JP2021089523A - Health-medical care data analysis system and health-medical care data analysis method - Google Patents

Abstract

To assist in creating a new policy for medical care, medical checkup or nursing care.SOLUTION: A health-and-medical care data analysis system for analyzing health-and-medical care data comprises: a first database for managing the first health-and-medical care data of a first group; a knowledge vector creation unit for outputting a correlation between the first health-and-medical care data as vector information; a second database for storing knowledge vector information; a third database for managing the second health-and-medical care data of a second group and policy information relating to the policy carried out in the second group; a first vector conversion unit for generating vector information by vector-converting the second health-and-medical care data using knowledge vector information; and a machine learning unit for outputting model information in which a relationship between the vector information and the policy information is modeled by machine learning and storing it in a fourth database. The second group is composed of low-order groups constituting the first group, and the policy information includes the subjects of the policy of health-and-medical care and the results of the policy in the second group.SELECTED DRAWING: Figure 1

Description

The present invention relates to a healthcare data analysis system and method for analyzing healthcare information such as medical care and long-term care.

In recent years, while healthy life expectancy and working life expectancy have been extended, changes in the population structure such as the declining birthrate and aging population have made increasing medical and long-term care costs a concern. For this reason, appropriate measures are required to prevent the occurrence and aggravation of health, medical care, long-term care, etc.

In order to decide measures for medical care, it is necessary to consider future forecasts and past achievements. For example, Patent Document 1 is known as a method for accurately predicting a future health condition. In Patent Document 1, an output vector indicating a prediction of the future state of the subject is obtained by normalizing the input vector expressing the consultation record of the subject and then inputting the input vector into the learning algorithm.

Further, Patent Document 2 is known as a method for examining past achievements. Patent Document 2 discloses a technique for obtaining a data representation that enables appropriate clustering of medical data in which different types of data such as medical examination values and interview results are mixed by using latent topic analysis. Has been done.

JP-A-2018-194904 Japanese Unexamined Patent Publication No. 2017-027307

With the above-mentioned conventional technology, it is possible to predict the health condition and classify medical information, but it is difficult to support the planning of measures for medical treatment, medical examination, or long-term care.

In particular, if predictions and classifications are made from healthcare data for measures implemented in a specific area (local government or region) or occupational area (group), the analysis results can be applied to that area, but they are specific. When measures implemented in a region are implemented in different regions, the same effect may not always be obtained, and the prediction model and classification method that are the results of analysis in the original region may not always be applicable. ..

In other words, the effects of measures related to health, medical care, prevention of long-term care, etc. are not always the same because the environment (climate, climate), population composition, and industrial structure differ from region to region. Therefore, with the above-mentioned conventional technology, it is difficult to support the creation of national medical measures and measures in other regions from the prediction model and classification method which are the analysis results of data in some regions. There was a problem.

Therefore, the present invention has been made in view of the above problems, and an object of the present invention is to support the creation of new measures for medical care, medical examination, or long-term care by using healthcare data for measures such as the community.

In the present invention, a computer system having a processor and a memory is a health care data analysis system that analyzes health care data, and has a first database that manages a first health care data of a first group, and the above-mentioned first. A knowledge vector creation unit that inputs 1 health care data, converts the mutual relationship between the first health care data into a vector and outputs it as knowledge vector information, and a second knowledge vector information storage unit. A third database that manages a database, a second health care data of the second group, and measure information regarding measures implemented in the second group, and a second health of the third database. A first vector transformation in which medical data and knowledge vector information of the second database are input, and vector information is generated by vector conversion of the second health care data of the second group using the knowledge vector information. A machine learning unit that inputs the converted vector information and the measure information of the third database and outputs model information that models the relationship between the vector information and the measure information by machine learning. It has a fourth database that manages the output model information, the second group is composed of lower groups constituting the first group, and the measure information is the second group. Includes the target persons of the health care measures in Japan and the results of the measures.

Therefore, the present invention calculates vector information from national (first group) health care data, and for this vector information and measures and measures implemented in a specific area (second group). Model information (predictive model) is generated from vector information of healthcare data. When the measures taken in a specific area (second group) are implemented in another area (third group) by applying the healthcare data of another area (third group) to this model information. It becomes possible to predict the effect of the above, and it becomes possible to support the creation of new measures for healthcare (medical care, medical examination or long-term care).

Details of at least one practice of the subject matter disclosed herein are set forth in the accompanying drawings and in the description below. Other features, aspects, and effects of the disclosed subject matter are manifested in the disclosures, drawings, and claims below.

It is a figure which shows the Example of this invention and shows the outline of the healthcare data analysis system. It is a block diagram which shows the Example of this invention and shows an example of the structure of the healthcare data analysis system. It is a flowchart which shows the Example of this invention and shows an example of the knowledge vector information creation processing performed by the knowledge creation apparatus. It is a flowchart which shows the Example of this invention and shows an example of the model generation processing performed by the model generation apparatus. It is a flowchart which shows the Example of this invention and shows an example of the simulation process performed by the simulation apparatus # 2. It is a flowchart which shows the Example of this invention and shows an example of the simulation process performed by the simulation apparatus # 1. It is a figure which shows the Example of this invention and shows an example of the medical examination information of database DB # 1. It is a figure which shows the Example of this invention and shows an example of the medical examination information of database DB # 3. It is a figure which shows the Example of this invention and shows an example of the medical examination information of database DB # 5. It is a figure which shows the Example of this invention and shows an example of the receipt information of the database DB # 1. It is a figure which shows the Example of this invention and shows an example of the receipt information of the database DB # 3. It is a figure which shows the Example of this invention and shows an example of the receipt information of the database DB # 5. It is a figure which shows the Example of this invention and shows an example of the care information of database DB # 1. It is a figure which shows the Example of this invention and shows an example of the care information of database DB # 3. It is a figure which shows the Example of this invention and shows an example of the care information of database DB # 5. It is a figure which shows the Example of this invention and shows an example of the measure information of the database DB # 3. It is a figure which shows the Example of this invention and shows an example of the measure effect information of the database DB # 3. It is a figure which shows the Example of this invention and shows an example of the matching information generated by the knowledge-creating apparatus. It is a figure which shows the Example of this invention and shows an example of the vector information generated by the knowledge-creating apparatus. It is a figure which shows the Example of this invention and shows an example of the knowledge vector information stored in the database DB # 1. It is a figure which shows the Example of this invention and shows an example of the model information stored in the database DB # 4. It is a figure which shows the Example of this invention and shows an example of the model selection screen. It is a figure which shows the Example of this invention and shows an example of the simulation result screen. It is a block diagram which shows the Example of this invention and shows an example of the structure of the simulation apparatus. It is a figure which shows the Example of this invention and shows an example of the generation of the vector representation of a term using a Skip-gram model.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

<Overview>
FIG. 1 is a diagram showing an embodiment of the present invention and showing an outline of a healthcare data analysis system. The healthcare data of this embodiment is information including at least one of medical information, medical examination information, and long-term care information.

In the healthcare data analysis system, the healthcare data collected by the country (or higher group) A1 is stored in the database (DB # 1) 1 including medical receipt information, medical examination information, and long-term care information. .. The healthcare data collected by Country A1 is an anonymous database, and it is not possible to identify an individual, but the receipt information, medical examination information, and long-term care information can be matched with the same person's information by a common personal ID. ..

Then, at the level of country A1, as will be described later, the knowledge vector information 21 and the vector information 23 are generated by vector-converting the healthcare data and stored in the database (DB # 2) 2. As an example of vector transformation, a distributed representation can be used.

On the other hand, the subordinate area (or subordinate group) A2 that constitutes the country (or group) A1 is the health care data (medical receipt information, medical examination information, care information) of the individual who lives or works in the area A2. , The information on the measures implemented in the area is stored in the database (DB # 3) 3.

In the area A2, the healthcare data of the database 3 and the measure information are vectorized using the knowledge vector information 21 held by the country A1 as described later. Then, in the region A2, machine learning is performed with the vectorized information to generate model information 41 for making predictions about healthcare, and the model information 41 is registered in the database (DB # 4) 4 managed by the country A1. The model information 41 is generated for each measure information implemented in the area A2, and can include the result (effect) of the measure.

In addition, another region (subgroup) A3 constituting the country (or group) A1 stores the healthcare data of the individual who lives or works in the region A3 in the database (DB # 5) 5.

The database 4 managed by country A1 is shared by region A2 and other regions A3. The model information 41 stored in the database 4 is used in the country A1 and other regions A3. In country A1, the model information 41 selected from database 4 can be used to simulate the healthcare data in database 1 and predict the effect of implementing the measures implemented in region A2 on a nationwide scale.

Similarly, in the other region A3, the model information 41 selected from the database 4 is used to simulate the healthcare data in the database 5 to predict the effect of implementing the measures implemented in the region A2 in the region A3. be able to.

As described above, in the healthcare data analysis system of this embodiment, the national level healthcare data owned by country A1 is vector-converted to generate knowledge vector information 21 and vector information 23, and the area A2 where the measures are implemented. Model information 41 (prediction model) is generated by machine learning the vectorized information using the knowledge vector information 21 for the healthcare data and the measure information of the above. The vector information 23 is an entity of the vector value, and the knowledge vector information 21 is a table that manages the vector information 23.

In country A1 and other regions A3, by implementing simulations with their respective healthcare data using this model information 41, when various measures implemented in region A2 are implemented in country A1 and other regions A3. The effect can be predicted.

In country A1 and other regions A3, by planning new measures and diverting existing measures based on the results of simulations, it is possible to efficiently plan and decide measures related to healthcare. In particular, country A1 and region A3 do not have data on the results of measures implemented in region A2, but by aggregating the forecast models created in region A2 into DB4 as model information, the country A1 and region A3 are concerned. It is possible to predict the effect of implementing the above measures. Further, when the prediction model is created in the region A2, the data in the region A2 is vectorized by using the knowledge vector information 21 created from the characteristics of the entire data such as the country A1. As a result, the prediction model created in the region A2 can create a model that reflects not only the data characteristics of the region A2 but also the tendency in the overall data of the country A1. Then, even when the model created in the region A2 is applied to the country A1 or the region A3, the data of the country A1 or the region A3 is vectorized by the same knowledge vector information 21 and then applied to the prediction model. It has the effect of being able to accurately predict the case where it is applied in the area A3. In this way, the healthcare data analysis system of this embodiment can support the planning and decision-making of measures related to healthcare.

In the above, an example is shown in which the highest group is country A1 and the lower group (A2, A3) is a region, but the present invention is not limited to this, and the highest group (A1) is the entire company. Then, the group (A2, A3) can be a business division or the like.

<System configuration>
FIG. 2 is a block diagram showing an example of the configuration of a healthcare data analysis system. The healthcare data analysis system of this embodiment has data center DC # 1 managed by country (group) A1, data center DC # 2 managed by region (group) A2, and data managed by region (group) A3. An example in which the center DC # 3 and the center DC # 3 are configured by a network (not shown) is shown.

The data center DC # 1 managed by the country (group) A1 is generated in the database 1 that stores anonymous healthcare data, the knowledge creation device 9 that generates the knowledge vector information 21 and the vector information 23, and the area A2. The database 4 for storing the model information 41 and the simulation device (# 1) 4 for selecting the model information 41 from the database 4 and performing the simulation with the healthcare data and the knowledge vector information 21 and the vector information 23 of the database 1 are included. ..

The data center DC # 2 managed by the area (group) A2 is a database 3 that stores healthcare data including personal information resident in the area A2 and measure information, and knowledge vector information 21 and vector information 23 of the data center DC # 1. Includes a model generator 7 that vectorizes healthcare data and measure information in region A2, generates a prediction model, and stores it in database 4 of data center DC # 1 as model information 41.

The data center DC # 3 managed by the area (group) A3 has a database 5 that stores healthcare data including personal information residing in the area A3, and model information 41 and knowledge selected from the database 4 of the data center DC # 1. Includes a simulation device 8 that performs simulations with healthcare data in database 5 using vector information 21 and vector information 23.

Database 1 of data center DC # 1 uses medical examination information 11 that anonymously stores information such as medical examinations, receipt information 12 that anonymously stores receipts when visiting a medical institution, and nursing care facilities. The nursing care information 13 and the care information 13 in which the information at the time of this is anonymously stored are stored.

The knowledge creation device 9 is a computer including a CPU 91, a memory 92, and a communication device (not shown). The knowledge vector creation unit 93 is loaded into the memory 92 as a program and executed by the CPU 91.

The knowledge vector creation unit 93 performs vector conversion by distributed representation using any one of the medical examination information 11, the receipt information 12, and the long-term care information 13 of the database 1, generates the vector information 23 and the knowledge vector information 21, and generates the database 2 Store in. In this embodiment, the number of vectorized dimensions of the vector information 23 is m-dimensional.

The medical examination information 11, the receipt information 12, and the long-term care information 13 of the database 1 are anonymized, but since the same person is given the same personal ID (identifier), the receipt information 12 and the medical examination information 11 And the personal IDs of the long-term care information 13 can be matched, and each information (table) can be combined before vector conversion can be performed.

As shown in FIG. 23, the simulation device (# 1) 6 is a computer including a CPU 61, a memory 62, an input device 67, an output device 68, a communication device 69, and a storage device 66.

The input device 67 is composed of a keyboard, a mouse, a touch panel, or the like. The output device 68 is composed of a display. The communication device 69 communicates with other computers via a network (not shown). The storage device 66 is composed of a non-volatile storage medium and stores data and programs.

The output unit 63, the vector conversion unit 64, and the estimation calculation unit 65 are loaded into the memory 62 as a program and executed by the CPU 61.

The simulation device (# 1) 6 is selected by the user from a plurality of model information 41 stored in the database 4, and then vectorized from the healthcare data and knowledge vector information 21 of the database 1 by the vector conversion unit 64. , The simulation is performed by the vector information 23 with the model information 41 selected by the estimation calculation unit 65.

The vector conversion unit 64 vectorizes at least one of the receipt information 12, the medical examination information 11, and the long-term care information 13 of the database 1 selected by the user. The vector conversion unit 64 can match the table selected by the user with the personal ID and then vectorize the table.

The estimation calculation unit 65 takes the vectorized data and the vector information 23 of the knowledge vector information 21 as inputs, performs a simulation with the selected model information 41, and outputs the result to the output device 68.

The CPU 61 operates as a functional unit that provides a predetermined function by executing a process according to a program of each functional unit. For example, the CPU 61 functions as the vector conversion unit 64 by executing the process according to the vector conversion program. The same applies to other programs. Further, the CPU 61 also operates as a functional unit that provides each function of a plurality of processes executed by each program. A computer and a computer system are devices and systems including these functional parts.

Next, the data center DC # 2 managed by the region (group) A2 constituting the country (group) A1 includes the database 3 and the model generator 7. Database 3 contains receipt information 32, medical examination information 31, long-term care information 33, measure information 34 implemented in the area, and measure effect information 3410 as healthcare data of individuals living or working in the area A2. , Is stored. The database 3 of the region A2, unlike the country A1, stores healthcare data including personal information.

The medical examination information 31 stores information such as a medical examination, the receipt information 32 stores a receipt when a medical institution is consulted, and the long-term care information 33 stores information when a long-term care facility or the like is used.

The measure information 34 stores the contents of the measures implemented in the area A2. As will be described later, the measure information 34 stores the target person of the health care measure and the measure result (effect of the measure).

The model generation device 7 is a computer including a CPU 71, a memory 72, and a communication device (not shown). The vector conversion unit 73 and the machine learning unit 74 are loaded into the memory 72 as a program and executed by the CPU 71.

The vector conversion unit 73 vectorizes the healthcare data (medical examination information 31, receipt information 32, care information 33) and the measure information 34 of the database 3 using the distributed representation, and the machine learning unit 74 machines the vectorized data. The training is performed to generate a prediction model, which is stored as model information 41 in the database 4 of the data center DC # 1.

Next, the data center DC # 3 managed by the region (group) A3 constituting the country (group) A1 includes the database 5 and the simulation device (# 2) 8. The database 5 stores the receipt information 52, the medical examination information 51, and the long-term care information 53 as the healthcare data of the individual who lives or works in the area A3. The database 5 of the region A3 is different from the country A1 in that it is healthcare data including personal information.

The medical examination information 51 stores information such as a medical examination, the receipt information 52 stores a receipt when a medical institution is consulted, and the long-term care information 53 stores information when a long-term care facility or the like is used.

The simulation device (# 2) 8 is a computer including a CPU 81, a memory 82, and a communication device (not shown). The output unit 83, the vector conversion unit 84, and the estimation calculation unit 85 are loaded into the memory 82 as a program and executed by the CPU 81.

The simulation device (# 2) 8 is selected by the user from a plurality of model information 41 stored in the database 4 of the data center DC # 1, and the healthcare data of the database 5 is vectorized by the vector conversion unit 84. Therefore, the simulation is performed by the model information 41 selected by the estimation calculation unit 85 and the vector information 23 of the knowledge vector information 21.

The vector conversion unit 84 vectorizes at least one of the receipt information 52, the medical examination information 51, and the long-term care information 53 of the database 1 selected by the user. The vector conversion unit 84 can match the table selected by the user with the personal ID and then vectorize the table.

The estimation calculation unit 85 takes the vectorized data and the vector information 23 of the knowledge vector information 21 as inputs, performs a simulation with the selected model information 41, and outputs the result to an output device (not shown). ..

In FIG. 2, two examples are shown in which the regions constituting the country A1 are not limited to these. Further, although an example in which the model generation device 7 is arranged in the area A2 and the simulation device 8 is arranged in the area A3 is shown, the present invention is not limited to this. For example, a model generator 7 and a simulation device 8 are arranged in each of the regions A2 and A3, and model information 41 based on the measures and healthcare data of the own region is generated and registered in the database 4 of the data center DC # 1. , The simulation can be carried out with the model information 41 of other regions.

Further, the knowledge creation device 9, the model generation device 7, and the simulation device (# 2) 8 have an input device and an output device as in the simulation device (# 1) 6 shown in FIG. 24, and are users. Is accepted by the input device and the processing result is displayed on the output device.

<Data>
Next, the data used in each of the data centers DC # 1 to DC # 3 will be described. FIG. 7 is a diagram showing an example of the medical examination information 11 of the database (DB # 1) 1. The medical examination information 11 is a table managed by the country A1.

The medical examination information 11 includes the medical examination ID 110, the area ID 111, the individual ID 112, the examination date 113, the age at reception 114, the abdominal circumference 115, the diastolic blood pressure 116, the systolic blood pressure 117, and the blood glucose level 118. Includes triglyceride 119 in one record.

The medical examination ID 110 stores an identifier indicating the institution that performed the medical examination. The area ID 111 stores an identifier indicating the area where the medical examination was performed. The personal ID 112 is given an identifier of the examinee. Measured values are stored in the abdominal circumference 115 to triglyceride 119.

The database 1 managed by the country A1 is an anonymous database, does not have personally identifiable information such as a name and an address, and the personal ID 112 functions as information for identifying whether or not the person is the same person.

Next, FIG. 8 is a diagram showing an example of the medical examination information 31 of the database (DB # 3) 3. The medical examination information 31 is a table managed by the area A2, and includes personal information that identifies an individual.

The medical examination information 31 includes a medical examination ID 310, an area ID 311 and an individual ID 312, a consultation date 313, a reception age 314, an abdominal circumference 315, a minimum blood pressure 316, a systolic blood pressure 317, and a blood glucose level 318. Includes triglyceride 319 in one record.

The medical examination ID 310 stores an identifier indicating the institution that performed the medical examination. The area ID 311 stores an identifier indicating the area where the medical examination was performed. The personal ID 312 is given an identifier of the examinee.

The database 3 managed by the area A2 is a database including personal information such as a name and an address, and the personal ID 112 is associated with the personal information. Measured values are stored in the abdominal circumference 315 to triglyceride 319.

Next, FIG. 9 is a diagram showing an example of the medical examination information 51 of the database (DB # 5) 5. The medical examination information 51 is a table managed by another area A3, and includes personal information that identifies an individual.

The medical examination information 51 includes a medical examination ID 510, an area ID 511, an individual ID 512, a consultation date 513, a reception age 514, an abdominal circumference 515, a minimum blood pressure 516, a systolic blood pressure 517, and a blood glucose level 518. One record contains triglyceride 519.

The medical examination ID 510 stores an identifier indicating the institution that performed the medical examination. The area ID 511 stores an identifier indicating the area where the medical examination was performed. The personal ID 512 is given an identifier of the examinee.

The database 3 managed by the area A2 is a database including personal information such as a name and an address, and the personal ID 112 is associated with the personal information. Measured values are stored in the abdominal circumference 515 to triglyceride 519.

FIG. 10 is a diagram showing an example of the receipt information 12 of the database (DB # 1) 1. Receipt information 12 is a table managed by country A1.

The receipt information 12 includes a receipt ID 120, an area ID 121, an individual ID 122, a medical treatment date 123, a medical institution number 124, an injury / illness name 125, a medical treatment practice 126, a drug 127, and a claim point 128. Included in one record.

The medical claim identifier is stored in the receipt ID 120. The area ID 111 stores an identifier indicating the area where the medical treatment was performed. The personal ID 112 is given an identifier of the examinee.

The medical institution number 124 stores the identifier of the medical institution that performed the medical treatment. The injury / illness name 125 stores the name or code of the injury / illness. The medical practice 126 stores a code of medical care content applied to the patient. The drug 127 stores the prescribed drug. Health points are stored in the billing points 128.

The database 1 managed by the country A1 is an anonymous database, does not have personally identifiable information such as a name and an address, and the personal ID 122 functions as information for identifying whether or not the person is the same person.

Next, FIG. 11 is a diagram showing an example of the receipt information 32 of the database (DB # 3) 3. The receipt information 32 is a table managed by the area A2, and includes personal information that identifies an individual.

The receipt information 32 includes a receipt ID 320, an area ID 321, an individual ID 322, a medical treatment date 323, a medical institution number 324, an injury / illness name 325, a medical practice 326, a drug 327, and a number of claims 328. Included in one record.

The medical claim identifier is stored in the receipt ID 320. The area ID 111 stores an identifier indicating the area where the medical treatment was performed. An identifier of the examinee is given to the personal ID 132.

The medical institution number 324 stores the identifier of the medical institution that performed the medical treatment. The name or code of the injury or illness is stored in the injury or illness name 325. The medical practice 326 stores a code of medical care content applied to the patient. The drug 327 stores the prescribed drug. Health points are stored in the billing points 328.

The database 3 managed by the area A2 is a database including personal information such as a name and an address, and the personal ID 322 is associated with the personal information.

Next, FIG. 12 is a diagram showing an example of the receipt information 52 of the database (DB # 5) 5. The receipt information 52 is a table managed by another area A3, and includes personal information that identifies an individual.

The receipt information 52 includes a receipt ID 520, an area ID 521, an individual ID 522, a medical treatment date 523, a medical institution number 524, an injury / illness name 525, a medical practice 526, a drug 527, and a number of claims 528. Included in one record.

The medical claim identifier is stored in the receipt ID 520. The area ID 111 stores an identifier indicating the area where the medical treatment was performed. The personal ID 152 is given an identifier of the examinee.

The medical institution number 524 stores the identifier of the medical institution that performed the medical treatment. The name or code of the injury or illness is stored in the injury or illness name 525. The medical practice 526 stores a code of medical care content applied to the patient. The drug 527 stores the prescribed drug. Health points are stored in the billing points 528.

The database 5 managed by the other area A3 is a database including personal information such as a name and an address, and the personal ID 522 is associated with the personal information.

FIG. 13 is a diagram showing an example of the long-term care information 13 of the database (DB # 1) 1. The long-term care information 13 is a table managed by the country A1.

The long-term care information 13 contains the long-term care information ID 130, the area ID 131, the personal ID 132, the date 133, the degree of long-term care 134, the business establishment number 135, the service code 136, and the billing amount 137 in one record. Including.

The long-term care information ID 130 stores the long-term care request identifier. The area ID 131 stores an identifier indicating the area where the long-term care service is provided. A caregiver identifier is assigned to the personal ID 132. The year and month 133 stores the year and month when the long-term care service was received. The degree of need for long-term care 134 stores the degree of need for long-term care determined by the certification for long-term care.

The business establishment number 135 stores the identifier of the business establishment that provided the long-term care service. The service code 136 stores the code of the implemented long-term care service. The billed amount of the long-term care service is stored in the billed amount 137.

The database 1 managed by the country A1 is an anonymous database, does not have personally identifiable information such as a name and an address, and the personal ID 132 functions as information for identifying whether or not the person is the same person.

FIG. 14 is a diagram showing an example of the long-term care information 33 of the database (DB # 3) 3. The long-term care information 33 is a table managed by the area A2.

The long-term care information 33 contains the long-term care information ID 330, the area ID 331, the personal ID 332, the date 333, the degree of long-term care 334, the business office number 335, the service code 336, and the billing amount 337 in one record. Including.

The long-term care information ID 330 stores the long-term care request identifier. The area ID 331 stores an identifier indicating the area where the long-term care service is provided. A caregiver identifier is assigned to the personal ID 332. The year and month 333 stores the year and month when the long-term care service was received. The degree of need for long-term care 334 stores the degree of need for long-term care determined by the certification for long-term care.

The business establishment number 335 stores the identifier of the business establishment that provided the long-term care service. The service code 336 stores the code of the long-term care service provided. The billed amount of the long-term care service is stored in the billed amount 337.

The database 3 managed by the area A2 is a database including personal information such as a name and an address, and the personal ID 332 is associated with the personal information.

FIG. 15 is a diagram showing an example of the long-term care information 53 of the database (DB # 5) 5. The long-term care information 53 is a table managed by another area A3.

The long-term care information 53 includes a long-term care information ID 530, an area ID 531, an individual ID 532, a date 533, a degree of long-term care 534, a business establishment number 535, a service code 536, and a billing amount 537 in one record. Including.

The long-term care information ID 530 stores an identifier of the long-term care information. The area ID 531 stores an identifier indicating the area where the long-term care service is provided. A caregiver identifier is assigned to the personal ID 532. The year and month 533 stores the year and month when the long-term care service was received. The degree of need for long-term care 534 stores the degree of need for long-term care determined by the certification for long-term care.

The business establishment number 535 stores the identifier of the business establishment that provided the long-term care service. The service code 536 stores the code of the long-term care service provided. The billed amount of the long-term care service is stored in the billed amount 537.

The database 5 managed by the other area A3 is a database including personal information such as a name and an address, and the personal ID 532 is associated with the personal information.

FIG. 16A is a diagram showing an example of the measure information 34 of the database (DB # 3) 3. The measure information 34 is a table managed by the area A2.

The measure information 34 includes the area ID 340, the measure ID 341, the measure name 342, the target person ID 343, and the goal achievement 344 in one record. The area ID 340 stores the identifier of the area where the measure is implemented. An identifier for identifying the measure is stored in the measure ID 341.

The name of the measure is stored in the measure name 342. The individual ID who participated in the measure is stored in the target person ID 343. In the goal achievement 344, a value indicating whether or not the target person has achieved the goal of the measure is stored, "1" is stored when the target person has achieved the goal, and "0" is stored when the target person cannot achieve the goal. To.

The target person ID 343 of the measure information 34 corresponds to the personal ID of the medical examination information 31, the receipt information 32, or the long-term care information 33 of the database 3. The measure information 34 can manage whether or not the individual who participated in the measure has reached the goal set for each measure.

FIG. 16B is a diagram showing an example of the measure effect information 3410 of the database (DB # 3) 3. The measure effect information 3410 is a table managed by the area A2.

The measure effect information 3410 includes the measure ID 3411, the measure name 3412, the implementation year 3413, the effect index 3414, and the actually measured effect 3415 in one record. The measure ID 3411 stores the measure identifier and corresponds to the measure ID 341 of the measure information 34.

The name of the measure is stored in the measure name 3412, and corresponds to the measure name 342 of the measure information 34. The year in which the measure was implemented is stored in the implementation year 3413. The effect index 3414 stores an index for measuring the effect of the measure. The measured effect 3415 stores the value of the index set in the effect index 3414. The measure effect information 3410 manages the actual measurement effect 3415 for each measure.

FIG. 17 is a diagram showing an example of the matching information 22 generated by the knowledge creation device 9 of the data center DC # 1. The knowledge creation device 9 can independently create the vector information 23 of the knowledge vector information 21 for the medical examination information 11, the receipt information 12, and the care information 13, but the medical examination information 11, the receipt information 12, and the care information 13 can be created independently. It is also possible to generate the knowledge vector information 21 and the vector information 23 after combining the above.

The example of FIG. 17 shows an example in which the medical examination information 11, the receipt information 12, and the long-term care information 13 are combined by matching the personal IDs 112, 122, and 132. In the case 221, the values of the personal IDs 112, 122, and 132 are stored.

The receipt information 222 stores the value of the record of the receipt information 12 corresponding to the personal ID (122) of the case 221. In the long-term care information 223, the value of the record of the long-term care information 13 corresponding to the personal ID (132) of the case 221 is stored. The medical examination information 224 stores the value of the record of the medical examination information 11 corresponding to the personal ID (112) of the case 221.

The knowledge creation device 9 performs vector conversion after generating the matching information 22 in which the medical examination information 11, the receipt information 12, and the nursing care information 13 are matched by the personal ID and combined. In this embodiment, the number n of the columns of the receipt information 222, the long-term care information 223, and the medical examination information 224 of the matching information 22 is set to the number of dimensions n before vector conversion.

The knowledge creation device 9 specifies a data source to be read from the database 1, a year (or period), a region, and the like when generating the vector information 23, and sets these designations in the knowledge vector information 21.

FIG. 18 is a diagram showing an example of the vector information 23 generated by the knowledge creation device 9. The vector information 23 is information obtained by converting the matching information 22 into a vector value by the knowledge vector creating unit 93 by a distributed representation.

The vector information 23 includes the knowledge ID 231 and the item 232 and the vector value 233 in one record. The knowledge ID 231 stores an identifier indicating a group managed by the knowledge vector information 21. Item 232 stores the field name of the matching information 22 (item name of receipt information 222, long-term care information 223, and medical examination information 224).

In the illustrated example, the n-dimensional matching information 22 shown in FIG. 17 is vectorized by an m-dimensional distributed representation.

FIG. 19 is a diagram showing an example of the knowledge vector information 21 generated by the knowledge creation device 9. The knowledge vector information 21 includes the knowledge ID 211, the data source 212, the year 213, and the region 214 in one record.

The knowledge ID 211 is an identifier indicating a group of vector information 23 assigned by the knowledge vector creation unit 93, and corresponds to the knowledge ID 231 in FIG. The data source 212 stores the information of the database 1 that is input when generating the vector value of the knowledge ID 211. In year 213, the year in which data is selected from the data source of database 1 is stored. The area 214 stores the area in which data is selected from the data source of the database 1.

The knowledge vector information 21 manages the data source, the year, and the region of the vector information 23 by the knowledge ID 231 for the group of the vector information 23. When the vector information 23 is used from the simulation devices 6 and 8, the desired knowledge ID 211 is selected from the data source 212, the year 213, the region 214, and the like with reference to the knowledge vector information 21, and the selected knowledge ID 211 is selected. The record of the knowledge ID 231 of the vector information 23 that matches the value of may be acquired.

In this embodiment, the knowledge vector information 21 and the vector information 23 are separated from each other, but the data source 212, the year 213, and the area 214 may be added to the vector information 23 to form one table.

FIG. 20 is a diagram showing an example of model information 41 generated by the model generation device 7. The model information 41 includes the model ID 411, the model 412, the knowledge ID 413, the measure ID 414, and the actual result 415 in one record.

The model ID 411 stores an identifier set by the model generation device 7. The model 412 stores the model generated by the model generation device 7. The knowledge ID 413 stores the value of the knowledge ID 211 of the knowledge vector information 21 acquired from the database 2 when the model is generated.

The measure ID 414 stores the identifier of the measure implemented in the area A2. The measure ID 414 corresponds to the measure ID 341 of the measure information 34 of FIG. 16A. The result of implementing the measure is stored in the result 415. The actual result 415 corresponds to the goal achievement 344 of FIG. 16A, and if the value of the goal achievement 344 is "1", "target achievement" is set for the actual result 415, and if the value of the goal achievement 344 is "0". "Unachieved target" is set for the achievement 415.

By referring to the model information 41, the simulation devices 6 and 8 can determine which measures have achieved the target and which measures have not.

<Processing>
FIG. 3 is a flowchart showing an example of the knowledge vector information 21 creation process performed by the knowledge creation device 9 of the data center DC # 1. This process is started in response to a command from the user of the knowledge creation device 9.

First, the knowledge creation device 9 acquires data specified by the user from the healthcare data (health care information) of the database 1 (S1). The knowledge creation device 9 receives a table to be read from the database 1, a year (or period), a target area, and the like, and acquires data (healthcare data) having a specified content from the database 2.

When a plurality of data sources are specified, the knowledge creation device 9 collates personal IDs (case IDs 221) and generates matching information 22 by combining a plurality of tables, as shown in FIG.

Next, the knowledge vector creation unit 93 of the knowledge creation device 9 vectorizes the mutual relationship between the health care data with respect to the data source (matching information 22) acquired from the database 2 by a distributed representation, and obtains the vector information 23. Generate (S2). As shown in FIG. 18, the knowledge creation device 9 calculates an m-dimensional vector value for the n-dimensional item 232 and assigns the knowledge ID 232 to each item 232.

The knowledge creation device 9 generates knowledge vector information 21 from the knowledge ID 231 of the generated vector information 23, data source, and input information such as year and region. Then, the knowledge creation device 9 stores the generated knowledge vector information 21 and vector information 23 in the database 2 (S3).

The Skip-gram model is known as an example of a method for vectorizing mutual relationships between health care data by a distributed representation. Also in this embodiment, the Skip-gram model can be used to generate a vector representation of terms.

FIG. 24 is a schematic diagram showing an example of generating a vector representation of terms using the Skip-gram model.

FIG. 24 shows a schematic diagram of the model used in this embodiment. The knowledge vector creation unit 93 sets the initial values of the parameters of this model. In the following, first, the model used in this embodiment will be described with reference to FIG. In this embodiment, the two-layer model shown in FIG. 24 is used.

The one-hot vector 401 indicated by x on the left side of FIG. 24 is a vector in which only one element indicating a term is 1 and all other elements are 0. Here, x is a column vector. The number of dimensions is equal to the number of terms for which the term representation vector is created. That is, if all the target terms are z1, z2, ..., Zn, the vector x is n-dimensional. The one-hot vector representing zp is a matrix in which only the p-th element is 1 and the other elements are 0. The vector x = (0, ..., 0, 1, 0, ..., 0) ^ T. Here, x ^ T represents the transpose of the vector x.

Now, for the sake of simplicity, assuming that the target terms are drug A, drug B, and drug C, the vector x is three-dimensional. Assuming that the corresponding elements of drug A, drug B, and drug C are the first element, the second element, and the third element, drug A is x = (1, 0, 0) ^ T, and drug B is x = ( 0, 1, 0) ^ T and drug C are represented as x = (0, 0, 1) ^ T.

In the above, three examples of terms are shown, but in reality, all the names of injuries and illnesses, drug names, and medical practice names are enormous, so the number of dimensions of the one-hot vector 401 is also large.

Now, one record is selected from the data source (matching information 22) acquired from the database 2. All terms of the injury / illness name, medicine, and medical practice of the receipt of the matching information 22 are w1, w2, ..., WT. Further, the one-hot vectors corresponding to each are x1, x2, ..., XT. At this time, FIG. 24 shows the one-hot vector xk of each term wi, and the terms w1, ..., wi-1, wi + 1, ..., wT one-hot vector x1, ..., xi-1, which co-occur with it. It is a model that predicts xi + 1, ..., XT. The matrices W and W'are the parameters of this model, and these parameters are updated based on the receipt data to improve the accuracy of the prediction.

In FIG. 24, first, the n-dimensional one-hot vector xi of the term wi is converted into the m-dimensional vector Wxi by the m × n matrix W. m is the number of dimensions of the term expression vector and is predetermined. It is recommended to set the size to several hundred dimensions. m is generally a small value compared to n, which is equal to the total number of terms, and Wxi can be regarded as encoding xi with an m-dimensional vector. Using this, the terms w1, ..., wi-1, wi + 1, ..., wT n-dimensional one-hot vector x1, ..., xi-1, xi + 1, ..., XT that co-occur with the term represented by xk are restored. Think about it. Therefore, it is converted into an n-dimensional vector W'Wx by the n × m matrix W'.

Here, the inner product of these is calculated as the degree of similarity between the one-hot vector xj of wj and W'Wxi. As for xi, it is assumed that the Iith element is 1 and the other elements are 0. Further, using the m-dimensional matrix vector vk, W = (v1, v2, ..., Vn). Further, using the m-dimensional matrix vector v'k, W'= (v'1, v'2, ..., V'n) ^ T. At this time, Wxi = vIi. Further, it is assumed that the Ijth element of xj is 1 and the other elements are 0. At this time, the inner product of xj and W'Wxj is equal to the inner product v'Ij · vIi of v'Ij and vIi. In this way, in order to calculate the inner product of xj and W'Wxi, the inner product v'Ij · vIi of the m-dimensional vectors may actually be calculated.

Now, using the softmax function, it is defined as follows. Here, Σ_p adds 1 to n with respect to p. Since this calculation needs to be performed for all terms, the amount of calculation is large, and in practice, it is preferable to use Hierarchical Softmax or NEG (Negative sampling).
p (wj | wi) = exp (v'Ij · vIi) / Σ_p exp (v'p · vIi)

In this model, the matching information 22 is used to update W and W'so as to maximize the following function, which is a logarithmic probability value that can reproduce co-occurrence terms. In the following, Σ_i adds 1 to T for i, and Σ_j ≠ i adds 1 to T excluding i for j.
(1 / T) Σ_i Σ_j ≠ i log {p (wj | wi)}

For each matching information 22, W and W'are repeatedly updated by the gradient descent method so as to maximize the above logarithmic probability value. At this time, when the one-hot vector of the term w is a vector x in which the i-th value is 1 and the other elements are 0, Wx = vi is the term expression vector of the term w.

Here, the first layer is called an input layer, the second layer is called an output layer, W is called an input layer matrix, W'is called an output layer matrix, x is called an input vector, and Wx is called a hidden vector.

By the above processing, vector information 23 having an m-dimensional vector value and knowledge vector information 21 as catalog information of the vector information 23 are generated from the designated healthcare data and stored in the database 2.

FIG. 4 is a flowchart showing an example of a model generation process performed by the model generation device 7 of the data center DC # 2. This process is started in response to a command from the user of the model generator 7.

The model generator 7 includes knowledge vector information 21 selected by the user from database 2 of data center DC # 1 and healthcare data information selected by the user from database 3 of data center DC # 2 (medical examination information 31, Receives receipt information 32 or care information 33) and measure information 34 (S11).

The model generation device 7 acquires the data of the knowledge ID 231 corresponding to the knowledge ID 211 of the knowledge vector information 21 of the data center DC # 1 from the vector information 23. In addition, the model generator 7 reads the table selected by the user and the healthcare data corresponding to the year and region from the database 3. When a plurality of healthcare data are specified, the model generation device 7 performs matching with an individual ID (case ID 221) and generates matching information 22 in which tables are combined, as shown in FIG. .. Further, the model generation device 7 acquires the measure ID 414 of the measure information 34 selected by the user.

Next, the vector conversion unit 73 of the model generation device 7 vectorizes the acquired healthcare data (health examination information 31, receipt information 32, nursing care information 33 or matching information 22) using the acquired vector information 23. (S12).

For example, the vector conversion unit 73 gives an m-dimensional vector value 233 to the item of healthcare data that matches the item of the vector information 23, and calculates the vector value by the distributed representation. The vectorization of the healthcare data (medical examination information 31, receipt information 32, care information 33) in the database 3 is not limited to this, and is vectorized using the vector information 23 (knowledge vector information 21). This may be done, and a well-known or known technique may be applied.

Next, the machine learning unit 74 of the model generation device 7 inputs the vectorized healthcare data to generate a model of the relationship between the healthcare data and the measure information 34 (S13). For example, the machine learning unit 74 executes learning and generates a model by using the ratio of the target achievement 344 of the measure information 34 (the ratio of “1”) and the actually measured effect 3415 of the measure effect information 3410 as objective variables.

The model generation device 7 stores the generated model in the database 4 of the data center DC # 1 as model information 41 (S14), and stores the model information 41 in the simulation devices 6 and 8 of the data centers DC # 1 and DC # 2. share it.

By the above processing, the model generator 7 uses the vector information 23 generated in the data center DC # 1 to vectorize the healthcare data stored in the database 2 of the data center DC # 2, and the machine learning unit 74 uses the vector information 23. Model information 41 can be generated from healthcare data vectorized using the designated measure information 34 as an objective variable.

Further, by applying the national level vector information 23 to the vectorization of the healthcare data of the region A2, the accuracy of the generated model information 41 (prediction model) can be improved.

FIG. 5 is a flowchart showing an example of simulation processing performed by the simulation device 8 of the data center DC # 3. This process is started in response to a command from the user of the simulation device 8.

The simulation device 8 includes knowledge vector information 21 and model information 41 selected by the user from the database 2 of the data center DC # 1, and healthcare data (medical examination information) selected by the user from the database 5 of the data center DC # 3. 51, receipt information 52 or care information 53) is accepted (S21).

The simulation device 8 acquires the vector information 23 of the knowledge ID 231 corresponding to the knowledge ID 211 of the knowledge vector information 21 of the data center DC # 1. In addition, the simulation device 8 reads the table selected by the user and the healthcare data corresponding to the year and region from the database 3.

When a plurality of data sources of the database 5 are specified, the simulation device 8 matches the personal IDs (case IDs 221) and generates the matching information 22 by combining the plurality of tables, as shown in FIG. To do. Further, the simulation device 8 acquires the selected model information 41 from the database 4 of the data center DC # 1. The model information 41 is acquired by using the model information 41 selected by the user of the simulation device 8 from the model selection screen 550 displayed on the output device (not shown).

FIG. 21 is a diagram showing an example of the model selection screen 550 displayed on the output device (not shown) by the simulation device 8. The model selection screen 550 includes an area 551 for inputting the code of the area for generating the model information 41, a year 552 for inputting the period of the database for generating the model information 41, a model selection area 560, and a simulation execution button 553. ..

The model selection area 560 has a check box 561, a measure ID 562, a measure name 563, an implementation year 564, an area number 565, an effect index 566, and an actual measurement effect 567. The user of the simulation device 8 can select the model information 41 by checking the check box 561 of the model of the desired measure ID 562.

The measure ID 562 corresponds to the measure ID 414 of the model information 41, the measure name 563 corresponds to the measure name 342 of the measure information 34, and the implementation year 564 corresponds to the implementation year 3413 of the measure effect information 3410. The number 565 corresponds to the area ID 340 of the measure information 34, the effect index 566 corresponds to the effect index 3414 of the measure effect information 3410, and the actual measurement effect 567 corresponds to the actual measurement effect 3415 of the measure effect information 3410.

Next, the vector conversion unit 84 of the simulation device 8 vectorizes the acquired healthcare data (or matching information 22) using the acquired vector information 23 (S22). This process is the same as the process of the vector conversion unit 73 in step S12.

Next, the estimation calculation unit 85 of the simulation device 8 inputs the vectorized healthcare data (or matching information 22) and executes a simulation using the acquired model information 41 (S23).

The output unit 83 of the simulation device 8 displays the simulation result of the estimation calculation unit 85 on the output device (not shown). By this processing, in the simulation device 8 of the area A3, it is possible to estimate the effect when the measure of the area A2 corresponding to the measure ID 414 of the model information 41 is implemented from the healthcare data of the database 5.

FIG. 22 is a diagram showing an example of a screen of the simulation result 610 displayed on the output device (not shown) by the simulation device 8. The simulation result 610 includes a region 611 that displays the region code, a year 612 that indicates the period of the data source, a graph area 620 that displays the target achievement rate for each measure, and a graph area 630 that displays the amount of restraint for each measure. And the area 640 for selecting the simulation result to be output to the graph area. Achieving the goal means, for example, the number of people who achieved the target weight of the weight loss program, the number of people who achieved smoking cessation in the smoking cessation program, the number of people who could improve the amount of activity by preventing rehabilitation and long-term care, etc. The number and percentage of people who achieved the target of the index.

Area 640 includes a check box 641, a measure ID 642, a measure name 643, an applicable number of people 644, a target achievement 645, an effect index 646, and a predicted value 647. The user of the simulation device 8 can display the graph of the simulation result in the graph areas 620 and 630 by checking the check box 641 of the simulation result of the desired measure ID 642.

The measure ID 642 corresponds to the measure ID 414 of the model information 41, the measure name 643 corresponds to the measure name 342 of the measure information 34, and the applicable number 644 corresponds to the number of healthcare data of the database 5 used for the simulation. Correspondingly, the target achievement (forecast) 645 corresponds to the number of people who can achieve the target among the applicable number of people 644, the effect index 646 corresponds to the effect index 3414 of the measure effect information 3410, and the predicted value 647 corresponds to the effect index. It corresponds to a value of 646 (amount of simulation result). In this way, as shown in FIGS. 21 and 22, when the weight loss program implemented in a certain area is applied in one's own area (or country), the effect of achieving the weight loss target and suppressing medical expenses is to what extent. You can check if you can get it.

FIG. 6 is a flowchart showing an example of simulation processing performed by the simulation device 6 of the data center DC # 1. This process is started in response to a command from the user of the simulation device 6.

The simulation device 6 includes knowledge vector information 21 and model information 41 selected by the user from the database 2 of the data center DC # 1, and healthcare data (medical examination information) selected by the user from the database 1 of the data center DC # 1. 11. Accepts receipt information 12 or care information 13) (S31).

The simulation device 6 acquires the vector information 23 of the knowledge ID 231 corresponding to the knowledge ID 211 of the knowledge vector information 21 of the data center DC # 1. In addition, the simulation device 6 reads the table selected by the user and the healthcare data corresponding to the year and region from the database 1.

When a plurality of data sources of the database 1 are specified, the simulation device 6 matches the personal IDs (case IDs 221) and generates the matching information 22 by combining the plurality of tables, as shown in FIG. To do. Further, the simulation device 6 selects the selected model information 41 from the database 4 of the data center DC # 1 in the same manner as in FIG. 21.

Next, the vector conversion unit 64 of the simulation device 6 vectorizes the acquired healthcare data (or matching information 22) using the acquired vector information 23 (S32). This process is the same as the process of the vector conversion unit 73 in step S12.

Next, the estimation calculation unit 65 of the simulation device 6 inputs the vectorized healthcare data (or matching information 22) and performs a simulation using the acquired model information 41 (S33).

The output unit 63 of the simulation device 6 displays the simulation result of the estimation calculation unit 65 on the output device (not shown) in the same manner as in FIG. 22. By this processing, in the simulation apparatus 6 of the country A1, the effect when the measure of the region A2 corresponding to the measure ID 414 of the model information 41 is implemented can be estimated from the healthcare data of the database 1.

As described above, the healthcare data analysis system of this embodiment calculates the vector information 23 and the knowledge vector information 21 from the national healthcare data, and the vector information 23 and the measure information 34 of the specific area A2 are used. Model information (prediction model) 41 is generated from healthcare data. By applying the healthcare data of another region A3 (or country A1) to this model information 41, it is possible to predict the effect when the measures of a specific region A2 are implemented in another region A3 (or country A1). Is possible. This makes it possible to support the creation of new measures for health care (medical care, medical examination or long-term care). In particular, country A1 and region A3 do not have data on the results of measures implemented in region A2, but by aggregating the forecast models created in region A2 into DB4 as model information, the country A1 and region A3 are concerned. It is possible to predict the effect of implementing the above measures. Further, when the prediction model is created in the region A2, the data in the region A2 is vectorized by using the knowledge vector information 21 created from the characteristics of the entire data such as the country A1. As a result, the prediction model created in the region A2 can create a model that reflects not only the data characteristics of the region A2 but also the tendency in the overall data of the country A1. Then, even when the model created in the region A2 is applied to the country A1 or the region A3, the data of the country A1 or the region A3 is vectorized by the same knowledge vector information 21 and then applied to the prediction model. It has the effect of being able to accurately predict the case where it is applied in the area A3.

In the above embodiment, there are three data centers DC # 1 in country A1 as the highest region (or group), data center DC # 2 in region A2 as a lower group, and data center DC # 3 in region A3. An example of having a data center is shown, but the present invention is not limited to this.

For example, a knowledge creation device 9, a simulation device 6, a model generation device 7, a simulation device 8, and each database may be arranged in one data center. Alternatively, the functions of the knowledge creation device 9, the simulation device 6, the model generation device 7, and the simulation device 8 are realized by one computer, and as the knowledge creation unit, the first simulation unit, the model generation unit, and the second simulation unit. May be good.

<Conclusion>
As described above, the healthcare data analysis system of the above embodiment can have the following configuration.

(1) A computer system having a processor and a memory is a health care data analysis system that analyzes health care data, and is the first health care data (medical examination information 11, receipt information) of the first group (country A1). 12. The first database (DB # 1) that manages the care information 13) and the first health care data are input, and the mutual relationship between the first health care data is converted into a vector. A knowledge vector creation unit (93) that outputs as knowledge vector information (21), a second database (DB # 2) that stores the knowledge vector information (21), and a second group (region A2). A third database (DB) that manages health and medical data (medical examination information 31, receipt information 32, care information 33) and measure information (34) regarding measures implemented in the second group (A2). # 3), the second health care data of the third database (DB # 3), and the knowledge vector information (21) of the second database (DB # 2) are input, and the second group The first vector conversion unit (73) that generates vector information by vector conversion of the second health care data of (A2) using the knowledge vector information (21), the converted vector information, and the first. The machine learning unit (74) that inputs the measure information (34) of the database (DB # 3) of 3 and outputs the model information (41) that models the relationship between the vector information and the measure information (34) by machine learning. ) And a fourth database (DB # 4) that manages the output model information (41), and the second group (A2) constitutes the first group (A1). The measure information (34) is a health data analysis system characterized by including the target persons of the health care measures in the second group (A2) and the results of the measures.

With the above configuration, the healthcare data analysis system calculates vector information 23 and knowledge vector information 21 from national healthcare data, and this vector information 23, measure information 34 of specific region A2, and healthcare of region A2. Model information (prediction model) 41 is generated from the data. By applying the healthcare data of another region A3 (or country A1) to this model information 41, it is possible to predict the effect when the measures of a specific region A2 are implemented in another region A3 (or country A1). Is possible. This makes it possible to support the creation of new measures for health care (medical care, medical examination or long-term care).

(2) The health care data analysis system according to (1) above manages the third health care data (health care information 51, receipt information 52, care information 53) of the third group (A3). A fifth database (DB # 5), a second vector conversion unit (84) that generates vector information by vector conversion of the third health care data using the knowledge vector information (21), and the vector. It further has a first estimation calculation unit (85) that applies the converted vector information to the model information (41) to perform simulation, and a first output unit (83) that outputs the result of the simulation. The third group (A3) is a lower group that constitutes the first group (A1), is composed of a group different from the second group (A2), and is the first estimation operation. The department (85) is a health care data analysis system characterized by estimating the result of executing the health care measures implemented in the second group (A2) in the third group (A3).

With the above configuration, the healthcare data analysis system uses vector information 23, measure information 34 of a specific area A2 (second group), and model information (prediction model) from healthcare data of area A2 (second group). 41 is generated. By applying the healthcare data of another region A3 (third group) to this model information 41, the measures of a specific region A2 (second group) are implemented in the other region A3 (third group). It is possible to predict the effect of this.

(3) In the health care data analysis system according to (1) above, a third vector conversion that generates vector information by vector conversion of the first health care data using the knowledge vector information (21). A unit (64), a second estimation calculation unit (62) that applies the vector-converted vector information to the model information (41) to perform a simulation, and a second output unit that outputs the result of the simulation. The second estimation calculation unit (65) further has (63) and, and the second estimation calculation unit (65) executed the health care measures implemented in the second group (A2) in the first group (A1). A health care data analysis system characterized by estimating results.

With the above configuration, the healthcare data analysis system uses vector information 23, measure information 34 of a specific area A2 (second group), and model information (prediction model) from healthcare data of area A2 (second group). 41 is generated. By applying the healthcare data of country A1 (first group) to this model information 41, the effect when the measures of a specific area A2 (second group) are implemented in country A1 (first group). Can be predicted.

(4) The health care data analysis system according to (1) above, wherein the first health care data of the first group (A1) is anonymous information. system.

With the above configuration, it is possible to promote the protection of personal information by concealing personal information and using the data in national level health care data.

(5) In the health and medical data analysis system according to (2) above, the model information (41) of the fourth database (DB # 4) includes the identifier of the measure information (34), and the first item. The estimation calculation unit (85) of 1 outputs the result of the measure for each of the plurality of model information (41), accepts the selection of the model information (41), and uses the selected model information (41) as the first. A health care data analysis system characterized by performing simulations of healthcare data of 3 groups (A3).

With the above configuration, the user of the healthcare data analysis system displays the result of the measure for each of the plurality of model information (41) on the output device, selects the desired model information (41), and selects the selected model. With information 41, it is possible to simulate the healthcare data of another region A3 (third group).

(6) In the health and medical data analysis system according to (3) above, the model information (41) of the fourth database (DB # 4) includes the identifier of the measure information (34), and the first item. The estimation calculation unit (65) of 2 outputs the result of the measure for each of the plurality of the model information (41), accepts the selection of the model information (41), and uses the selected model information (41). A health and medical data analysis system characterized by performing simulations.

With the above configuration, the user of the healthcare data analysis system displays the result of the measure for each of the plurality of model information (41) on the output device, selects the desired model information (41), and selects the selected model. Information 41 can be used to simulate country A1 (first group) healthcare data.

The present invention is not limited to the above-described examples, and includes various modifications. For example, the above-described embodiment is described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to the one including all the configurations described. Further, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, for a part of the configurations of each embodiment, any of addition, deletion, or replacement of other configurations can be applied alone or in combination.

Further, each of the above configurations, functions, processing units, processing means and the like may be realized by hardware by designing a part or all of them by, for example, an integrated circuit. Further, each of the above configurations, functions, and the like may be realized by software by the processor interpreting and executing a program that realizes each function. Information such as programs, tables, and files that realize each function can be stored in a memory, a hard disk, a recording device such as an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

In addition, the control lines and information lines indicate those that are considered necessary for explanation, and do not necessarily indicate all the control lines and information lines in the product. In practice, it can be considered that almost all configurations are interconnected.

1, 2, 3, 4, 5 Database 6, 8 Simulation device 7 Model generation device 9 Knowledge creation device 11, 31, 51 Medical examination information 12, 32, 52 Receipt information 13, 33, 53 Nursing information 21 Knowledge vector information 23 Vector information 34 Measure information 41 Model information 61, 71, 81, 91 CPU
52, 62, 72, 82, 82 Memory 63, 83 Output unit 64, 84 Vector conversion unit 65, 82 Estimation calculation unit 93 Knowledge vector creation unit

Claims (12)

A computer system with a processor and memory is a health care data analysis system that analyzes health care data.
A first database that manages the first health care data of the first group,
A knowledge vector creation unit that inputs the first health care data, converts the mutual relationship between the first health care data into a vector, and outputs it as knowledge vector information.
A second database that stores the knowledge vector information,
A third database that manages the second health care data of the second group and the measure information regarding the measures implemented in the second group.
The second health care data of the third database and the knowledge vector information of the second database are input, and the second health care data of the second group is vector-converted using the knowledge vector information. The first vector conversion unit that generates vector information,
A machine learning unit that inputs the converted vector information and the measure information of the third database and outputs model information that models the relationship between the vector information and the measure information by machine learning.
It has a fourth database that manages the output model information, and has.
The second group is composed of lower groups constituting the first group.
The measure information is a health care data analysis system characterized by including the target persons of the health care measures in the second group and the results of the measures. The health care data analysis system according to claim 1.
A fifth database that manages the third health care data of the third group,
A second vector conversion unit that generates vector information by vector conversion of the third health care data using the knowledge vector information, and
A first estimation calculation unit that applies the vector-converted vector information to the model information to perform a simulation, and
It further has a first output unit that outputs the result of the simulation, and
The third group is
It is a lower group that constitutes the first group, and is composed of a group different from the second group.
The first estimation calculation unit is
A health care data analysis system characterized in that the results of implementing the health care measures implemented in the second group in the third group are estimated. The health care data analysis system according to claim 1.
A third vector conversion unit that generates vector information by vector conversion of the first health care data using the knowledge vector information,
A second estimation calculation unit that applies the vector-converted vector information to the model information to perform a simulation, and
It further has a second output unit that outputs the result of the simulation.
The second estimation calculation unit is
A health care data analysis system characterized in that the results of implementing the health care measures implemented in the second group in the first group are estimated. The health care data analysis system according to claim 1.
The first health care data of the first group is a health care data analysis system characterized in that it is anonymous information. The health care data analysis system according to claim 2.
The model information of the fourth database includes an identifier of the measure information.
The first estimation calculation unit is
Health and medical data characterized by outputting the results of measures for each of a plurality of model information, accepting the selection of the model information, and performing the simulation of the third health and medical data with the selected model information. Analysis system. The health care data analysis system according to claim 3.
The model information of the fourth database includes an identifier of the measure information.
The second estimation calculation unit is
Health medical care characterized in that the results of measures for each of the plurality of the model information are output, the selection of the model information is accepted, and the simulation of the first health care data is performed with the selected model information. Data analysis system. A computer with a processor and memory is a health care data analysis method that analyzes health care data.
The computer inputs the first health care data from the first database that manages the first health care data of the first group, and vectorizes the mutual relationship between the first health care data. A knowledge vector creation step of converting and outputting as knowledge vector information and storing the knowledge vector information in a second database, and
The computer inputs the second health care data from the third database that manages the second health care data of the second group and the measure information regarding the measures implemented in the second group. , The first vector conversion step of inputting knowledge vector information from the second database and generating vector information by vector conversion of the second health care data of the second group using the knowledge vector information.
The computer inputs the converted vector information and the measure information of the third database, and stores the model information in which the relationship between the vector information and the measure information is modeled by machine learning in the fourth database. Including machine learning steps,
The second group is composed of lower groups constituting the first group.
The measure information is a health care data analysis method characterized by including the target persons of the health care measures in the second group and the results of the measures. The health care data analysis method according to claim 7.
The computer inputs the third health care data from the fifth database that manages the third health care data of the third group, inputs the knowledge vector information from the second database, and the third The second vector conversion step of generating vector information by vector conversion of the health and medical data of the above using the knowledge vector information, and
A first estimation calculation step in which the computer applies the vector-converted vector information to the model information to perform a simulation.
The computer further includes a first output step that outputs the result of the simulation.
The third group is
It is a lower group that constitutes the first group, and is composed of a group different from the second group.
The first estimation calculation step is
A method for analyzing health care data, which estimates the result of implementing the health care measures implemented in the second group in the third group. The health care data analysis method according to claim 7.
A third vector conversion step in which the computer generates vector information by vector conversion of the first health care data using the knowledge vector information.
A second estimation calculation step in which the computer applies the vector-converted vector information to the model information to perform a simulation.
The computer further includes a second output step, which outputs the result of the simulation.
The second estimation calculation step is
A method for analyzing health care data, which estimates the result of implementing the health care measures implemented in the second group in the first group. The health care data analysis method according to claim 7.
A method for analyzing health care data, wherein the first health care data of the first group is anonymous information. The health care data analysis method according to claim 8.
The model information of the fourth database includes an identifier of the measure information.
The first estimation calculation step is
Health medical care characterized in that the results of measures for each of the plurality of the model information are output, the selection of the model information is accepted, and the simulation of the third health care data is performed with the selected model information. Data analysis method. The health care data analysis method according to claim 9.
The model information of the fourth database includes an identifier of the measure information.
The second estimation calculation step is
Health medical care characterized in that the results of measures for each of the plurality of the model information are output, the selection of the model information is accepted, and the simulation of the first health care data is performed with the selected model information. Data analysis method.

Images