JP7145261B2 - Care plan creation support system - Google Patents

Description

The present invention relates to a care plan creation support system.

A care plan for a support-requiring person or a care-requiring person (hereinafter also collectively referred to as a "care recipient") who receives care or support from others in daily life including bathing, toileting, eating, etc. is called a care plan. The care plan includes specific details of care services and a service plan schedule that defines dates and times for providing the care services.

A care plan is usually prepared by a care manager, taking into account care benefits determined according to the level of care required, which is certified based on the situation of the care recipient. When a person who intends to be certified for the level of long-term care needs applies for certification of long-term care level, the long-term care insurer who receives the application has a certified investigator of the certification research organization investigate the applicant's physical condition. The above-mentioned survey is conducted by selecting from two or more options suitable for the applicant's situation, or by inputting the items in text form. The survey results are input to a computer or the like and used for the primary determination of the degree of care required by a prescribed calculation method. The long-term care insurer has the Long-Term Care Certification Examination Board, which is its establishment organization, certify the applicant's degree of long-term care need based on the above primary judgment, other special notes of the care recipient, and the opinion of the attending physician. The upper limit of the nursing care benefit to the person receiving care is determined according to the level of nursing care certified in this way.

After the upper limit of care benefits is determined, the care manager prepares a care plan suitable for the care recipient in consideration of the care benefits. At this time, the care manager or a person entrusted by the care manager newly evaluates the care-receiver's condition at the time of preparing the care plan (hereinafter also simply referred to as "assessment"). After that, the care manager prepares the care plan based on the evaluated current situation of the care-receiver (hereinafter simply referred to as "assessment result").

In Patent Documents 1 to 6, assessment is performed for the purpose of streamlining the creation of care plans and homogenizing the care plans provided by suppressing deterioration in the quality of care plans due to the competence of care managers. A method and system for implementing the method are proposed for assisting in the development of a care plan from the results.

By the way, the number of care recipients tends to increase due to reasons such as the declining birthrate and aging population that have progressed in recent years. In addition, since there are very few cases where nursing care improves the level of care required, the number of care recipients with a high level of care need is also increasing along with the increase in the number of care recipients. Some of the funds necessary for nursing care benefits are covered by nursing care benefits, which are a type of social security expenses. Therefore, due to the increase in the number of care recipients and the accompanying increase in the number of care recipients with a high degree of need for care, nursing care benefit costs are also increasing year by year. For example, the population of so-called elderly people aged 65 and over in Japan is about 34 million in 2015 (1 in 3.7 people), and about 36.5 million in 2025 (population of 3.3 people). 1 person per year), and nursing care benefit costs are expected to reach 21 trillion yen in 2025, compared to 10.1 trillion yen in 2015.

In addition, an increase in the number of care recipients and an increase in the number of care recipients with a high degree of need for care has a great impact on society. For example, problems such as so-called separation from work to care for relatives who are forced to leave their jobs to care for close relatives have become apparent. or reduce the labor force, which may also bring about social losses.

JP-A-11-149499 JP-A-2000-3404 JP-A-2000-3391 Japanese Patent Application Laid-Open No. 2001-101279 JP 2006-146410 A JP 2010-205263 A

As described above, the financial and social burdens are increasing due to the increase in the number of care recipients, and it is required to control these burdens. In particular, as the level of nursing care requirement increases, the amount of nursing care benefit costs and social burden increase, so it is desirable to suppress an increase in the number of care recipients with a high level of nursing care requirement. However, no effective method has been established to reduce the number of care recipients with a high degree of need for care.

For example, if care plans are devised to improve the level of nursing care needs of care recipients, it is thought that an increase in the number of care recipients with high levels of care need can be suppressed. However, there are only a few examples of care plans that have improved the level of long-term care need, and even if these examples are applied to other care recipients in different situations, similar improvements in the level of long-term care need can be expected. Not always possible.

In addition, in order to implement the care plan devised above, it is considered important to improve the efficiency of care plan creation and to homogenize the created care plans. On the other hand, Patent Documents 1 to 6 state that by having a computer support the creation of care plans, the objectives of improving the efficiency of creating care plans and homogenizing the created care plans can be achieved. (Hereinafter, a system that supports care plan creation is also simply referred to as a "care plan creation support system." A care plan creation support system is a system that can output useful information for the creation of a care plan by a care manager. Well, it may have only the function of outputting only nursing care services to be included in the care plan, or it may have the function of simultaneously outputting a service plan schedule that defines the dates and times for providing the above nursing care services.) . However, in these documents, no consideration is given to creating a care plan that can improve the level of nursing care required.

The present invention has been made in view of the above problems, and creates a care plan by outputting a recommended care plan that can be expected to improve the level of care required according to the situation of each care recipient. The purpose is to provide a care plan creation support system that can support

One embodiment of the present invention for solving the above problems is an input reception unit that receives input of the results of assessment performed on a care recipient, a storage unit that stores a plurality of candidate care plans, and the care recipient. a determination unit that determines a change in the degree of care need of the care recipient after executing the candidate care plan for the person; and a processing change unit that changes the probabilistic inference model based on the determination result of the determination unit. and a probability output unit that outputs a probability that the situation of the care-receiver will improve by implementing the candidate care plan for the care-receiver, wherein the probability output unit outputs the probabilistic inference model to the The present invention relates to a care plan creation support system for inputting an assessment result and the candidate care plan and outputting a probability that the level of care need of the care recipient will improve.

The present invention provides a care plan creation support system capable of assisting creation of a care plan by outputting a recommended care plan that can be expected to improve the level of care required according to the situation of each care recipient. be done.

FIG. 1 is a block diagram showing an example configuration of a care plan creation support system according to an embodiment of the present invention. FIG. 2 is a flowchart showing an example of processing by the care plan preparation support system. FIG. 3 is a flow chart showing specific processing included in care plan output (step S100). FIG. 4 is a schematic diagram showing the configuration of the intermediate unit in the first embodiment. FIG. 5 is a flowchart showing specific processes included in the learning process (step S200). FIG. 6 is a flow chart showing the configuration of a care plan creation support system according to another embodiment of the present invention. FIG. 7 is a block diagram showing the configuration of a care plan creation support system according to still another embodiment of the present invention. FIG. 8 is a block diagram showing the configuration of a care plan creation support system according to still another embodiment of the present invention.

[First embodiment]
FIG. 1 is a block diagram showing an example configuration of a care plan creation support system 100 according to an embodiment of the present invention. The care plan creation support system 100 according to the present embodiment includes a care plan output unit 110, a determination unit 120, a probability calculation unit 130, and a processing change unit 140 functioning as processing units, a storage unit 150 functioning as a storage medium, and a display unit. A unit 160 and an input unit 170 are provided.

Although not shown, the care plan creation support system 100 includes, for example, a CPU (Central Processing Unit) as a processor, a storage medium such as a ROM (Read Only Memory) storing a control program, a RAM (Random Access Memory), etc. It has a working memory and a communication circuit. In this case, the functions of the respective processing units described above are realized by the CPU executing a control program (corresponding to the "care plan creation support program" of the present invention). At least part of the program for executing the processes by the care plan output unit, the determination unit, the probability calculation unit, and the processing change unit is stored in a server. , at least part of the program may be stored in a cloud server.

The care plan output unit 110 uses a preset care plan to provide care to a cared person (hereinafter also simply referred to as "providing a care plan"), and when the cared person's degree of need for care is determined. It is a processing unit that acquires the probability of improvement for each care plan set in advance, and outputs the care plan with the higher acquired probability as a recommended care plan. For example, the care plan output unit 110 inputs the care-receiver's condition of the input care-receiver into a probabilistic inference model (hereinafter simply referred to as "care plan output support model"), thereby providing a care plan that is set in advance. Obtaining the probability that the care recipient's degree of need for care improves when the plan is provided. Note that the probabilistic inference model is an inference value (in this embodiment, the level of care required is improved) for each predetermined attribute (type of care plan in this embodiment) by performing a predetermined operation on the input value. It is an inference model that outputs the probability of After that, the care plan output unit 110 refers to the probability that the level of care required that is output for each care plan improves, and selects, for example, a care plan with a higher probability that the level of care required improves, It is a processing unit that outputs a care plan recommended for the cared person.

The determination unit 120 is a processing unit that determines whether or not the current level of care need of the care receiver has improved from the past level of care need of the care receiver. For example, the determination unit 120 determines whether or not the degree of need for care of the care receiver has improved after the care receiver has been cared for using the care plan output from the care plan output unit 110 . Note that the determination unit 120 may determine whether or not the care-required level of the care-receiver has improved after the care-receiver has been cared for using a care plan separately prepared by a care manager.

The probability calculation unit 130 uses the result of the determination by the determination unit 120 to determine whether the care-receiver will It is a processing unit that newly calculates the probability that the degree of care required of the patient will improve. For example, the probability calculation unit 130 calculates the total number of records stored in the storage unit 150 in which the situation of the care receiver and the care plan used for care are the same, and the degree of care required included in the records. is determined to have improved, and the ratio of the number of records determined to have improved the level of care need to the total number of records is calculated. probability that the level of nursing care required for a person will improve.

The processing change unit 140 includes the care recipient's past situation, the care plan created from the care recipient's past situation and provided to the care recipient in the past, and the judgment result of the judgment unit 120. is used as teacher data to change the care plan output support model. Specifically, the processing change unit 140 determines that the combination of the situation of the care recipient and the care plan has the same probability of being acquired for a given care plan when the situation of the care recipient is input to the care plan output unit 110. The care plan output support model is changed so as to approach the probability calculated by the probability calculation unit 130 for .

Details of processing by these processing units will be described later.

One of the features of the care plan creation support system 100 is that the care plan output support model is changed by learning processing. According to the learning process, it is possible to mechanically extract features contained in past data and change the care plan output support model so that inputs with similar features will output similar results. is. Therefore, for each care plan provided, it is determined whether or not the degree of care need has improved for the situation of a certain care recipient, and these data (the situation of the care recipient, the care plan, and the above care plan) Change the care plan output support model so that by performing the above learning process using the above judgment results), a care plan with a high probability of improving the level of care needed even for situations with similar characteristics can be output. can. As a result, the care plan creation support system 100 can create a care plan with a higher probability of improvement after the need for care, according to the characteristics of the situation of the care recipient, even for individual care recipients who have various situations. can be output as a recommended care plan. In addition, by preparing a large amount of such data and repeating the above learning process, the accuracy of outputting a care plan with a higher probability of improving the level of care required for a given care recipient's situation will increase. .

The storage unit 150 is a writable and readable storage medium.

The storage unit 150 can be a storage medium such as a database. In the storage unit 150, cared person information, date or time, situation of the cared person, care plan, degree of care need of the cared person, and care need after care is performed using the care plan A record in which the level of care required or the change in the level of care required as a result of providing the care plan is linked. A unique record number may be assigned to the record.

The above date or date and time is the date or date and time when the cared person was assessed.

The cared person information is information for uniquely identifying the cared person receiving care, and can include an ID number unique to the cared person, the cared person's name, sex, date of birth, and the like. .

The condition of the care-receiver can be an assessment result obtained by an assessment of the care-receiver by a care manager or a person entrusted by the care manager. In addition, as the condition of the care-receiver, it is possible to use the contents described in the report of the attending physician, such as the results of general surveys on the use of services, the results of surveys on special matters, and the names of diseases.

The care plan is a care plan created by a care manager based on the situation of the cared person and provided to the cared person.

The care-required degree of the care-receiver is the care-required degree of the care-receiver certified by the care certification review board when the care plan is created.

The level of care required after care is provided using the above care plan is the level of care required for the same care recipient, which has been recertified by the care certification review board after the care plan has been provided to the same care recipient. degree.

The changes in the level of care required are tabulated data determined by determination by determination unit 120 . For example, a record containing a care plan that was determined to have improved the level of care required would be marked with "1", and a record containing a care plan that was determined to have not improved the level of care needed would be marked with "0". Granted.

In the storage unit 150, at the time of initial setting, the above records are prepared in advance and stored in such a number that the probability calculation unit 130 can calculate the above probabilities.

Display unit 160 is a display device such as a smartphone, PC, or TV carried by a care manager, and displays a recommended care plan according to a display control signal input from care plan output unit 110 . The display unit 160 can be used by a care manager who views the displayed care plan to create a care plan to be provided to the care receiver.

The input unit 170 includes various operation keys such as a numeric keypad and a start key, an audio reception unit, a video reception unit such as a camera, and the like. Data is output to care plan output unit 110 and determination unit 120 . The input unit 170 is various signal receiving devices configured to receive input data electrically output from another medium, such as a receiving connector, and output the input data to the care plan output unit 110 and the determination unit 120. may The input data output from the input unit 170 is output to the care plan output unit 110 and used to output a care plan, and is also output to the determination unit 120 to output the care plan by the probability calculation unit 130 and the processing change unit 140. Used to change the support model.

FIG. 2 is a flowchart showing an example of processing by the care plan creation support system 100. As shown in FIG. The care plan creation support system 100 receives input of data regarding the care recipient's condition and outputs a recommended care plan for the care recipient based on a predetermined probabilistic inference model (step S100). Then, the care plan creation support system 100 performs a learning process for changing the probabilistic inference model so as to output a care plan with a higher probability of improving the degree of care required of the care receiver (step S200).

FIG. 2 shows a mode in which learning processing (step S200) is performed after care plan output (step S100). Care plan output (step S100) is performed after learning processing (step S200). Alternatively, care plan output (step S100) and learning processing (step S200) may be performed in parallel.

Input data used for processing by care plan creation support system 100 will be described below, and then more specific processing included in care plan output (step S100) and learning processing (step S200) will be described.

(Input data)
The input data is information related to the assessment performed on the care recipient in order to create a new care plan. and the level of care need of the care recipient on the date or time of the above assessment. The cared person information included in the input data and the cared person information stored in the storage unit 150 preferably consist of only the same items, but the determination unit 120 can make the same determination of the cared person. As long as there is, only some items may be duplicated.

The care recipient's situation is the one that best matches the care recipient's situation from two or more choices of "Yes (Y)" or "No (N)" for multiple survey items. can be the result of selecting The types of the plurality of survey items and options are not particularly limited, and can be arbitrarily determined. For example, survey items and options determined for use in certifying the level of care need can be used. The survey in the certification of the level of long-term care needs includes a general survey and a basic survey, but either or both of these may be used, and the items and options of the basic survey (74 items as of July 2015) may be used only. At this time, when the certification criteria for the level of care required are changed, it is preferable that the survey items and options can be changed accordingly. As for the situation of the cared person, the result of the general condition survey, the opinion of the attending physician, etc. may be input as a text.

For example, as of July 2015, the survey items used to certify the level of nursing care required are "paresis of the left upper limb (choose from ``yes'' and ``no'')," roll over ("can do without holding onto it", items related to the care recipient's physical function and movement (choose from "able to hold on to something" and "cannot"); Items related to the life function of the care recipient, such as "requires partial assistance" and "requires total assistance"), "transmission of intentions" ("subjects can communicate their intentions to others", "sometimes care recipient's cognitive abilities, such as "able to communicate", "hardly communicable" and "cannot"), and "saying date of birth and age" (selecting from "can" and "cannot"). Items related to mental and behavioral disorders of care recipients, such as “becoming victimized by things being stolen” (choose from “No,” “Sometimes,” and “Yes”) management (choose from "not canceled", "partially unlocked" and "fully unlocked")" and "daily decision making ("can (even in special cases)", "special items related to adaptation to social life, such as "possible unless necessary", "difficult in daily life" and "impossible"), and "dialysis (select from "yes" and "no") It consists of 74 items such as items related to medical care received in the past 14 days. These items may be used as they are, some of them may be selected and used, or other items may be arbitrarily registered and used.

From the viewpoint of saving the time and effort of inputting, the number of items of the cared person's situation can be 5 or more and 20 or less, further 5 or more and 15 or less. The situation of the care receiver at this time preferably consists of survey items used when evaluating (monitoring) activities of daily living (ADL) and instrumental activities of daily living (IADL) performed on the care receiver. As the survey items used for the above ADL evaluation, use the evaluation items used in evaluation methods such as the Basel Index, the performance status established by the Eastern Cooperative Oncology Group (ECOG), and the Functional Independence Scale (FIM). can be done.

(Step S100: Output of Care Plan)
In the care plan output (step S100), the care plan output unit 110 inputs the input condition of the care recipient into the care plan output support model, and acquires the probability of improving the level of care required for each care plan. do. After that, the care plan output unit 110 outputs a care plan recommended for the care recipient based on the probability that the level of care need that is output for each care plan will improve, and displays the care plan on the display unit 160 .

The care plan includes at least care services. Nursing care service is the specific content of nursing care service provided to the care receiver. The care service may be selected from those that can be provided by care providers. The care plan may include a service plan schedule that defines the date and time for providing the care service, in addition to the specific contents of the care service.

FIG. 3 is a flow chart showing specific processing included in care plan output (step S100).

First, the care plan output unit 110 inputs the care-receiver's condition into the care plan output support model, and obtains the probability of improving the level of care required for each care plan (step S310). After that, the care plan output unit 110 outputs one or more care plans with a high probability of improving the level of care required as recommended care plans (step S320). For example, the care plan output unit 110 outputs a care plan with a high output probability (for example, a care plan with the highest probability or a care plan corresponding to the above 10th output unit with probability) as a recommended care plan. do.

After that, the care plan output unit 110 outputs the output care plan to the display unit 160 as a display control signal, and causes the display unit 160 to display the care plan (step S330). At this time, if the care plan is changed as in the third embodiment described later, the care plan output unit 110 returns to step S310 and causes the display unit 160 to display the corrected care plan.

Finally, care plan output unit 110 stores the care receiver information, date or time, care receiver status and level of care need, and output or changed care plan included in the input data as new records in storage unit 150. (step S340).

The care plan output support model described above is a probabilistic inference model that outputs a certain inference value for each type of care plan that is set in advance in response to the input of the situation of the care recipient. configured to be modifiable. The inferred value is the probability that the degree of need for care of the care-receiver will improve when each care plan is provided to the care-receiver having the input situation. Care plan output support models include, for example, neural networks, Boltzmann machines, and decision trees.

Hereinafter, the configuration of the care plan output support model when the care plan output support model includes a neural network will be exemplified.

The neural network (hereinafter also simply referred to as "plan output network") includes an input layer consisting of a plurality of input units, one or more intermediate layers each consisting of a plurality of intermediate units, and an output consisting of a plurality of output units. have layers. The plan output network is a multi-level neural network with N layers, which has multiple intermediate layers from the input layer side to the output layer side. Information received from one or more intermediate units (or input units) included in the layer is output to one or more intermediate units (or output units) included in the next intermediate layer. The number of layers (N) that the plan output network has may be 3 or more, and can be arbitrarily determined according to the number of assessment items and the like. For example, the number of layers (N) can be 3 or more and 25 or less, preferably 5 or more and 20 or less, and more preferably 8 or more and 16 or less.

The input layer (N=1) has the same number of input units as the survey items included in the cared person's situation. The input unit corresponds to one of the plurality of survey items, and outputs an input value corresponding to each option of the survey item as it is to one or more intermediate units.

The intermediate layers (N=k (1<k<N)) each have an arbitrary number of intermediate units. The intermediate unit receives an output from one or more intermediate units possessed by the intermediate layer of the previous hierarchy (or one or more input units possessed by the input layer), and any or a plurality of intermediate units possessed by the intermediate layer of the next hierarchy. Output to an intermediate unit (or an output unit that the output layer has). The intermediate layer may be a fully connected layer in which intermediate units included in the intermediate layer are connected to all intermediate units (or input units) of the previous layer, or intermediate units included in the intermediate layer may be intermediate units of the previous layer. (or input units) may be convolutional layers that only connect to some of the input units.

At this time, as shown in FIG. 4, the intermediate unit Z of the k-th layer receives outputs w1, w2, w3 and w4 from the intermediate units of the k-1th layer (four in the figure). After that, the intermediate unit Z adds a predetermined bias b to the sum of the respective outputs multiplied by predetermined weights y1, y2, y3, and y4 to obtain a value u1, as shown in equation (1). calculate. Note that the weights and biases described above can be changed by the processing change unit 140 .
u1=w1×y1+w2×y2+w3×y3+w4×y4+b (1)

The intermediate unit further outputs a value z1 obtained by applying a predetermined activation function f to u1 to the intermediate unit (or output unit) of the k+1 layer. The type of activation function f is not particularly limited, and a sigmoid function, a logistic function, a Rectified Linear Unit (ReLU), a softmax function, or the like can be used. Among these, the activation function f of the intermediate unit is preferably ReLu from the viewpoint of increasing the accuracy of output and speeding up processing.

The output layer (N=N) has the same number of output units as the preset number of care plans that can be output. Each output unit receives the output from one or more N-1-th intermediate units, and the value obtained by performing the same processing as the intermediate unit is input to the input layer. It is output as an inferred value for the situation (probability that the level of nursing care will improve). At this time, the activation function f possessed by the output unit is such that the sum of the inference values output from all the output units is 1, and the output inference values can be used as-is with a probability distribution. Functions are preferred.

At this time, each of the output units is made to correspond to each of preset care plans that can be output by the care plan output unit 110 on a one-to-one basis. Then, the inference value output by each of the output units is regarded as the probability that the degree of need for care of the care recipient having the above situation will improve when each care plan is provided. Therefore, the care plan output unit 110 can select and output the care plan corresponding to the output unit with the larger inference value as the care plan recommended for the care recipient having the above situation.

The weights and biases of each unit in the intermediate layer and the output layer are set by using a combination of the care recipient's past situation and the care plan selected by the care manager for the care recipient as teacher data at the time of initial setting. , is preferably set in advance by learning processing by the processing change unit 140 .

(Step S200: learning process)
In the learning process (step S200), the determination unit 120, the probability calculation unit 130, and the processing change unit 140 change the care plan output support model described above by learning processing. Specifically, the determination unit 120 determines whether or not the level of care required included in the input data has improved from the previous level of care required of the same care-receiver registered in the storage unit 150 . After that, the probability calculation unit 130, for any combination of the care-receiver's situation and the care plan contained in the storage unit 150, determines the care-receiver having the care-receiver's situation that constitutes the above-described combination based on the result of the above determination. Calculate the probability that the level of care required will improve when the above care plan is provided. After that, the processing change unit 140 changes the care plan output support model using the situation of the care recipient and the care plan that constitute the above combination and the probability of improving the level of care required for the combination as teacher data. For example, the processing change unit 140 may be configured such that when the situation of the cared person is input to the neural network, the inference value output from each output unit is the request when the care plan corresponding to the output unit is provided. The weight and bias values of each intermediate unit and output unit are changed so as to approach the values calculated as the probability of care level improvement.

It should be noted that the past degree of need for care, the past situation of the care recipient, and the past care plan, which will be described later, are stored in the same care recipient's record registered in the storage unit 150, and the date or time of the input data. , the level of care required, the status of the care receiver, and the care plan, which are included in the record at the time of going back a predetermined period of time. Specifically, the date or date registered in the date or date of assessment included in the record is included in the record that is the point in time before the date or date of assessment included in the input data by a predetermined period. The items are the past degree of care need, the past care plan, and the past situation of the care receiver. The predetermined period can be a period from when a care plan is created until when a new care plan is created, and can be six months, for example.

FIG. 5 is a flowchart showing specific processes included in the learning process (step S200). The determination unit 120 collates the care-receiver information included in the input data with the care-receiver information registered in the storage unit 150, and determines whether a record of the same care-receiver as the input data is registered in the storage unit 150. (step S510).

As a result of the determination, if the same cared person's record is registered (step S510: YES), the determining unit 120 determines whether the cared person has been cared for using the past care plan. A change in the care recipient's level of care requirement (whether or not the level of care requirement has improved) is evaluated (step S520). Specifically, the determining unit 120 determines whether the level of care required of the care recipient (current level of care required) included in the input data and the past record of the same care recipient registered in the storage unit 150 It is determined whether or not the current level of care need has improved (became lower) than the past level of care need. If the same cared person's record is not registered (step S510: NO), the input data cannot be used for the learning process, so the process in FIG. 5 ends.

As of July 2016, the level of long-term care required in Japan consists of seven levels: support level 1, support level 2, long-term care level 1, long-term care level 2, long-term care level 3, long-term care level 4, and long-term care level 5. It is judged that the level of care required is low, and the latter is higher. Therefore, when the level of care need shifts from the latter classification to the former classification, for example, although the care recipient's past care need level was "3" in the past, the care need level included in the input data is "required". Care 2”, it is determined that the degree of need for care of the care recipient has improved.

This determination result is stored in the record containing the past care recipient situation and the past care plan used in the above evaluation among the records of the same care recipient contained in the storage unit 150, and the item "change in degree of care need". is registered as the value of " (step S530).

After that, the probability calculation unit 130 counts the number of records in which the value of the item "change in level of care required" is newly registered in the storage unit 150 (step S540). As a result, when it is determined that the predetermined number or more of data have been newly registered (step S540: YES), the probability calculation unit 130 obtains the probability that the care recipient's level of care need is improved (step S550). For example, the probability calculation unit 130 calculates the total number of records, among the records stored in the storage medium 150, in which the situation of the care recipient and the care plan used for care are the same, and is determined to have improved, and the ratio of the number of records determined to have improved in the care level to the total number of records is calculated. If it is not determined that the predetermined number or more of data have been newly registered (step S540: NO), the process in FIG. 5 ends.

Thereafter, the processing change unit 140 performs learning processing to change the care plan output support model (step S560).

For example, the processing change unit 140 may set the inference value output from the output unit corresponding to the care plan when the situation of the care-receiver is input to the plan output network described above so as to be the calculated ratio. , to change the above plan output network.

Specifically, the processing change unit 140 obtains the difference (error) between the inference value output by the plan output network when the situation of the cared person is input and the calculated ratio. Next, the processing change unit 140 adjusts the weights and biases (y1, y2, y3 in the above example of FIG. 4) possessed by either the intermediate units or the output units of the plan output network so that the error becomes small. , y4 and b) are changed.

A method for changing the evaluation network and the plan output network is not particularly limited, and a known method such as the gradient descent method using backpropagation can be used. Also, the type of error function used to express the error in the gradient descent method is not particularly limited, and known formulas such as squared error and cross entropy formulas can be used.

With such a configuration, the care plan output by the care plan output unit 110 through the process including the execution of the care plan output support model including the plan output network has a high probability of improving the care recipient's degree of need for care. become a thing. In particular, by inputting a large amount of input data and performing a large number of learning processes, the above combination with a high probability of improving the level of care required for a care recipient having an arbitrary care recipient situation is configured. A care plan can be selected and provided to the care recipient.

In addition, by allowing a computer to support the creation of care plans, it is possible to achieve the objectives of improving the efficiency of care plan creation and homogenizing the created care plans.

In the conventional care plan preparation method, one of the factors that hindered the improvement of the quality of the prepared care plan was whether or not the care recipient who was provided with the care plan included in the care plan was lively. The quality of the selection of the care plan has been evaluated based on judgment criteria that are subject to change depending on the subjectivity of the observer. On the other hand, if a care plan is evaluated based on evaluation criteria that can be objectively judged whether or not the degree of care need improves, it is possible to eliminate as much as possible the observer's subjectivity that affects the evaluation of the care plan. .

In addition, when the care plan output support model is configured to include a multi-layered neural network as in this embodiment, when a specific care plan is provided among the survey items included in the input care recipient's situation The care plan output support model can be changed so that the probability of outputting the specific care plan is increased when a model having a characteristic that the degree of care need is likely to improve is input to . Therefore, even when a similar care-receiver situation having survey items having the above characteristics is input, the specific care plan is output with a high probability, so the care plan output unit 110 is not included in the teacher data. It is possible to output a care plan with a high probability of improving the degree of need for care even for the situation of the care-receiver (or the number of data is small).

[Second embodiment]
FIG. 6 is a flow chart showing specific processing included in the learning processing (step S200) in another embodiment of the present invention. This embodiment differs from the first embodiment in that a care plan evaluation model, which is a probabilistic inference model different from the care plan output support model, is used to change the care plan output support model by learning processing. Since other processing and system configuration are the same as those of the first embodiment, overlapping descriptions are omitted.

The care plan evaluation model is a probabilistic inference model that outputs the probability that the level of care need will be improved by implementing the care plan for the input of the combination of the care recipient's situation and care plan, and is changed by learning processing. configured as possible. Care plan evaluation models include, for example, neural networks, Boltzmann machines and decision trees.

A configuration in which the care plan evaluation model includes a neural network (hereinafter also simply referred to as "evaluation network") will be exemplified below. At this time, the care plan evaluation model has an input layer consisting of a plurality of input units, one or a plurality of middle layers each consisting of a plurality of middle units, and an output layer consisting of one output unit. The evaluation network is a multi-level neural network having N layers, with multiple intermediate layers from the input layer side to the output layer side. outputs information received from one or more intermediate units (or input units) included in the next hierarchy to one or more intermediate units (or output units) included in the next intermediate layer. The number of layers (N) that the evaluation network has may be 3 or more, and can be arbitrarily determined according to the number of combinations of care-receiver situations and care plans. For example, the number of layers (N) can be 3 or more and 25 or less, preferably 5 or more and 20 or less, and more preferably 8 or more and 16 or less.

In the evaluation network, the input item to the input layer is a combination of the care recipient's situation and care plan, and the inference value output from the output layer is the probability that the level of nursing care will improve with respect to the above combination. It can be configured similarly to the plan output network described above.

At this time, probability calculation unit 130 determines that a sufficient amount of data has been newly registered in storage unit 150 in step S540 described above (step S540: YES). , the total number of records in which the situation of the care recipient and the care plan used for care are the same, and the number of records in which the level of care need was judged to have improved, included in the above records, and the total number of records , the ratio of the number of records determined to have improved care level (step S550). After that, the processing change unit 140 changes the evaluation network so that the ratio is output when the combination is input to the care plan evaluation model (step S570).

Specifically, the processing change unit 140 obtains the difference (error) between the probability of improvement in the level of care need output by the evaluation network when the combination is input and the calculated ratio. Next, the processing changer 140 changes one or more of the weights and biases of each intermediate unit and output unit of the evaluation network so that the error is reduced.

After that, the processing change unit 140 performs the same processing as in the first embodiment so that the inference value output by each output unit when the care-receiver's situation is input to the plan output network is changed to the care-receiver's situation. and the care plan corresponding to the output unit are input, the plan output network is changed so that the probability that the level of care need output from the evaluation network will improve (step S560-2).

A method for changing the evaluation network and the plan output network is not particularly limited, and a known method such as the gradient descent method using backpropagation can be used. Also, the type of error function used to express the error in the gradient descent method is not particularly limited, and known formulas such as squared error and cross entropy formulas can be used.

With such a configuration, the probability that the level of care need has improved, which is more accurate than the "percentage of records in which the level of care need has improved" in the first embodiment, is set as the target output when the care plan output support model is changed. can be used as a value. Therefore, the recommended care plan output from the care plan output support model can have a higher probability of improving the level of care required.

With such a configuration, it is also possible to extract and use only the output from the evaluation network to predict the prognosis of the cared person.

[Third embodiment]
FIG. 7 is a block diagram showing the configuration of a care plan creation support system according to still another embodiment of the present invention. A care plan creation support system 700 according to this embodiment includes a care plan changing unit 710 that changes the care plan displayed on the display unit 160 .

The care plan changing unit 710 includes various operation keys such as a numeric keypad and a start key, an audio receiving unit, a video receiving unit such as a camera, and the like, and accepts various input operations by the care manager for changing the output care plan. and transmits the generated input data to the care plan output unit 110 . The care plan output unit 110 receives input data and modifies the output care plan. Although FIG. 7 shows an example in which the care plan changing unit 710 is configured separately from the display unit 160, for example, the display unit 160 may be configured such that the operation part is integrated with the display screen, such as a touch panel. Alternatively, it may be configured to be connected to a hard key such as a button or key, and function as the care plan changing unit 710 at the same time.

The changed care plan is registered in the storage unit 150 as in the first and second embodiments.

At the same time, the display unit 160 may connect to the server of the care provider and display information such as the availability of the care service included in the output care plan and whether it is available or not. With such a configuration, the care manager can change the care plan from the display unit 160 when it is determined that the displayed care service cannot be reserved.

In addition, the care plan change unit 710 may be configured to connect to a server of a care provider and arrange care services by input from the touch panel or hard keys.

[Fourth embodiment]
FIG. 8 is a block diagram showing the configuration of a care plan creation support system according to still another embodiment of the present invention. A care plan creation support system 800 according to this embodiment has a cloud server 810 that is a virtual server that cooperates with various devices via the Internet. Cloud server 810 includes a program for executing the processes by care plan output unit 110, determination unit 120, probability calculation unit 130, and processing change unit 140 described above, data stored in a storage medium such as storage unit 150, and the like. are preserved in whole or in part.

The cloud server 810 is owned, managed, and operated by an operating company that operates a nursing care business or a consignee entrusted by the operating company (hereinafter also simply referred to as an “operating company, etc. 830”). The operating company or the like 830 accesses the cloud server 810 from one or a plurality of processing units 835, such as a PC connectable to the cloud server 810, and causes the processing unit 835 to execute the above-described processing.

At this time, the input unit 170 can be a device (for example, a smart phone, a PC, a TV, etc.) that can be connected to the Internet owned by a group 820 such as an assessment company that assesses care recipients. Input data input from the device as the input unit 170 is transmitted from the device to the cloud server 810 . At this time, the input unit 170 is preferably configured to automatically delete the personal information such as the cared person's name, date of birth, and cared person's situation after transmission.

Also, at this time, the display unit 160 can be a device (for example, a smart phone, a PC, a TV, etc.) that can be connected to the Internet owned by a group 840 such as a care manager who creates a care plan. The care plan output by the care plan output unit 110 is transmitted to the device as the display unit 160 and displayed. When group 840 modifies the care plan from display 160 , the modified data is sent to cloud server 810 .

Group 820 and group 840 may be the same business operator or individual.

It should be noted that each of the above-described embodiments is merely an example of implementation of the present invention, and the technical scope of the present invention should not be construed to be limited by these. Thus, the invention may be embodied in various forms without departing from its spirit or essential characteristics.

For example, in each of the above-described embodiments, data in which the level of care need has changed by one level and data in which the level of care need has changed by two or more levels are treated in the same way as data indicating that the level of care need has improved. The degree of improvement may be weighted according to how many times the degree of care has improved, and used for learning processing.

Further, in each of the above embodiments, the storage unit 150 was described as a component incorporated in the care plan creation support system, but the storage unit 150 is configured independently of the care plan creation support system, and the care plan creation support system may be configured to be able to refer to data for performing each of the above processes by connecting from. Also at this time, in the fourth embodiment, the storage unit 150 may be stored in the cloud server.

Further, in each of the above embodiments, the result of determination by the determination unit 120 is registered in the storage unit 150, and the probability calculation unit 130 uses the registered determination result to newly calculate the probability that the level of care required improves. However, the past nursing care level and the nursing care level included in the input data are registered in the storage unit 150, and the determination unit 120 compares the two nursing care levels registered in the storage unit 150 and compares the above-mentioned A change in the level of care required may be determined. Also at this time, the storage unit 150 may separately register a change in the level of care required as a result of providing care using the care plan.

In the third embodiment, the care plan output unit 110 outputs a plurality of care plans as recommended care plans, and one care plan selected by the care manager is input from the care plan change unit 710. , the selected and input care plan may be registered in the storage unit 150 .

In each of the above-described embodiments, the display unit 160 is described as a smart phone or tablet PC that can be carried by the care manager. You may create a care plan while confirming the care plan. It should be noted that the output unit 160 may read out the care plan by voice instead of displaying it on the screen.

This application is an application claiming priority based on Japanese Application No. 2016-155859 filed on August 8, 2016, and the contents described in the claims, specification and drawings of the application are incorporated in the application.

According to the care plan creation support system of the present invention, it is possible to output a care plan that has a high probability of improving the degree of care required according to the situation of each care recipient, and to increase the accuracy of the output by learning processing. be able to. Therefore, if a care plan prepared using the care plan creation support system of the present invention is provided to a care recipient, improvement in the need for care of the care recipient can be expected. If the level of nursing care needs of people receiving care improves, it is expected that nursing care benefits will be reduced and the number of people who leave their jobs for nursing care will decrease.

100, 700, 800 care plan creation support system 110 care plan output unit 120 determination unit 130 probability calculation unit 140 processing change unit 150 storage unit 160 display unit 170 input unit 710 care plan change unit 810 cloud server 820 group 830 operating company, etc. 835 Processing unit 840 group

Claims (4)

an input reception unit that receives input of the results of the assessment performed on the care recipient;
a storage unit that stores a plurality of candidate care plans ;
a determining unit that determines a change in the level of care required of the care recipient after executing the candidate care plan for the care recipient;
a processing change unit that changes the probabilistic inference model based on the result of determination by the determination unit;
a probability output unit that outputs a probability that the situation of the care-receiver will improve by implementing the candidate care plan for the care-receiver;
The probability output unit inputs the result of the assessment and the candidate care plan to the probabilistic inference model, and outputs the probability that the level of care required of the care recipient will improve.
Care plan creation support system. 2. The care plan creation according to claim 1, wherein the probability output unit calculates, as the probability of improvement, a ratio of care recipients whose level of care need has been improved by the candidate care plan to the number of implementations of the candidate care plan. support system. a probability calculation unit that newly calculates a probability that the care-required level of the care-receiver will be improved by the candidate care plan based on the determination result of the determination unit;
The processing change unit changes the probabilistic inference model so that the probability that the level of care need of the care recipient, output by the probabilistic inference model, improves approaches the probability calculated by the probability calculation unit.
The care plan creation support system according to claim 1 or 2 . 4. The care plan creation support system according to any one of claims 1 to 3 , wherein said probability output unit outputs a probability that said care recipient's level of care required improves.

Images