicted value of the future eGFR.

3. The information processing device according to claim 1 , wherein the future trend of eGFR change includes a magnitude of change in eGFR over a predetermined period of time or an amount of change in eGFR on a yearly basis.

4. The information processing device according to claim 1 , wherein the time series of eGFR and the time series of measurement values ​​other than eGFR include measurement values ​​at each medical checkup in three consecutive years.

5. The information processing device according to claim 1 , wherein the time-series measurement values ​​other than eGFR include at least one of a time-series renal function-related value and a time-series risk factor.

6. The information processing device according to claim 1 , wherein the machine learning model further uses information other than the measurement values ​​of the time series of the specific person.

7. The information processing device according to claim 1 , wherein the machine learning model further uses a time-series medical interview result of the specific person, the medical interview result in response to a medical interview sheet in a health checkup.

8. The information processing device according to claim 1, characterized in that the machine learning model further uses at least one of the amount of change in eGFR on an annual basis obtained by statistical processing of the specific person's past eGFR, and fluctuations in the time series of health checkup values ​​other than eGFR over a predetermined period of time in the past.

9. The information processing device according to claim 1, characterized in that the prediction means predicts a tendency of future changes in the eGFR of the specific person whose eGFR is 45 mL / min / 1.73 m2 or more and less than 60 mL / min / 1.73 m2.

10. 6 . The information processing device according to claim 5 , wherein the time-series renal function-related values ​​include a time-series hematocrit value, and the time-series risk factors include a time-series HbA1c value.

11. The information processing device according to claim 1 , wherein the acquiring means acquires the time-series medical insurance information of the specific person from a database that stores personal medical insurance information collected and managed by a public system.

12. An information processing method executed by an information processing device, acquiring time-series health insurance information of a particular person; A prediction step of predicting a trend of change in future estimated glomerular filtration rate (eGFR) of the specific person based on the time-series medical insurance information by a trained machine learning model, The information processing method is characterized in that the machine learning model predicts the future trend of eGFR changes for the specific person using a time series of eGFR and a time series of measurement values ​​other than eGFR contained in the time series of medical insurance information.

13. A program for causing a computer to function as each of the means of the information processing device according to any one of claims 1 to 11.

Images