Estimation system, control method thereof, and control program

Abstract

The present invention addresses the problem of achieving an estimation system that can estimate warfarin dosage with higher accuracy. An estimation system (100) solves this problem by including an acquisition unit (12) that acquires subject information and target information, and an estimation unit (14) that estimates an estimated value using an estimation model (21) that uses at least any of subject information, target information, and subject comparison information as explanatory variables and an estimated value of warfarin dosage as a response variable.

Description

Estimation system, control method thereof, and control program

[0001] The present invention relates to an estimation system, a control method thereof, and a control program thereof.

[0002] Warfarin exerts anticoagulant and antithrombotic effects by antagonizing the action of vitamin K and inhibiting the biosynthesis in the liver of vitamin K-dependent blood coagulation factors, namely prothrombin, factor VII, factor IX, and factor X. Warfarin also exerts anticoagulant and antithrombotic effects by increasing the prothrombin precursor released into the blood, which is induced by vitamin K deficiency or vitamin K antagonists.

[0003] The dose of warfarin is determined based on a blood coagulation test. Here, the blood coagulation test measures blood clotting time. Examples of blood clotting time measurements include prothrombin time (PT), activated partial thromboplastin time, or diluted Russell's viper venom time. For example, the International Normalized Ratio (PT-INR) is commonly used to express PT measurements. The PT-INR is calculated by dividing the PT measurement (seconds) of the patient's plasma by the mean PT measurement (seconds) of normal plasma, and correcting the result by the International Sensitivity Index (ISI) attached to the reagent used in the blood coagulation test.

[0004] For example, when warfarin is used for the prevention and treatment of thromboembolism, blood coagulation management is necessary, and maintaining an appropriate therapeutic range is important. The appropriate therapeutic range of warfarin is determined for each patient using PT-INR. Sensitivity to warfarin differs from patient to patient and even varies within the same patient. Therefore, blood coagulation tests must be performed periodically for each patient, and the warfarin dosage must be adjusted as necessary.

[0005] In order to suppress thrombus formation in patients while reducing the risk of bleeding, it is important to determine a warfarin dose that maintains an appropriate therapeutic range. However, determining an appropriate warfarin dose places a considerable burden on both medical professionals such as physicians and patients. Conventionally, the warfarin dose is determined by, for example, hospitalizing the patient, orally administering a predetermined initial dose once daily, conducting blood coagulation tests over several days, and adjusting the dose to fall within the appropriate therapeutic range.

[0006] Therefore, a technique for easily determining an appropriate warfarin dosage has been proposed. Patent Document 1 describes a method for predicting a patient's PT-INR or a desired warfarin dosage using a logit function based on the patient's measured PT-INR. Non-Patent Documents 1 and 2 describe a technique for estimating a warfarin dosage to be administered to a patient using a model generated by machine learning using data on patients admitted to a Korean hospital. For example, the technique described in Non-Patent Document 1 uses the patient's gender, age, weight, height, and body surface area (BSA), the amount of warfarin actually administered over four days, and the PT-INR actually measured over four days to predict the patient's PT-INR on day five. Furthermore, Non-Patent Document 3 describes a technique for predicting a warfarin dosage using the absolute value of the difference in warfarin dosage between doses, the absolute value of the difference in INR measurements, etc.

[0007] Special Publication No. 2012-505910

[0008] Heemoon Lee et al., “Development of a system to support warfarin dose decisions using deep neural networks”, Scientific Reports 11:14745, 20, 2021.Heejung Choi et al., “Machine learning models to predict the warfarin discharge dosage using clinical information of inpatients from South Korea”, Scientific Reports 13:22461, 2023.Lindsay Dryden, et al Evaluation of Machine Learning Approaches for Predicting Warfarin Discharge Dose in Cardiac Surgery Patients. JMIR Cardio, 2023 Dec 6, e47262, doi:10.2196 / 47262

[0009] However, there is still room for improvement in the technology for determining the appropriate dose of warfarin.

[0010] An object of one aspect of the present invention is to realize an estimation system that can estimate a warfarin dosage with higher accuracy.

[0011] An estimation system according to one aspect of the present invention includes an acquisition unit that acquires, for a subject, (1) subject information including a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information including a target value for the blood clotting time on a target date, which is a first predetermined number of days after the reference date, set for the subject by a medical professional; and an estimation unit that outputs the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the subject information, the target information, and subject comparison information regarding the difference between the first measurement value and the target value, and has, as a dependent variable, an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time on the target date becomes the target value.

[0012] A control method for an estimation system executed by one or more computers according to one aspect of the present invention includes an acquisition step of acquiring, for a subject, (1) subject information including a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information including a target value of the blood clotting time on a target date, which is a day a first predetermined number of days after the reference date, set for the subject by a medical professional; and an estimation step of outputting the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has as a dependent variable an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time will be the target value on the target date.

[0013] According to one aspect of the present invention, an estimation system can be realized that can estimate a warfarin dosage simply and with higher accuracy.

[0014] FIG. 1 is a schematic diagram showing the configuration of an estimation system according to one embodiment of the present invention. FIG. 2 is a functional block diagram showing an example of an estimation device according to one embodiment of the present invention. FIG. 3 is a flowchart showing an example of a processing flow in which an estimation system according to one embodiment of the present invention estimates a dose of warfarin to be administered to a subject. FIG. 4 is a diagram showing an example of a display screen of a user terminal provided in the estimation system according to one embodiment of the present invention. FIG. 5 is a diagram showing an example of a display screen of a user terminal provided in the estimation system according to one embodiment of the present invention. FIG. 6 is a schematic diagram showing the configuration of an estimation system according to one embodiment of the present invention. FIG. 7 is a functional block diagram showing an example of an estimation device according to one embodiment of the present invention.

[0015] One embodiment of the present invention will be described below, but the present invention is not limited thereto. Furthermore, unless otherwise specified in this specification, the expression "A to B" representing a range of numerical values ​​means "A or more and B or less."

[0016] [Embodiment 1] An overview of an estimation system 100 according to one embodiment of the present invention will be described using Fig. 1. Fig. 1 is a schematic diagram showing an example of the configuration of the estimation system 100 according to one embodiment of the present invention. As shown in Fig. 1, the estimation system 100 may include a user terminal 50 and an estimation device 10 connected via a wide area communication network 40.

[0017] The estimation device 10 acquires subject information and target information. The subject information acquired by the estimation device 10 includes a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject. The target information acquired by the estimation device 10 also includes a target value for the blood clotting time on a target date, which is the day a first predetermined number of days has elapsed since the reference date. The estimation device 10 uses a trained estimation model 21 that includes, as explanatory variables, at least one of the first measurement value, the target value, and target comparison information regarding the difference between the first measurement value and the target value, and that uses, as a response variable, an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time on the target date becomes the target value. The estimation device 10 may output the target value to the user terminal 50 in addition to the estimated value.

[0018] As used herein, the term "subject" is not particularly limited as long as the subject requires administration of warfarin due to any disease and / or symptom. Preferably, the subject requires continuous administration of warfarin.

[0019] The user terminal 50 is a terminal equipped with a control unit 51 for a medical professional to input and view subject information and goal information. The subject information and goal information input to the user terminal 50 are acquired by the estimation device 10. The user terminal 50 receives the estimated value of the warfarin dosage transmitted (output) by the estimation device 10.

[0020] The user terminal 50 is not particularly limited as long as it is a terminal capable of communication. Examples of the user terminal 50 include a smartphone, a mobile phone, a tablet, a laptop computer, and a personal computer. In one embodiment, the user terminal 50 may be used by a subject other than a medical professional to check subject information, target information, and estimated values.

[0021] The user terminal 50 further includes a presented information output unit 52 that outputs presented information including the first measurement value and the target value. This configuration facilitates confirmation of information related to the subject. In one embodiment, the presented information output unit 52 may further output at least one of subject comparison information and the estimated value.

[0022] According to the above configuration, in the estimation system 100, the estimation device 10 can estimate the dose of warfarin to be administered to the subject on the administration date so that the subject's blood clotting time on the target date will be the target value, and output the estimated value, based on information transmitted from the user terminal 50. The estimated value is transmitted to the user terminal 50 via the wide area communication network 40, so that anyone who owns the user terminal 50 can easily check the estimated value.

[0023] Furthermore, warfarin is a high-risk drug that is subject to special drug management guidance premiums, etc. Therefore, estimating the appropriate dose of warfarin with high accuracy is useful for patients (subjects) who require warfarin administration and medical professionals.

[0024] In one embodiment, estimation system 100 may be used by a medical professional treating a subject to help determine the dosage of warfarin to be administered to the subject. In one embodiment, estimation system 100 may be used by a pharmacist prescribing a warfarin-containing medication to a subject to confirm that the prescribed dosage is appropriate. In one embodiment, estimation system 100 may be used by a subject to confirm their own condition.

[0025] The wide area communication network 40 is not particularly limited and may be the Internet, a telephone line network, a mobile communication network, a CATV communication network, a satellite communication network, etc. The wide area communication network 40 may further be connected to a cloud server that stores information transmitted by the user terminal 50 and estimated values ​​output by the estimation device 10. When the wide area communication network 40 is connected to a cloud server, the estimation device 10 may be realized as the cloud server. The user terminal 50 and the estimation device 10 may be connected via a local area network connection, LTE communication, etc., without going through a provider, etc.

[0026] 2 is a functional block diagram showing an example of the configuration of the estimation device 10. The estimation device 10 includes a control unit 1 that controls each unit of the estimation device 10, and a storage unit 2 that stores various data used by the estimation device 10, but is not limited to this configuration. For example, the storage unit 2 may be an external device attached to the estimation device 10. Furthermore, the estimation device 10 may be connected to the wide area communication network 40 as described above.

[0027] <Controller 1> First, we will explain the controller 1. The controller 1 includes an acquisition unit 12, a calculation unit 13, an estimation unit 14, and an output controller 15. Furthermore, some of the blocks included in the controller 1 may be omitted from the controller 1 by assigning their functions to another device that can communicate with the estimation device 10.

[0028] The acquisition unit 12 acquires the subject information and the target information transmitted from the user terminal 50. The acquisition unit 12 may store the acquired information in the storage unit 2.

[0029] The calculation unit 13 calculates subject comparison information regarding the difference between the first measurement value included in the subject information and the target value included in the target information. The calculation unit 13 may store the calculated subject comparison information in the storage unit 2.

[0030] The estimation unit 14 inputs at least one of the subject information, the target information, and the subject comparison information regarding the difference between the first measurement value and the target value into the estimation model 21 to estimate the warfarin dose of the subject. The information input by the estimation unit 14 to the estimation model 21 is preferably the same type of information as the information used as the explanatory variables during training of the estimation model 21. In one embodiment, the estimation unit 14 may further estimate the PT-INR value of the subject after warfarin administration. The estimation unit 14 may store the estimated value in the estimation result 25.

[0031] The output control unit 15 outputs the estimated value output by the estimation unit 14 to the user terminal 50 via the wide area communication network 40. The output control unit 15 may further output the subject information and target information acquired by the acquisition unit 12 to the user terminal 50. The output control unit 15 may transmit the estimated value directly to the user terminal 50, or may transmit the estimated value to a web page or the like on the wide area communication network 40 that is accessible by the user terminal 50. The output control unit 15 may also be another device that is capable of acquiring the estimated value from the estimation device 10 and has an output control function. In this case, the estimation device 10 may transmit the result estimated by the estimation unit 14 via the other device. In one embodiment, the output control unit 15 may further have a function as a communication unit.

[0032] <Storage Unit 2> Next, the storage unit 2 will be described. The storage unit 2 may store an estimation model 21. Furthermore, subject information 22, target information 23, subject comparison information 24, and estimation results 25 may also be stored as necessary.

[0033] The estimation model 21 includes as explanatory variables at least one of subject information, target information, and subject comparison information regarding the difference between the first measurement value and the target value, and uses as a dependent variable an estimated value of the dose of warfarin to be administered to the subject on the administration date in order to achieve the target blood clotting time on the target date.

[0034] Preferably, the estimation model 21 further includes, as an explanatory variable, a second measurement value indicating a blood clotting time measured a second predetermined number of days before the reference date, thereby improving the estimation accuracy of the estimation system 100.

[0035] In one embodiment, the estimation model 21 is trained by machine learning using training data. The training data includes the following explanatory variables and dependent variables. The explanatory variables include, for each of a plurality of sample subjects, at least one of a first sample measurement value indicating the measured blood clotting time, a second sample measurement value indicating the blood clotting time measured on a day a first predetermined number of days after the measurement date on which the first sample measurement value was measured, and sample comparison information regarding the difference between the first sample measurement value and the second sample measurement value. Note that, in this specification, "a plurality of sample subjects" may include patients who have previously been administered warfarin and patients who are currently administered warfarin. The plurality of sample subjects may also include the subject. The dependent variable includes the warfarin dose administered to each of the plurality of sample subjects on the day before or the day of the measurement date. Here, the sample comparison information is not particularly limited as long as it is a difference between the first sample measurement value and the second sample measurement value, and may be, for example, a difference or a ratio. In one embodiment, when the sample comparison information indicates the difference between the first sample measurement value and the second sample measurement value, it may be a signed difference value rather than an absolute value. With the above configuration, the estimation model 21 can be used to estimate an appropriate amount of warfarin to be administered to a subject.

[0036] In one embodiment, the dependent variable included in the training data and used to train the estimation model 21 may further include a measured value of the blood clotting time after administration of warfarin. In this case, the estimation model 21 can further estimate the blood clotting time (e.g., PT-INR).

[0037] The explanatory variables included in the training data and used to train the estimation model 21 may further include sample information. In one embodiment, the sample information is information related to the sample, of the type that may be included in the subject information described below. Specifically, examples include the sample's treatment information, physique information, second measurement value, age information, examination information, the total warfarin dose for the sample, the sample's international normalized ratio (INR) measured after warfarin administration, the sample's reactivity derived from (ΔINR / Δwarfarin dose), and other arbitrary information. When the sample information is expressed numerically, it is preferably a signed differential value. If the sample information is a signed differential value, it is possible to calculate the increase or decrease in the warfarin dose over time, thereby further estimating the fluctuation in the PT-INR value due to fluctuations corresponding to the increase or decrease in the warfarin dose.

[0038] The estimation model 21 may be generated using a known machine learning algorithm. Examples of machine learning algorithms that can be used to generate the estimation model 21 include a k-nearest neighbor method, logistic regression, a support vector machine (SVM), a gradient boosting method (e.g., XGBoost), an LSTM, and a neural network.

[0039] The subject information 22 is data related to a subject receiving warfarin, transmitted from the user terminal 50. The subject information 22 includes at least a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject. In one embodiment, the reference date is the day on which blood was drawn from the subject and the blood clotting time was measured.

[0040] The blood coagulation time is not particularly limited as long as it is a commonly used standard, and may be any of prothrombin time (PT), international normalized ratio of prothrombin time (PT-INR), activated partial thromboplastin time (APTT), and diluted Russell's viper venom time (dRVVT). Among these, PT-INR is preferred from the viewpoint of versatility.

[0041] From the viewpoint of improving the accuracy of the estimation system 100, the subject information 22 may further include information related to the subject. The subject information 22 preferably further includes a second measurement value indicating a blood clotting time measured a predetermined number of days (second predetermined number of days) before the reference date. The second predetermined number of days is preferably 1 to 3 days. When the subject information 22 includes the second measurement value, the estimation model 21 further includes, as an explanatory variable, second comparison information regarding the difference between the second measurement value and the target value.

[0042] From the viewpoint of improving the accuracy of the estimation system 100, it is preferable that the subject information 22 further includes treatment information including the date of the medical treatment if a medical treatment has been performed on the subject in the past. When the subject information 22 includes treatment information, the estimation model 21 further includes the treatment date or the number of days between the treatment date and the reference date as an explanatory variable.

[0043] In order to improve the accuracy of the estimation system 100, the subject information 22 preferably includes age information relating to the subject's age, physique information relating to the subject's physique, and test information including test results of blood collected from the subject. In one embodiment, the subject information 22 may also include information such as the subject's gender and race.

[0044] The physique information preferably includes height and weight from the viewpoint of improving the accuracy of the estimation system 100. In one embodiment, the physique information may include body-surface area (BSA), which can be calculated from height and weight.

[0045] From the viewpoint of improving the accuracy of the estimation system 100, the test information preferably includes at least one of serum protein concentration, eGFR, alkaline phosphatase (ALP), serum creatinine concentration, and serum bilirubin concentration. In one embodiment, the serum protein may be one or more selected from albumin, C-reactive protein (CRP), and fibrinogen.

[0046] In one embodiment, the subject information 22 may further include information on the blood concentration of warfarin and any information that affects the blood concentration of warfarin. Examples of such information include information on warfarin metabolic activity, vitamin K metabolic activity, concomitant medications, dietary intake, health foods, and supplements. The warfarin metabolic activity and vitamin K metabolic activity may be information based on genetic polymorphisms. The concomitant medications are not particularly limited as long as they can be used in combination with warfarin, and examples include drugs that inhibit or activate enzymes involved in warfarin metabolism and drugs that inhibit or activate enzymes involved in vitamin K metabolism. More specifically, the concomitant medication may be amiodarone. The diets are not particularly limited, and examples include foods that affect warfarin metabolism and foods that affect vitamin K metabolism.

[0047] The target information 23 includes a target value for the blood clotting time on a target date, which is set by a medical professional for the subject and is a day a predetermined number of days (first predetermined number of days) has passed since the reference date. In one embodiment, the first predetermined number of days is 1 to 4 days. The blood clotting time in the target information 23 is preferably the same type as the blood clotting time in the subject information 22.

[0048] In one embodiment, information for linking the subject information and / or the target information with the estimated value may be transmitted from the user terminal 50. Preferably, the subject information and the target information may be associated with an ID and a password for identifying the subject.

[0049] The target comparison information is information regarding the difference between the first measurement value and the target value. The target comparison information may be the value obtained by subtracting the target value from the first measurement value (or the value obtained by subtracting the first measurement value from the target value), or the value obtained by dividing the first measurement value by the target value (or the value obtained by dividing the first measurement value from the target value). The target comparison information is preferably the value obtained by dividing the first measurement value by the target value (or the value obtained by dividing the first measurement value from the target value). In one embodiment, the target comparison information may be a signed difference value rather than an absolute value.

[0050] The estimation result 25 stores an estimated value of the warfarin dose estimated by the estimation unit 14. In one embodiment, the estimation result 25 may include an estimated value of the blood clotting time according to the warfarin dose.

[0051] [Processing Performed by Estimation System] An outline of a control method for the estimation system 100 according to one embodiment of the present invention will be described with reference to Fig. 3. Fig. 3 is a flowchart showing an outline of the control method for the estimation system 100.

[0052] In one embodiment, a screen shown in Fig. 4 may be displayed to identify the subject before step S1. Fig. 4 is an example of a screen that may be displayed on the user terminal 50 when the estimation system 100 is used as an application. By entering the subject's ID and password on the screen shown in Fig. 4, the estimated value can be easily confirmed without having to enter the subject information and target information that have already been entered multiple times.

[0053] In step S1, the acquisition unit 12 acquires (1) subject information for the subject, including a first measurement value indicating the blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information, including a target value for the blood clotting time on a target date, which is the day a first predetermined number of days has elapsed since the reference date, set by a medical professional for the subject (acquisition step).

[0054] Fig. 5 shows an example of a display screen of the user terminal 50 when the estimation system 100 is used as an application. Fig. 5 shows an input screen for subject information and goal information. The subject information and goal information input by the user of the estimation system 100 are acquired by the estimation device 10 in step S1. Note that in Fig. 5, PT-INR is used as the blood clotting time. In one embodiment, the presence or absence of concomitant medication and the presence or absence of intake of a restricted diet may also be input as subject information.

[0055] In step S2, the calculation unit 13 calculates the target comparison information based on the difference between the first measurement value and the target value (calculation step).

[0056] In step S3, the estimation unit 14 outputs an estimated value using a trained estimation model 21 that includes as explanatory variables at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has as the objective variable an estimated value of the dose of warfarin to be administered to the subject on the administration date in order to bring the blood coagulation time to the target value on the target date (estimation step).

[0057] In step S4, the output control unit 15 transmits (outputs) the estimated value output by the estimation unit 14 to the user terminal 50 via the wide area communication network 40 based on the subject information. The estimated value transmitted to the user terminal 50 may be output by the presentation information output unit.

[0058] 6 shows an example of a display screen of the user terminal 50 in step S4 when the estimation system 100 is used as an application. In step S4, not only the estimated value but also subject information and target information may be output to the user terminal 50 at the same time, as shown in FIG. 6. In addition to the target information, an estimated value of the blood clotting time in the case where warfarin is not administered appropriately may also be output to the user terminal 50. Note that PT-INR is used as the blood clotting time in FIG. 6.

[0059] Sensitivity to warfarin varies from patient to patient and even varies within the same patient. Therefore, it is necessary to periodically perform blood coagulation tests for each patient and adjust the warfarin dose as necessary. By using the estimation system 100, estimation is performed using the estimation model 21 trained based on sample information, target information, and the warfarin dose, so that the warfarin dose can be estimated with high accuracy.

[0060] [Embodiment 2] Hereinafter, an overview of an estimation system 100a according to another embodiment of the present invention will be described with reference to Fig. 7. Fig. 7 is a schematic diagram showing an example of the configuration of the estimation system 100a. Note that descriptions of matters that have already been described will be omitted.

[0061] 7 , in the estimation system 100a, the user terminal 50 and the estimation device 10a directly transmit and receive information to and from each other. The estimation device 10a outputs an estimated value of the warfarin dose based on the acquired subject information and target information. That is, in the estimation system 100a, the estimation device 10a may be integrated with the user terminal 50. In addition, in the estimation system 100a, the estimation device 10a and the user terminal 50 may be connected via a local area network connection, LTE communication, or the like, without going through a provider or the like.

[0062] 8 is a functional block diagram showing an example of the configuration of an estimation system 100a different from that of embodiment 1. The estimation system 100a includes an estimation device 10a and a user terminal 50.

[0063] The estimation device 10 a includes a control unit 1 that controls all the components of the estimation device 10 a , a storage unit 2 that stores various data used by the estimation device 10 a , and an input unit 11 .

[0064] The input unit 11 is for accepting various input operations from a user and may be, for example, a keyboard, a mouse, a touch panel, etc. The input unit 11 may be used to input information including, for example, subject information and target information.

[0065] With the above configuration, the estimation device 10a is directly connected to the user terminal 50 without going through a network, so the time lag until an estimated value is obtained is reduced.

[0066] [Example of implementation by software] The functions of the estimation device 10, 10a (hereinafter referred to as "device") can be realized by a control program that causes a computer to function as the device, and a control program that causes a computer to function as each control block of the device (particularly each part included in the control unit 1).

[0067] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The functions described in each of the above embodiments are realized by executing the program using the control device and storage device.

[0068] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0069] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0070] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0071] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0072] [Summary] The estimation system according to aspect 1 of the present invention comprises an acquisition unit that acquires, for a subject, (1) subject information including a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information set for the subject by a medical professional, including a target value for the blood clotting time on a target date, which is the day on which a first predetermined number of days have elapsed since the reference date; and an estimation unit that outputs the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has, as a dependent variable, an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time on the target date becomes the target value.

[0073] In the estimation system according to aspect 2 of the present invention, in aspect 1, the estimation model may include, as explanatory variables, at least one of a first sample measurement value indicating the measured blood clotting time for each of a plurality of sample subjects, a second sample measurement value indicating the blood clotting time measured on the day the first predetermined number of days has elapsed since the measurement date on which the first sample measurement value was measured, and sample comparison information regarding the difference between the first sample measurement value and the second sample measurement value, and may be generated by machine learning using training data in which the objective variable is the dose of warfarin administered to each of the plurality of sample subjects on the day before or the day of the measurement date.

[0074] In the estimation system of aspect 3 of the present invention, in aspect 1 or 2, the estimation model may further include as an explanatory variable a second measurement value indicating a blood coagulation time measured a second predetermined number of days before the reference date.

[0075] In the estimation system of aspect 4 of the present invention, in any of aspects 1 to 3, the subject information may further include a second measurement value indicating a blood coagulation time measured a second predetermined number of days before the reference date, and the estimation model may further include second comparison information regarding the difference between the second measurement value and the target value as an explanatory variable.

[0076] In the estimation system of aspect 5 of the present invention, in any of aspects 1 to 4, the subject information may further include treatment information including the date on which the medical treatment was performed if a medical treatment was performed on the subject in the past, and the estimation model may further include the date of the treatment or the number of days between the treatment date and the reference date as an explanatory variable.

[0077] In the estimation system of aspect 6 of the present invention, in any of aspects 1 to 5, the subject information may include age information regarding the subject's age, physique information regarding the subject's physique, and test information including test results of blood taken from the subject.

[0078] In the estimation system according to aspect 7 of the present invention, in any one of aspects 1 to 6, the physical information may include height and weight.

[0079] In the estimation system according to aspect 8 of the present invention, in any one of aspects 1 to 7, the test information may include at least one of serum protein concentration, alkaline phosphatase (ALP), eGFR, serum creatinine concentration, and serum bilirubin concentration.

[0080] In the estimation system according to aspect 9 of the present invention, in any one of aspects 1 to 8, the blood coagulation time may be any one of prothrombin time (PT), international normalized ratio of prothrombin time (PT-INR), activated partial thromboplastin time (APTT), and diluted Russell's viper venom time (dRVVT).

[0081] The estimation system according to Aspect 10 of the present invention is in any one of Aspects 1 to 9, and may further include a presentation information output unit that outputs presentation information including the first measurement value and the target value.

[0082] A control method for an estimation system according to aspect 11 of the present invention is a control method for an estimation system executed by one or more computers, and includes an acquisition step of acquiring, for a subject, (1) subject information including a first measurement value indicating a blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information including a target value of the blood clotting time on a target date, which is a day on which a first predetermined number of days have elapsed since the reference date, set for the subject by a medical professional; and an estimation step of outputting the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has as a dependent variable an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time will be the target value on the target date.

[0083] A control program according to a twelfth aspect of the present invention is a control program for causing a computer to function as the estimation system of any one of the first to tenth aspects, and causes the computer to function as the acquisition unit and the estimation unit.

[0084] A study was conducted at Osaka University Hospital on 11,949 patients with cardiovascular disease who provided informed consent. 2,838 of the patients were randomly selected to serve as the training data group. 1,216 of the patients were used as the validation data group. For the patients included in the training data group, machine learning using XGBoost was performed with multiple parameters as explanatory variables and the patient's warfarin dose as the objective variable to obtain an estimation model. The explanatory variables used were as follows:

[0085] First measurement value: PT-INR value on the predicted date, PT-INR value one day before the predicted date Target value: target INR value Subject comparison information: difference between the average target INR and the INR value on the predicted date, difference between the average target INR and the INR value one day before the predicted date Using the obtained estimation model, warfarin dosage was estimated for 1,216 patients in the validation data group. The obtained results were compared with the actual results to calculate the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). Note that the smaller the MSE and RMSE values, the smaller the error in the estimated results. Furthermore, the closer the R2 value is to 1, the more closely the estimated results fit the actual data.

[0086] The results were MAE = 0.334, RMSE = 0.791, and R = 0.919. Therefore, the estimation system according to one embodiment of the present invention makes it possible to estimate the dose of warfarin to be administered to a patient with higher accuracy.

[0087] 10, 10a Estimation device 12 Acquisition unit 14 Estimation unit 21 Estimation model 52 Presentation information output unit 100 Estimation system

Claims

1. An estimation system comprising: an acquisition unit that acquires, for a subject, (1) subject information including a first measurement value indicating the blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information set for the subject by a medical professional, including a target value for the blood clotting time on a target date, which is the day a first predetermined number of days has elapsed since the reference date; and an estimation unit that outputs the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has, as a dependent variable, an estimated value of the dose of warfarin to be administered to the subject on the administration date so that the blood clotting time on the target date becomes the target value.

2. The estimation system of claim 1, wherein the estimation model includes, for each of a plurality of sample subjects, at least one of a first sample measurement value indicating the measured blood clotting time, a second sample measurement value indicating the blood clotting time measured on the day the first specified number of days has elapsed since the measurement date on which the first sample measurement value was measured, and sample comparison information regarding the difference between the first sample measurement value and the second sample measurement value as explanatory variables, and is generated by machine learning using training data in which the dose of warfarin administered to each of the plurality of sample subjects on the day before or the day of the measurement date is used as the objective variable.

3. The estimation system according to claim 1, wherein the estimation model further includes, as an explanatory variable, a second measurement value indicating a blood clotting time measured a second predetermined number of days before the reference date.

4. The estimation system described in claim 1, wherein the subject information further includes a second measurement value indicating a blood clotting time measured a second predetermined number of days before the reference date, and the estimation model further includes second comparison information regarding the difference between the second measurement value and the target value as an explanatory variable.

5. The estimation system of claim 1, wherein the subject information further includes treatment information including the date on which the medical treatment was performed if the subject has previously undergone medical treatment, and the estimation model further includes the date of the treatment or the number of days between the date of the treatment and the reference date as an explanatory variable.

6. The estimation system of claim 1, wherein the subject information includes age information regarding the subject's age, physique information regarding the subject's physique, and test information including test results of blood taken from the subject.

7. The estimation system according to claim 6, wherein the physical information includes height and weight.

8. The estimation system according to claim 6, wherein the test information includes at least one of serum protein concentration, alkaline phosphatase (ALP), eGFR, serum creatinine concentration, and serum bilirubin concentration.

9. The estimation system according to claim 1, wherein the blood coagulation time is any one of prothrombin time (PT), international normalized ratio of prothrombin time (PT-INR), activated partial thromboplastin time (APTT), and diluted Russell's viper venom time (dRVVT).

10. The estimation system according to claim 1, further comprising a presentation information output unit that outputs presentation information including the first measurement value and the target value.

11. A control method for an estimation system executed by one or more computers, comprising: an acquisition step of acquiring, for a subject, (1) subject information including a first measurement value indicating the blood clotting time measured on a reference date, which is the day before or the day on which warfarin is administered to the subject, and (2) target information set for the subject by a medical professional, including a target value for the blood clotting time on a target date, which is the day on which a first predetermined number of days have elapsed since the reference date; and an estimation step of outputting the estimated value using a trained estimation model that includes, as explanatory variables, at least one of the first measurement value, the target value, and subject comparison information regarding the difference between the first measurement value and the target value, and that has as a response variable an estimated value of the dose of warfarin to be administered to the subject on the administration date in order to bring the blood clotting time to the target value on the target date.

12. A control program for causing a computer to function as the estimation system according to claim 1, the control program causing a computer to function as the acquisition unit and the estimation unit.

Images