JP6737488B1 - Information processing apparatus, information processing method, program, and learned model - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing device and the like capable of accurately estimating the state of the future degree of need for nursing care of a subject. An information processing device 1 includes an acquisition unit that acquires care-related information of a target person, care-related information of the target person, care-related information at a first time point of a plurality of persons, and the first time point. Based on a learned model that has already learned the degree of need for nursing care of the plurality of persons at a second time point after a predetermined period from It is characterized by being provided. [Selection diagram] Figure 1

Description

The present invention relates to an information processing device, an information processing method, a program, and a learned model.

In Japan's long-term care insurance system, if you become bedridden or in need of constant care (need to care) or need support for daily life such as housework, care prevention services are particularly effective ( If you need support, you can receive nursing care service. Whether or not the insured person is in a state requiring long-term care or a state requiring support, and to what extent nursing care or support is required, at each local government that is an insurer, certification for long-term care (including certification required for support; the same applies below) Is being reviewed. Various methods have been proposed to support this examination of certification for long-term care.

For example, in Patent Document 1, the Euclidean distance between various parameters obtained by surveying or measuring the target person and parameters of a person requiring long-term care certification (hereinafter referred to as “certification person”) is calculated to calculate the Euclidean distance between the two parameters. There is disclosed a treatment data estimation method or the like for performing pattern matching and estimating the degree of long-term care of the target person.

JP, 2001-34688, A

However, the invention according to Patent Document 1 estimates the degree of need for nursing care by simple pattern matching, and it cannot be said that sufficient accuracy is guaranteed.

In one aspect, it is an object to provide an information processing device or the like that can accurately estimate the future state of the degree of need for nursing care of a subject.

The information processing apparatus according to one aspect is the degree of long-term care required by a certified person who has been certified as requiring long-term care, the doctor's determination result regarding the state of the authorized person, and the determination of the visiting investigator regarding the state of the authorized person. The result, and care-related information including at least one of the payment performances of the benefits related to the care of the authorized person, and the degree of care required of the authorized person at the second time point after a predetermined period from the first time point. Machine learning is performed using the associated teacher data, and a storage unit that stores a learned model that receives the care-related information at the first time point as an input and outputs the degree of long-term care at the second time point. inputs to the care and acquisition unit for acquiring relevant information, the learned model the care-related information acquired by the acquiring unit stored in the storage unit of the subject at the present time, the learned model outputs the An estimation unit that estimates the degree of long-term care of the target person after the predetermined period of time by acquiring the degree of long- term care .

According to one aspect, it is possible to accurately estimate the future state of the degree of need for nursing care of the subject.

It is a schematic diagram which shows the structural example of a nursing care degree estimation system. It is a block diagram which shows the structural example of a server. It is explanatory drawing which shows an example of a record layout of local government DB and object DB. It is explanatory drawing regarding an estimation model. It is explanatory drawing which shows the example of a display screen of the estimation result of a nursing care degree. It is a flow chart which shows the procedure of generation processing of an estimation model. It is a flowchart which shows the procedure of the estimation process of the state of a nursing care degree. FIG. 9 is an explanatory diagram showing an outline of the second embodiment. It is explanatory drawing which shows the example of a display screen of the estimation result of the number of authorized persons. 9 is a flowchart showing a procedure of generation processing of an estimation model according to the second embodiment. It is a flowchart which shows the procedure of the estimation process of the number of authorized persons. It is a functional block diagram which shows the structural example of the server of the form mentioned above.

Hereinafter, the present invention will be described in detail with reference to the drawings showing an embodiment thereof.
(Embodiment 1)
FIG. 1 is a schematic diagram showing a configuration example of a nursing care degree estimation system. In the present embodiment, a degree-of-care-necessity estimation system that estimates a future state of the degree of long-term care of the target person based on the care-related information of the target person that correlates with the certification for long-term care will be described. The subject is any insured person having care related information, including but not limited to authorized persons. The care related information will be described later. The care level estimation system includes an information processing device 1 and terminals 2, 2, 2,... Each device is communicatively connected via a network N such as the Internet.

The information processing device 1 is a device that performs various types of information processing and information transmission/reception, and is, for example, a server device, a personal computer, or the like. In the present embodiment, the information processing device 1 is assumed to be a server device, and will be read as server 1 below for simplicity. The server 1 estimates the future state of the degree of need for care of the target person at the second time point after a predetermined period (for example, three years) from the first time point, based on the care-related information of the target person at the first time point. Perform processing to Specifically, as will be described later, the server 1 estimates the future state of the degree of need for care of the target person by using the estimation model 141 (see FIG. 4) generated by machine learning. The future state of the degree of long-term care of the target person is the future degree of long-term care of the target person, the change in the future degree of long-term care of the target person, and the transition probability of the degree of long-term care of the target person by future long-term care degree. Including.

Note that the "degree of need for nursing care" referred to in this specification is not limited to the state of need for nursing care represented by the five stages of "necessity of nursing care 1" to "necessity of nursing care 5", but also "support required 1" or "support required 2". It also includes the support required status. Further, the “degree of need for nursing care” is not limited to the index defined by law or the like, and may be a unique index.

In the present embodiment, a series of explanations are given as a system for local governments, but for example, as a system for estimating the future state of the degree of long-term care for other groups or individuals (such as the target person and their caregivers). It may be configured.

The terminal 2 is a terminal device of each local government that is an insurer, and is, for example, a personal computer or the like. The server 1 receives the information of the target person from each local government via the terminal 2, estimates the future state of the degree of long-term care, and outputs the estimation result to the terminal 2.

FIG. 2 is a block diagram showing a configuration example of the server 1. The server 1 has a control unit 11, a main storage unit 12, a communication unit 13, and an auxiliary storage unit 14.
The control unit 11 has an arithmetic processing unit such as one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), and GPUs (Graphics Processing Units), and stores the programs P stored in the auxiliary storage unit 14. By reading and executing, various information processing, control processing, and the like are performed. The main storage unit 12 is a temporary storage area such as an SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), and a flash memory, and temporarily stores data necessary for the control unit 11 to execute arithmetic processing. Remember. The communication unit 13 is a communication module for performing processing related to communication, and transmits/receives information to/from the outside.

The auxiliary storage unit 14 is a non-volatile storage area such as a hard disk or a large-capacity memory, and stores the program P and other data necessary for the control unit 11 to execute processing. In addition, the auxiliary storage unit 14 stores an estimation model 141, a local government DB 142, and a subject DB 143. The estimation model 141 is a classifier generated by machine learning, and is a learned model in which care-related information regarding the authorized person has been learned. The estimation model 141 is assumed to be used as a program module that is a part of artificial intelligence software. The local government DB 142 is a database that stores information on each local government that uses this system. The subject DB 143 is a database that stores information about the subject.

The auxiliary storage unit 14 may be an external storage device connected to the server 1. Moreover, the server 1 may be a multi-computer including a plurality of computers, or may be a virtual machine virtually constructed by software.

Further, in the present embodiment, server 1 is not limited to the above configuration, and may include, for example, an input unit that receives an operation input, a display unit that displays an image, and the like. Further, the server 1 includes a reading unit for reading a portable storage medium 1a such as a CD (Compact Disk)-ROM and a DVD (Digital Versatile Disc)-ROM, and reads and executes the program P from the portable storage medium 1a. You can Alternatively, the server 1 may read the program P from the semiconductor memory 1b.

FIG. 3 is an explanatory diagram showing an example of record layouts of the local government DB 142 and the subject DB 143.
The local government DB 142 includes a local government ID column, a local government name column, and a local government information column. The local government ID column stores the local government ID for identifying each local government. The local government name column and the local government information column store the names of the local governments (municipalities, etc.) and information about other local governments.

The subject DB 143 includes a subject ID column, a name column, an insurer column, a history number column, an application date column, and a care related information column. The subject ID column stores subject IDs for identifying each subject. The target person ID is, for example, the insured person number or my number, but is not limited to this. The name column, the insurer column, the history number column, the application date column, and the care related information column are associated with the target person ID, and the name of the target person and the insurer (local government) that provides the care service to the target person. It stores the local government ID, the history number of the care-related information of the target person submitted to the local government, the application date of the long-term care certification application, and the care-related information. In the care-related information column, for example, the basic information of the target person (insured person information), statement of position information indicating the contents of the statement of opinion created by the doctor interviewing the target person and his/her family, and the survey results by the visiting investigators Is stored, and benefit record information indicating the benefit record of long-term care insurance is stored.

FIG. 4 is an explanatory diagram regarding the estimation model 141. The estimation model 141 will be described with reference to FIG.
The estimation model 141 is a machine learning model whose parameters have been learned. For example, the server 1 uses the nursing care related information of the certified person who has already been certified as requiring nursing care as teacher data to generate the estimation model 141. In the present embodiment, the server 1 generates a decision tree model as the estimation model 141, as shown in FIG.

For example, the server 1 generates the estimation model 141 using a method of gradient boosting, which is a type of ensemble learning. That is, the server 1 generates one weak discriminator (decision tree) using the teacher data, and based on the error (precisely, the residual) between the predicted value and the actual measurement value by the generated weak discriminator, Weak classifiers are sequentially generated. The server 1 sequentially generates weak classifiers by referring to the gradient of the loss function defined by the error between the predicted value and the actual measurement value so as to consider the learning result of the previous weak classifier, and finally classifies the weak classifiers. To generate an estimation model 141.

Although gradient boosting is used as a learning method in the present embodiment, other ensemble learning such as bagging for generating a plurality of weak classifiers in parallel may be performed. The estimation model 141 may be generated by a learning method other than the ensemble learning.

The estimation model 141 is not limited to the decision tree model, and may be a neural network, a random forest, an SVM (Support Vector Machine), or another model.

As described above, the server 1 uses the care-related information of the authorized person as teacher data to generate the estimation model 141. The care-related information is any information that correlates with the degree of long-term care required of the target person, and includes at least one of the above-mentioned basic information, statement of opinion information, visit survey information, and benefit record information. Further, the care-related information may include at least one of medical receipt information, answer information of the basic checklist provided by the Ministry of Health, Labor and Welfare, specific medical examination information, and specific medical examination guidance information.

The basic information is, for example, the age and gender of the certified person, and the degree of long-term care required by the certified person at the time when the local government receives an application for certification of required nursing care (first time) (previous application The information on the degree of long-term care required by) is included. The statement of opinion information includes information such as the bedridden degree, the degree of dementia, the degree of short-term memory, and the cognitive ability of the authorized person judged by the doctor. The visit survey information includes information such as the degree of soliloquy of the authorized person, the degree of laughter and the degree of maladaptation to the group determined by the visit surveyor. Statement of position information and visit survey information are created when an application for certification for long-term care is made. The benefit record information includes, for example, the amount of benefits per month, the number of receipts per month, and the like. The benefit record information is created when long-term care is certified. The above-mentioned various parameters are all examples, and the care-related information may be an input parameter suitable for estimating the degree of long-term care.

In addition, in the present embodiment, the information of the certified person who has been certified as requiring long-term care is used as teacher data, but it is not possible for persons other than the certified person (for example, a person who has applied for certification for long-term care but was not certified). Information may be included in the teacher data. Further, the information of the authorized person used as the teacher data may include not only the person who has been certified as requiring nursing care, but also the information of a person who has been certified as requiring nursing care in the past (for example, a dead person). ..

The server 1 acquires the care-related information (Excel (registered trademark) file, CSV file, etc.) in which the above various information is described from the terminal 2 of each local government, uses it as the input parameter for the teacher, and outputs the correct answer parameter. The estimation model 141 is generated so as to approximate the value. It should be noted that the server 1 can also acquire the care-related information from other than the local government.

Here, the server 1 uses the degree of long-term care required at a time point (second time point) when a predetermined period has passed from the time point, not when the local government receives an application for certification of long-term care (first time point), as a correct answer value. Learning is performed by generating the estimation model 141 that estimates the state of the degree of long-term care required after a predetermined period from the time corresponding to the care-related information. The period is a period corresponding to the reapplication period for certification for long-term care, and is, for example, 3 years. Since the certification for long-term care can be recognized for a maximum of two years in one application, the server 1 estimates the degree of long-term care required three years after the certification person applies for certification for long-term care, for example. It should be noted that the server 1 only needs to be able to estimate the state of the degree of long-term care after a predetermined period from the time of application, and the period is not limited to 3 years. In addition, the server 1 is capable of estimating the state of the degree of long-term care, such as one year, two years, three years, four years, and five years after the application, so that a plurality of servers corresponding to each period can be estimated. Each of the estimation models 141 may be generated.

For example, as in the example of FIG. 4, the server 1 generates, as the estimation model 141, the estimation model 141A that estimates the transition probability for each degree of long-term care required 3 years after the application. The estimation model 141A is a learned model that receives care-related information at a certain time point and outputs transition probabilities (probabilities) for each degree of care required three years later. The server 1 generates an estimation model 141A by learning a parameter in a machine learning model (decision tree model or the like) using the degree of long-term care required three years after the input care-related information as a correct value. More specifically, the server 1 learns the parameters in the machine learning model so that the transition probability of the correct answer value output when the care-related information is input is high and the transition probability of the other degree of long-term care is low. By doing so, the estimation model 141A is generated.

When the nursing-related information for teachers is data for 2015, the server 1 may use the degree of long-term care required for 2018, which is three years later, as the correct answer value. By generating the estimation model 141A that estimates the transition probability for each degree of long-term care after a predetermined period, the state of the future degree of long-term care of the target person can be estimated in detail and accurately.

Further, the server 1 may generate an estimation model 141B that estimates a change in the degree of long-term care required 3 years after the application. The estimation model 141B is a learned model that receives care-related information at a certain time and outputs a change in the degree of long-term care after 3 years. The change in the degree of long-term care is represented by, for example, a binary value of “worse” and “not deteriorate”, or a three value of “worse”, “not change”, and “improve”. The server 1 prepares a change in the degree of long-term care required (such as “worse” or “not worsen”) by comparing the degree of long-term care required at a certain point in the teacher data three years later, and changes in the degree of long-term care required. The estimated model 141B can be generated by making the machine learning model (decision tree model or the like) learn the parameters with the correct value as.

In addition, the server 1 adds, to the downstream of the estimation model 141A, the transition probabilities for each nursing care level output by the estimation model 141A according to the change in the nursing care level, and adds a layer that outputs the maximum change. By doing so, the estimation model 141B can be generated. For example, when the care-related information of 2015 of a certain target person is input, and the degree of long-term care required in 2015 of the target person is the long-term care required 3, the above layer is the long-term care required 3 output by the estimation model 141A. The following transition probabilities are summed as the probability of "not worsening", the transition probabilities of requiring care of 4 or more are summed as the probability of "being worse", and the one with the larger total may be output as the change in the degree of long-term care after 3 years. .. In this way, by coarsening the granularity of the estimation result, it is possible to more accurately estimate the future state of the degree of need for nursing care of the subject.

The server 1 may also generate an estimation model 141C that estimates the degree of long-term care required 3 years after the application. The estimation model 141C is a learned model that receives care-related information at a certain time and outputs the degree of long-term care required after three years. The server 1 may generate the estimation model 141C by adding a layer that outputs the maximum degree of nursing care required among the transition probabilities for each degree of nursing care output by the estimation model 141A downstream of the estimation model 141A. it can. For example, in the above layer, if the transition probability of the long-term care 3 is the maximum among the transition probabilities for each degree of long-term care output from the estimation model 141A, if long-term care 3 is output as the degree of long-term care 3 years later. Good. In this way, by outputting the degree of long-term care required as the state of the degree of long-term care required, the state of the degree of long-term care required can be estimated specifically and accurately.

As shown in the lower part of FIG. 4, the server 1 adds the output of the estimation model 141 to the seven stages of “necessary care 1” to “necessary care 5”, “support required 1”, and “support required 2”. Therefore, 9 levels may be added by adding 2 levels of “loss” and “not applicable”. “Loss” refers to the case where the certification is lost or the certification for long-term care is lost due to other reasons. Further, “not applicable” indicates a case in which an application for certification for long-term care is made, but as a result of the examination, it is determined that the person is not a certified person. In addition to the seven levels of long-term care required in accordance with laws and regulations, we also made an estimate including cases where the certification for long-term care is lost and cases where it is not applicable Since the estimation can be performed, the estimation accuracy can be further improved. The degree of long-term care need not include “lost” and “not applicable”, or may include only one of them.

The server 1 generates at least one of the above estimation models 141A, 141B, and 141C as the estimation model 141, and uses the generated estimation model 141 to estimate the state of the degree of long-term care required of the subject 3 years later. To do. When the server 1 estimates the state of the degree of long-term care required of the target person by the estimation model 141, the estimation result is output to the terminal 2. The terminal 2 displays the estimation result acquired from the server 1 on a display device (such as a liquid crystal display).

FIG. 5: is explanatory drawing which shows the example of a display screen of the estimation result of a nursing care degree. The terminal 2 receives the output of the estimation result from the server 1 and displays the screen shown in FIG.
The display screen of FIG. 5 is a screen for displaying the state of the degree of long-term care of the target person. The server 1 displays the display screen shown in FIG. 5 on the terminal 2 based on the care-related information of the target person uploaded from the terminal 2 and the estimation result of the state of the future degree of long-term care required based on the care-related information. ..

For example, the terminal 2 displays basic information such as the name and age of the target person, and also uses the estimation model 141B to determine whether the degree of long-term care required at the time of previous certification (current) and whether the degree of long-term care changes after a predetermined period. The estimated estimation result (sign of deterioration) is displayed. In addition, the terminal 2 displays the degree-of-care-necessity estimation graph 51 that represents the transition probabilities for each degree of care-necessity estimated by the estimation model 141A. Further, the terminal 2 displays the statement of opinion information, the visit survey information, the benefit record information, and the like included in the care-related information according to the operation by the user (local government staff).

In addition, although the estimation result of the degree of long-term care is only displayed on the terminal 2 in the above, for example, when the server 1 estimates that there is a high possibility that the degree of long-term care will deteriorate after a predetermined period, the server 2 displays the result on the terminal 2. You may output an alert. For example, if the expected value of the future degree of long-term care required estimated by the estimation model 141A is higher than the current degree of long-term care requirement, the server 1 estimates that the future degree of long-term care required is “deteriorated” by the estimation model 141B. Alternatively, an alert may be output to the terminal 2 when the future degree of long-term care required estimated by the estimation model 141C is higher than the current degree of long-term care required. As a result, it is possible to notify the local government of the target person who should be focused on.

As described above, according to the present embodiment, by inputting the care-related information of the authorized person to the learned estimation model 141, it is possible to accurately estimate the state of the degree of long-term care required of the target person. .. The local government can provide appropriate care to each subject by referring to the estimated state of the degree of long-term care required for each subject.

FIG. 6 is a flowchart showing the procedure of the generation process of the estimation model 141. The contents of the process of performing the machine learning to generate the estimation model 141 will be described with reference to FIG.
The control unit 11 of the server 1 includes teacher data including the care-related information at the first time point of a plurality of authorized persons (persons) and the state (correct value) of the degree of long-term care required for each authorized person at the second time point. Is acquired (step S11). The care-related information includes basic information of the authorized person, written opinion information from the medical institution, visit survey information indicating the visit survey result, and benefit record information indicating the benefit record. The state of the future degree of long-term care required, which is included in the teacher data as a correct value, is the state of degree of long-term care required (for example, three years later) from the time of application for the corresponding long-term care-related information (degree of long-term care or its change). is there.

The control unit 11 uses the teacher data to generate the estimation model 141 (step S12). For example, the control unit 11 performs learning by using the gradient boosting method as described above, and generates a decision tree model that estimates the state of the future degree of long-term care from the care-related information. That is, the control unit 11 sequentially generates the weak discriminator (decision tree) from the gradient of the loss function defined by the error (residual difference) between the predicted value and the actually measured value, and finally generates the estimation model 141. In this case, the control unit 11 uses the state of the degree of long-term care required at the time (second time) when a predetermined period has passed from the application time (first time) corresponding to the care-related information, as a correct answer value, and requires long-term care. An estimation model 141 for estimating the probability that the degree of transition to the state of each degree of long-term care after a predetermined period has been generated is generated. The control unit 11 ends the series of processes.

FIG. 7 is a flowchart showing the procedure of the process of estimating the degree of need for nursing care. Based on FIG. 7, the content of the process of estimating the state of the future degree of need for nursing care from the care-related information of the target person will be described.
The control unit 11 of the server 1 acquires the care-related information of the target person from the terminal 2 of the local government (step S31). The control unit 11 inputs the acquired care-related information of the target person into the estimation model 141 and estimates the state of the future degree of need for care of the target person (step S32). Specifically, the control unit 11 estimates the transition probability of each subject's degree of long-term care after a predetermined period, the change in the degree of long-term care of the subject after a predetermined period, or the degree of long-term care of the subject after the predetermined period. To do. The control unit 11 outputs the estimation result to the terminal 2 (step S33) and ends the series of processes.

In addition, although the state of the degree of long-term care required after a predetermined period from the time of application for certification of long-term care (first time) has been described above, the present embodiment is not limited, and the estimation model 141A or The estimation model 141C may be used to estimate the probability at the time of application, that is, the degree of long-term care required or the degree of long-term care.

As described above, according to the first embodiment, it is possible to accurately estimate the state of the future degree of care required of the target person.

Further, according to the first embodiment, by estimating the change in the degree of long-term care, the state of the future degree of long-term care of the target person can be estimated more accurately.

Further, according to the first embodiment, by estimating the transition probabilities for each degree of long-term care, it is possible to estimate the state of the future degree of long-term care of the target person in detail and accurately.

Further, according to the first embodiment, by estimating the degree of long-term care itself, the state of the future degree of long-term care of the target person can be estimated specifically and accurately.

Further, according to the first embodiment, by estimating whether or not the certification for long-term care is lost (and whether or not it is not applicable), the future degree of long-term care required of the target person More accurately estimate the state of.

Further, according to the first embodiment, by estimating the state of the degree of long-term care required after a predetermined period from the time of application, it is possible to more appropriately support the local government and the like.

(Embodiment 2)
In the present embodiment, a form in which the number of authorized persons in the future is estimated for each degree of long-term care will be described. In addition, the same reference numerals will be given to the same contents as those in the first embodiment and the description thereof will be omitted.
FIG. 8 is an explanatory diagram showing the outline of the second embodiment. FIG. 8 conceptually illustrates how to estimate the number of authorized persons (the number of persons) by the degree of long-term care in the estimation target year based on the care-related information of a plurality of years. The outline of the present embodiment will be described with reference to FIG.

In the present embodiment, the server 1 estimates the number of authorized persons for a certain future year according to the degree of long-term care based on the care-related information for a plurality of years and presents it to the local government. For example, as shown in FIG. 8, the server 1 generates three types of estimation models 141A for estimating the transition probabilities for each degree of long-term care after three years, four years, and five years, From the care-related information for the past three years including the estimation time, estimate the number of certified persons in the estimation target year three years after the estimation time. In the example of FIG. 8, since X month 2018 is the estimation time, based on the care related information of 2016, the care related information of 2017, and the care related information until X month of 2018, the 2021 (estimated) The number of certified people in the target year) is estimated.

The reason why the estimation model 141A for three years later, four years later, and five years later is prepared is that certification for long-term care can be accepted for a maximum of two years with one application. It is sufficient that the number of certified persons in the estimation target year can be estimated from the care-related information of multiple years, and the period from the year including the estimation time point to the estimation target year is not limited to 3 years.

The server 1 performs learning using the care-related information of the authorized person as teacher data, and generates three types of estimation models 141A (three years later, four years later, and five years later estimation models 141A). Keep it. Then, the server 1 estimates the number of persons by degree of nursing care in the estimation target year using the three types of estimation models 141A. That is, the server 1 selects one of the estimation models 141A for three years later, four years later, and five years later according to the application time (first time) of the care-related information of each target person. The care-related information is input to the selected estimation model 141A, and the transition probability for each degree of long-term care in the estimation target year is estimated for each target person.

In this case, since the long-term care authorization is recognized for a maximum of two years, the care-related information of the same target person may be duplicated. In the example of FIG. 8, the care-related information of the target person A is duplicated in 2016 and 2018. In this case, the server 1 selects the care-related information of the target person A in the latest year, and discards (does not use) the old care-related information other than the latest year of the target person A. The server 1 inputs the care-related information of the latest year (2018) of the target person A into the estimation model 141A to estimate the transition probabilities of the target year of the target person A (2021) by degree of long-term care. As a result, it is possible to prevent the same subject from being counted twice.

FIG. 9 is an explanatory diagram showing a display screen example of the result of estimation of the number of authorized persons. The server 1 estimates the number of authorized persons after a predetermined period of time in the local government for each degree of long-term care based on the transition probability for each degree of long-term care estimated for each target person, and outputs the estimation result to the terminal 2. In response to the output from the server 1, the terminal 2 displays the screen shown in FIG. For example, as shown in FIG. 9, the terminal 2 displays a graph 91 by the degree of long-term care required showing the number of authorized persons by degree of long-term care, and also displays the time-series transition of the total number of authorized persons in the local government, etc. Present it to the local government staff).

It should be noted that the terminal 2 not only displays the number of persons classified by long-term care, but also other predicted values that can be predicted according to the estimation result, such as the nursing care cost required to provide the nursing care service to the authorized persons of that number. May be displayed. This makes it possible to more appropriately support the operation of the local government.

In addition, although not specifically described above, not only the number of persons subject to certification for long-term care (number of certified persons) but also certification for long-term care, including “loss” and “not applicable” as output parameters of the estimation model 141A Of course, it is possible to estimate the number of target persons who lose and the number of target persons who are not applicable.

FIG. 10 is a flowchart showing the procedure of the generation process of the estimation model 141A according to the second embodiment.
The control unit 11 of the server 1 manages the care-related information of a plurality of authorized persons (persons) at the first time point, and the state of the degree of long-term care required for each authorized person three years after the first time point (second time point). (Correct answer value) and teacher data for three years later including is acquired (step S201).

The control unit 11 uses the teacher data to generate the estimation model 141 for three years after, which estimates the state of the degree of long-term care after three years (step S202).

Next, the control unit 11 of the server 1 manages a plurality of authorized persons (persons) at a first point in time with respect to nursing care information, and four years after the first point of time for each authorized person (second point). The teacher data for four years later including the degree state (correct value) is acquired (step S203).

The control unit 11 uses the teacher data to generate the estimation model 141 for four years after, which estimates the state of the degree of long-term care after four years (step S204).

Subsequently, the control unit 11 of the server 1 provides the care-related information of the plurality of authorized persons (persons) at the first time point and the need for nursing care of each authorized person five years after the first time point (second time point). The teacher data for five years including the degree state (correct value) is acquired (step S205).

The control unit 11 uses the teacher data to generate the estimation model 141 for five years after, which estimates the state of the degree of long-term care required after five years (step S206). After that, the control unit 11 ends the series of processes.

FIG. 11 is a flowchart showing the procedure of the process of estimating the number of authorized persons. Based on FIG. 11, the content of the process of estimating the number of authorized persons for each degree of long-term care will be described.
The control unit 11 of the server 1 acquires care-related information of a plurality of subjects for a plurality of years (step S221). For example, the control unit 11 acquires the care-related information of each target person submitted to the same local government, and the care-related information for a plurality of years.

The control unit 11 extracts the care-related information of one target person (step S222). The control unit 11 determines whether or not the extracted care-related information of the target person is duplicated, that is, whether or not there are a plurality of care-related information of the target person (step S223). When it determines with overlapping (S223:YES), the control part 11 selects the care related information of the newest year, and abandons care related information other than the latest year (step S224).

After executing NO in step S223 or the process of step S224, the control unit 11 selects one of the estimation models 141A according to the application time point (first time point) of the care-related information of the target person. A selection is made (step S225).

Specifically, the control unit 11 selects the estimation model 141A capable of estimating the transition probability for each degree of long-term care required of the target person in the estimation target year. For example, the control unit 11 selects the estimation model 141A for three years later when the estimation target year is 2021 and the application time of the latest care-related information of the selected target person is 2018. Moreover, the control part 11 selects the estimation model 141A for four years after, when the estimation object year is 2021 year and the application time of the latest nursing care related information of the selected object person is 2017 year. In addition, the control unit 11 selects the estimation model 141A for 5 years later when the estimation target year is 2021 and the application time of the latest care-related information of the selected target person is 2016.

The control unit 11 inputs the care-related information of the target person to the selected estimation model 141A and estimates the transition probability of the target person for each degree of long-term care required in the estimation target year (step S226).

The control unit 11 determines whether or not the transition probabilities by the degree of long-term care required in the estimation target year have been estimated for all the target persons (step S227). When it is determined that the transition probabilities by the degree of long-term care in the estimation target year have not been estimated for all the target persons (S227: NO), the control unit 11 returns the process to step S222. When it is determined that the transition probabilities by the degree of long-term care in the estimation target year have been estimated for all the target persons (S227: YES), the control unit 11 outputs the estimation result output from each estimation model 141A (estimation of each target person). Based on the transition probabilities for each degree of care required in the target year), the number of target persons for each degree of care required in the estimation target year is estimated (step S228). Specifically, the control unit 11 may sum the transition probabilities of each target person for each degree of long-term care in the estimated target year for each degree of long-term care, and perform fractional processing. The control unit 11 outputs the estimation result to the terminal 2 (step S229) and ends the series of processes.

As described above, according to the second embodiment, it is possible to estimate the number of authorized persons for each degree of long-term care required in the estimation target year based on the care-related information of the plurality of persons in the plurality of years. As a result, the number of authorized persons for each degree of long-term care can be accurately estimated. The reason is as follows.

If the current year is X month of 2018, and if the number of certified people in 2021 is estimated based only on the care related information in 2018, application for certification for long-term care will be made in 2017 and done in 2018. Care-related information for those who are not certified will not be reflected in the estimated number of certified persons. As described above, certification for long-term care can be granted for a maximum of two years per application, so there are many such certifications. Therefore, when the number of certified persons in 2021 is estimated based on only the care related information in 2018, the estimation accuracy is reduced by the care related information not reflected in the estimation. In other words, when estimating the number of certified persons in the 2021 year, the accuracy of estimating the number of certified persons can be improved by using the care related information in the 2017 fiscal year together with the 2018 fiscal year.

In addition, it is the X month of 2018, and when estimating the number of certified people in 2021 based only on the care related information in 2018 and 2017, application for certification for long-term care is made in 2016, Care-related information of authorized persons who will reapply after X month will not be reflected in the estimated number of authorized persons. As described above, certification for long-term care can be granted for a maximum of two years per application, so there are many such certifications. Therefore, if the number of certified persons in 2021 is estimated based on only the care-related information in 2018 and 2017, the estimation accuracy will be reduced by the care-related information that is not reflected in the estimation. In other words, when estimating the number of certified persons in 2021, the estimation accuracy of the number of certified persons can be further improved by using the care-related information from 2018 to 2016.

As described above, by using the care-related information of a plurality of years, it is possible to accurately estimate the number of authorized persons for each degree of long-term care in the estimation target year. In particular, when the longest period required for long-term care is Y years, the accuracy of estimating the number of certified persons can be maximized by using the care-related information for (Y+1) years.

Further, according to the second embodiment, the number of certified persons can be estimated more appropriately by preparing a plurality of estimation models 141A according to the period between the application time and the estimation target year.

Further, according to the second embodiment, it is possible to more accurately estimate the number of authorized persons by omitting the redundant portion of the care-related information for the same target person.

Although the case where the number of certified persons is estimated using a plurality of estimation models 141A has been described above, the number of certified persons may be estimated using a plurality of estimation models 141C. In this case, the control unit 11 may estimate the degree of long-term care required for each target person by the estimation model 141C and total the number of target persons for each degree of long-term care required.

(Embodiment 3)
FIG. 12 is a functional block diagram showing a configuration example of the server 1 having the above-described form. When the control unit 11 reads out and executes the program P, the server 1 operates as follows.
The acquisition unit 121 acquires the care-related information of the target person. The estimation unit 122 learns the care-related information of the target person, the care-related information of the plurality of persons at the first time point, and the degree of need for care of the person at the second time point after a predetermined period from the first time point. A future state of the degree of need for nursing care of the subject is estimated based on the already learned model.

The embodiments disclosed this time are to be considered as illustrative in all points and not restrictive. The scope of the present invention is shown not by the above meaning but by the scope of the claims, and is intended to include meanings equivalent to the scope of the claims and all modifications within the scope.

1 server (information processing device)
11 control unit 12 main storage unit 13 communication unit 14 auxiliary storage unit P program 141 estimation model 142 local government DB
143 Target DB

Claims (7)

Regarding the degree of long-term care required for the certified person who has been certified as requiring long-term care, the judgment result of the doctor regarding the condition of the certified person, the judgment result of the visiting researcher regarding the state of the certified person, and the care of the certified person Machine learning using the teacher data in which the nursing care related information including at least one of the payment results of the benefits and the degree of long-term care of the authorized person at the second time point after the predetermined time period from the first time point are associated with each other. And a storage unit that stores a learned model that receives the care-related information at the first time point as an input and outputs the degree of long-term care at the second time point,
An acquisition unit that acquires the care-related information of the subject at the present time,
The care-related information acquired by the acquiring unit and input to the learning Sumimo del stored in the storage unit, the predetermined from the moment of the subject by the trained model obtained the level of care required for outputting An information processing device, comprising: an estimation unit that estimates the degree of long-term care required after the period . The learned model, the outputs of the level of care required by the transition probability after the predetermined period, the information processing apparatus according to claim 1. The learned model outputs the change in level of care required after the predetermined period, the information processing apparatus according to claim 1. The information processing apparatus according to any one of claims 1 to 3 , wherein the degree of long-term care includes at least one of a loss of authorization for long-term care and non-applicability. Regarding the degree of long-term care required for the certified person who has been certified as requiring long-term care, the judgment result of the doctor regarding the condition of the certified person, the judgment result of the visiting researcher regarding the state of the certified person, and the care of the certified person Machine learning using the teacher data in which the nursing care related information including at least one of the payment results of the benefits and the degree of long-term care of the authorized person at the second time point after the predetermined time period from the first time point are associated with each other. Is performed, using the learned model that receives the care-related information at the first time point as an input and outputs the degree of long-term care at the second time point,
To get the care-related information of the subject at the present time,
The process of inputting the acquired care-related information into the learned model and estimating the degree of long-term care of the target person after the predetermined period from the present time by obtaining the degree of long-term care required output by the learned model An information processing method characterized by being executed by. Regarding the degree of long-term care required for the certified person who has been certified as requiring long-term care, the judgment result of the doctor regarding the condition of the certified person, the judgment result of the visiting researcher regarding the state of the certified person, and the care of the certified person Machine learning using the teacher data in which the nursing care related information including at least one of the payment results of the benefits and the degree of long-term care of the authorized person at the second time point after the predetermined time period from the first time point are associated with each other. Is performed, using the learned model that receives the care-related information at the first time point as an input and outputs the degree of long-term care at the second time point,
To get the care-related information of the subject at the present time,
The process of inputting the acquired care-related information into the learned model and estimating the degree of long-term care of the target person after the predetermined period from the present time by obtaining the degree of long-term care required output by the learned model A program characterized by being executed by. Regarding the degree of long-term care required for the certified person who has been certified as requiring long-term care, the judgment result of the doctor regarding the condition of the certified person, the judgment result of the visiting researcher regarding the state of the certified person, and the care of the certified person Machine learning using the teacher data in which the nursing care related information including at least one of the payment results of the benefits and the degree of long-term care of the authorized person at the second time point after the predetermined time period from the first time point are associated with each other. is a learning Sumimo Dell has been made,
On the computer,
Accepts as input the care-related information of the subject at the present time,
A learned model characterized by causing a process of outputting a degree of long-term care required of the target person after the predetermined period of time , based on the inputted care-related information.

Images