Information processing method, information processing system, information processing apparatus, and program

Abstract

To grasp the changes in future ADL evaluation of a care receiver, and utilize them for maintaining or improving health condition of the care receiver.SOLUTION: An information processing method causes a processor to execute: a probability calculation step of outputting, for each of a plurality of ADL prediction models, probability of the changes in ADL evaluation of a person to be evaluated after a predetermined period, as ADL change probability, using an input information group related to the person to be evaluated, as input; and a prognostic evaluation prediction step of predicting the changes in ADL evaluation of the person to be evaluated after the predetermined period, using the plurality of ADL change probabilities output for each ADL prediction model, and storing the changes as ADL prognostic prediction results. The input information group includes basic attributes of the person to be evaluated and a plurality of features regarding health condition. The ADL prediction model is a model that has learned correlations between past input information group regarding a plurality of persons to be evaluated and the changes in ADL evaluation after the predetermined period.SELECTED DRAWING: Figure 7

Description

Technical Field

[0001] The present disclosure relates to an information processing method, an information processing system, an information processing apparatus, and a program.

Background Art

[0002] Social security costs related to nursing care in Japan have been continuously increasing, and the government is aiming to transform into a sustainable social security system by promoting self-support and prevention of exacerbation. In the future, the long-term care insurance system will accelerate the shift to an outcome-based reward system that emphasizes results, and care facilities will need to improve the quality of rehabilitation in order to achieve results. However, there are few care facilities that have rehabilitation professionals, and especially in day care, about 90% are in a situation where it is difficult to provide "effective rehabilitation" due to the absence of rehabilitation professionals. In order to propose effective rehabilitation, it is important to appropriately grasp the condition and problems of the care recipient. As an index for evaluating the condition of the care recipient, Activities of Daily Living (ADL) is generally used.

[0003] ADL (Activities of Daily Living) refers to the daily actions that are minimally necessary to live a daily life. The evaluation items of ADL include (1) eating, (2) transfer (transfer to the toilet, transfer to the bed, etc.), (3) grooming (brushing teeth, shaving, applying makeup, etc.), (4) toilet actions, (5) bathing (washing the body, washing hair, transferring to the bathtub, etc.), (6) walking (walking with a cane or a walker, etc.), (7) ascending and descending stairs, (8) changing clothes (putting on and taking off pants, jackets, underwear, etc.), (9) bowel control, and (10) bladder control. These 10 items are also called the Barthel Index (BI). In the medical and nursing care fields, it is grasped how much the care recipient, who is the user, can perform these daily living actions by themselves, and the service plan (contents to be supported) is considered.

[0004] In response to such problems, Patent Document 1 discloses a technique for estimating daily living activities and the like from limited movements of a subject by photographing a predetermined movement of the subject to obtain a movement video, extracting the movement of the subject from the movement video, assigning a reference movement value corresponding to the movement data to generate a subject reference movement value, assigning a reference body information value corresponding to predetermined body information of the subject to generate a subject reference body information value, combining the subject reference body information value with the subject reference movement value to estimate the motor function of the subject, and outputting an estimation result.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] However, although the technique disclosed in Patent Document 1 can know the current state and problems of the care recipient, it has not been studied for appropriately predicting and taking preventive measures against the future state and problems of the care recipient.

[0007] Therefore, the present disclosure has been made to solve the above problems, and an object thereof is to provide an information processing method, an information processing system, an information processing apparatus, and a program for grasping changes in future ADL evaluations of a care recipient and using them to maintain or improve the health state of the care recipient.

Means for Solving the Problems

[0008] In order to achieve the above object, an information processing method according to the present invention is an information processing method executed by an information processing apparatus including a processor and a storage unit. The processor is caused to execute a probability calculation step of outputting, for each of a plurality of ADL prediction models, a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period as an ADL change probability, using, as an input, an input information group regarding the evaluation target person, and a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction result as an ADL prognosis prediction result. The input information group is a plurality of feature amounts regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period for a plurality of past evaluation target persons. In order to achieve the above object, an information processing system according to the present invention is an information processing system including an information processing apparatus including a processor and a storage unit and a user terminal. By executing a program, a probability calculation step of outputting, for each of a plurality of ADL prediction models, a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period as an ADL change probability, using, as an input, an input information group regarding the evaluation target person, and a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction result as an ADL prognosis prediction result are performed. The input information group is a plurality of feature amounts regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period for a plurality of past evaluation target persons. In order to achieve the above object, in the information processing apparatus according to the present invention, by executing a program, an input information group regarding an evaluation target person is used as an input, and a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period is output for each of a plurality of ADL prediction models as an ADL change probability, and a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction result as an ADL prognosis prediction result are performed. The input information group is a plurality of feature amounts regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between the input information group regarding a plurality of past evaluation target persons and the change in the evaluation of the ADL after a predetermined period. In order to achieve the above object, the program according to the present invention is a program for causing a computer including a processor and a storage unit to execute. The program causes the processor to use an input information group regarding an evaluation target person as an input, output a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period as an ADL change probability for each of a plurality of ADL prediction models, and a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction result as an ADL prognosis prediction result are executed. The input information group is a plurality of feature amounts regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between the input information group regarding a plurality of past evaluation target persons and the change in the evaluation of the ADL after a predetermined period.

Advantages of the Invention

[0009] According to the present disclosure using the above means, it is possible to take measures early by grasping the change in the future ADL evaluation of the care recipient, improve the quality of rehabilitation by proposing rehabilitation according to the change in the future ADL evaluation of the care recipient, and contribute to the maintenance or improvement of the health status of the care recipient.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Embodiments for Carrying Out the Invention

[0011] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0012] <System Configuration> The information processing system 1 in the present disclosure predicts how the ADL evaluation of an evaluation target person managed by a user will change in the future (prognosis), and proposes a rehabilitation menu for improving or maintaining the ADL evaluation of the evaluation target person based on the prognosis prediction result. It is an information processing system that provides an ADL prognosis prediction service.

[0013] Users who use the ADL prognosis prediction service are mainly businesses, local governments, individuals, etc. that provide care and assistance services to the evaluation target person. In the following description, the user will be described as a care business that provides care services. The evaluation target person is mainly a user (care recipient) who uses the care services of the user and is an elderly person.

[0014] FIG. 1 is a diagram showing the overall configuration of the information processing system 1 according to an embodiment of the present disclosure. The information processing system 1 includes information processing devices such as an information processing server 10 and a user terminal 20 connected via a network NW.

[0015] Each information processing device is configured by a computer including an arithmetic device (processor) and a storage device. The basic hardware configuration of the computer and the basic functional configuration of the computer realized by the hardware configuration will be described later.

[0016] The network NW is a communication network such as the Internet, VPN (Virtual Private Network), intranet, short-range wireless communication, etc. For simplicity of explanation, FIG. 1 shows one user terminal 20, but the information processing server 10 can be connected to two or more user terminals 20 via the network NW. Also, one user terminal 20 may be used by multiple users.

[0017] The operator of the ADL prognosis prediction service uses the information processing system 1 to provide the ADL prognosis prediction service to users. Specifically, the information processing system 1 provides an application that operates on the user terminal 20 used by the user. The user uses the application to access the information processing server 10 and requests the generation of an ADL prognosis prediction result that predicts the future change in the ADL evaluation of the person to be evaluated. In response to the user's request, the information processing server 10 performs a prognosis prediction of the ADL evaluation of the person to be evaluated using the information regarding the attributes and health status of the person to be evaluated stored in the information processing server 10, and generates an ADL prognosis prediction result (hereinafter sometimes simply referred to as the prognosis prediction result). Further, the information processing server 10 selects an appropriate rehabilitation menu based on the ADL prognosis prediction result. The ADL prognosis prediction result and the selected rehabilitation menu are transmitted to the user terminal 20.

[0018] Figure 2 is a block diagram showing the functional configuration of the information processing server 10. The information processing server 10 (hereinafter sometimes simply referred to as the server 10) is an information processing device that provides the ADL prognosis prediction service. The server 10 includes a storage unit 101 and a control unit 103.

[0019] <Configuration of the storage unit of the information processing server> The storage unit 101 of the server 10 is a storage medium or storage device that stores information used for providing the ADL prognosis prediction service by the information processing system 1. Although shown integrated with the server 10 in FIG. 2, it may be a storage device physically independent of the server 10. The storage unit 101 includes an application program 1011, a user information table 1012, a person-to-be-evaluated table 1013, a learning information table 1014, a menu table 1015, an ADL prediction model 1021, and a menu proposal model 1022. The storage unit 101 also stores other information that does not belong to these storage units.

[0020] FIG. 3 is a diagram showing an example of information stored in the storage unit 101 of the information processing server 10. Hereinafter, each table in the storage unit 101 will be described with reference to FIG. 3.

[0021] The user information table 1012 is a table that stores and manages information of users (user members) who use the ADL prognosis prediction service. The user information table 1012 includes, as items of information to be stored, a user ID and user basic information. By registering for use of the service, the information of the user is stored in a new record of the user information table 1012. As a result, the user can use the ADL prognosis prediction service according to the present disclosure. The user ID is an item that stores user identification information for identifying a user. The user identification information is an item for which a unique value is set for each user. The user basic information is an item that stores basic information such as the name, address, contact information, etc. of the user.

[0022] The evaluation target person table 1013 is a table that stores and manages information of evaluation target persons who are targets for evaluating ADL using the ADL prognosis prediction service. The evaluation target person table 1013 includes, as information to be stored, an evaluation target person ID, evaluation target person basic information, scientific care indicators, day care nursing plan information, functional training plan information, various evaluation information, feature data (input information group), ADL prediction information, and menu proposal information. Among these, the evaluation target person basic information, scientific care indicators, day care nursing plan information, functional training plan information, and various evaluation information are mainly information registered, edited, and managed by users of the ADL prognosis prediction service, and are collectively referred to as user management information. In the present embodiment, a care business operator who is a user registers the user management information of the evaluation target person on the application of the ADL prognosis prediction service, but it may be possible to link and refer to the already registered similar information of the evaluation target person registered in another system or service used by the care business operator.

[0023] The subject ID is an item that stores the subject identification information for identifying the subject. The subject identification information is an item for which a unique value is set for each subject.

[0024] The basic information of the subject includes the name, gender, date of birth, and management user ID of the subject. The management user ID stores the user ID of the care business operator to which the subject belongs.

[0025] The scientific care indicators include the evaluations of the subject's degree of care need, bedridden degree, cognitive ability, ADL, nutrition, oral cavity, dementia, DBD, motivation, etc., and the information on the evaluation dates. The scientific care indicators are the data submitted as the calculation requirements of the "Acceleration of the Scientific Care Promotion System" of the Scientific Care Information System (LIFE) operated by the Ministry of Health, Labour and Welfare. DBD (Dementia Behavior Disturbance Scale) is a cognitive and behavioral disorder scale. For each ADL evaluation item, the actual ADL evaluation results, which are the actual evaluation values, are stored. The ADL evaluation items in this embodiment are (1) diet, (2) transfer, (3) grooming, (4) toilet behavior, (5) bathing, (6) walking, (7) stair climbing and descending, (8) dressing, (9) bowel control, and (10) bladder control, a total of 10 items.

[0026] The day care nursing plan information includes the information on the subject's health status, medical risks and precautions in care, and the background up to the use of day care nursing.

[0027] The functional training plan information includes the disease name, co-existing disease name, precautions in implementing functional training, program frequency, program implementation time, program support code, etc., of the subject.

[0028] The various evaluation information includes the information on the evaluation of the subject's interests and the information on the evaluation of physical functions. The information on the evaluation of physical functions includes the left and right grip strength, sit-to-stand, long-sitting forward bend, left and right single-leg standing, TUG (Timed Up & Go Test), left and right FRT (Functional Reach Test), 5m walking.

[0029] Feature data is a set of various features obtained by converting the characteristics of the person to be evaluated into a form that allows statistical or analytical processing such as numerical, symbolic, or categorical data. The features included in the feature data are classified into feature basic information, scientific care information, disease information, functional training information, interest information, and physical function information. The feature data is generated by performing data preprocessing described later using the above-mentioned user management information of the person to be evaluated.

[0030] The feature basic information includes one or more features for each of the age group, gender, care level, bedridden degree, and awareness level of the person to be evaluated. The scientific care information includes one or more features for each of the ADL, BMI, oral cavity, cognition, and motivation of the person to be evaluated. ADL includes features for each evaluation item. For the evaluation item "transfer", the evaluation can be any value from a four-level evaluation of "independent", "under supervision", "can sit but cannot move", and "total assistance". For the evaluation item "walking on flat ground", the evaluation can be any value from a four-level evaluation of "independent", "walker, etc.", "wheelchair operation possible", and "total assistance". For other evaluation items, the evaluation can be any value from a three-level evaluation of "independent", "partial assistance", and "total assistance". The disease information includes one or more features for each of the disease and number of diseases of the person to be evaluated. The functional training information includes one or more features for each of the training content and time, and total training time of the person to be evaluated. The interest information includes one or more features for each of the type of interest and number of contents of the person to be evaluated. For the interest information, for example, the evaluation results of the "Interest and Concern Check Sheet" provided by the Ministry of Health, Labour and Welfare can be used. The evaluation items for interest and concern include daily life, housework, hobbies, sports, social participation, etc. The number of contents is an index indicating the degree of interest and concern step by step, and is selected from, for example, "doing", "wanting to try", "interested", and "not applicable". It is known as an empirical rule that the amount and degree of the person to be evaluated's interest and concern affect the maintenance and improvement of the ADL evaluation. The physical function information includes one or more feature quantities for each of the physical function evaluation flag, left and right grip strength, sitting and standing, long-sitting forward bend, left and right single-leg standing, TUG, left and right FRT, and 5m walking of the person to be evaluated.

[0031] The ADL prediction information includes the prediction execution date indicating the date and time when the ADL prediction process described later was executed for the person to be evaluated, the prediction results generated by the process, and information on SHAP values. The prediction results include information on the ADL increase probability, ADL decrease probability, ADL prognosis prediction results (improvement, maintenance, deterioration) for each evaluation item, and the comprehensive prediction result.

[0032] The menu proposal information includes items such as the proposal execution date indicating the date and time when the menu proposal process described later was executed for the person to be evaluated, and the proposed menu output by the process. The proposed menu is stored as the menu ID in the menu table 1015.

[0033] The learning information table 1014 is a table that stores and manages the teacher data for generating the ADL prediction model and menu proposal model, and the learning history of the models. The learning information table 1014 includes items such as teacher data for the ADL prediction model, teacher data for the menu proposal model, and learning history as items of the information to be stored.

[0034] The menu table 1015 is an item in which information on rehabilitation menus to be proposed according to the ADL prognosis prediction results of the person to be evaluated is stored. The menu table 1015 includes items such as menu ID, menu basic information, and rehabilitation content as items of the information to be stored. The menu ID is an item that stores menu identification information for identifying rehabilitation menus. The menu identification information is an item in which a unique value is set for each rehabilitation menu. The menu basic information is an item that stores basic information on rehabilitation menus. Specifically, it includes the name of the rehabilitation, the target part, the difficulty level, the number of times or sets to be performed, precautions, other remarks, etc. Rehabilitation content is an item that stores content explaining rehabilitation content composed of images, explanatory text, videos, explanatory audio, or combinations thereof related to a rehabilitation menu.

[0035] Returning to FIG. 2, each model of the storage unit 101 will be described.

[0036] The ADL prediction model 1021 is a model that infers changes in the evaluation of the ADL of an evaluation target person after a predetermined period using the feature data (input information group) of the evaluation target person as input data. Details of the ADL prediction process using the ADL prediction model 1021 will be described later.

[0037] The ADL prediction model 1021 of the present embodiment is, for example, one kind or a combination thereof such as machine learning, artificial intelligence, deep learning model, rule-based, etc. The ADL prediction model 1021 does not necessarily have to be a single model, and may be realized by combining or switching a plurality of independent models of a plurality of types. The ADL prediction model 1021 is generated for each ADL evaluation item. Also, for each ADL evaluation item, an ADL increase prediction model and an ADL decrease prediction model are generated. The ADL increase prediction model outputs, as a value between 0 and 1, the probability that the change in the evaluation of the ADL after a predetermined period is "increase (improvement)". The ADL decrease prediction model is a model that outputs, as a value between 0 and 1, the probability that the change in the evaluation of the ADL after a predetermined period is "decrease (deterioration)". For example, for the "facial appearance" item of the ADL, a facial appearance increase model and a facial appearance decrease model are generated. That is, in the present embodiment, a total of 20 models of increase and decrease models are generated for the 10 items of the ADL, and each is used in the ADL prediction process described later. Also, the ADL prediction model 1021 does not necessarily have to be a single model, and may be realized by combining or switching a plurality of independent models of a plurality of types.

[0038] The menu proposal model 1022 is a model that selects and outputs one or more rehabilitation menus using the ADL prognosis prediction result of the person to be evaluated, the SHAP value described later in the ADL prognosis prediction result, and the feature importance of the ADL prediction model described later as input data. The menu proposal model 1022 is, for example, one type or a combination thereof such as machine learning, artificial intelligence, deep learning model, rule-based, etc. The menu proposal model 1022 does not necessarily have to be a single model, and may be realized by combining or switching a plurality of independent models of a plurality of types.

[0039] <Configuration of the control unit of the information processing server> The control unit 103 of the server 10 includes, as functional units, a model management unit 1031, an ADL prediction unit 1032, a menu proposal unit 1033, a data preprocessing unit 1034, and an input / output unit 1035. The control unit 103 is a computer equipped with a CPU (processor). The control unit 103 realizes each functional unit by executing the application program 1011 stored in the storage unit 101. The control unit 103 executes the ADL prediction model learning process, menu proposal model learning process, ADL prediction process, menu proposal process, and data preprocessing described later by coordinating each functional unit. Although not shown, the control unit 103 also has a function of controlling communication with external devices connected to the server 10 via a network NW or the like.

[0040] The model management unit 1031 generates and updates each model in the storage unit 101. Note that the generation and update of the model include not only new learning but also re-learning. Details of the learning process of the ADL prediction model and the learning process of the menu proposal model will be described later.

[0041] The ADL prediction unit 1032 mainly performs ADL prediction processing to predict the changes in ADL evaluation after a predetermined period for each ADL evaluation item of the person to be evaluated from the feature amount data of the person to be evaluated using the ADL prediction model 1021, generates an ADL prognosis prediction result, calculates its SHAP value, and stores the result in the ADL prediction information in the evaluation target person table 1013 of the storage unit 101. Basically, the latest data group is used for the feature amount data, but the user may specify a predetermined time and extract and use the data group closest to that time. Details of the ADL prediction processing will be described later.

[0042] The menu proposal unit 1033 executes menu proposal processing to select and output one or more rehabilitation menus using the menu proposal model 1022 based on the ADL prognosis prediction result of the person to be evaluated, the SHAP value, and the feature amount importance (to be described later) of the ADL prediction model. Details of the menu proposal processing will be described later.

[0043] The data preprocessing unit 1034 executes data preprocessing to generate feature amount data from the user management information in the evaluation target person table 1013 of the storage unit 101. The feature amount data is used as data input to the ADL prediction model and as teacher data for the ADL prediction model. Details of the data preprocessing will be described later.

[0044] The input / output unit 1035 outputs the ADL prognosis prediction result and the selected rehabilitation menu to the user terminal 20. An example of the screen output to the user terminal 20 will be described later.

[0045] Figure 4 is a block diagram showing the functional configuration of the user terminal 20. The user terminal 20 includes a storage unit 201, a control unit 203, an input device 205, and an output device 207. The user terminal 20 is an information processing device operated by a user who uses the service. The user terminal 20 may be, for example, a mobile terminal such as a smartphone or a tablet, or may be a stationary PC (Personal Computer) or a laptop PC. Further, it may be a wearable terminal such as an HMD (Head Mount Display) or a wristwatch-type terminal.

[0046] <Configuration of the storage unit of the user terminal> The storage unit 201 of the user terminal 20 is a storage medium or a storage device that can store information used for providing the ADL prognosis prediction service by the information processing system 1. Although it is shown integrated with the user terminal 20 in FIG. 4, it may be a storage device physically independent of the user terminal 20. The storage unit 201 includes an application program 2011 and a user ID 2012. The storage unit 201 also stores other information that does not belong to these storage units.

[0047] The application program 2011 may be pre-stored in the storage unit 201, or may be configured to be downloaded from a web server or the like operated by a service provider via a communication IF. The application program 2011 includes applications such as a web browser application. The application program 2011 includes an interpreter-type programming language such as JavaScript (registered trademark) that is executed on a web browser application stored in the user terminal 20.

[0048] The user ID 2012 stores the user ID of the user who uses the user terminal 20 as identification information for the user to use the ADL prognosis prediction service. The user transmits the user ID from the user terminal 20 to the server 10. The server 10 identifies the user by querying the received user ID in the user information table 1012, and provides the ADL prognosis prediction service according to the present disclosure to the user. Note that the user ID 2012 may include information such as a session ID temporarily assigned from the server 10 when identifying the user who uses the user terminal 20.

[0049] <Configuration of the control unit of the user terminal> The control unit 203 of the user terminal 20 includes an input control unit 2031 and an output control unit 2032 as functional units. The control unit 203 is a computer equipped with a CPU. The control unit 203 realizes each functional unit by executing the application program 2011 stored in the storage unit 201. Although not shown, the control unit 203 also has a function of controlling communication with an external device connected to the user terminal 20 via a network NW or the like.

[0050] The input control unit 2031 controls the input device 205. The input device 205 of the user terminal 20 includes a camera 2051, a microphone 2052, a position information sensor 2053, a motion sensor 2054, a touch device 2055, and the like. The input device 205 is various input devices that enable input and selection of information by the user's operation. The input and selection of information are, for example, input by entering numerical values, characters, and symbols, selection of items, and input for determining processing.

[0051] The output control unit 2032 controls the output device 207. The output device 207 of the user terminal 20 includes a display 2071, a speaker 2072, and the like. The output device 207 is an output device that can display, reproduce, and notify information to the user.

[0052] <ADL prediction model learning process> FIG. 5 is a flowchart showing the operation of the learning process of the ADL prediction model. The learning process of the ADL prediction model is mainly a series of processes for learning a set of feature quantity data of an evaluation target person at a predetermined time and the change in the ADL evaluation (for example, improvement, maintenance, deterioration in three stages) of the ADL after a predetermined period has elapsed from the predetermined time of the evaluation target person as teacher data. The predetermined period is the submission interval of LIFE, and it is preferably about six months when the effects of rehabilitation (functional training) and the PDCA cycle of countermeasures appear to a certain extent, but it can also be an integral multiple thereof. Note that the data group of the measured values of the change in the ADL evaluation after a predetermined period has elapsed from the predetermined time of the evaluation target person often becomes unbalanced data that does not follow a normal distribution. For example, when the predetermined period is from several months to several years and the evaluation is performed in three stages of improvement, maintenance, and deterioration, maintenance is about 90%, and improvement and deterioration are each about 5%. The longer the predetermined period, the more improvement and maintenance decrease, and deterioration increases. Hereinafter, the operation of the learning process of the ADL prediction model will be described according to the same flowchart. The learning process of the ADL prediction model is executed separately for each ADL evaluation item. In addition, the learning process of the ADL prediction model is executed separately for each of the ADL increase prediction model and the ADL decrease prediction model. Note that basically the same dataset can be used for the ADL increase prediction model and the ADL decrease prediction model. As the teacher data used for the ADL increase prediction model, a set in which the ADL evaluation result is "improvement" may be used. Also, as the teacher data used for the ADL decrease prediction model, a set in which the ADL evaluation result is "deterioration" may be used. Hereinafter, as an example, the process of the ADL increase prediction model for the ADL evaluation item "facial appearance" will be described as being executed.

[0053] In step S101, the model management unit 1031 of the server 10 acquires the teacher data (learning data) for the ADL prediction model in the learning information table 1014 of the storage unit 101.

[0054] In step S102, the model management unit 1031 generates an ADL increase prediction model for the ADL item "cosmetic surgery" using the teacher data for the ADL prediction model obtained in step S101, and stores it as the ADL prediction model 1021 in the storage unit 101. In the case of re-learning, the ADL prediction model 1021 is updated.

[0055] In step S103, the model management unit 1031 calculates the feature importance for the ADL prediction model 1021 of the ADL item "cosmetic surgery" generated in step S102. The feature importance is the contribution rate of each feature in the inference calculated globally from the entire learning data. The feature importance is calculated for each of the ADL increase prediction model and the ADL decrease prediction model. That is, for the ADL increase prediction model, the increase feature importance is calculated for each feature of the feature data, and is stored in the ADL prediction model 1021 in the storage unit 101 as the attached information of the generated ADL increase prediction model for the ADL item "cosmetic surgery". Similarly, the model management unit 1031 calculates the decrease feature importance for each feature of the feature data for the ADL decrease prediction model of the ADL item "cosmetic surgery", and stores it in the ADL prediction model 1021 in the storage unit 101 as the attached information of the generated ADL decrease prediction model for the ADL item "cosmetic surgery". As the means for calculating the feature importance (increase feature importance and decrease feature importance), an appropriate method can be selected and used from the Embedded Methods, Wrapper Methods, Filter Methods, Permutation Importance, etc. according to the machine learning model and data. Then, this flow ends.

[0056] <Menu Proposal Model Learning Process> FIG. 6 is a flowchart showing the operation of the learning process of the menu proposal model. The learning process of the menu proposal model is mainly a series of processes for training, as teacher data, a set of the user management information at a predetermined past time point related to the performance of the evaluation target person who has a past record of maintaining / improving ADL, and the rehabilitation implementation history from the predetermined time point related to the performance to after a predetermined period has elapsed. The operation of the menu proposal model learning process will be described below according to the same flowchart. The menu proposal model learning process is executed for each evaluation item of ADL. In the following, it will be described as if the process is being executed for the evaluation item "cosmetic surgery" of ADL as an example.

[0057] In step S201, the model management unit 1031 of the server 10 acquires the teacher data (learning data) for the menu proposal model from the learning information table 1014 of the storage unit 101. Specifically, it acquires, as teacher data, a set of the evaluation of "cosmetic surgery" at a predetermined past time point of the evaluation target person, the rehabilitation implementation history from the predetermined time point to after a predetermined period has elapsed, and the evaluation of "cosmetic surgery" after the predetermined period has elapsed.

[0058] In step S202, the model management unit 1031 generates a menu proposal model for the ADL item "cosmetic surgery" using the teacher data for the menu proposal model acquired in step S101, stores it as the menu proposal model 1022 in the storage unit 101, and ends the flow. In the case of re-learning, the menu proposal model 1022 is updated.

[0059] <ADL Prediction Process> FIG. 7 is a flowchart showing the operation of the ADL prediction process. The ADL prediction process is mainly a series of processes for predicting the change in the ADL evaluation of the evaluation target person after a predetermined period using the ADL prediction model 1021 from the feature amount data of the evaluation target person, and generating and outputting the ADL prognosis prediction result. The predetermined period is, for example, 6 or 7 months from the time when the most recent ADL evaluation was performed on the evaluation target person, such as immediately after admission or subsequent regular evaluations. The operation of the ADL prediction process will be described below according to the same flowchart. Steps S302 to S304 of the ADL prediction process are executed for all evaluation items for each evaluation item of ADL. In the following description, for steps S302 to S304, it will be described as if the process is being executed for "cosmetic surgery" as an example among the 10 items of ADL.

[0060] In step S301, the ADL prediction unit 1032 of the server 10 acquires, as input data to be used for predicting the ADL, the feature amount data (input information group) associated with the subject to be evaluated in the evaluation target person table 1013 of the storage unit 101.

[0061] In step S302, the ADL prediction unit 1032 inputs the feature amount data acquired in step S301 into the ADL prediction model 1021 for "cosmetic surgery", and as output, acquires an ADL change probability which is the probability regarding the change in the evaluation of "cosmetic surgery" after a predetermined period has elapsed (probability calculation step). Specifically, the ADL prediction unit 1032 inputs the feature amount data acquired in step S301 into the ADL increase prediction model for "cosmetic surgery", and as output, acquires the ADL increase probability of "cosmetic surgery" after a predetermined period has elapsed (increase probability calculation step). Then, it stores it in association with the subject to be evaluated in the ADL prediction information of the evaluation target person table 1013. Also, the ADL prediction unit 1032 inputs the feature amount data acquired in step S301 into the ADL decrease prediction model, and as output, acquires the ADL decrease probability of "cosmetic surgery" after a predetermined period has elapsed (decrease probability calculation step). Then, it stores it in association with the subject to be evaluated in the ADL prediction information of the evaluation target person table 1013 of the storage unit 101.

[0062] In step S303, the ADL prediction unit 1032 predicts the change in the ADL evaluation of "cosmetic surgery" for the subject to be evaluated after a predetermined period has elapsed based on the ADL change probability of "cosmetic surgery" acquired in step S302, and stores it in association with the subject to be evaluated in the ADL prediction information of the storage unit 101 as the ADL prognosis prediction result for "cosmetic surgery" (prognosis evaluation prediction step). Specifically, the ADL prediction unit 1032 predicts the change in the ADL evaluation of "cosmetic surgery" for the subject to be evaluated after a predetermined period has elapsed based on the ADL increase probability and the ADL decrease probability of "cosmetic surgery" acquired in step S302, and stores it in association with the subject to be evaluated in the ADL prediction information of the storage unit 101 as the ADL prognosis prediction result for "cosmetic surgery". The ADL prognosis prediction result is any one of "improvement", "maintenance", and "deterioration" regardless of the ADL evaluation item.

[0063] In step S304, the ADL prediction unit 1032 calculates the SHAP value for each feature of the input feature data with respect to the ADL change probability of "cosmetic surgery" obtained in step S302, and stores it in the ADL prediction information of the evaluation target person table 1013 as the SHAP value of the ADL change probability of "cosmetic surgery" for the evaluation target person (SHAP value calculation step). Specifically, the ADL prediction unit 1032 calculates the SHAP value for each feature of the input feature data with respect to the ADL increase probability of "cosmetic surgery" obtained in step S302, and stores it in the ADL prediction information of the evaluation target person table 1013 as the SHAP value (first SHAP value) of the ADL increase probability of "cosmetic surgery" for the evaluation target person. Further, the ADL prediction unit 1032 calculates the SHAP value for each feature of the input feature data with respect to the ADL decrease probability of "cosmetic surgery" obtained in step S302, and stores it in the ADL prediction information of the evaluation target person table 1013 as the SHAP value (second SHAP value) of the ADL decrease probability of "cosmetic surgery" for the evaluation target person. Note that SHAP (SHapley Additive exPlanations) calculates how much the predicted value calculated by the machine learning model is affected by each variable by using the Shapley value used in cooperative game theory.

[0064] In step S305, the ADL prediction unit 1032 determines whether the processing from step S302 to step S304 has been completed for all evaluation items. If the determination is true, that is, if it is determined that the processing has been completed for all evaluation items, the process proceeds to step S306. On the other hand, if the determination is false, that is, if it is determined that the processing has not been completed for all evaluation items, the process returns to step S302 and continues the processing for the unprocessed evaluation items.

[0065] In step S306, the ADL prediction unit 1032 generates a comprehensive prediction result from the ADL prognosis prediction results of all evaluation items of the evaluation target person generated in step S303 after a predetermined period (comprehensive prediction result generation step). The comprehensive prediction result is, for example, "deterioration", "maintenance", or "improvement" corresponding to the average score of the three-level evaluation of the ADL prognosis prediction result.

[0066] In step S307, the ADL prediction unit 1032 presents the ADL prognosis prediction result of the subject to be evaluated after a predetermined period generated in step S303 and the comprehensive prediction result of the subject to be evaluated after a predetermined period generated in step S306 to the user terminal 20 via the input / output unit 1035.

[0067] <Menu Proposal Process> The menu proposal process is mainly a process for selecting one or more rehabilitation menus using the menu proposal model 1022 from the ADL prognosis prediction result of the subject to be evaluated, the SHAP value in the ADL increase probability, the SHAP value in the ADL decrease probability, and the feature importance of the ADL prediction model, and presenting them to the user terminal 20. FIG. 8 is a flowchart showing the operation of the menu proposal process. The operation of the menu proposal process will be described below according to the flowchart.

[0068] In step S401, the menu proposal unit 1033 of the server 10 refers to the feature importance of the ADL increase prediction model and the ADL decrease prediction model of all evaluation items stored in the ADL prediction model 1021 of the storage unit 101.

[0069] In step S402, the menu proposal unit 1033 refers to the ADL increase probability, the ADL decrease probability, the ADL prognosis prediction result, and the SHAP value of all evaluation items of the subject to be evaluated from the ADL prediction information in the subject to be evaluated table 1013 of the storage unit 101.

[0070] In step S403, the menu proposal unit 1033 inputs the feature importance referred to in step S401, the ADL prognosis prediction result and the SHAP value referred to in step S502 into the menu proposal model, and as output, obtains the menu ID of the rehabilitation menu and the increase / decrease coefficient. The menu proposal unit 1033 associates and stores the obtained menu ID and increase / decrease coefficient, and the proposal execution date indicating the date when the process was executed, with the menu proposal information in the evaluation target person table 1013 for the evaluation target person. The presented rehabilitation menu changes according to the importance of the feature quantity and / or the influence degree of the feature quantity indicated by the SHAP value. For example, when considering the influence degree of the feature quantity related to functional training, the training time and exercise intensity can be increased (adjustment of the increase / decrease coefficient). Also, when considering the influence degree of the feature quantity related to physical function, a rehabilitation menu for improving physical function can be proposed. Also, when considering the influence degree of the feature quantity related to oral cavity / cognition, rehabilitation menus such as contact swallowing training and cognitive function training can be proposed. Also, when considering the influence degree of the feature quantity related to interest, a rehabilitation menu related to the interest can be proposed.

[0071] In step S404, the menu proposal unit 1033 refers to the menu table 1015 based on the menu ID obtained in step S403, displays the rehabilitation menu on the user terminal 20, and ends the flow.

[0072] <Data preprocessing> Data preprocessing is mainly a process for generating feature quantity data, which is a set of feature quantities used for various models and learning, from various information (basic information, scientific care indicators, day care nursing plan information, functional training plan information, various evaluation information) of the evaluation target person registered by the user. Figure 9 is a flowchart showing the operation of data preprocessing. The operation of data preprocessing will be described below according to the same flowchart. Data preprocessing is basically executed when data is added to or updated in part or in whole of various information of the person to be evaluated, but it can also be executed at any timing according to the user's request. Also, data preprocessing is executed for each person to be evaluated.

[0073] In step S501, the data preprocessing unit 1034 of the server 10 acquires the user management information of the person to be evaluated from the evaluation target person table 1013 in the storage unit 101.

[0074] In step S502, for various information included in the user management information of the person to be evaluated acquired in step S501, the data preprocessing unit 1034 compares with the essential items of the feature data and the specifications of each item, and performs data verification and formatting. For data verification, for example, it is checked whether the measured value or evaluation value is within the range of the specification (upper limit, lower limit) of a certain feature amount, and if it is outside the range, it is set as a Null value, etc. For data formatting, for example, text data is morphologically analyzed to extract keywords, classified into a predetermined category, or the number of corresponding categories is calculated, etc.

[0075] In step S503, the data preprocessing unit 1034 calculates or generates, as additional feature amounts, feature amounts not covered by the user management information using the user management information of the person to be evaluated acquired in step S501. For example, the current age is calculated from the date of birth included in the basic information of the person to be evaluated in the user management information, and further category data of age groups is generated.

[0076] In step S504, the data preprocessing unit 1034 stores, as a group of feature data of the feature amounts verified, formatted, and generated from step S502 to step S503, in association with the person to be evaluated and the generation date, as the feature data in the evaluation target person table 1013, and ends the flow.

[0077] <Output Example> Figure 10 is an example of a screen showing the prediction evaluation of ADL for all users of the care facility operated by the user. Figure 11 is another example of a screen showing the prediction evaluation of ADL for all users. Figure 12 is another example of a screen showing the prediction evaluation of ADL for each individual user. Figure 13 is another example of a screen showing the prediction evaluation of ADL for each individual user. Figure 14 is a diagram showing an example of a list display of the rehabilitation menus presented. Figure 15 is a diagram showing an example of the content of the rehabilitation menus presented. These may be displayed on the display 2071 of the user terminal 20 or printed out on paper. In these screen examples, the evaluation target for which the ADL prediction was made is shown as "user", and the integrated prognosis prediction result of ADL is shown as "prediction result" or "prediction". By providing the ADL prognosis prediction result to caregivers such as occupational therapists, physical therapists, care managers, and staff in a manner like these screens, the user can easily identify the evaluation targets that require attention for the future care of the evaluation target, use it as a reference for creating a rehabilitation plan, and easily achieve mutual common understanding.

[0078] In the screen example shown in Figure 10, the prognosis prediction results of the ADL evaluation for all users belonging to the user's care facility are displayed. That is, it shows the outcome prediction of the care facility that is the user. In the screen example shown in Figure 10, on the left, a donut pie chart showing the percentage of the prognosis prediction results for all users is arranged, and on the right, a list table of the prognosis prediction results for each user is arranged. In the list table, the prognosis prediction results are sorted in the order of "deterioration", "maintenance", and "improvement", and the users with the prognosis prediction result of "deterioration" are grouped and displayed at the top. This makes it easier for the user, the staff to whom the screen is presented by the user, occupational therapists, physical therapists, caregivers, etc. to identify the users who need attention in the future.

[0079] In the screen example shown in FIG. 11, the prognosis prediction results and the actual results of the ADL evaluations of all users are displayed in a comparable manner. In the upper part of the screen example shown in FIG. 11, a bar graph showing the ratio of the prognosis prediction results to the actual results is displayed for all the subjects who had their initial monthly evaluations during the period from April 1, 2023 to September 30, 2023. Also, in the lower part, for each user, the evaluation time, prognosis prediction results, actual results, difference between prediction and actual results, and the ADL evaluation items that are likely to have a decreased ADL evaluation are displayed in a list format. In addition, in the center, the ADL gain calculated based on the actual results of the change in the ADL evaluation of the subject is displayed. In this way, by presenting the difference between the prediction and the actual results numerically and graphically, the staff of the nursing facility can be made to feel the results of their daily work, which can lead to an improvement in motivation. Also, by presenting the evaluation items that are likely to decrease, the corresponding measures that should be prioritized become clear.

[0080] In the screen example shown in FIG. 12, the prognosis prediction results of the ADL evaluation for an individual user (User A) are displayed. In the screen example shown in FIG. 12, on the left side, a bar graph is arranged to show the current measured value and the predicted value six months later for the ADL evaluation of User A in a comparable manner, and on the right side, a list of the prognosis prediction results six months later for each ADL evaluation item of User A is arranged.

[0081] In the screen example shown in FIG. 13, the prognosis prediction results and the actual results of the ADL evaluation for an individual user (User A) are displayed in a comparable manner. In the screen example shown in FIG. 13, on the left side, a bar graph is arranged to show the measured value at the time of the previous measurement, the predicted value six months after the previous measurement predicted at the time of the previous measurement, and the measured value (actual value) measured six months after the previous measurement (this time) for the ADL evaluation of User A in a comparable manner. Also, on the right side, a list showing the prognosis prediction results six months after the previous measurement predicted for each ADL evaluation item of User A and the measured value (actual results) in this measurement is arranged. In the list, when the prediction and the actual results are different, the background color is changed so that it can be easily recognized.

[0082] In the screen examples shown in FIGS. 14 and 15, a list or details of the rehabilitation menu are displayed for the ADL prognosis prediction result of user A.

[0083] As described above, the information processing system 1 according to the present disclosure inputs data related to the person to be evaluated up to now into the ADL prediction model 1021 that predicts changes in future evaluations of ADL, thereby presenting whether the ADL of the person to be evaluated is likely to deteriorate, be maintained, or improve in the future. In addition, an appropriate rehabilitation menu is presented for the presented future change in ADL. That is, the information processing system 1 can take countermeasures early by grasping the change in the future ADL evaluation of the person to be evaluated, and can improve the quality of rehabilitation by proposing rehabilitation according to the change in the future ADL evaluation of the person to be evaluated, and can contribute to the maintenance or improvement of the health status of the person to be evaluated. From the perspective of improving the quality of rehabilitation, by using the ADL prediction model 1021 for each evaluation item, it is possible to easily identify the complex and diverse problems of the elderly. In addition, by outputting the ADL prognosis prediction results for the entire users of the nursing facility and for each individual user, it is possible to smoothly share information and discuss risks among the facility staff, and it is possible to easily cooperate and communicate with occupational therapists and physical therapists. Also, instead of one model that outputs the probabilities of the three values of the future change in ADL evaluation, "deterioration (decrease)", "maintenance", and "improvement (increase)", and also instead of three models that output the respective probabilities, by dividing and training into two models, one that predicts a decrease and one that predicts an increase, the prediction accuracy of the change in the ADL evaluation of the person to be evaluated can be improved. The reason for this is that the case where the future ADL evaluation is "maintained" accounts for more than half. In addition, by using the feature importance of the ADL prediction model 1021 and the SHAP value specific to the subject to be evaluated obtained from the ADL prognosis prediction result, a more appropriate rehabilitation menu can be proposed. The feature importance is the importance of each feature calculated globally from the entire learning data, and the SHAP value calculated for each subject to be evaluated is the contribution degree of the feature to the predicted value. For example, in the ADL decline prediction model, when the importance of multiple features related to physical function is high, by identifying those with low evaluations for the corresponding features of the SHAP value, the weaknesses of the subject to be evaluated can be grasped, and a rehabilitation menu for improving or maintaining those weaknesses can be proposed. Also, when there are multiple rehabilitation menus for the same physical function, a more appropriate rehabilitation menu can be proposed by considering the past performance of other users and the interests of the subject to be evaluated.

[0084] <Program> FIG. 16 is a schematic block diagram showing the configuration of the computer 801. The computer 801 includes a CPU 802 (processor), a main storage device 803, an auxiliary storage device 804, an interface 805, and an image processing device (GPU) 806.

[0085] Here, the details of the program for realizing each function constituting the server 10 according to the above embodiment will be described.

[0086] The server 10 is implemented on the computer 801. And the operations of the respective components of the server 10 are stored in the auxiliary storage device 804 in the form of a program. The CPU 802 reads the program from the auxiliary storage device 804 and expands it in the main storage device 803, and executes the above ADL prediction model learning process, menu proposal model learning process, ADL prediction process, menu proposal process, and data preprocessing according to the program. Also, the CPU 802 secures a storage area corresponding to the above-described storage unit 101 in the main storage device 803 or the auxiliary storage device 804 according to the program.

[0087] The program related to the ADL prediction process among the programs is a program for causing a computer to execute steps S301 to S307 above. Specifically, the program is a program for causing a computer including a processor and a storage unit to execute, and the program causes the processor to use, as an input, an input information group regarding an evaluation target person, and output, for each of a plurality of ADL prediction models, a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period as an ADL change probability; and a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction result as an ADL prognosis prediction result. The input information group is a plurality of feature quantities regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period regarding a plurality of past evaluation target persons. Also, the program related to the ADL prediction model learning process among the programs is a program for causing a computer to execute steps S101 to S103 above. The program related to the menu proposal model learning process among the programs is a program for causing a computer to execute steps S201 to S202 above. The program related to the menu proposal process among the programs is a program for causing a computer to execute steps S401 to S404 above. The program related to the data preprocessing among the programs is a program for causing a computer to execute steps S501 to S504 above.

[0088] Note that the auxiliary storage device 804 is an example of a non-transitory tangible medium. Other examples of non-transitory tangible media include magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, etc. connected via the interface 805. Also, when this program is distributed to the computer 801 via the network NW, the receiving computer 801 may expand the program in the main storage device 803 and execute the above processing.

[0089] Also, the program may be for realizing a part of the functions described above. Further, the program may be one that realizes the functions described above in combination with other programs already stored in the auxiliary storage device 804, so-called differential files (differential programs). It may be.

[0090] Although the embodiments of the present invention have been described above, these embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are to be included in the scope and gist of the invention, and are also to be included in the invention described in the claims and the equivalent scope thereof.

[0091] Although the description of the embodiments of the present invention ends here, the aspects of the present invention are not limited to this embodiment.

[0092] For example, the ADL prediction model 1021 in the above embodiment included a combination of two models, an ADL increase prediction model and an ADL decrease prediction model, but it may also be a combination of one or more models. That is, it may output "increase", "maintain", and "decrease" with one prediction model, or output "increase (improvement)", "maintain", and "decrease (deterioration)" respectively to each of a plurality of three models: an ADL increase prediction model that predicts an increase in the ADL evaluation, an ADL maintenance prediction model that predicts the maintenance of the ADL evaluation, and an ADL decrease prediction model that predicts a decrease in the ADL evaluation.

[0093] In addition, although the change in the ADL evaluation after the elapse of the predetermined period in the above embodiment was in three stages of improvement, maintenance, and deterioration, it may be in two stages of improvement and deterioration, or may be in four or more stages. In this case, the number of prediction models may be different from the number of stages.

[0094] Further, as input data to the menu proposal model 1022 in the menu proposal process of the above embodiment, the feature importance of the ADL prediction model may be omitted.

[0095] In addition, in the ADL prediction process of the above embodiment, the output of step S305 is performed for each evaluation item, but after predicting all the evaluation items, all the results may be collectively output.

[0096] <Supplementary Note> The configuration of this embodiment is illustrated as follows. [1] An information processing method executed by an information processing apparatus including a processor and a storage unit, wherein the processor uses an input information group regarding an evaluation target person as input, and outputs, for each of a plurality of ADL prediction models, a probability regarding a change in the ADL evaluation of the evaluation target person after a predetermined period as an ADL change probability in a probability calculation step; uses the plurality of ADL change probabilities output for each of the ADL prediction models to predict a change in the ADL evaluation of the evaluation target person after the predetermined period and stores it as an ADL prognosis prediction result in a prognosis evaluation prediction step; and causes to execute, wherein the input information group is a plurality of feature quantities regarding the basic attributes and health status of the evaluation target person, and the ADL prediction model is a model that has learned the correlation between an input information group regarding a plurality of past evaluation target persons and a change in the ADL evaluation after a predetermined period. Information processing method. Thus, even when learning an ADL prediction model based on unbalanced data such as changes in ADL evaluation, the characteristics of the data are considered as multiple ADL change probabilities to predict changes in the ADL evaluation of the subject to be evaluated, so that the prediction accuracy can be improved. Therefore, it becomes possible to more accurately grasp changes in the future ADL evaluation of the subject to be evaluated, and it becomes clear who among the people around should pay attention in daily interactions with the subject to be evaluated (such as the way of living in a nursing facility or during functional training), or it becomes possible to seek advice from occupational therapists or physical therapists preferentially and intensively for subjects with a high risk, etc., and take countermeasures at an early stage. In addition, by proposing rehabilitation according to changes in the future ADL evaluation of the subject to be evaluated, the quality of rehabilitation can be improved, which can contribute to maintaining or improving the health status of the subject to be evaluated. [2] The ADL includes a plurality of evaluation items, The ADL prediction model is a model learned for each of the evaluation items, The probability calculation step calculates the ADL change probability for each of the evaluation items, The prognosis evaluation prediction step generates a prognosis evaluation prediction for each of the evaluation items. The information processing method according to [1]. Thus, it is possible to confirm changes in future evaluations for each ADL evaluation item. [3] The plurality of ADL prediction models include an ADL increase prediction model and an ADL decrease prediction model, The ADL increase prediction model is a model that learns the correlation between the input information group regarding the plurality of past subjects to be evaluated and an increase in the ADL evaluation after a predetermined period, The ADL decrease prediction model is a model that learns the correlation between the input information group regarding the plurality of past subjects to be evaluated and a decrease in the ADL evaluation after a predetermined period, The probability calculation step is an increase probability calculation step of inputting the input information group regarding the subject to be evaluated into the ADL increase prediction model and outputting an ADL increase probability indicating the probability that the ADL evaluation of the subject to be evaluated will increase after the predetermined period, A decline probability calculation step of inputting the input information group of the evaluation target into the ADL decline prediction model and outputting an ADL decline probability indicating the probability that the evaluation of the ADL of the evaluation target will decline after the predetermined period; is a step including The prognosis evaluation prediction step is a step of predicting a change in the evaluation of the ADL of the evaluation target after the predetermined period using the ADL increase probability and the ADL decline probability and storing it as the ADL prognosis prediction result. [1] or the information processing method according to [2]. Thereby, the prediction accuracy can be improved compared with the case of using one model. In addition, since it is only necessary to generate a minimum number of multiple models, the prediction accuracy can be improved while suppressing the time for model learning and the development cost. [4] The change in the evaluation of the ADL of the evaluation target after the predetermined period is a three-stage evaluation of improvement, maintenance, and deterioration. [1] or the information processing method according to any one of [2]. Thereby, the relevant parties can easily grasp the ADL state of future elderly people, can more appropriately formulate a plan for providing care services, and can reduce the care burden. [5] The change in the evaluation of the ADL of the evaluation target after the predetermined period is a three-stage evaluation of improvement, maintenance, and deterioration. [3] or the information processing method according to [2]. Thereby, the relevant parties can easily grasp the ADL state of future elderly people, can more appropriately formulate a plan for providing care services, and can reduce the care burden. In addition, for the three-stage evaluation, the prediction accuracy can be improved with only two prediction models. [6] Furthermore, to the processor, a comprehensive prediction result generation step of generating a comprehensive prediction result indicating the evaluation of the entire evaluation item from the prognosis evaluation prediction for each evaluation item; The information processing method according to any one of [2] to [5] for executing. As a result, the overall state can be easily grasped as the cumulative value of individual items of ADL. [7] The processor is further caused to perform a SHAP value calculation step of calculating the SHAP value of the person to be evaluated in the ADL change probability for each evaluation item; An information processing method according to any one of [2] to [6] for causing the above to be executed. As a result, for each person to be evaluated, the contribution rate of each feature amount to the ADL change probability (ADL increase probability or ADL decrease probability) can be grasped. [8] The processor is further caused to perform a rehabilitation menu proposal step of proposing a rehabilitation menu for the person to be evaluated based on the SHAP value; An information processing method according to [7] for causing the above to be executed. As a result, a more effective rehabilitation can be proposed in consideration of the contribution rate of the feature amount unique to the person to be evaluated. [9] The processor is further caused to perform a feature amount importance calculation step of calculating the contribution rate of each feature amount of the input information group in the ADL prediction model as the feature amount importance; causing the above to be executed, The rehabilitation menu proposal step further proposes the rehabilitation menu based on the feature amount importance. An information processing method according to [8]. As a result, the combination of the minimum necessary feature amounts and the necessary values for maintaining and improving all ADLs from the ascending feature amount importance and / or the descending feature amount importance are calculated as the minimum set of the future ideal state, and a rehabilitation menu is proposed from these values. That is, a more appropriate rehabilitation menu can be proposed in consideration of the feature amounts that affect the estimation.

[10] The input information group includes the past ADL evaluation of the person to be evaluated, The processor is further caused to An ADL gain calculation step for calculating a predicted value of the ADL gain of the evaluation subject after the predetermined period from the past ADL evaluation of the evaluation subject and the ADL prognosis prediction result is executed. The information processing method according to any one of [1] to [9]. By comparing the prediction with the actual result, the result of the ADL improvement efforts as a business office can be visualized numerically, which becomes an issue for improving the motivation of the business office staff, improving the awareness regarding functional training, and improving the quality of rehabilitation such as service and facility review. In addition, it becomes possible to appeal to home care support businesses such as care managers. Furthermore, since it is possible to predict "ADL maintenance addition" in LIFE, profit prediction can be performed.

[11] The plurality of evaluation items are eating, transfer, grooming, toileting, bathing, walking on flat ground, ascending and descending stairs, changing clothes, bowel control, and bladder control. The information processing method according to any one of [2] to

[10] . Thereby, the ADL evaluation corresponding to each item of the Barthel index can be grasped.

[12] An information processing system comprising an information processing device and a user terminal having a processor and a storage unit, which, by executing a program, A probability calculation step of taking an input information group regarding an evaluation subject as an input and outputting, for each of a plurality of ADL prediction models, a probability regarding a change in the ADL evaluation of the evaluation subject after a predetermined period as an ADL change probability; A prognosis evaluation prediction step of predicting a change in the ADL evaluation of the evaluation subject after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the result as an ADL prognosis prediction result; are performed, The input information group is a plurality of feature amounts regarding the basic attributes and health status of the evaluation subject, The ADL prediction model is a model that has learned the correlation between the input information group and the change in the ADL evaluation after a predetermined period for a plurality of past evaluation subjects. Information processing system.

[13] By executing a program, using the input information group regarding the person to be evaluated as input, a probability calculation step that outputs, for each of a plurality of ADL prediction models, the probability regarding the change in the evaluation of the ADL of the person to be evaluated after a predetermined period as the ADL change probability, a prognosis evaluation prediction step that predicts the change in the evaluation of the ADL of the person to be evaluated after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and stores it as an ADL prognosis prediction result, is performed, the input information group is a plurality of feature amounts regarding the basic attributes and health status of the person to be evaluated, the ADL prediction model is a model that has learned the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period regarding a plurality of past persons to be evaluated, an information processing apparatus.

[14] A program for causing a computer including a processor and a storage unit to execute, the program causes the processor to, using the input information group regarding the person to be evaluated as input, a probability calculation step that outputs, for each of a plurality of ADL prediction models, the probability regarding the change in the evaluation of the ADL of the person to be evaluated after a predetermined period as the ADL change probability, a prognosis evaluation prediction step that predicts the change in the evaluation of the ADL of the person to be evaluated after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and stores it as an ADL prognosis prediction result, execute, the input information group is a plurality of feature amounts regarding the basic attributes and health status of the person to be evaluated, the ADL prediction model is a model that has learned the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period regarding a plurality of past persons to be evaluated, a program.

Explanation of Signs

[0097] 1: Information processing system 10: Information processing server 101: Memory unit 1011: Application program 1012: User information table 1013: Evaluated person table 1014: Learning information table 1015: Menu table 1021: ADL prediction model 1022: Menu proposal model 103: Control unit 1031: Model management unit 1032: ADL prediction unit 1033: Menu proposal unit 1034: Data preprocessing unit 1035: Input / output unit 20: User terminal

Claims

1. An information processing method executed by an information processing apparatus including a processor and a memory unit, the method comprising causing the processor to: a probability calculation step of taking, as input, an input information group regarding an evaluation target person and outputting, for each of a plurality of ADL prediction models, a probability regarding a change in the evaluation of the ADL of the evaluation target person after a predetermined period as an ADL change probability; a prognosis evaluation prediction step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the prediction as an ADL prognosis prediction result; execute; wherein the input information group is a plurality of feature quantities regarding the basic attributes and health status of the evaluation target person; wherein the ADL prediction model is a model obtained by learning the correlation between the input information group and the change in the evaluation of the ADL after a predetermined period for a plurality of past evaluation target persons; information processing method.

2. wherein the ADL includes a plurality of evaluation items; wherein the ADL prediction model is a model learned for each of the evaluation items; wherein the probability calculation step calculates the ADL change probability for each of the evaluation items; wherein the prognosis evaluation prediction step generates a prognosis evaluation prediction for each of the evaluation items; The information processing method according to claim 1.

3. wherein the plurality of ADL prediction models include an ADL increase prediction model and an ADL decrease prediction model; wherein the ADL increase prediction model is a model obtained by learning the correlation between the input information group regarding the plurality of past evaluation target persons and the increase in the ADL evaluation after a predetermined period; wherein the ADL decrease prediction model is a model obtained by learning the correlation between the input information group regarding the plurality of past evaluation target persons and the decrease in the ADL evaluation after a predetermined period; wherein the probability calculation step includes: an increase probability calculation step of inputting the input information group regarding the evaluation target person into the ADL increase prediction model and outputting an ADL increase probability indicating the probability that the evaluation of the ADL of the evaluation target person will increase after the predetermined period; a decrease probability calculation step of inputting the input information group of the evaluation target person into the ADL decrease prediction model and outputting an ADL decrease probability indicating the probability that the evaluation of the ADL of the evaluation target person will decrease after the predetermined period; and wherein the prognosis evaluation prediction step includes: a step of predicting a change in the evaluation of the ADL of the evaluation target person after the predetermined period using the ADL increase probability and the ADL decrease probability and storing the prediction as the ADL prognosis prediction result. The information processing method according to claim 1 or 2.

4. The change in the evaluation of the ADL of the person to be evaluated after the predetermined period is a three-level evaluation of improvement, maintenance, and deterioration. The information processing method according to claim 1.

5. The change in the evaluation of the ADL of the person to be evaluated after the predetermined period is a three-level evaluation of improvement, maintenance, and deterioration. The information processing method according to claim 3.

6. Further, to the processor, a comprehensive prediction result generation step of generating a comprehensive prediction result indicating the evaluation of the entire evaluation item from the prognosis evaluation prediction for each evaluation item; The information processing method according to claim 2, which causes the processor to execute.

7. Further, to the processor, a SHAP value calculation step of calculating the SHAP value of the person to be evaluated in the ADL change probability for each evaluation item; The information processing method according to claim 2, which causes the processor to execute.

8. Further, to the processor, a rehabilitation menu proposal step of proposing a rehabilitation menu for the person to be evaluated based on the SHAP value; The information processing method according to claim 7, which causes the processor to execute.

9. Further, to the processor, a feature importance calculation step of calculating the contribution rate of each feature amount of the input information group in the ADL prediction model as the feature importance; causes the processor to execute, The rehabilitation menu proposal step further proposes the rehabilitation menu based on the feature importance. The information processing method according to claim 8.

10. The input information group includes the past ADL evaluation of the person to be evaluated. Further, to the processor, an ADL gain calculation step of calculating a predicted value of the ADL gain of the person to be evaluated after the predetermined period from the past ADL evaluation and the ADL prognosis prediction result of the person to be evaluated; The information processing method according to claim 1.

11. The plurality of evaluation items are eating, transfer, grooming, toileting, bathing, walking, stair climbing and descending, dressing, bowel control, and bladder control. The information processing method according to claim 2.

12. An information processing system including an information processing device and a user terminal including a processor and a storage unit, which, by executing a program, a probability calculation step of outputting, for each of a plurality of ADL prediction models, a probability regarding a change in the evaluation of the ADL of the person to be evaluated after a predetermined period as an ADL change probability using an input information group regarding the person to be evaluated as an input; A prognosis evaluation prediction step of predicting a change in the ADL evaluation of the subject after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the result as an ADL prognosis prediction result. Perform The input information group is a plurality of feature quantities related to the basic attributes and health status of the subject. The ADL prediction model is a model that learns the correlation between the input information group and the change in the ADL evaluation after a predetermined period for a plurality of past subjects. Information processing system.

13. By executing a program, A probability calculation step of taking an input information group related to a subject as an input and outputting, for each of a plurality of ADL prediction models, a probability related to a change in the ADL evaluation of the subject after a predetermined period as an ADL change probability. A prognosis evaluation prediction step of predicting a change in the ADL evaluation of the subject after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the result as an ADL prognosis prediction result. Perform The input information group is a plurality of feature quantities related to the basic attributes and health status of the subject. The ADL prediction model is a model that learns the correlation between the input information group and the change in the ADL evaluation after a predetermined period for a plurality of past subjects. Information processing device.

14. A program for causing a computer including a processor and a storage unit to execute, The program causes the processor to A probability calculation step of taking an input information group related to a subject as an input and outputting, for each of a plurality of ADL prediction models, a probability related to a change in the ADL evaluation of the subject after a predetermined period as an ADL change probability. A prognosis evaluation prediction step of predicting a change in the ADL evaluation of the subject after the predetermined period using the plurality of ADL change probabilities output for each of the ADL prediction models and storing the result as an ADL prognosis prediction result. Execute The input information group is a plurality of feature quantities related to the basic attributes and health status of the subject. The ADL prediction model is a model that learns the correlation between the input information group and the change in the ADL evaluation after a predetermined period for a plurality of past subjects. Program.

Images