JP2019061632A - Service effect evaluation system - Google Patents

Abstract

To provide a service effect evaluation system capable of modeling a relationship between a service that a user receives and user's state classification, and quantitatively evaluating a change in the state classification even when the service changes.SOLUTION: A state change index for evaluating a degree of change in a service user's state is defined as explained variables, and a service provider index for evaluating whether a service that a service provider provides is good or bad and an auxiliary index to be calculated from a service provider accompanying index accompanying the service provider index are defined as explanation variables to estimate a model explaining the state change index. The service provider index and the auxiliary index are applied to the model to calculate a first state change index, and the service provider index and the auxiliary index updated by applying a target auxiliary index set by a target auxiliary index setting part to the service provider index and the auxiliary index are applied to the model to calculate a second state change index.SELECTED DRAWING: Figure 1

Description

  The embodiment of the present invention relates to a service effect evaluation system for evaluating the effect of a service provided to a user in a nursing care business or the like.

  Care insurance business is implemented nationwide, and each municipality operates as an insurer. In this long-term care insurance business, when residents of the municipality wish to use the nursing care service and apply for the certification for long-term care to the municipality, authorized investigators visit the applicant and perform / authorized survey.

  Accreditation survey results are input to the computer, the standard time for certification of long-term care etc. is calculated, and the degree of long-term care required (non-applicability, required level 1 and 2, required 1 to 5 levels required) is output as primary judgment result Be done. After this, a secondary judgment is made at the nursing care certification examination committee, and based on the judgment, the municipality who is the insurer performs the certification for long-term care.

  When the certification for long-term care is carried out, a care plan is created by the home long-term care support office or the regional comprehensive support center, including the mental and physical condition including the degree of long-term care required, nursing power, economic power, etc. Use of the service is started based on that. That is, a person (user) who is identified as needing care is identified as one of the required 1 and 2 needs and the 7 required 1 to 5 needs (degree of need of nursing care) services according to this condition It can be received from a care provider. As this service, there are various services such as regular care services, regular rehabilitation services, and dementia compatible communal living care services.

  By the way, recently, with the progress of the aging, the cost of long-term care has been increasing year by year, and there is a background to wanting to suppress the increase of the cost of long-term care by providing appropriate service to the user of the care service. However, at present, it has not been possible to quantitatively evaluate the effect of the service provided by a care provider. In particular, it is not possible to quantitatively evaluate the effect of the change service when changing the service to be provided.

  As the prior art related to the nursing care business, those described below have been proposed.

  By managing the progress in the implementation of the care prevention instruction, the care prevention work can be supported (see Patent Document 1). However, this prior art can not evaluate the quality of the service provided by the care prevention service.

  The score between the presence or absence of the medical examination item was scored according to the region of the human body organ or region to allow evaluation of the relationship between health and medical dependence (see Patent Document 2). However, with this prior art, it is not possible to evaluate the direction of the change in the degree of health of the patient and the period until the change. That is, in the nursing care business, it is desirable to be able to quantitatively evaluate the direction of change of the care need status category and the period until reaching the change based on the provided service, but in this conventional technology, such evaluation It can not be done.

  The correlation between the profit and loss information indicating the degree of profit and loss for each of a plurality of patients and the contents of a plurality of treatment items indicating a treatment for a patient can be analyzed (see Patent Document 3). However, this prior art can not evaluate the change in the degree of profit and loss when the treatment for the patient is changed. That is, when the service is changed, the effect of the service can not be quantitatively evaluated.

  Based on static data related to diseases and groups, and dynamic data on groups and individual objects, it was possible to model diseases occurring in the groups (see Patent Document 4). In this prior art, an infectious disease is assumed as a target disease, and dynamic data with rapidly changing parameters such as physiological parameters are used for modeling. On the other hand, it is not suitable for modeling using the index of the service which changes with a certain time interval, and the index associated with the service related to the care business.

Japanese Patent Application Publication No. 2003-323501 JP, 2006-114027, A JP, 2012-108791, A JP, 2013-508859, A

  As mentioned above, at present, it has not been possible to quantitatively evaluate the effect of the service provided by the care provider. In addition, it is not possible to quantitatively evaluate the effect of the change service when the service to be provided is changed.

  The present invention models the relationship between the service received by the user and the status category of the user, and quantitatively evaluates the effect of the service based on the model. Even if there is a change in the service, the change is It is an object of the present invention to provide a service effect evaluation system capable of quantitatively evaluating changes in status divisions that occur along with the above.

  A service effect evaluation system according to an embodiment of the present invention is a state change index which is an index for evaluating the degree of change of a state of a service user, and a service which is an index for evaluating the quality of service provided by a service provider. Providing a service based on a learning data storage unit storing, as learning data, data consisting of a provider index and a service provider incident index which is an index attached to the service provider index, and the learning data Using the learning data preprocessing unit that calculates an assistance indicator for the service provider including the service implementation rate from the person-accompanying indicator, and the state change indicator as an explained variable based on the learning data and the assistance indicator, The state change finger describes a model that describes the state change indicator using the service provider index and the auxiliary index as explanatory variables. A model learning unit that estimates each time, a model storage unit that stores the estimated model, a target assistance indicator setting unit that sets a target assistance indicator that is a target assuming the assistance indicator, the target assistance indicator A service provider index and a target index estimation unit which estimates the service provider index and the auxiliary index updated by applying the service provider index and the auxiliary index, and the model of the service provider index and the auxiliary index To calculate a first state change index, and applying the target index to the model to calculate a second state change index, a state change index prediction unit, the first state change index, and And a prediction result storage unit for storing the second state change index.

  According to the above configuration, since the relationship between the service received by the user and the status category of the user is modeled and the effect of the service is quantitatively evaluated based on the model, even if there is a change in the service, It is possible to quantitatively evaluate the change of the status category that accompanies the change.

It is a figure of a functional block showing composition of a service effect evaluation system concerning a 1st embodiment of the present invention. It is a flow figure explaining the presumption processing of the model in a 1st embodiment. It is a flow figure explaining prediction processing of a state change index in a 1st embodiment. It is a figure explaining data composition of learning data in a 1st embodiment. It is a figure explaining the duration and the aggravation-improvement gradient which are state change indicators in a 1st embodiment. It is a figure of a functional block which shows the composition of the service effect evaluation system concerning a 2nd embodiment of the present invention. It is a flow figure explaining prediction processing of a target auxiliary indicator effect in a 2nd embodiment. It is a figure explaining the data composition of the reference data in a 2nd embodiment. FIG. 12 is a diagram for explaining the case of prediction: improvement, simulation: improvement in the second embodiment. Prediction in the second embodiment: Deterioration, simulation: a diagram for explaining the case of deterioration. It is a figure explaining the case of prediction: Deterioration, simulation: improvement in a 2nd embodiment. FIG. 16 is a diagram for explaining another case of prediction: improvement, simulation: improvement in the second embodiment. FIG. 12 is a diagram for explaining the case of prediction: improvement, simulation: deterioration in the second embodiment. FIG. 16 is a diagram for explaining another case of prediction: improvement, simulation: deterioration in the second embodiment. It is a figure of the functional block which shows the structure of the service effect evaluation system which concerns on the 3rd Embodiment of this invention. It is a flowchart explaining the selection process of the model in 3rd Embodiment. It is a figure explaining the data composition of the modeling index in a 3rd embodiment. It is a figure of the functional block which shows the structure of the service effect evaluation system which concerns on the 4th Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
First Embodiment

  The service effect evaluation system according to the embodiment of the present invention is constructed as a computer system including a computer and its peripheral devices, and the functional blocks shown in FIG. 1 represent functions executed by this computer system.

  As described above, the service effect evaluation system models the relationship between the service received by the user and the status category of the user, and quantitatively evaluates the effect of the service based on the model. As shown in the figure, the learning data storage unit B01, the learning data preprocessing unit B08, the model learning unit B02, the model storage unit B03, the state change indicator prediction unit B04, the target indicator estimation unit B05, the target auxiliary indicator setting unit B06, and prediction It has a result storage unit B07.

  The learning data storage unit BO1 stores learning data for constructing a model. The learning data is defined in an evaluation target, and is configured in such a form that three types of data such as a state change index, a service provider index, and a service provider incident index are given for each target. For example, in the learning data shown in FIG. 4, establishments A, B,... Z are given as targets of evaluation, and the above-mentioned state change indicator, service provider indicator, and service provider incidental Indicators are given.

  Here, the state change index is an index for evaluating the degree to which the value indicating the state of the service user changes, such as the condition requiring care category. Also, the service provider index is an index for evaluating the quality of a specific service provided by the service provider. The service provider incidental index is an index associated with a specific service provided by the service provider, and is an index associated with the service provider index described above. The details of these will be described later.

  As the above-mentioned state change index, in this embodiment, the duration and the deterioration-improvement gradient are considered. Here, as shown in FIG. 5, with the continuation period, the current state of the service user (here, described as the care requiring condition category) changes (deteriorates or improves) to a different care requiring condition category It is an index that evaluated the length. The deterioration-improvement gradient is an index that evaluates the difference between the current care need status category and the different care need status category.

As a service provider index, a service Pα, 1 provided to a user of a specific care need status category α provided by a company operating a care facility Let's consider an indicator that evaluates the quality of each of Pα, n_α for each service. However, it is assumed that n_α represents the total number of services provided to the user of the nursing need status category α.

Here, the above-mentioned service Pα, 1 For example, a service provision amount is used as an index for evaluating the quality of Pα and n_α. Specifically, the service type is provided / month, the service type provided / month, the service type provided total number of units / month, and so on. Such data may be used because service provision data provided for each user, nursing care receipt data, etc. are aggregated in each unit, such as user, office, area, etc.

  The service provider incidental index is an index incidental to the above-described service provider index, and is, for example, the number of users of the service. That is, the number of users of the service Pα, i (i ∈ [1, n_α]), which is the total number of people using a specific service provided by a specific enterprise who operates a care facility, Consider the number of service users, which is the total number of people using any of the services provided by the business.

  The learning data preprocessing unit B08 has a function of calculating an auxiliary index (here, the service implementation rate) for the service provider from the above-described service provider associated index based on the learning data. Details of the calculation of the auxiliary index will be described later.

  The model learning unit B02 estimates a model that describes the state change indicator based on the learning data and the auxiliary indicator described above. That is, the state change index is used as the explained variable, and the service provider index and the auxiliary index (service implementation rate) calculated from the service provider associated index are estimated for each state change index as the explanatory variable. The estimated model is stored in the model storage unit B03.

  The state change index prediction unit B04 has a function of calculating a state change index by applying the service provider index and the auxiliary index to the model using the model stored in the model storage unit B03. The state change index prediction unit B04 assumes that the first state change index calculated by applying the current data based on the learning data stored in the learning data storage unit B01 to the model and that a certain measure is to be implemented. The target values of the service provider index and the auxiliary index, which are estimated, are applied to the model to obtain the second state change index calculated as the simulation result.

  The target auxiliary index setting unit B06 sets a value of an auxiliary index that is estimated to change by implementing the above-described measure as a target auxiliary index. The target index estimation unit B05 estimates, as the target index, the service provider index and the auxiliary index, which are updated by applying the target auxiliary index set by the target auxiliary index setting unit B06 to the service provider index and the auxiliary index. It is a thing. The service provider index and the auxiliary index estimated as the target index are used by the state change index prediction unit B04 to calculate the second state change index.

  The first state change index and the second state change index calculated by the state change index prediction unit B04 are stored in the prediction result storage unit B07.

  In the service effect evaluation system configured as described above, first, the flow of processing for estimating a model for explaining a change in the state change index from learning data will be described using the flow chart of FIG. The process of estimating this model is a process based on a specific period and a specific state.

  In FIG. 2, in step S01, the learning data stored in the learning data storage unit B01 is read into the learning data preprocessing unit B08. As described above, it is assumed that three types of data are provided for each target, such as a state change indicator, a service provider indicator, and a service provider incident indicator as described above.

  In step S02, the learning data preprocessing unit B08 calculates an auxiliary index by referring to the service provider associated index of the read learning data. For example, in the case where the number of users of service Pα, i and the number of service users are considered as service provider incidental indexes, the number of service Pα, i users is the service user as shown in the following equation (1) By dividing by the number of people, the service Pα, i implementation rate can be calculated as an auxiliary index corresponding to the service Pα, i.

  However, when the number of users is not given, the service P α, i implementation rate is calculated as 0 by determining that there is no user using the service.

  In step S03, the model learning unit B02 extracts a target state change index. At this time, if there is a state change index to be taken out, it is determined that the process is successful, and the process proceeds to step S04. On the other hand, when the state change index has already been extracted and there is no state change index to be extracted, it is determined that the processing has failed, and the processing in this flow is ended. For example, let us consider two state change indicators, such as duration and deterioration-improvement gradient. At this time, if step S04 is performed after the extraction of the two indices is completed, step S03 determines that the process has failed, and the process of this flow ends.

In step S04, the model learning unit B02 learns a model that describes the extracted state change index based on the service provider index and the service assistance index. For example, as shown below, a linear regression equation ( _denoted variables y ₁ and y ₂ as state change indicators, service provider indicators and service auxiliary indicators as explanatory variables x _k, and regression coefficients a _k and b _k ) 2) Consider equations (3).

被説明変数ｙ_１、ｙ_２は以下のとおりである。
ｙ_１：継続期間
ｙ_２：悪化-改善勾配 Here, the explanatory variable x _k is determined as follows.
x ₁ : Value of service P α, 1 Provider index x ₂ : Value of service P α, 2 provider index
|
x _{n_α} : service Pα, n_α Provider index value x _{n_α + 1} : service Pα_1 implementation rate x _{n_α + 2} : service Pα_2 implementation rate
|
x 2 _{n — α} : service Pα, n — α implementation rate The regression coefficients a _k and b _k are determined as follows.

The explained variables y ₁ and y ₂ are as follows.
y ₁ : duration y ₂ : worse-improvement gradient

The coefficients a _k and b _k of the linear regression equation described above can be calculated by applying the value of the state change index, the value of the service provider index, and the value of the service assistance index given to each business operator. It is. Here, the coefficients of the state change index representing the duration are modeled as a _k ( _k ∈ [0, 2n_α]) and the coefficients b _k ( _k指標 [0, 2n_α]) of the state change index representing the deterioration-improvement gradient. Be able to However, since there is a service provider index and a service assistance index for each service, the total number of explanatory variables is twice the number n_α of services in a specific care need status category α, and the total number of coefficients is a constant term And 2n_α + 1.

  In step S05, the model learning unit B02 stores the learned model corresponding to the extracted state change index in the model storage unit B03. Also, the model learning unit B02 returns the process to step S03. If the model is represented by a linear regression equation, 2n_α + 1 coefficients will be stored in the model storage unit.

  By doing as described above, a model based on a specific period and a specific care requiring condition category is created and stored for each condition change index.

  Next, the flow of the process of estimating the effect of the measure using a model will be described along the flow of FIG. However, in this flow as well, the effects of measures will be estimated on the basis of a specific period and a specific state.

  In step S11, the state change index prediction unit B04 reads the model learned from the model storage unit B03.

  In step S12, as in step S01, the learning data preprocessing unit B08 reads the learning data stored in the learning data storage unit B01. Further, in step S13, as in step S02, the learning data preprocessing unit B08 calculates an auxiliary index.

  In step S14, the target auxiliary index setting unit B06 receives an input of a specific target auxiliary index, and sets a value of a target auxiliary index that can be achieved when a specific measure is performed. For example, it is assumed that the target service implementation rate is set as the target service implementation rate, which is the target achievable when the specific measure is implemented, as the target auxiliary index.

  In step S15, the target indicator estimation unit B05 is updated based on the target auxiliary indicator set by the target auxiliary indicator setting unit B06, the service provider indicator, and the auxiliary indicator calculated from the learning data. Estimate the target indicators, which are service provider indicators and supplementary indicators. For example, assuming that a target service implementation rate is given as a target auxiliary index, consider a case where the value of each service provider index and the implementation rate of each service are estimated as a target index when the target service implementation rate is achieved. Make it

  In this case, it is assumed that among the users belonging to the specific condition requiring long-term care, users who do not use the service targeted by the measure come to use the service at random. At this time, the service implementation ratio by period, which is the ratio of users who have already used a specific service in a specific period, can be calculated as shown by the following equation (4). For this reason, the service-related index (the number of service users) who will newly use a specific service can be estimated by the following equation (5).

  Since the value of the estimated service incidental index is randomly distributed to each business operator, each business operator estimates the service according to the ratio of users who do not receive the service. Values of incidental indicators will be allocated. Therefore, the value of the term-specific / target-based service supplementary index, which is a service supplementary index (service implementation rate) in a specific period and a specific establishment, can be estimated by the following formulas (6) and (7).

  However, Formula (6) is a formula in case there are at least one user who is using the said service in a specific establishment, and the service-related index of the said service is not missing. Moreover, (7) Formula is a formula in the case of defect | deletion.

  On the other hand, it is assumed that the user who has newly used the service receives the average service that the user who has used the service has received so far. Therefore, the value of the service provider index indicating the quality of the service can be estimated by the average value of the service provider index in the service. That is, it is assumed that the service of the quality equivalent to the average service provider index calculated by equation (8) is received. For this reason, the value of the target service provider index, which is the service proponent index updated as a result of implementing the policy, can be estimated by the equations (9) and (10).

  However, Formula (9) is a formula in case there is at least one user who is using the said service in a specific establishment, and the service-related index of the said service is not missing, and Formula (10) is It is an expression in the case of missing.

  In step S16, the state change index prediction unit B04 extracts a state change index in the same manner as the model learning unit in step S03. At this time, when there is a state change index to be extracted, the process proceeds to step S17. On the other hand, when the state change index to be extracted does not exist, the process proceeds to step S19. For example, if there are two conditions, ie, the duration and the deterioration-improvement gradient, as the state change index, the process proceeds to step S17 until the processing of the duration and the deterioration-improvement gradient, and after processing the two, step S19. Go to

In step S17, the state change index prediction unit B04 applies, to a model corresponding to the extracted state change index, the service provider index provided as learning data and the auxiliary index calculated from the service provider associated index. As a result, the value of the first state change index, which is the first predicted value of the state change index, is predicted. For example, if the service provider indicator value and service implementation rate obtained for each service are applied to two models corresponding to duration and deterioration-improvement gradient, the applied service provider indicator and service implementation The duration _{y1, pred, A} and the worse-improved slope y2 _{, pred, A} depending on the rate can be predicted.

In step S18, the state change index prediction unit B04 is a second predicted value of the state change index by applying the target index extracted from the target index estimation unit B05 to the model corresponding to the extracted state change index. Predict the value of the second state change indicator. For example, when applying the value of the target indicator obtained for each service to the two models corresponding to the duration and the deterioration-improvement gradient, the duration y _{1, sim, A} corresponding to the applied target indicator And worse-improved slope y _{2, sim, A} can be predicted.

  In step S19, the state change index prediction unit B04 stores the predicted value of the first state change index and the value of the second state change index in the prediction result storage unit B07.

  By implementing the flow of FIG. 3 as described above, the effects of the measures when various target supplementary indicators are set can be evaluated by the value of the first state change indicator and the value of the second state change indicator. .

  Besides the above, various modifications can be made without changing the spirit of the present invention.

  For example, in the model learning unit B02, the model is described by a linear regression model, but the present invention is not limited to this. That is, any model that uses the state change index as the explained variable and the service provider index and the auxiliary index as the explanatory variable to describe the state change index can be described by a non-linear regression model. Furthermore, it can also be described by a model such as a neural network.

  In addition, although target auxiliary index setting part B06 set target auxiliary index assuming that measure is implemented to all establishments, setting target auxiliary indicator to a specific establishment is also possible. it can. At this time, the target index estimation unit B05 estimates the target index only for the relevant business operator, and the state change index prediction unit B04 calculates the effect of the business establishment as a final effect.

  In addition, the learning data preprocessing unit B08 can also perform processing such as correction of the value of learning data such as missing value complementation and normalization prior to calculation of the auxiliary index.

  Moreover, although the granularity of the object which comprises learning data is set as a business establishment, modeling and a prediction based on a user can also be performed by regarding one user as one business establishment.

As described above, according to the first embodiment, it is possible to model the service provided by the care provider and the change in the state of the service user, and providing the model by using the model It is possible to evaluate the change of the state change indicator accompanying the change of service.
Second Embodiment

  Next, a service effect evaluation system according to a second embodiment of the present invention will be described with reference to a functional block diagram shown in FIG. 6 and a flowchart shown in FIG. However, the description is abbreviate | omitted regarding the same part as 1st Embodiment.

  In the configuration shown in FIG. 6, a target auxiliary index effect prediction unit B09 and a reference data storage unit B10 are newly added to the configuration of the first embodiment shown in FIG.

  The reference data storage unit B10 is associated with the status category of the service user, and stores status category indicators set for each category. As the condition classification indicator, the condition classification of the service user, for example, the condition requiring care needs to be used. That is, as shown in FIG. 8, it is necessary to consider the eight categories of support unnecessary, required support 1, required support 2, required care 1 to required care 5 as persons in need of support as a need of care status division. Do. And the value of the state division index is set for each of these divisions.

  As the value of the condition classification index, for example, a care cost set for each condition classification is used. In this case, the average value of the care expenses generated for the users belonging to the specific care requiring condition category α is considered. That is, as shown in FIG. 8, the care which occurs in the user of the corresponding care division about each of 8 divisions from support unnecessary, support required 1, support required 2, care required 1 to care 5 as care requiring condition division The average cost is set. In addition, since the user who belongs to support unnecessary is not using the service, care cost is not generated, and it is assumed that 0 is given as an average value.

  The target auxiliary indicator effect prediction unit B09 generates the first condition change indicator and the second condition change indicator based on the first condition change indicator and the second condition change indicator, and the condition classification indicator stored in the reference data storage unit B10. From the difference from the condition change indicator of 2, calculate the effect by applying the target auxiliary indicator. When the care cost set according to the state classification is used as the value of the state classification index, the effect described above is an increase or decrease in the care cost.

  A process of predicting the target auxiliary index effect will be described with reference to FIG. In FIG. 7, in step S <b> 30, the target supplementary index effect prediction unit B <b> 09 reads reference data composed of state classification indicators from the reference data storage unit B <b> 10. The condition classification index is an index related to the condition of the service user as described above, and in the case of the reference data shown in FIG. And the reference data is read.

  In step S31, the target supplementary indicator effect prediction unit B09 determines the target based on the estimated value of the first condition change indicator, the value of the second condition change indicator, and the value of the condition category indicator given as reference data. Evaluate the effects of implementing measures to reach the value of the supplementary indicator as the value of the state classification indicator. For example, when the duration and the deterioration-improvement gradient are given as the state change indicator, the effect is based on the difference between the estimated two durations and the difference in the care need status category based on the deterioration-improvement gradient Evaluate

  9 to 14 are "prediction" corresponding to the first state change index predicted based on the current data, and "simulation" corresponding to the second state change index predicted based on the target auxiliary index setting. It is an explanatory view for predicting an effect based on a difference with. That is, for each of the “prediction” and the “simulation”, the effects of the duration divided and the “deterioration” / “improvement” of the state according to the positive / negative value of the deterioration-improvement gradient It is a figure explaining the trial calculation method. However, if the deterioration-improvement gradient is 0, it is regarded as "deterioration" of the condition.

The effects of each of FIGS. 9 to 14 will be described below using the corresponding equations (11) to (16). The symbols used in the equations (11) to (16) will be described below.

（１１）式は、予測及びシミュレーションが改善する場合における事業所Ａの効果の例を示す式である。 FIG. 9 shows that “prediction” and “simulation” are “improvement”, and “simulation” indicates that the duration of the long-term care condition category is shorter. In this case, "simulation" can obtain a positive effect by moving to the lower care need status category more quickly. The effect in this case is calculated by the following equation (11).

Equation (11) is an equation showing an example of the effect of the establishment A when prediction and simulation improve.

（１２）式は、予測及びシミュレーションが悪化する場合における事業所Ａの効果の例を示す式。 FIG. 10 shows that "prediction" and "simulation" are "deterioration", and the "simulation" has a longer duration of the care requiring condition category, so that the transition to the higher care requiring condition category is made more positive. It shows that the effect of can be obtained. The effect in this case is calculated by the following equation (12).

The equation (12) shows an example of the effect of the establishment A when the prediction and the simulation deteriorate.

  FIG. 11 and FIG. 12 show the positive effects when “prediction” deteriorates and “simulation” improves. That is, FIG. 11 shows the effect of shifting to the lower care need status category earlier when the duration of “simulation” is shortened. Also, FIG. 12 shows the effect of shifting to a lower care requiring status category later when the duration of the “simulation” becomes longer. The effect of FIG. 11 is calculated by equation (13), and the effect of FIG. 12 is calculated by equation (14).

（１３）式は、予測が悪化し、シミュレーションが改善する場合で、予測の継続期間が長い場合における事業所Ａの効果の例を示す式である。

（１４）式は、予測が悪化し、シミュレーションが改善する場合で、予測の継続期間が短い場合における事業所Ａの効果の例を示す式である。

Expression (13) is an expression showing an example of the effect of the establishment A when the prediction is deteriorated and the simulation is improved and the duration of the prediction is long.

Expression (14) is an expression showing an example of the effect of the establishment A when the prediction is deteriorated and the simulation is improved and the duration of the prediction is short.

  FIG. 13 and FIG. 14 show negative effects when “prediction” is improved and “simulation” is deteriorated. That is, FIG. 13 shows the effect of transitioning to a higher care requiring status category earlier when the duration of the “simulation” is shorter, and FIG. 12 is later when the duration of the “simulation” is longer. It shows the effect of transitioning to a high care need status category. The effect of FIG. 13 is calculated by equation (15), and the effect of FIG. 14 is calculated by equation (16).

（１５）式は、予測が改善し、シミュレーションが悪化する場合で、予測の継続期間が長い場合における事業所Ａの効果の例を示す式である。

（１６）式は、予測が改善し、シミュレーションが悪化する場合で、予測の継続期間が短い場合における事業所Ａの効果の例を示す式である。

Expression (15) is an expression showing an example of the effect of the establishment A in the case where the prediction improves and the simulation deteriorates and the duration of the prediction is long.

Expression (16) is an expression showing an example of the effect of the establishment A when the prediction improves and the simulation deteriorates and the duration of the prediction is short.

  Thus, based on FIG. 9 to FIG. 14, the difference of the duration and the difference of the classification of the care requiring condition category obtained corresponding to the deterioration-improvement gradient are integrated by the equations (11) to (16) The effect according to the case of each figure can be calculated as a care cost.

  In Equations (11) to (16), [] is a floor function that calculates the largest integer that is less than or equal to the specified real value, and is a function introduced because the nursing care status category is given as a discrete value It is. In addition, min () and max () are in need of nursing care status division 0 (support not required), 1 (support required 1), 2 (support required 2), 3 (care required 1), 4 (care required 2), When expressed as 5 (necessary care 3), 6 (necessary care 4), 7 (necessary care 5), the smaller value introduced to limit the value to the range of 0 to 7, and It is a function to select the larger value.

A measure to reach the value of the target support indicator by integrating the effects _{fe and ZA} (A) for each business operator calculated in this way, as the target support index effect prediction unit B09 indicates by equation (17) Can be evaluated as the value of the state division indicator.

  In step S32, the target assistance index effect prediction unit B09 stores the effect of the evaluated policy in the prediction result storage unit B07.

  By implementing the flow of FIG. 7 as described above, it is possible to evaluate the effect of the measure in the case where various target supplementary indicators are set as the value of the state division indicator.

  Besides the above, various modifications can be made without changing the spirit of the present invention. For example, although the target auxiliary index effect prediction unit B09 estimates the effect based on the difference between the predicted state change indexes, the effect can be obtained for each period by explicitly specifying the length of the period from the outside. It is also possible to estimate the effect by estimation.

  Furthermore, although the target auxiliary index effect prediction unit B09 calculates the effect by equally weighting and integrating the effects for each business site, it is also possible to integrate the effects using the number of users of the service at the business site as a weight. it can.

  Moreover, in the target supplementary indicator effect prediction part B09, after calculating the care requiring condition classification individually in the first condition change indicator and the second condition change indicator, the difference of the care requiring condition classification is calculated, but the difference Can be discretized after taking

  In addition, although the target auxiliary index effect prediction unit B09 does not take into consideration the cost generated with the implementation of the measure, the final effect is calculated by subtracting it from the effect obtained by the implementation of the measure. It can also be done.

As described above, according to the second embodiment, it is possible to model the service provided by the care provider and the state change of the service user, and providing the model by using the model Can be evaluated as the value of the status indicator.
Third Embodiment

  Next, a service effect evaluation system according to a third embodiment of the present invention will be described with reference to a functional block diagram shown in FIG. 15 and a flowchart shown in FIG. However, the description is abbreviate | omitted regarding the same part as 1st Embodiment.

  In the configuration shown in FIG. 15, a learning data selection unit B11 and a modeling range data storage unit B12 are newly added to the configuration of the first embodiment shown in FIG.

  The modeling range data storage unit 12, as shown in FIG. 17, is a modeling that specifies an object-specific modeling index that describes features of each object related to the modeling range of learning data, and a modeling range. It consists of indicators and

  For example, as shown in FIG. 17 (a), the target classified modeling index is a target classified modeling index relating to “area”, “scale”, and “form” for each of the establishments A, B,. A value has been set. That is, in the present embodiment, since modeling is performed with the business establishment as an object, values of target-based modeling indexes that characterize the business establishment are set. When the user is regarded as an object, characteristics related to the user such as "sex" and "age" are set as the object-specific modeling index.

  As the modeling index, for example, as shown in FIG. 17B, the values of the modeling index regarding “area”, “scale”, and “form” are set.

  The learning data selection unit B11 selects an object to be used for learning a model based on the object-specific modeling index and the modeling index shown in FIG. Then, based on the selection result, the learning data storage unit B01 has a function of causing only the selected learning data to be output to the learning data preprocessing unit B08. Hereinafter, this model selection process will be described with reference to FIG.

  In FIG. 16, in step S41, the learning data selection unit B11 reads an object-specific modeling index out of data stored in the modeling range data storage unit B12. Assuming that the data shown in FIG. 17A is stored in the modeling range data storage unit B12 as the object type modeling index, the object type modeling index relating to “area”, “size”, and “form” The value will be read for each establishment.

  In step S42, the learning data selection unit B11 reads a modeling index among the data stored in the modeling range data storage unit B12. For example, if the data shown in FIG. 17B is stored in the modeling range data storage unit B12 as a modeling index, the values of the modeling index relating to “area”, “scale”, and “form” are It will be read.

  In step S43, the learning data selection unit B11 selects a target to be used for modeling by applying the modeling index to the target-based modeling index. For example, consider a case where only business establishments having values of the target-based modeling index completely matching the values of the modeling index are to be modeled. At this time, assuming that an object-specific modeling index and a modeling index shown in FIGS. 17A and 17B are given, each of “area”, “scale”, and “form” specified by the modeling index is provided. Establishments in which the value and each value of the targeted modeling index all match will be selected for modeling.

  In step S44, the learning data preprocessing unit B08 reads, from the learning data storage unit B01, learning data corresponding to the target selected by the learning data selection unit B11.

  By implementing the flow of FIG. 16 in this manner, the effects of the measures when various target supplementary indicators are set for the objects existing in the designated modeling range, the value of the first state change indicator, and (2) It can be evaluated by the value of the state change index.

As described above, according to the third embodiment, it is possible to model the service provided by the care provider and the state change of the service user from the learning data existing in the designated range, By using the model, it is possible to evaluate the change of the state change indicator accompanying the change of the service to be provided.
Fourth Embodiment

  Next, a service effect evaluation system according to a fourth embodiment of the present invention will be described according to a functional block diagram shown in FIG. However, the description is abbreviate | omitted about the same part as 3rd Embodiment.

  The configuration shown in FIG. 18 is obtained by adding the target auxiliary index effect prediction unit B09 and the reference data storage unit B10 described in FIG. 6 to the configuration of the third embodiment shown in FIG.

  The flow of processing in the fourth embodiment can be realized by combining FIG. 7 and FIG. That is, by performing the process of the flowchart of FIG. 16 after the process of the flowchart of FIG. 7 is completed, the process of the fourth embodiment can be realized.

  According to the fourth embodiment, both effects obtained in the second embodiment and the third embodiment can be obtained.

In addition to the features in each of the embodiments described above, the other features of the present invention include the following.
・ We designed a model that describes the change in the state of needing care as a gradient that indicates the duration and the degree of change of the specific state of needing care.
・ We devised a device to explain the change in the state of need of nursing care by the index indicating the goodness of the service and the index accompanying the service such as the service implementation rate.
・ The average service is provided to the users of the randomly selected service, the value of the index that indicates the quality of the service that changes when the service is changed, and the value of the index that accompanies the service We made a device to estimate with the assumption of

  While certain embodiments of the present invention have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and the gist of the invention, and are included in the invention described in the claims and the equivalent scope thereof.

B01 ... learning data storage unit B02 ... model learning unit B03 ... model storage unit B04 ... state change index prediction unit B05 ... target index estimation unit B06 ... target auxiliary index setting unit B07 ... prediction result storage unit B08 ... learning data preprocessing unit B09 ... Target auxiliary index effect prediction unit B10 ... Reference data storage unit

Claims (5)

A state change index which is an index for evaluating the degree of change of the state of the service user, a service provider index which is an index for evaluating the quality of the service provided by the service provider, an index accompanying the service provider index A learning data storage unit for storing data constituted by a certain service provider incidental index as learning data for each target;
A learning data preprocessing unit that calculates, from the service provider incidental index, an auxiliary index for the service provider including the service implementation rate based on the learning data;
Based on the learning data and the auxiliary index, a model for explaining the state change index is estimated for each of the state change indices, with the state change index as an explained variable and the service provider index and the auxiliary index as an explanatory variable. Model learning unit,
A model storage unit for storing the estimated model;
A target supplementary indicator setting unit that sets a target supplementary indicator that is a target assuming the supplementary indicator;
A target indicator estimation unit that estimates, as a target indicator, the service provider indicator and the auxiliary indicator updated by applying the target auxiliary indicator to the service provider indicator and the auxiliary indicator;
A state change index prediction that calculates a first state change index by applying the service provider index and the auxiliary index to the model, and calculates a second state change index by applying the target index to the model Department,
A prediction result storage unit that stores the first state change index and the second state change index;
The service effect evaluation system characterized by having. A reference data storage unit that stores a state classification index, which is an index set according to the classification, related to the classification of the state of the service user;
A difference between the first state change index and the second state change index based on the first state change index and the second state change index, and the state classification index stored in the reference data storage unit A target auxiliary indicator effect prediction unit that calculates an effect of applying the target auxiliary indicator from the
The service effect evaluation system according to claim 1, further comprising: A modeling range data storage unit configured of an object-specific modeling index describing features of each object related to the modeling range of the learning data, and a modeling index specifying the range of the modeling;
The target to be used for learning the model is selected based on the target classified modeling index and the modeling index, and only the learning data selected from the learning data storage unit is selected based on the selection result. A learning data selection unit to be output to the learning data preprocessing unit;
The service effect evaluation system according to claim 1 or 2, further comprising:   The service effect evaluation system according to claim 1, wherein the service provider index for evaluating the quality of the service is the provision amount of the service, and the service provider incident index is the number of users of the service.   The service effect evaluation system according to claim 2, wherein the condition classification index is a care cost set according to the condition classification, and the effect is an increase or decrease in care cost.

Images