Computer program, drug therapy support device, drug therapy support method, and learning model generation method

Abstract

The present invention provides a computer program, a drug therapy support device, a drug therapy support method, and a learning model generation method capable of estimating the condition of a patient after the start of drug therapy.　The computer program causes a computer to execute processing for: acquiring first administration information for a first drug used in drug therapy for a patient; acquiring patient information of a patient and test information of a test performed on the patient; acquiring vital signs of the patient; inputting the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimation information of the patient condition after the start of the drug therapy when the first administration information of the first drug used in the drug therapy, the patient information, the test information, and the vital signs have been input; and estimating the patient condition related to side effects associated with administration of the first drug after the start of the drug therapy for the patient.

Description

Computer program, drug therapy support device, drug therapy support method, and learning model generation method

[0001] The present disclosure relates to a computer program, a pharmacotherapy support device, a pharmacotherapy support method, and a learning model generation method.

[0002] Anticancer drug treatment (drug therapy) uses drugs that directly attack cancer cells. Anticancer drugs inhibit the growth of cancer cells and promote their death, but because they act not only on cancer cells but also on normal cells, they can cause harmful reactions (side effects) in normal cells. Endocrine therapy (hormone therapy) is also used for prostate cancer and breast cancer, suppressing the secretion and function of male and female hormones, which are involved in the growth of cancer cells. The antihormonal drugs used in this endocrine therapy disrupt the hormonal balance in the body, which can cause harmful reactions (side effects) such as hot flashes (hot flashes, flushing, sweating), osteoporosis, and liver dysfunction.

[0003] Patent Document 1 discloses a side effect reducing agent that reduces various side effects during drug therapy and improves the rate at which drug therapy is completed.

[0004] On the other hand, cancer patients in outpatient drug therapy rooms at medical institutions may experience highly urgent adverse reactions such as hypersensitivity reactions (HSR) and infusion reactions (IR) during drug therapy. HSR is characterized by anaphylaxis as the main symptom and is accompanied by fever, while IR is characterized by fever and headache as the main symptoms and may be accompanied by anaphylaxis.

[0005] Japanese Patent Application Laid-Open No. 2019-203025

[0006] For patients at high risk of side effects, we carefully monitor their condition by observing their physical condition and regularly measuring vital signs such as blood pressure, body temperature, and oxygen saturation during treatment in the drug therapy room, but we cannot predict the occurrence of side effects in advance.

[0007] The present disclosure has been made in consideration of the above circumstances, and aims to provide a computer program, a drug therapy support device, a drug therapy support method, and a learning model generation method that can estimate a patient's condition after starting drug therapy.

[0008] (1) A computer program according to the present disclosure causes a computer to execute a process of acquiring first administration information of a first drug used in a drug therapy for a patient, acquiring patient information and test information of tests on the patient, acquiring vital signs of the patient, and inputting the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimated information of the patient's condition after initiation of the drug therapy when the first administration information of the first drug used in the drug therapy, the patient information, test information, and vital signs are input, thereby estimating the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy for the patient. Here, an embodiment of the present disclosure is as follows: (2) In the computer program of (1) above, the vital signs are measured from time to time during the drug therapy. (3) In the computer program of (1) or (2) above, the learning model outputs estimated information from a required time from the current time during the drug therapy, thereby estimating the patient's condition from a required time from the current time during initiation of the drug therapy for the patient. (4) Any one of the computer programs described in (1) to (3) above causes a computer to execute a process of further acquiring second administration information of a second drug used in supportive therapy for the patient, and further inputting the acquired second administration information into the learning model, which outputs estimated information of the patient's condition after the start of the drug therapy when the second administration information of the second drug used in supportive therapy is further input, thereby estimating the patient's condition after the start of the drug therapy for the patient. (5) Any one of the computer programs described in (1) to (4) above causes a computer to execute a process of classifying the estimated information output by the learning model into one class from multiple classes indicating the risk of the patient's condition. (6) Any one of the computer programs described in (1) to (5) above causes a computer to execute a process of outputting an instruction to change first administration information of a first drug used in the drug therapy according to the estimated patient condition. (7) Any one of the computer programs described in (1) to (6) above causes a computer to execute a process of accepting a setting of the required time.(8) In any one of the computer programs (1) to (7) above, the first administration information includes at least one of a drug name, a dosage, an administration rate, an administration cycle, and an administration route. (9) In any one of the computer programs (1) to (8) above, the vital signs include a pulse rate, a respiratory rate, a body temperature, a blood pressure, and an oxygen saturation level. (10) In any one of the computer programs (1) to (9) above, the estimated information includes a blood pressure or a respiratory rate. (11) In any one of the computer programs (1) to (10) above, the patient information includes age, sex, weight, and cancer type. (12) In any one of the computer programs (1) to (11) above, the test information includes a blood test result before the start of drug therapy. (13) In any one of the computer programs described in (1) to (12), the estimated information output by the learning model is an estimated value obtained by estimating a numerical value from the current time point during the drug therapy for at least one of the vital signs input to the learning model, the estimated value output by the learning model is classified into one class from among a plurality of classes that classify the risk of the patient's condition into multiple stages based on a comparison result with a predetermined value, and the current numerical value, the output estimated value, and the classified class are displayed on a display in association with each other for the at least one of the vital signs input to the learning model. (14) In the computer program described in (13), the computer is caused to execute a process of generating alert information when the estimated value output by the learning model is classified into the highest-risk class of the plurality of classes, and displaying the generated alert information on a display in association with the current numerical value, the output estimated value, and the classified class.(15) The computer program of (1) above causes a computer to execute the following processes: the learning model outputs estimated information about the patient's condition from the current time point to a required time point during the drug therapy; based on the estimated information output by the learning model, classify the patient's condition into one class from multiple classes that classify the risk of the patient's condition into multiple stages; input the vital signs acquired at a first interval during the drug therapy into the learning model to acquire estimated information about the patient's condition from the current time point to a required time point during the drug therapy at the first interval; classify the risk of the patient's condition at the first interval based on the estimated information acquired at the first interval; if the patient is classified into a high-risk class among the multiple classes, acquire the vital signs at a second interval that is shorter than the first interval; input the vital signs acquired at the second interval into the learning model to estimate estimated information about the patient's condition from the current time point to a required time point during the drug therapy at the second interval. (16) The computer program of (15) causes a computer to execute a process of transmitting, to a measuring device for measuring the vital signs, an instruction to measure the vital signs at the second interval when the subject is classified into a high-risk class among the plurality of classes. (17) The computer program of (15) causes a computer to execute a process of displaying, on a display, recommendation information recommending that the vital signs be measured at the second interval when the subject is classified into a high-risk class among the plurality of classes. (18) A drug therapy support device according to the present disclosure includes a control unit, which acquires first administration information of a first drug used in drug therapy for a patient, acquires patient information of the patient and test information of tests on the patient, acquires vital signs of the patient, and inputs the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimated information of the patient's condition after initiation of drug therapy when the first administration information of a first drug used in drug therapy, patient information, test information, and vital signs are input, thereby estimating the patient's condition regarding side effects associated with the administration of the first drug after initiation of the drug therapy for the patient.(19) A drug therapy support method according to the present disclosure acquires first administration information of a first drug used in drug therapy for a patient, acquires patient information and test information of tests on the patient, acquires vital signs of the patient, and inputs the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimated information of the patient's condition after initiation of drug therapy when the first administration information of a first drug used in drug therapy, patient information, test information, and vital signs are input, thereby estimating the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy for the patient. (20) A learning model generation method according to the present disclosure acquires training data including the first administration information of a first drug used in drug therapy, patient information, test information, vital signs, and estimated information of the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy, and generates a learning model based on the acquired training data so as to output estimated information of the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy when the first administration information of a first drug used in drug therapy, patient information, test information, and vital signs are input.

[0009] According to the present disclosure, it is possible to estimate the condition of a patient after starting drug therapy.

[0010] FIG. 1 is a diagram showing a first example of the configuration of a pharmacotherapy support system of this embodiment. FIG. 2 is a diagram showing an example of patient information. FIG. 3 is a diagram showing an example of blood test information. FIG. 4 is a diagram showing an example of administration information. FIG. 5 is a diagram showing an example of vital signs. FIG. 6 is a diagram showing an example of a method for estimating a patient's condition using a learning model. FIG. 7 is a diagram showing a first example of a patient's condition estimation result screen. FIG. 8 is a diagram showing a second example of a patient's condition estimation result screen. FIG. 9 is a diagram showing a third example of a patient's condition estimation result screen. FIG. 10 is a diagram showing a first example of a patient's condition estimation processing procedure by a pharmacotherapy support device. FIG. 11 is a diagram showing a second example of the configuration of a pharmacotherapy support system of this embodiment. FIG. 12 is a diagram showing an example of a learning model generation processing procedure by a pharmacotherapy support device. FIG. 13 is a diagram showing a second example of a patient's condition estimation processing procedure by a pharmacotherapy support device.

[0011] Embodiments of the present disclosure will be described below. Fig. 1 is a diagram showing a first example of the configuration of a pharmacotherapy support system according to this embodiment. The pharmacotherapy support system according to this embodiment includes a pharmacotherapy support device 50. The pharmacotherapy support system may include a medication device 10, a measuring device 20, a terminal device 30, and a display device 40. The medication device 10, the measuring device 20, the terminal device 30, the display device 40, and a data server 100 are connected to the pharmacotherapy support device 50 via a communication network 1.

[0012] The medication device 10 is a device for administering drugs such as anticancer drugs to patients in drug therapy. Examples of such devices include medical pumps, such as syringe pumps for injecting relatively small amounts of drugs and infusion pumps for injecting relatively large amounts of drugs. The drug therapy support device 50 acquires drug administration information for drugs used in drug therapy from the medication device 10. Examples of such drugs include anticancer drugs such as platinum preparations and taxane preparations, molecular targeted drugs that target molecules involved in cancer cell proliferation, metastasis, and invasion, immune checkpoint inhibitors, and antihormonal drugs used in endocrine therapy. Details of the administration information will be described later. A drug used in drug therapy is also referred to as a first drug, and administration information for the first drug is also referred to as first administration information. A drug used in supportive care is also referred to as a second drug, and administration information for the second drug is also referred to as second administration information. In this embodiment, details will be described using anticancer drug therapy, one type of drug therapy, as an example.

[0013] The measuring device 20 is one or more devices that measure the vital signs of a patient undergoing drug therapy, and measures the vital signs every moment during drug therapy. The drug therapy support device 50 acquires the vital signs from the measuring device 20. Details of the vital signs will be described later.

[0014] The terminal device 30 is a management terminal for managing the medical therapy support device 50, and is used by an operator or manager who operates the medical therapy support system.

[0015] The display device 40 displays the estimated results of the patient's condition after the start of drug therapy by the drug therapy support device 50. The display device 40 is installed in a treatment room where drug therapy is being performed, etc., so that medical personnel can constantly monitor it.

[0016] The data server 100 includes a patient information DB 101 and a test information DB 102. The drug therapy support apparatus 50 acquires the patient information and test information from the data server 100. The patient information and test information will be described in detail later.

[0017] The medical therapy support device 50 includes a control unit 51 that controls the entire device, a communication unit 52, a memory 53, a storage unit 54, and a recording medium reading unit 57.

[0018] The control unit 51 may be configured by incorporating a required number of central processing units (CPUs), micro-processing units (MPUs), graphics processing units (GPUs), etc. The control unit 51 may also be configured by combining digital signal processors (DSPs), field-programmable gate arrays (FPGAs), etc.

[0019] The communication unit 52 includes a communication module and has the function of communicating with the medication device 10 , the measuring device 20 , the terminal device 30 , the display device 40 , and the data server 100 via the communication network 1 .

[0020] The storage unit 54 can be configured with a semiconductor memory, a hard disk, or the like, and stores a computer program 55 (program product), a learning model 56, and required information. The learning model 56 will be described in detail later.

[0021] The computer program 55 can be stored in the storage unit 54 by reading the computer program 55 recorded on a storage medium (for example, an optically readable disk storage medium such as a CD-ROM) M using the storage medium reading unit 57. The computer program 55 may also be downloaded from an external device via the communication unit 52 and stored in the storage unit 54.

[0022] The memory 53 can be configured with semiconductor memory such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, etc. A computer program 55 can be loaded into the memory 53, and the control unit 51 can execute the computer program 55. The control unit 51 can execute processing defined by the computer program 55. In other words, processing by the control unit 51 is also processing by the computer program 55.

[0023] The drug therapy support device 50 may be configured as a single device. Alternatively, the drug therapy support device 50 may be configured as multiple devices with distributed functions. The computer program 55 may be deployed to run on a single computer, or on multiple computers located at one site, or distributed across multiple sites and interconnected by a communication network.

[0024] FIG. 2 is a diagram showing an example of patient information. Patient information is stored in the patient information DB 101 for each patient. The drug therapy support device 50 can acquire the patient information of a required patient from the data server 100. The patient information includes a patient ID, the patient's name, date of birth, age, sex, blood type, weight, medical history (including type of cancer), treatment history, etc. The treatment history is historical information indicating what treatments and medications were administered for the patient's illness, when, and for how long. The patient information is information collected in advance before drug therapy is administered to the patient.

[0025] FIG. 3 is a diagram showing an example of blood test information. Blood test information is stored for each patient in the test information DB 102. In addition to blood test information, the test information includes various tests (e.g., CT scans, MRI scans, ultrasound scans, etc.) depending on the type of cancer the patient has. However, before administering an anticancer drug, a doctor performs a blood test on the patient to check the test values ​​before deciding to administer the anticancer drug, and then performs a blood test each time anticancer drug treatment (pharmacotherapy) is administered to check the test values ​​and confirm whether the treatment is proceeding smoothly. Therefore, in this embodiment, blood test information is used as the test information.

[0026] The drug therapy support device 50 can acquire required blood test information of a patient from the data server 100. The blood test information includes, for example, test results for various test items related to general blood, liver, gallbladder, pancreas, renal function, electrolytes, etc. General blood information includes, for example, WBC (white blood cell count), SEG (neutrophils among white blood cells), RBC (red blood cell count), HGB (a pigment formed by the combination of iron and protein in red blood cells), PLT (which acts to clot blood and stop bleeding), etc.

[0027] Hepatobiliary and pancreatic functions include, for example, total bilirubin, AST (distributed in the liver, heart, kidneys, and muscles), ALT (an enzyme primarily found in the liver), LDH (elevated with hepatocellular damage), γ-GTP (elevated with excessive intake of alcohol, fat, etc.), ALP (elevated with obstruction or stenosis of the biliary system), and AMY (elevated with pancreatic inflammation). Renal function includes, for example, urea nitrogen and creatinine. Electrolytes include sodium, potassium, chloride, magnesium, etc. Blood test information also includes total protein and CRP (elevated with inflammation and tissue destruction). Blood test information (test information) is information collected in advance of drug therapy for the patient. It is preferable that the blood test information include test values ​​associated with risk factors for hypersensitivity and infusion reactions. Specifically, for the antibody drug rituximab, which is the first drug used in drug therapy, a high number of tumor cells in the blood is considered a risk factor. For this reason, it is preferable that blood test information includes test values ​​for WBC (white blood cell count), which increases due to the amount of tumor in the blood.

[0028] FIG. 4 shows an example of administration information. The administration information is registered (set) in a drug library or the like included in the administration device 10. The drug therapy support device 50 can acquire the administration information for a patient from the administration device 10 each time drug therapy is administered to the patient. The administration information includes, for example, the anticancer drug (drug name), dosage, administration rate (base flow rate), administration cycle (number of times), drug holiday period, administration route (peripheral vein, central vein), drugs used in combination with or alone than the anticancer drug (e.g., drugs used in supportive care), and the name of the regimen. A typical anticancer drug treatment cycle lasts 3 to 4 weeks, and this cycle is repeated several times followed by a drug holiday period. A regimen is a chronological treatment plan combining anticancer drugs, infusions, supportive care drugs, etc., and includes the dosage, administration schedule, and treatment period of the anticancer drug.

[0029] 5 is a diagram showing an example of vital signs. The vital signs are measured by the measuring device 20 from time to time while the patient is receiving drug therapy (at least from the start to the end). The drug therapy support device 50 can acquire the patient's vital signs from the measuring device 20 from time to time while the patient is receiving drug therapy. The vital signs are measured as time-series data continuously or at a required sampling interval while the drug therapy is being performed.

[0030] Vital signs include pulse rate, respiratory rate, body temperature, blood pressure (systolic, diastolic, mean), oxygen saturation, etc. Systolic blood pressure is the highest blood pressure, and diastolic blood pressure is the lowest blood pressure, and mean blood pressure can be calculated using the formula: (systolic blood pressure - diastolic blood pressure) ÷ 3 + diastolic blood pressure. A normal pulse rate is about 60 to 100 beats per minute. A normal respiratory rate for adults is about 12 to 20 beats per minute.

[0031] FIG. 6 illustrates an example of a method for estimating a patient's condition using a learning model 56. The learning model 56 is generated (trained) to output estimated information about the patient's condition after the start of drug therapy when it receives input information about the administration of drugs used in drug therapy, patient information, test information (blood test information), and vital signs during drug therapy. The estimated information about the patient's condition after the start of drug therapy is, specifically, estimated information for a required time from the current time while drug therapy is being performed. That is, the learning model 56 always predicts estimated information for a required time (e.g., 5 minutes, 10 minutes, 20 minutes, 30 minutes, etc.) from the current time from the start to the end of drug therapy (e.g., 2 to 3 hours). The estimated information includes, for example, blood pressure, respiratory rate, etc. The estimated information may also include pulse rate, body temperature, and oxygen saturation. This estimated information is patient information related to side effects such as hypersensitivity and infusion reactions. The estimated information may also include the type of side effect, such as hypersensitivity or infusion reaction, and the probability of its occurrence.

[0032] The learning model 56 receives input of patient information, test information (blood test information), administration information, and vital signs during drug therapy for a patient undergoing drug therapy. The learning model 56 also receives input of the required time. Based on the input required time, the learning model 56 outputs estimated information for the required time from the current time into the future. By changing the input required time, the prediction range of the estimated information can be changed.

[0033] As described above, the control unit 51 acquires administration information (first administration information) of a drug (first drug) used in a drug therapy for a patient, acquires patient information and test information (blood test information) of tests performed on the patient, and acquires vital signs of the patient. The control unit 51 can estimate the patient's condition after the start of drug therapy by inputting the acquired administration information, patient information, test information, and vital signs into the learning model 56, which outputs estimated information of the patient's condition after the start of drug therapy when the administration information of a drug used in drug therapy, patient information, test information, and vital signs are input.

[0034] The above-described configuration allows for estimation of the patient's condition after the start of drug therapy, which makes it possible to predict the occurrence of highly urgent adverse reactions (adverse reactions) such as hypersensitivity reactions (HSR) and infusion reactions (IR), allowing for prompt treatment of the patient during drug therapy before the occurrence of side effects, thereby preventing the occurrence of side effects.

[0035] As described above, the learning model 56 outputs estimated information for a required time from the current time during drug therapy, and the control unit 51 can estimate the patient's condition for a required time from the current time during the start of drug therapy. This allows for prediction of estimated information for a required time from the current time, making it possible to predict in advance whether or not a highly urgent side effect (adverse reaction) such as hypersensitivity reaction (HSR) or infusion reaction (IR) will occur in the future. This allows for prompt treatment of the patient during drug therapy before the side effect occurs, thereby preventing the occurrence of side effects in advance.

[0036] The control unit 51 can also receive a setting for the required time and input the received required time into the learning model 56. This allows the time range for predicting estimated information to be set according to the type of drug therapy, the patient's risk, etc. For example, for high-risk patients, the prediction time range can be shortened to detect sudden changes in the estimated information.

[0037] 6 , the vital signs during drug therapy input into the learning model 56 can be current values ​​(latest values), but are not limited to current values. For example, time-series data of vital signs from a past time to the current time (e.g., the past 5 minutes, the past 10 minutes, etc.) may be used, or time-series data of vital signs from the start of medication to the current time may be used. Furthermore, the difference between vital signs before and after the start of medication (change in vital signs) may be input into the learning model 56.

[0038] As described above, the control unit 51 acquires administration information (first administration information) of a drug (first drug) used in a drug therapy for a patient, acquires patient information about the patient and test information about tests performed on the patient (blood test information), and acquires time-series data of the patient's vital signs. The control unit 51 can estimate the patient's condition after the start of drug therapy by inputting the acquired administration information, patient information, test information, and time-series data of the vital signs into the learning model 56, which outputs estimated information about the patient's condition after the start of drug therapy when the administration information of a drug used in drug therapy, patient information, test information, and time-series data of the vital signs are input.

[0039] The control unit 51 also acquires administration information (first administration information) of a drug (first drug) used in the drug therapy for the patient, acquires patient information about the patient and test information about tests on the patient (blood test information), and acquires differential data on the patient's vital signs before and after the start of drug administration. The control unit 51 can estimate the patient's condition after the start of drug therapy by inputting the acquired administration information, patient information, test information, and differential data on the vital signs before and after the start of drug administration into a learning model 56 that outputs estimated information on the patient's condition after the start of drug therapy when it receives administration information about the drug used in drug therapy, patient information, test information, and differential data on the vital signs before and after the start of drug administration.

[0040] 6, the learning model 56 may be generated (trained) so as to output estimated information on the patient's condition after the start of drug therapy when administration information of drugs used in drug therapy, patient information, test information (blood test information), vital signs during drug therapy, and administration information of drugs used in supportive therapy are input. In this case, the control unit 51 can further acquire administration information (second administration information) of drugs (second drugs) used in supportive therapy for the patient and input the acquired administration information into the learning model 56 to estimate the patient's condition after the start of drug therapy.

[0041] This will enable us to obtain estimated patient information, taking into account the drugs used in supportive care to prevent side effects and complications from cancer treatment, including biologics (antibody drugs) for autoimmune diseases.

[0042] The learning model 56 may be configured with a machine learning model such as a deep neural network (DNN), a support vector machine (SVM), a logistic regression, or a decision tree. The learning model 56 may be generated (trained) as follows, for example. The control unit 51 collects training data including administration information of drugs used in drug therapy, patient information, test information, vital signs, and estimated information on the patient's condition after the start of drug therapy. The estimated information on the patient's condition after the start of drug therapy is training data. The estimated information on the patient's condition after the start of drug therapy is estimated information from the current time during drug therapy after a required time has elapsed, and the estimated information can be categorized by predetermined required time. The control unit 51 can generate the learning model 56 based on the acquired training data for each required time, so that when administration information of drugs used in drug therapy, patient information, test information, and vital signs are input, the learning model 56 outputs estimated information on the patient's condition after the start of drug therapy.

[0043] Furthermore, assuming that the required time is, for example, 5 minutes, 10 minutes, or 20 minutes, training data is first collected, including administration information of drugs used in drug therapy, patient information, test information, vital signs, and estimated information on the patient's condition 5 minutes from the current time point during the start of drug therapy. Based on the acquired training data, the control unit 51 inputs the administration information of drugs used in drug therapy, patient information, test information, and vital signs into the learning model 56. When 5 minutes is input into the learning model 56 as the required time point, the learning model 56 can be generated so as to output estimated information on the patient's condition 5 minutes from the current time point during the start of drug therapy. The same applies to the case where the required time point is 10 minutes or 20 minutes. Alternatively, a separate learning model 56 may be generated for each set required time point. For example, a unique learning model 56 may be generated for each required time point of 5 minutes, 10 minutes, and 20 minutes.

[0044] The learning model 56 may be generated by the drug therapy support device 50, and the generated learning model 56 may be stored in the memory unit 54, or may be generated by another learning processing device, and the learning model 56 generated from the learning processing device may be acquired and stored in the memory unit 54.

[0045] The control unit 51 can display the estimated result of the patient's condition on the display device 40.

[0046] The vital signs during drug therapy used as training data for generating the learning model 56 may be current values ​​(latest values), but are not limited to current values. For example, time-series data of vital signs from a past time to the current time (e.g., the past 5 minutes, the past 10 minutes, etc.) may be used, or time-series data of vital signs from the start of medication to the current time may be used. Furthermore, the difference between vital signs before and after the start of medication (change in vital signs) may be used.

[0047] FIG. 7 is a diagram showing a first example of a patient condition estimation result screen. The estimation result screen displays patient information (e.g., patient ID or name), the names of drugs used in drug therapy, the start time of drug therapy, the elapsed time since the start, the patient's condition, etc. The patient condition displays estimated information, the current value of the estimated information, the predicted value after the required time (20 minutes after in the example shown), the class, and an alert. In the example of FIG. 7, the estimated information displayed is blood pressure and respiratory rate, but is not limited to blood pressure and respiratory rate.

[0048] The patient's current condition is that the blood pressure (systolic blood pressure) is 130 mmHg and the respiratory rate is 15 breaths per minute, and it is estimated that 20 minutes from now the blood pressure will be 125 mmHg and the respiratory rate will be 16 breaths per minute.

[0049] The control unit 51 can classify the estimated information output by the learning model 56 into predetermined classes and output the classified classes. The predetermined classes can be, for example, two, "safe" and "danger." The estimated information can be classified into a predetermined class by comparing the value of the estimated information with a predetermined threshold. For example, if blood pressure (systolic blood pressure) is less than a threshold (e.g., 90 mmHg), the estimated information can be classified as "danger," and if blood pressure (systolic blood pressure) is greater than or equal to a threshold (e.g., 90 mmHg), the estimated information can be classified as "safe." Note that instead of classifying the estimated information into two, "safe" and "danger," the estimated information may be classified into three, such as "safe," "caution," and "danger."

[0050] In the example of Fig. 7, the predicted value of the estimated information after 20 minutes is classified as "safe." Because it is classified as "safe," no alert is displayed.

[0051] As described above, the estimated information output by the learning model 56 is an estimated value that estimates the numerical value from the current time point during drug therapy after a required time has elapsed for at least one of the vital signs input to the learning model 56, and the control unit 51 classifies the estimated value output by the learning model 56 into one class from among multiple classes that classify the risk of the patient's condition into multiple stages based on the comparison result with a predetermined value, and can display on the display the current numerical value, the output estimated value, and the classified class for at least one of the vital signs input to the learning model 56 in correspondence with each other.

[0052] FIG. 8 is a diagram showing a second example of a patient condition estimation result screen. In the second example, the patient's current condition is estimated to be a blood pressure (systolic blood pressure) of 125 mmHg and a respiratory rate of 15 breaths per minute, and five minutes from the current time, the blood pressure will be 85 mmHg and the respiratory rate will be 13 breaths per minute. The control unit 51 classifies the blood pressure value five minutes from now as "dangerous" and the respiratory rate as "safe." In the example of FIG. 8, the estimated time is five minutes from now, but this is not limited to five minutes from now and may be any time, such as 10 minutes or 20 minutes from now.

[0053] The control unit 51 can output instructions to change the administration information of drugs used in drug therapy depending on the estimated patient condition. In the example of FIG. 8 , the estimated blood pressure after five minutes is 85 mmHg, which is below the threshold of 90 mmHg, so it is classified as "dangerous" and an alert such as "Please temporarily suspend medication" is output. The alert may be output as a voice, or the color of the display screen may be changed to red, or the alert text may flash. Note that the alert is not limited to the example of FIG. 8 , and may be an alert such as "Please check vital signs" or "Please report to your doctor or the doctor on-site at the outpatient clinic."

[0054] As described above, the control unit 51 can generate alert information when the estimated value output by the learning model 56 is classified into the class with the highest risk among multiple classes, and display the generated alert information on the display in association with the current numerical value, the output estimated value, and the classified class.

[0055] 9 is a diagram showing a third example of a patient condition estimation result screen. In the third example, the patient's current condition is estimated to be a blood pressure (systolic blood pressure) of 130 mmHg and a respiratory rate of 15 breaths per minute, and 20 minutes from the current time, the blood pressure will be 85 mmHg and the respiratory rate will be 13 breaths per minute. The control unit 51 classifies the blood pressure value 20 minutes from now as "dangerous" and the respiratory rate as "safe." In the example of FIG. 9, the estimated time is 20 minutes from now, but this is not limited to 20 minutes from now and may be any time, such as 5 minutes or 10 minutes from now.

[0056] The control unit 51 can output instructions to change the administration information of drugs used in drug therapy according to the estimated patient condition. In the example of FIG. 9 , the estimated blood pressure after 20 minutes is 85 mmHg, which is below the threshold of 90 mmHg, and is therefore classified as "dangerous." An alert such as "Please reduce the administration rate" is output. The alert is not limited to the example of FIG. 9 , and may be, for example, an alert such as "Please temporarily suspend administration," "Please check vital signs," or "Please report to your doctor or a doctor on-site at the outpatient clinic." The alert may be output as a voice message, a warning message such as by changing the display screen color to red, or by flashing the alert text.

[0057] This makes it possible to predict the occurrence of highly urgent side effects (adverse reactions) such as hypersensitivity reactions (HSR) and infusion reactions (IR) in advance, allowing patients undergoing drug therapy to be treated quickly before side effects occur, and preventing the occurrence of side effects in advance.

[0058] 10 is a diagram showing a first example of a patient condition estimation process performed by the pharmacotherapy support device 50. The control unit 51 acquires patient information about the patient receiving pharmacotherapy (S11), acquires administration information about the drugs used in the pharmacotherapy (S12), and acquires test information about the patient (S13). The test information is, for example, blood test information.

[0059] The control unit 51 determines whether or not drug therapy has started (S14). Whether or not drug therapy has started can be determined, for example, by acquiring information on whether or not drug administration has started from the medication device 10. If drug therapy has not started (NO in S14), the control unit 51 continues the processing of step S14. If drug therapy has started (YES in S14), the control unit 51 acquires the patient's vital signs (S15).

[0060] The control unit 51 inputs the acquired patient information, administration information, test information, and vital signs into the learning model 56 to estimate the patient's condition a required time from the current time (S16). The learning model 56 outputs the estimated information for a required time from the current time. The control unit 51 classifies the patient's condition into one class from multiple classes that classify the patient's condition into multiple stages based on the estimated information output by the learning model 56 (S17), and determines whether the estimated information has been classified as "dangerous" (S18). If the estimated information has been classified as "dangerous" (YES in S18), the control unit 51 outputs an alert (S19) and performs the processing of step S20, which will be described later.

[0061] If the condition is not classified as "dangerous" (NO in S18), the control unit 51 determines whether or not the drug therapy has ended (S20). Whether or not the drug therapy has ended can be determined, for example, by obtaining information from the medication device 10 as to whether or not the administration of the drug has ended. If the drug therapy has not ended (NO in S20), the control unit 51 performs the processing from step S15 onwards. If the drug therapy has ended (YES in S20), the control unit 51 ends the processing.

[0062] Fig. 11 is a diagram showing a second example of the configuration of the pharmacotherapy support system of this embodiment. The difference from the first example shown in Fig. 1 is that a display device 70 is connected to the data server 100. The pharmacotherapy support device 50 (controller 51) can display an alert (the alert output in step S19 in Fig. 10 ) on the medication device 10 and the display device 70 connected to the data server 100 via the communication network 1.

[0063] This allows an alert to be displayed on the medication device 10 installed in a treatment room where drug therapy is being performed, allowing medical personnel to monitor the situation. In addition, by sending an alert to the data server 100 and displaying the alert on the display device 70, the alert can be displayed on the department system and electronic medical record.

[0064] 12 is a diagram showing an example of a processing procedure for generating a learning model 56 by the drug therapy support device 50. The control unit 51 acquires training data including patient information, test information (blood test information), administration information of drugs used in drug therapy, the patient's vital signs, and estimated information on the patient's condition after the start of drug therapy (S31). Based on the acquired training data, the control unit 51 generates a learning model so as to output estimated information on the patient's condition after the start of drug therapy when the patient information, test information (blood test information), drug administration information, and the patient's vital signs are input, and then ends the processing.

[0065] In the training data, the estimated information on the patient's condition after the start of drug therapy corresponds to the teacher data. That is, when the control unit 51 inputs patient information, test information (blood test information), drug administration information, and the patient's vital signs into the learning model 56, the learning model 56 can be generated by adjusting the parameters of the learning model 56 so that the estimated information on the patient's condition after the start of drug therapy output by the learning model 56 matches the estimated information on the patient's condition after the start of drug therapy as the teacher data.

[0066] 13 is a diagram showing a second example of the procedure for estimating a patient's condition by the pharmacotherapy support device 50. The control unit 51 acquires patient information of a patient to be administered pharmacotherapy (S41), acquires administration information of drugs used in the pharmacotherapy (S42), and acquires examination information of the patient (S43).

[0067] The control unit 51 determines whether drug therapy has started (S44). Whether drug therapy has started can be determined, for example, by obtaining information from the medication device 10 indicating whether medication has started. If drug therapy has not started (NO in S44), the control unit 51 continues the processing of step S44. If drug therapy has started (YES in S44), the control unit 51 acquires the patient's blood pressure at predetermined intervals (S45). Specifically, the control unit 51 can acquire blood pressure at predetermined intervals (e.g., 10-minute intervals) from the patient's blood pressure data continuously measured by the measurement device 20 that performs non-invasive blood pressure measurement. Note that when the measurement device 20 measures the patient's blood pressure at predetermined intervals, the control unit 51 can acquire blood pressure at the predetermined intervals by synchronizing with the measurement operation of the measurement device 20.

[0068] The control unit 51 inputs the acquired patient information, administration information, test information, and blood pressure into the learning model 56 to estimate the patient's condition a required time from the current time (S46). The learning model 56 outputs the estimated information for a required time from the current time. The control unit 51 classifies the patient's condition into one class from multiple classes that classify the risk level of the patient's condition into multiple stages based on the estimated information output by the learning model 56 (S47), and determines whether the estimated information has been classified as "Caution" (S48). If the estimated information is not classified as "Caution" (NO in S48), the control unit 51 continues processing from step S45 onwards.

[0069] If the control unit 51 classifies the result as "Caution" (YES in S48), it outputs an alert (S49). The alert may be, for example, "Pay attention to changes in blood pressure," but is not limited thereto. The alert may be output as a voice message, by changing the display screen color to yellow, or by flashing the alert text. The alert may also be issued by changing the color of the "Caution" text or blood pressure value displayed on the display screen to yellow. While the alert continues to be output, the control unit 51 acquires the patient's blood pressure at intervals shorter than the predetermined interval (S50). Specifically, if the measuring device continuously measures the patient's blood pressure, the control unit 51 acquires blood pressure data measured continuously by the measuring device at intervals shorter than the predetermined interval. If the measuring device intermittently measures the patient's blood pressure, the control unit 51 instructs the measuring device to measure blood pressure at intervals shorter than the predetermined interval, and acquires blood pressure data measured by the measuring device at intervals shorter than the predetermined interval in accordance with the instruction. The short interval may be, for example, every five minutes. The control unit 51 inputs the blood pressure acquired at intervals shorter than the predetermined interval into the learning model 56 along with separately acquired patient information, administration information, and test information to re-estimate the patient's condition after a required time from the current time (S51). The control unit 51 re-classifies the estimated information into one class based on the estimated information output by the learning model 56 (S52) and determines whether the estimated information has been classified as "Caution" (S53). If the controller 51 classifies the estimated information as "Caution" (YES in S53), it continues to output an alert (S54). If the drug therapy has not ended (NO in S55), the control unit 51 continues the processing from step S50 onwards. If the estimated information is not classified as "Caution" in step S53 (NO in S53), the control unit 51 stops outputting the alert, returns to step S45, and again acquires the patient's blood pressure at predetermined intervals. Here, the control unit 51 repeats the processing from step S50 to step S55 until the estimated information is no longer classified as "Caution" in step S53, thereby enabling the control unit 51 to estimate the patient's condition after a required time from the current time at intervals shorter than the predetermined interval. This allows for more detailed monitoring of the patient's condition.

[0070] If the drug therapy has ended (YES in S55), the control unit 51 ends the process.

[0071] As described above, when the estimation information is classified as "caution," estimating the patient's condition by acquiring blood pressure at intervals shorter than the predetermined interval increases the number of data sets including administration information of drugs used in drug therapy, patient information, test information, vital signs (blood pressure), and estimated information on the patient's condition from the present time to the end of the required time during the drug therapy, compared to when estimating the patient's condition by acquiring blood pressure at predetermined intervals. By collecting such an increased number of data sets as training data, the amount of training data increases, making it possible to improve the estimation accuracy of the learning model 56.

[0072] The learning model 56 outputs estimated information about the patient's condition after a required time from the current time during drug therapy, and the control unit 51 classifies the patient's condition into one class from multiple classes that classify the risk of the patient's condition into multiple stages based on the estimated information output by the learning model 56, inputs vital signs acquired at a first interval during drug therapy into the learning model 56 to acquire estimated information about the patient's condition after a required time from the current time during drug therapy at the first interval, classifies the risk of the patient's condition at the first interval based on the estimated information acquired at the first interval, and if the patient is classified into a high-risk class among the multiple classes, acquires vital signs at a second interval that is shorter than the first interval, inputs the vital signs acquired at the second interval into the learning model 56, and acquires estimated information about the patient's condition after a required time from the current time during drug therapy at the second interval.

[0073] When the risk level is classified into a high risk class among the multiple classes, the control unit 51 may send an instruction to the measuring device for measuring the vital signs to measure the vital signs at a second interval.

[0074] Furthermore, when the risk level is classified into a high risk class among the multiple classes, the control unit 56 may display recommendation information on the display recommending that vital signs be measured at a second interval.

[0075] As described above, according to this embodiment, the condition of a patient after starting drug therapy can be estimated, and therefore, medical personnel can grasp the risk of the patient from the predicted estimated information and respond promptly to treatment.

[0076] REFERENCE SIGNS LIST 1 Communication network 10 Medication administration device 20 Measuring device 30 Terminal device 40, 70 Display device 50 Drug therapy support device 51 Control unit 52 Communication unit 53 Memory 54 Storage unit 55 Computer program 56 Learning model 57 Recording medium reading unit 100 Data server 101 Patient information DB 102 Examination information DB

Claims

1. A computer program that causes a computer to execute the following processes: acquire first administration information of a first drug used in a drug therapy for a patient; acquire patient information and test information of tests on the patient; acquire vital signs of the patient; and estimate the patient's condition regarding side effects associated with the administration of the first drug after the start of the drug therapy for the patient by inputting the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimated information of the patient's condition after the start of the drug therapy when the first administration information of a first drug used in the drug therapy, patient information, test information, and vital signs are input.

2. The computer program according to claim 1, wherein the vital signs are measured from time to time during the drug therapy.

3. The computer program according to claim 1, wherein the learning model outputs estimated information for a required time period from the present time during the drug therapy, and estimates the patient's condition for a required time period from the present time during the start of the drug therapy.

4. The computer program according to claim 1, which causes a computer to execute the following process: further acquiring second administration information of a second drug used in supportive therapy for the patient; and estimating the patient's condition after initiation of drug therapy by further inputting the acquired second administration information into the learning model, which outputs estimated information on the patient's condition after initiation of drug therapy when further input of second administration information of a second drug used in supportive therapy.

5. A computer program according to any one of claims 1 to 4, which causes a computer to execute a process of classifying estimated information output by the learning model into one class from among multiple classes indicating the risk of the patient's condition.

6. The computer program according to any one of claims 1 to 4, which causes a computer to execute a process of outputting an instruction to change first administration information of a first drug used in said drug therapy in accordance with the estimated patient condition.

7. The computer program according to any one of claims 1 to 4, which causes a computer to execute a process of accepting a setting of the required time.

8. A computer program according to any one of claims 1 to 4, wherein the first administration information includes at least one of a drug name, a dosage, an administration rate, an administration cycle, and an administration route.

9. The computer program of any one of claims 1 to 4, wherein the vital signs include pulse rate, respiratory rate, body temperature, blood pressure, and oxygen saturation.

10. A computer program according to any one of claims 1 to 4, wherein the estimated information includes blood pressure or respiratory rate.

11. The computer program according to any one of claims 1 to 4, wherein the patient information includes age, sex, weight, and type of cancer.

12. A computer program according to any one of claims 1 to 4, wherein the test information includes blood test results before the start of drug therapy.

13. A computer program as described in any one of claims 1 to 4, which causes a computer to perform the following processing: the estimated information output by the learning model is an estimated value obtained by estimating the numerical value from the current time point during the drug therapy after a required time has elapsed for at least one of the vital signs input to the learning model; the estimated value output by the learning model is classified into one class from a plurality of classes that classify the risk of the patient's condition into multiple stages based on the result of comparison with a predetermined value; and the current numerical value, the output estimated value, and the classified class for at least one of the vital signs input to the learning model are displayed on a display in correspondence with each other.

14. A computer program as described in claim 13, which causes a computer to execute the following process: when the estimated value output by the learning model is classified into the class with the highest risk among the multiple classes, generate alert information; and display the generated alert information on a display in association with the current numerical value, the output estimated value, and the classified class.

15. The computer program of claim 1, which causes a computer to perform the following processes: the learning model outputs estimated information about the patient's condition after a required time from the current time during the drug therapy; classifying the patient's condition into one class from a plurality of classes that classify the risk of the patient's condition into multiple stages based on the estimated information output by the learning model; inputting the vital signs acquired at a first interval during the drug therapy into the learning model to acquire estimated information about the patient's condition after a required time from the current time during the drug therapy at the first interval; classifying the risk of the patient's condition at the first interval based on the estimated information acquired at the first interval; if the patient is classified into a high-risk class among the plurality of classes, acquiring the vital signs at a second interval that is shorter than the first interval; inputting the vital signs acquired at the second interval into the learning model to estimate estimated information about the patient's condition after a required time from the current time during the drug therapy at the second interval.

16. The computer program according to claim 15, which causes a computer to execute a process of transmitting, to a measuring device for measuring the vital signs, an instruction to measure the vital signs at the second intervals if the vital signs are classified into a high-risk class among the plurality of classes.

17. The computer program according to claim 15, which causes a computer to execute a process of displaying, on a display, recommendation information that recommends measuring the vital signs at the second intervals when the subject is classified into a high-risk class among the plurality of classes.

18. A drug therapy support device comprising a control unit which: acquires first administration information of a first drug used in drug therapy for a patient; acquires patient information and test information of tests on the patient; acquires vital signs of the patient; and estimates the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy for the patient by inputting the acquired first administration information, patient information, test information, and vital signs into a learning model which outputs estimated information of the patient's condition after initiation of the drug therapy when the first administration information of a first drug used in drug therapy, patient information, test information, and vital signs are input.

19. A drug therapy support method comprising: acquiring first administration information of a first drug to be used in drug therapy for a patient; acquiring patient information and test information of a test on the patient; acquiring vital signs of the patient; and inputting the acquired first administration information, patient information, test information, and vital signs into a learning model that outputs estimated information of the patient's condition after initiation of drug therapy when first administration information of a first drug to be used in drug therapy, patient information, test information, and vital signs are input, thereby estimating the patient's condition regarding side effects associated with administration of the first drug after initiation of the drug therapy for the patient.

20. A learning model generation method comprising: acquiring training data including first administration information of a first drug used in drug therapy, patient information, test information, vital signs, and estimated information on the patient's condition regarding side effects associated with the administration of the first drug after the start of the drug therapy; and generating a learning model based on the acquired training data so as to output estimated information on the patient's condition regarding side effects associated with the administration of the first drug after the start of the drug therapy when the first administration information of the first drug used in drug therapy, patient information, test information, and vital signs are input.

Images