JP7218850B2 - Record management system, record management device, document approval device and document creation device, method, program and recording medium recording the program - Google Patents

Description

The present invention provides a record management system, a record management device, a document approval device, a document creation device, a record management method, a document approval method, a document creation method, a record management program, a document approval program, a document creation program, and these programs. It relates to a recorded recording medium.

Facilities that provide services such as nursing care, medical care, and childcare to the elderly, people with disabilities, and children are always required to provide better and more appropriate services. In particular, it is preferable to be able to detect abnormal behavior of the subject in advance and deal with it.

In the care system described in Patent Document 1, various sensors are used to detect abnormal situations of care recipients, and reference is made to care information accumulated in a database that has been performed in the past to inform the caregiver of appropriate work details. It is. The compliance evaluation support system described in Patent Literature 2 attempts to evaluate possible risks in advance at the stage of preparing work instructions and care plans.

JP-A-2002-149825 JP 2012-175170 A

On the other hand, in the above-mentioned facilities, various records such as case records, nursing care diaries, nursing diaries, care plans, etc. are kept using a computer system generally called a record management system. These records contain a large number of examples and can be a living teaching aid for proper service. Much can be learned from these records, especially for inexperienced caregivers and others, provided that appropriate advice is given by experienced practitioners.

In this invention, an experienced staff (chief, manager, etc.) always gives approval (electronically seals) to nursing care diaries, work diaries such as nursing diaries, and other report documents created by the record management system. ), I always look at the documents (records such as diaries) that have been created when entering approval, so I will check the contents of the records (status of the subject, environment, etc.). It is possible to give advice on the possibility of situations that can occur in the target person, environment, etc. by looking at important or specific keywords representing such as, etc.). is.

In one form, the record management system according to the present invention comprises an approval request display means for displaying an approval request document including an approval column, an advice input corresponding to a keyword included in the displayed approval request document, and an approval request document. Approval means for outputting, in response to an approval input, a training data set containing document content including the keyword of the requested document and the received advice, and corresponding to the keyword based on the learning data set output from the approval means Learning means for learning the degree of adequacy of advice, document creation means for outputting a document on the screen, and outputting advice matching keywords contained in the document output from the document creation means based on the learning results of the learning means and the document preparation means displays advice (guidance) output from the guidance means on a document output screen.

In another form, the record management system according to the present invention includes an approval request display means for displaying an approval request document including an approval column, an advice input corresponding to a keyword included in the displayed approval request document, and an approval request document. Approval means for outputting a learning data set containing document contents including keywords of a requested document and accepted advice to a learning means for learning the degree of conformance of advice corresponding to keywords in response to an approval input; Document creation means for outputting a document on the screen and guidance means for outputting advice matching keywords contained in the document output from the document creation means based on the learning results of the learning means, the document creation means The advice (guidance) output from the guidance means is displayed on the document output screen.

In still another form, the record management system (apparatus) according to the present invention comprises: approval request display means for displaying an approval request document including an approval column; and outputs a learning data set containing the document content including the keywords of the approval request document and the accepted advice to a learning means that learns the degree of relevance of the advice corresponding to the keywords in response to the approval input. In addition to outputting the approval means and the document displayed on the screen to the guidance means, the advice (guidance) output from the guidance means based on the learning results in the learning means that matches the keywords contained in the document is displayed on the screen. It is provided with the document preparation means displayed above.

In any case, a document creator (staff) creates (inputs) various record documents such as daily reports, monthly reports, and reports on a computer. In this document, the actions, behaviors, conditions, environmental problems, etc. of subjects (care recipients, persons with disabilities, hospitalized patients, children, etc.) are recorded in characters (words, sentences). When a document requiring approval (approval) from the superior is created, an approval column is created and an approval request is made to the superior on the system (approval workflow works) (approval request display means). When approving, the supervisor reads the document to be approved, looks at the contents described therein, and considers the possibility of occurrence regarding the target person, environment (equipment, organization, individual staff), etc., and matters that staff should pay attention to. Enter advice and give approval (decision) (approval means). In response to the input of approval by the supervisor, the approved document and the input advice are sent to the learning means (learning system, learning server). Advice is positioned as teacher data for learning. Specific terms in the approved document (keywords representing the subject's condition, behavior, behavior, environmental problems, etc.) and advice, which is training data, are used as a learning data set in machine learning methods to determine the suitability of advice. degree is learned. Learning results are stored in a learning database. If there are multiple approvers, preferably the document and advice is sent to the learning means upon all approvals or when the highest approver has provided approval input.

Staff routinely create various documents on computer screens, or display and read documents created in the past or created by others (including document creation means and document viewing). The document prepared by the staff is sent to the guidance means when the user is inputting the document, browsing the document, completing the input, or pressing the guidance button. Based on specific terms (keywords) contained in documents created by staff, the guidance means retrieves advice or guidance that matches the keywords from the learning means (learning database), and the computer where the staff inputs the documents. (guidance means). The transmitted advice or guidance is displayed on the computer. In this way, the staff can receive appropriate advice (regarding the target person, environment, etc. recorded in the document) for the document they are inputting (inputted or browsed). An approver (staff) who gives approval is a skilled person with a lot of experience and knowledge. Advice input by an expert is used as teacher data for learning. Furthermore, even in the case of inexperienced staff, appropriate guidance or advice can be provided by learning results in relation to keywords related to the target person, environment, etc. in the documents that the staff is inputting or inputting or viewed. will be given.

In this way, according to the record management system of the present invention, it is possible to input teaching data for learning at the timing of giving approval, and since the person giving approval is a skilled person, appropriate teaching data can be obtained. will be entered. On the other hand, when other staff (including approvers) are inputting documents, etc., appropriate advice and guidance based on learning results are displayed, so even inexperienced staff can Able to learn or acquire knowledge in input and consider precautions. In addition, staff can obtain detailed guidance information, including relevant manuals, if needed. In this way, knowledge and experience can be transferred from those with a lot of experience and knowledge to inexperienced staff.

The document approval apparatus according to the present invention comprises: approval request display means for displaying an approval request document including an approval column; Learning including the document content including keywords of the displayed approval request document and the received advice in response to the second input of the series of approval inputs while receiving the advice input on the advice input screen. and learning data output means for outputting the data set for learning to the learning means. This document approval device can be used by an approver to input advice as teacher data.

The document preparation apparatus according to the present invention comprises prepared document display means for displaying a prepared document on the screen, prepared document output means for outputting the document displayed on the screen to the guidance means, and learning means based on the learning results. An advice display means is provided for displaying on a screen advice (guidance) that is output from the guidance means and matches the keyword contained in the created document. The document preparation devices (including document viewing devices) allow staff to obtain appropriate advice or guidance while entering, viewing, or entering documents. The present invention also provides a record management method, a document approval method, a document creation method, a program, and a recording medium recording the program, which have the same features as those described above.

1 is a block diagram showing the entire record management system; FIG. 4 is a flow chart showing record document creation processing in the record management client. 10 is a flow chart showing the flow of decision (approval) processing and learning processing in the entire record management system; 10 is a flow chart showing the overall flow of learning processing (processing 1 to 3) in the learning server; 4 is a flowchart showing details of learning processing 1 in the learning server; 10 is a flowchart showing details of learning processing 2 in the learning server; 10 is a flowchart showing details of learning processing 3 in the learning server; 4 shows an example of a document creation screen in a record management client. An example of the document creation screen in the record management client is shown, showing the state of approval by the supervisor. An example of a document creation screen in the record management client is shown, showing a state in which a comment input screen is displayed when the manager approves the document. An example of a document creation screen in a record management client is shown, showing a state in which a section manager is inputting comments for approval. An example of the document creation screen in the record management client is shown, showing the state of approval by the manager. 4 shows an example of data transmitted from a recording management client to a study guidance server (morphological analysis server); An example of a keyword master in a learning server (learning database) is shown. An example of a class classification master in a learning server (learning database) is shown. 4 shows an example of a learning data group (table) in a learning server (learning database); 10 is a flow chart showing the flow of document creation (document viewing) processing and associated guidance display in the entire record management system; 10 is a flowchart showing the first half (process 4) of guidance creation processing in the guidance server; FIG. 10 is a flowchart showing the second half (processing 5) of guidance creation processing in the guidance server; FIG. An example of a document creation (browsing) screen in the record management client is shown. An example of a document creation (browsing) screen in a record management client is shown, showing how guidance is displayed in the guidance area. An example of the document creation (browsing) screen in the record management client is shown, showing how the manual is displayed in the window. 4 shows an example of data sent from a record management client to a morphological analysis server; 4 shows an example of data (messages) sent from the guidance server to the record management client;

FIG. 1 shows the configuration of the entire record management system.

In general, a plurality (or even one) of record management clients (computers) (systems) 11 are provided in the record management system 10 . These record management clients 11 can mutually communicate with the record management server 12 . The record management server 12 has a record management database 13 .

The record management client 11 functions as approval request display means, approval means (advice input guidance means, learning data output means), document creation means (document creation screen display means, created document output means, advice display means) and the like. The record management client 11 is also called a document approval device or a document creation device. The records management client 11 alone or together with the records management server 12 and the records management database 13 can also be thought of as a records management system or a records management device. The record management server 12 may form part of the approval means (learning data output means), or may form part of the document creation means (created document output means).

The entire record management system 10 includes a morphological analysis server 14, a learning server (learning means) 15 and a guidance server (guidance means) 17. The learning server 15 is accompanied by a learning database 16, and the guidance server 17 Manual database 18 is attached. Learning database 16 is accessible from learning server 15 and guidance server 17 . Guidance server 17 can access manual database 18 . A set of these servers 14 , 15 , 17 and databases 16 , 18 is sometimes called a learning guidance server (system) 19 .

In one embodiment, the study guidance server 19 can be implemented using a group of servers on the Internet called cloud (cloud computing). The record management server 12 and record management database 13 can also be implemented using cloud computing. In this case, the later-described processing program executed by each server is uploaded to the cloud. The morphological analysis server 14 and the learning server 15 can also construct this record management system using an existing morphological analysis server and learning server, respectively.

Details of the functions or operations achieved by executing the programs installed in each server that constitutes the record management system 10 are described below.

First, the recording document creation process will be described with reference to FIG.

Recording document creation is a process (operation) that is routinely performed by staff (persons in charge such as caregivers) using the record management client 11 in nursing care facilities, hospitals, nursery schools, and the like. An example of the record document created is shown in FIG. Referring also to FIG. 8, the staff launches a record edit screen on the record management client 11, and selects a report to the care manager (care manager report) from the form menu in the right column of the screen. The staff selects the target person's name (user: Taro Shizuoka), their own name (reporter: Ichiro Tanaka), date, etc. from the pull-down list on the screen using the mouse or inputs it from the keyboard, etc. Then, enter the sentence (text data) of the "remarks" (Fig. 2, S11). A special note is the content of the report (document content). Note that boxes (buttons) for adding, saving, duplicating, deleting, etc. are also displayed at the bottom of this record editing screen.

Documents created by staff include documents that require approval by superiors (administrators, etc.) and documents that do not. An approval column (approval column) A is displayed on the record editing screen for a document that requires approval. Documents that require approval include various work diaries and various reports. The approval column is not displayed for documents that do not require approval. The approval field is controlled by the approval field control (program) as described below.

The number of approvers displayed in the approval column A differs depending on the type of document. Here, for simplicity, a box is provided in which two approvers can be specified. The staff inputs Sato, the chief, and Suzuki, the section manager, as approvers (Fig. 2, S12) (creation and display of an approval request document). By inputting and specifying the approver's name in this way, the document on the record editing screen becomes an approval request document. If there are multiple approvers, the order of approval (approval order) is determined. In the example of FIG. 8, the order is the chief (Sato) and then the section chief (Suzuki). Approval workflow works in this order of approval.

Approval workflow itself is a well-known technology. For example, an approval request document is sent to the record management server 12, and the approver and approval order data are saved. The approval request document is stored in the record management database 13 . Then, when the approver logs in from the record management client 11, the record management server 12 sends the record management client 11 a notification that the approver has been designated as the approver for the specific document. The approver confirms whether or not there is a document to be approved by the approval status viewer in the record management client 11 . The approver activates the record editor and displays the document requiring approval on the records management client 11 . Staff members and approvers each have their own record management clients, and if these record management clients are connected via a network, the staff who created the approval request document will be notified that they have been designated as approvers. The client 11 may directly transmit to the client 11 of the approver.

Next, with reference to FIG. 3, machine learning processing including approval processing in the record management client 11 and server 12 and morphological analysis processing by the morphological analysis server 14 and learning server 15 will be described.

When the approver, Chief Sato, clicks the box (seal column) a1 in the approval column A on the record management client 11 and confirms the approval, Sato's approval seal is displayed in the box a1 as shown in FIG. This completes the approval (approval) by Chief Sato (Fig. 3, S21, S22). As will be described later, it is assumed that Chief Sato has tagged (inputted advice) "risk of dementia" by looking at the sentences of special notes (contents of the document).

Since the order of approval is Chief Sato and Manager Suzuki, when Chief Sato completes approval, the record management client 11 of Manager Suzuki is notified that there is a document to be approved. Section Manager Suzuki causes the document to be approved to be displayed on the display screen of the record management client 11 (Fig. 3, S21) (display of approval request document). Section Manager Suzuki then clicks the box (seal column) a2 under his name in the approval column A. Then, a comment (advice) input screen B is displayed as shown in FIG. 10 (advice input guidance (means)).

This comment input screen includes an approval status pull-down list box b1, an approver name display box b2, an approval date box b3, a tag pull-down list box b4, an add button b5, a confirm button b6, a cancel button b7, and the like. Is displayed. Approver Manager Suzuki selects approval in the approval status box b1. Naturally, the approver reads the special notes (report content, document content) of the report to the care manager and considers what advice should be given. In this example, we focus on the terms "phlegm-related" and "choking" (referred to as keywords as described later), and select "dysphagia risk" (see Figure 11) from the tag pull-down list. Thus, when the approver reads the content of the report and selects one or a plurality of tags (advice) that are appropriate comments (advice), the selected tags are accepted by the client 11 as training data for machine learning. If there is no appropriate comment (advice) term in the tag list, the comment can be added by pressing the add button b5. Added comments are also accepted as teacher data. When the approver presses the confirmation button b6, the report to the care manager is approved (settled) (Fig. 3, S22). The advice includes not only the target person (care recipient, etc.) but also various environments (equipment, organization, individual staff), etc. according to the content of the approval request document.

As a result, the comment input screen B disappears, and as shown in FIG. Since all the members displayed as approvers in the approval column A have entered approval (YES in S23 in FIG. 3), the document (report to care manager) displayed on the record edit screen is approved (approved). It becomes a completed document.

In response to the approval input by the last (all) approvers, the record management client 11 sends the approved document and the input tag to the record management server 12 (Fig. 3, S24) (Approval means, learning data output means). In response to this, the record management server 12 transmits the approved document and the input tag (these constitute the learning data set) to the morphological analysis server 14 (learning guidance server 19) (approval means, learning data output means) (Fig. 3, S25).

An example of transmission data transmitted from the record management server 12 to the morphological analysis server 14 (study guidance server 19) is shown in FIG. This transmission data includes the destination, document type (care manager report = report to care manager), and the total number of recorded documents that have been returned so far (3000 as an example, so the document (record) in Figure 13 is 3001). ), the content of the report ("special notes", document content), the name of the approver (authorizer), the position, the tag (comment) entered by the approver, and the weight value of the approver . In the previous example, Manager Suzuki has entered the advice (comment) (tag) "Risk of dysphagia" and Chief Sato has entered "Risk of dementia." The weighting values are the weights of the approvers, with a weight of 0.7 for the “risk of swallowing disorder” entered by Manager Suzuki and a weight of 0.3 for the “risk of authentication” entered by Chief Sato. be done. The weighting represents the likelihood of the advice input by the approver, is determined in advance in consideration of the experience and knowledge of the approver, and is stored in the record management client 11 or server 12 for each approver.

Reports to care managers are listed as approved documents, but there are also intake sheets (interview sheets), face sheets (detailed personal information), care plan sheets, daily reports, monthly work diaries, etc. .

The morphological analysis server 14 decomposes the text of the document content (the document content of FIG. 13, i.e., the special notes of FIG. 8, etc.) in the approved document sent from the record management server 12 into morphemes, and determines the parts of speech, etc. It is the server that does the processing. Using the document content of FIG. 13 as an example, verbs and nouns (compounds of these if necessary) are extracted. For example, the terms include vital measurement, tachycardia, stable blood pressure, participating (participating), phlegm-related, choking, complete eating, paper pants, failure to excrete, and the like. These terms are called keywords (hereinafter, as examples of keywords, for the sake of simplicity, it is assumed that phlegm-related, choking, tachycardia, and excretion failure are extracted from the document in FIG. 13). The morphological analysis server 14 is provided with a dictionary containing terms frequently used in fields related to the facility where this record management system is placed (nursing care, medical care, kindergarten education, etc.). It may be one that extracts only. In addition, since verbs and nouns, their combination words, and compound words may have the same meaning (for example, phlegm-related and phlegm-related), a processing function is provided to unify words with the same meaning into either business cards or verbs. You can have it.

The document content containing the above keywords and the tags (comments, advice) (teaching data) entered at the time of approval for it constitute the learning data set. A set of tags (comments, advice) may be called a learning data set. These learning data sets are sent to the learning server 15 (Fig. 3, S26).

The learning server 15 performs machine learning with teacher data. Here, as an example, using Bayes' theorem, the classification with the highest probability of occurring exceed).

Bayes' theorem is expressed by the following equation.

If P(A) > 0, then
P(B|A)=P(A|B)×P(B)/P(A) ‥‥Formula (1)

here,
P(A) : Probability that event A will occur P(B) : Probability that event B will occur P(A|B) : Probability that event A will occur after event B has occurred P(B|A) : Event A has occurred is the probability that event B will occur later.

Applying this Bayes' theorem to a learning data set containing keyword A and advice B yields the following.

P(A) : Percentage of keywords (words) A in all P(B) : Percentage of tags (advice) B in all P(A|B) : When tag (advice) B applies contains the keyword A P(B|A) : Probability (appropriate) that the tag (advice) B applies when the keyword A is included

Here, the total refers to the total (number of) approved (settled) documents (records).

As an example, when the word "choking" is included, the probability P(B|A) of being a "record (document) regarding the risk of dysphagia" can be obtained by the following calculation.

Probability that the word “choking” is included in “records on the risk of dysphagia” P(A|B) × Proportion of “records on the risk of dysphagia” in all records P(B) ÷ all Percentage P(A) of records containing the word choking

14 to 16 show the keyword master (FIG. 14), class classification master (FIG. 15) and learning data group (table) (FIG. 16) stored in the learning database 16. FIG. The numerical values representing each ratio described in these masters and data groups are values calculated using past 3000 records (documents). A record means an approved (approved) document, and the number of records is the number of documents.

Referring to FIG. 14, the keyword master stores the ratio P(A) of the approved (approved) records (documents) containing the keyword corresponding to the keyword.

Referring to FIG. 15, the class classification master stores the rate P(B) of approved (approved) records (documents) tagged for each class classification. A tag (advice, comment) is called a class classification here, and a class classification code is assigned to each tag. Nursing care facilities are divided into categories according to the typical or representative (managed) behavior or condition of care recipients, and each category is assigned a category code. A plurality of classifications exist within each division. Further, the class classification master describes the URL of the site where the manual describing the useful information (action policy, precautions, etc.) to be provided to the staff for each class classification is stored.

Figure 16 shows a data group during or obtained by machine learning. The classification code, class classification code, and the percentage P(B) of tagged approved records are the same as those shown in FIG. Corresponding to the class classification, each keyword (A), the percentage of approved records containing the keyword P(A), and the keyword (A) included in the approved records tagged with the class classification (B) The rate P(A|B) and the rate P(B|A) with which the class classification (B) is tagged when the keyword (A) is included are stored.

Details of the processing in the learning server 15 of S27 shown in FIG. 3 are shown in FIGS. Details of process 1 of S31 in FIG. 4 are shown in FIG. 5, details of process 2 of S32 are shown in FIG. 6, and details of process 3 of S33 are shown in FIG.

Processing 1 (S31) in FIG. 4 is processing for updating the keyword master in FIG. Referring to FIG. 5, keywords are obtained one by one from the keyword list obtained as the analysis result of the morphological analysis server 14, and the processing of S42 to S45 is repeated (S41). The keyword list includes keywords such as phlegm-related, choking, etc., as described above. It is checked whether the acquired keyword exists in the keyword master (FIG. 14) (S42, S43), and if there is a corresponding keyword, the ratio P(A) of records containing the keyword is taken out and the value is updated. For example, the P(A) of the keyword "choking" is 0.0020, which is the value up to 3000 total records. That is, among the 3000 records, the number of records in which the keyword "choking" appears is 6 by multiplying 3000 by 0.0020. This time, since there was a record (document) containing the keyword "choking", P(A) was newly
P(A) = (6 + 1) cases / 3001 cases = 0.0023 Formula (2)
becomes.

In the keyword master, P(A) of the keyword "choking" is updated from 0.0020 to 0.0023 (S45).

If the corresponding keyword is not found in the keyword master (unregistered), the keyword is added to the keyword master, the ratio P(A) of the included records is calculated, and written to the keyword master (S44). .

As a result of the morphological analysis, for example, if the keyword "has tachycardia" is extracted, the keyword is not registered in the keyword master of Fig. 14, so the ratio P(A) is P(A) = 1 case / 3001 = 0.0003 ‥‥Formula (3)
, and this ratio of 0.0003 is registered in the keyword master together with the keyword "has tachycardia."

Processing 2 (S32) in FIG. 4 is processing for updating the class classification master in FIG. Referring to FIG. 6, all class classifications (comments, advice) included in data transmitted from record management server 12 are obtained, and the following processing of S52 to S55 is repeated for each. The received class classification is as follows.

Dysphagia risk Manager Suzuki Weighted value 0.7
Risk of dementia Chief Sato Weighted value 0.3

Of the above classifications, "risk of dysphagia" is included in the classification master. That is, "risk of dysphagia, 0.1222" (S52, S53). First, the number of cases of dysphagia risk is calculated as 3000 cases × 0.1222 = 366.6 cases. Next, a new ratio P(B) is obtained as follows. At this time, a weighting value of 0.7 is used.

P(B) = (366.6 cases + 1 case x 0.7) ÷ 3001 cases = 0.1224 Formula (4)

The dysphagia risk 0.1222 in the class classification master is updated with the newly calculated value 0.1224 (S55).

Since the risk of dementia is not registered in the classification master, P(B) is calculated as follows. Use a weighting value of 0.3.

Dementia risk ratio P(B) = 1 case × 0.3 ÷ 3001 cases = 0.0001 Formula (5)

A new risk of dementia is added to the class classification master, and its ratio P(B)=0.0001 is registered (S54).

Process 3 (S33) in FIG. 4 is a process of updating the learning data group in FIG. Referring to FIG. 7, for all combinations of all tagged class classifications (comments, advice) received from record management server 12 and all keywords obtained by analysis at morphological analysis server 14, S62 , the processing of S65 is repeated.

Records of all combinations of the above class classifications and keywords are extracted from the learning data group (FIG. 16) (S62). In this example all records are:

Dysphagia risk + choking P(A|B) = 0.0151
Risk of dysphagia + tachycardia No record Risk of dysphagia + voiding failure No record ‥‥‥‥ ‥‥ ‥‥
Dementia risk + choking No record ‥‥‥ ‥‥ ‥‥

Since the first combination exists, a new ratio P(B|A) is calculated for that combination as follows (S65).

That is, P(A|B) = 0.0151 for the risk of swallowing + "choking", the total number of records is 3000, and the weight is 0.7 (Mr. Suzuki).

Number of records tagged with previous dysphagia risk = 3000 × 0.1222 = 366.6 Number of records tagged with new dysphagia risk = (366.6 + (1 × 0.7)) = 367.3 new The number of records tagged with the risk of severe dysphagia containing “choking” = 366.6 × 0.0151 = 5.5357 Therefore P(A|B) = (5.5357 + 1) / 367.3) = 0.0178
From Bayes' theorem, P(B|A)=P(A|B)×P(B)÷P(A), so P(A)=0.0023 (see formula (2)), P(B)= Using 0.1224 (see formula (4)) P(A|B) = 0.0178 × 0.1224 ÷ 0.0023 = 0.9473 ‥‥Formula (7)

In the learning data group (Fig. 16), these data are updated using the values represented by the formulas (6) and (7) for the combination of the risk of dysphagia and choking (S65).

If the class classification and keyword combination record does not exist in the learning data group of FIG. 16, P(A|B) and P(B|A) are calculated and added to the learning data group (S64).

For example, the risk of dysphagia + “tachycardia” is P(A|B) = 0, the total number of records is 3000, and the weight is 0.7 (Mr. Suzuki).

Number of records tagged with previous dysphagia risk = 3000 × 0.1222 = 366.6 Number of records tagged with new dysphagia risk = (366.6 + (1 × 0.7)) = 367.3 new The number of records tagged with the risk of severe dysphagia including “tachycardia” = 366.6 x 0 = 0 Therefore, P(A|B) = (0 + 1) / 367.3) = 0.0027 ‥‥Formula (8)

As for P(B|A), assuming that P(A) and P(B) are calculated as P(A)=0.0003 and P(B)=0.0001,
P(B|A)=P(A|B)×P(B)÷P(A)=0.0027×0.0003÷0.0001=0.0009 ‥‥Formula (9)

In the training data group (Fig. 16), a new record is created for dysphagia risk + tachycardia, and the values calculated by equations (8) and (9) are registered.

Although machine learning using Bayes' theorem has been described in detail, other machine learning, such as fuzzy inference, may be performed with supervised data.

Next, based on the results of the machine learning described above, when the staff inputs documents in the record management client system 11, when they are inputting documents, when they have finished inputting, etc. , nursing, childcare, etc.

FIG. 20 shows an example of a screen displayed on the record management client system 11 (document creation, document creation means). When the staff member selects, for example, breakfast from the form menu, a meal record input (edit) screen (window) C appears, and the staff member enters the amount of food eaten for breakfast, the appearance of the meal, and oral care using numbers, sentences, checks, and the like. To go. At the bottom of the screen, a guidance display button D is provided in addition to normal buttons such as delete, copy, save and add. The staff (recorder) who is inputting is displayed as Ichiro Tokkyo, and the user of the facility (person receiving care) is displayed as Goro Shizuoka.

When the recording person (staff) presses the guidance display button D or the save button D, the input recording contents are sent from the recording management client 11 (which may be via the recording management server 12) to the morphological analysis server 14 as will be described later. (created document output, created document output means). The morphological analysis server 14 extracts keywords (nouns, verbs) highly relevant to nursing care (nursing, childcare) and the like from the recorded contents. For example, in the example of the input (recording) screen (window C) in FIG.

The guidance server 17 refers to the learning data group (Fig. 16) in the learning database 16 for the keywords sent from the morphological analysis server 14, extracts advice that is strongly (highly) related to these keywords, and creates guidance. and send it to the record management client 11. As shown in FIG. 21, the client 11 displays the guidance transmitted from the guidance server in the guidance area E (advice display, advice display means).

In the example shown in FIG. 21, in the guidance area E, the risks of sleep disorders and dysphagia selected by the guidance server 17 are displayed. guidance items) are displayed. When the staff (recorder) selects (clicks) any of the detailed guidance items, the detailed content is displayed on the client 11 screen. In FIG. 22, the aspiration prevention manual is displayed in window F as a result of selecting "aspiration prevention manual". In this manual, the staff selects further details (points to prevent aspiration, eating posture, meal content, oral care), reads the displayed commentary, studies aspiration, and deepens their understanding (nursing care etc.).

17 to 19 for the processing according to the programs of the client 11, the morphological analysis server 14, and the guidance server 17 that produce the displays shown in FIGS. are detailed below.

The staff (including administrators such as chiefs and section chiefs) selects the type of form on the screen of the client 11 and inputs records accordingly. Alternatively, the staff pulls out a prescribed document (form) (including documents created by other staff) from the record management server 12 (record management database 13) and displays it on the screen of the client 11 for viewing (Fig. 17, S71). ).

The client 11 transmits the displayed recorded document to the morphological analysis server 14 (S72). The timing at which the client 11 transmits the document is when the guidance display button D is pressed as described above, when the save button is pressed, or at any other time (for example, at regular intervals after the start of input, when there is no time input, etc.). An example of a message (data) transmitted from the recording management client 11 to the morphological analysis server 14 is shown in FIG. The transmission destination may be a morphological analysis server instead of the learning guidance server (system). The type of the document is a breakfast record, and the content of the document is the text entered and displayed in the window C, and the like.

The morphological analysis server 14 performs morphological analysis on the content of the received document, refers to the necessary technical term dictionaries (nursing care, nursing, medical care, childcare, etc.), and extracts keywords (nouns, verbs, etc.) specific to the field (facility). ). In the above example, somnolence, dizziness (with dizziness), choking, etc. are extracted. The extracted keywords are sent from the morphological analysis server 14 to the guidance server 17 (S73).

When the guidance server 17 receives the keyword, it performs processing 4 (S74) and processing 5 (S75). Details of processing 4 and processing 5 are shown in FIGS. 18 and 19, respectively.

Process 4 searches the learning data group of the learning database 16 using all the keywords sent from the morphological analysis server 14, and extracts the class classification (advice) with the highest probability among the class classifications (advice) corresponding to the matching keywords. (Fig. 17, S74). That is, referring to FIG. 18, first, in the learning data group (FIG. 16), a record having a keyword that matches the received keyword is extracted (FIG. 18, S81). In the example of FIG. 16, records having the keywords of choking, somnolence, and dizziness are extracted. Next, among the extracted keyword records, records having P(B|A) equal to or greater than a predetermined threshold value (for example, 0.5) are extracted (FIG. 18, S82). In the learning data group in Fig. 16, the P(B|A) value for the keyword "choking" is 0.9226, the P(B|A) value for the keyword "drowsiness" is 0.6452, and the P(B| Since A) is 0.3047, the keywords having P(B|A) above the threshold are "choking" and "somnolence". The threshold will be determined by trial and error to give proper guidance.

Some of the records with keywords extracted in this way may have duplicate classifications, so remove the duplicate classifications and create a list of classifications contained in the extracted keyword records ( Figure 18, S83). In the above example, the keywords "choking" and "somnolence" were extracted, but since the classifications included in the records for these keywords are the risk of dysphagia and the risk of sleep disorders, there is no duplication. A list of two class classifications is created.

In process 5, the guidance server 17 refers to the class classification master (Fig. 15) for all class classifications in the list of class classifications created in this way, links useful manuals, and sends them to the record management client 11. Send (Fig. 17, S75) (guidance, guidance means). That is, in FIG. 19, referring to the class classification master (FIG. 15), various manuals (all manuals or manuals selected under specific conditions) corresponding to the class classification selected in FIG. 18, S83 are acquired. , in a response message (response data), edited into HTML format (FIG. 19, S91, S92), and transmitted to the recording management client 11 (FIG. 19, S93). An example of a message sent from the guidance server 17 to the record management client 11 is shown in FIG. The class classification and the URL of the corresponding manual are edited and included in the display HTML format.

Returning to FIG. 17, the record management client 11 that has received the reply from the guidance server 17 displays the class classification and a menu of useful manuals in the guidance area E in the display screen (S76). Alternatively, when a detailed guidance item is clicked, the corresponding manual is obtained from the website using the URL, and a guidance window F is created and displayed therein (S77).

10 Records management system
11 Records Management Client (System)
12 Records Management Server
13 Records Management Database
14 Morphological Analysis Server
15 learning servers
16 learning databases
17 Guidance Server
18 manual database
19 Study guidance server (system)

Claims (21)

a document approval device, a learning device, and a document creation device, the document approval device comprising:
Approval request display means for displaying an approval request document including an approval column, and approval input means are provided, and advice input corresponding to a keyword included in the displayed approval request document is received and the keyword of the approval request document is included. Approval means for outputting a learning data set containing document contents and accepted advice in response to an approval input from the approval input means;
The learning device is
learning means for learning the degree of conformity of the advice corresponding to the keyword based on the learning data set output from the approval means;
The document preparation device
document creation means for outputting a document on a screen; and guidance means for outputting advice matching keywords contained in the document output from the document creation means based on the learning result of the learning means,
the document creation means displays the advice output from the guidance means on a document output screen;
Record keeping system. approval request display means for displaying an approval request document including an approval column;
Approval input means is provided, advice input corresponding to the keyword included in the displayed approval request document is accepted, and the learning data set including the document content including the keyword of the approval request document and the received advice is input to the approval means for outputting in response to approval input from the approval input means;
learning means for learning the degree of conformity of advice corresponding to a keyword based on the learning data set output from the approval means;
document creation means for outputting a document on a screen; and guidance means for outputting advice matching keywords contained in the document output from the document creation means based on the learning result of the learning means,
the document creation means displays the advice output from the guidance means on a document output screen;
Record keeping system. approval request display means for displaying an approval request document including an approval column;
Approval means comprising approval input means for accepting advice input according to the displayed approval request document and outputting the content of the approval request document and the accepted advice in response to the approval input from the approval input means ，
keyword extraction means for extracting keywords included in the approval request document output from the approval means;
learning means for learning the degree of matching of the advice corresponding to the keyword based on a learning data set containing the keyword extracted by the keyword extracting means and the advice output from the approval means; document creation means for outputting a created document to guidance means, and displaying on a screen advice matching keywords contained in the created document, output from said guidance means based on the learning result of said learning means;
records management system. the means for acknowledging
Advice input guidance means for displaying an advice input screen in response to a first input of a series of approval inputs associated with the approval field; learning data output means for outputting to the learning means a learning data set containing document contents including the keywords of the displayed approval request document and the accepted advice in response to the input of
3. The record management system according to claim 1 or 2. The document preparation means
created document display means for displaying the created document on the screen;
Created document output means for outputting a document displayed on a screen to said guidance means; Provided with advice display means for displaying on the screen,
A record management system according to any one of claims 1 to 3. approval request display means for displaying an approval request document including an approval column;
Approval input means for inputting approval for the displayed approval request document;
advice input guidance means for displaying an advice input screen in response to the approval input from the approval input means;
teaching data output means for accepting advice input on the advice input screen and outputting the accepted advice as teaching data for machine learning;
keyword extracting means for extracting keywords contained in the approval request document displayed on the approval request display means and for which approval is input from the approval input means; and teacher data output from the teacher data output means and the learning means for learning the degree of matching of the advice corresponding to the keyword based on the learning data set containing the keyword extracted by the keyword extracting means;
document approval device. 7. The document approval apparatus according to claim 6, wherein said teacher data output means outputs an identification code of an approver who inputs approval from said approval means together with the teacher data. 7. The document approval device according to claim 6, wherein said training data output means outputs weighting data of approvers together with training data. Obtained by performing machine learning based on a learning data set consisting of keywords included in an approval request document to which approval has been given by the document approval device and advice as teacher data input when approving the approval request document. storage means for storing learned data representing the degree of adaptation of the advice corresponding to the keyword,
created document display means for displaying the created document on the screen;
Created document output means for outputting the created document displayed on the screen to the guidance means; and Keywords included in the created document output from the guidance means based on the learned data stored in the storage means. Advice display means for displaying on the screen advice that conforms to
A document preparation device comprising a The advice display means displays detailed information output from the guidance means together with the advice.
A document creation device according to claim 9 . the approval request display means of the document approval device displays the approval request document including the approval column;
The approval means of the document approval device accepts advice input corresponding to the keyword included in the displayed approval request document, and a learning data set including the document content including the keyword of the approval request document and the received advice. is output in response to the approval input from the approval input means,
learning means learns the degree of conformity of the advice corresponding to the keyword based on the learning data set output from the approval means;
The document creation means of the document creation device outputs the document on the screen,
Guidance means of the document creation device outputs advice that matches keywords contained in a document output from the document creation means based on the learning result of the learning means, and the document creation means is output from the guidance means. display advice on the document output screen,
record keeping method. an approval request display means for displaying an approval request document including an approval column;
The approval means accepts the advice input corresponding to the keywords included in the displayed approval request document, and transfers the learning data set including the document contents including the keywords of the approval request document and the accepted advice to the approval input means. outputs in response to authorization input from
learning means learns the degree of conformity of the advice corresponding to the keyword based on the learning data set output from the approval means;
Document creation means outputs a document on the screen,
Guidance means outputs advice matching keywords contained in a document output from the document creation means based on learning results in the learning means, and Document creation means outputs advice output from the guidance means to a document. displayed on the output screen,
record keeping method. an approval request display means for displaying an approval request document including an approval column;
the approval means accepts advice input according to the displayed approval request document, outputs the content of the approval request document and the accepted advice in response to the approval input from the approval input means;
A keyword extraction means extracts a keyword included in the approval request document output from the approval means,
learning means learns the degree of matching of the advice corresponding to the keyword based on the learning data set containing the keyword extracted by the keyword extraction means and the advice output from the approval means; outputs the prepared document displayed on the screen to the guidance means, and displays advice matching the keywords contained in the prepared document output from the guidance means based on the learning result in the learning means on the document display screen. display above,
record keeping method. The approval request means displays the approval request document including the approval column,
the advice input guidance means displays an advice input screen in response to the approval input from the approval input means;
teaching data output means accepts advice input on the advice input screen, outputs the accepted advice as teaching data for machine learning,
A keyword extracting means extracts a keyword included in an approval request document displayed on the approval request display means and for which approval input is received from the approval input means, and a learning means extracts a keyword from the teacher output from the teacher data output means. learning the degree of adaptation of advice corresponding to the keyword based on a learning data set containing the data and the keyword extracted by the keyword extracting means;
document approval method. A storage means performs machine learning based on a learning data set consisting of keywords included in an approval request document to which approval has been given by the document approval device and advice as teacher data input when approving the approval request document. store the learned data representing the degree of adaptation of the advice corresponding to the keyword obtained by performing
Created document display means displays the created document on the screen,
The prepared document output means outputs the prepared document displayed on the screen to the guidance means, and the advice display means outputs from the guidance means based on the learned data stored in the storage means. display advice on the document creation screen that matches the keywords contained in the created document;
document preparation method. The approval request display means displays the approval request document including the approval column, and the approval means accepts the advice entered corresponding to the keyword included in the displayed approval request document and displays the document including the keyword of the approval request document. a document approval program for controlling a computer of the document approval device to output a learning data set including content and accepted advice in response to an approval input from an approval input means;
a learning program for controlling the computer of the learning device so that the learning means learns the degree of conformity of the advice corresponding to the keyword based on the learning data set output from the approval means;
Document creation means outputs a document on the screen,
Guidance means outputs advice matching keywords contained in a document output from said document creation means based on learning results in said learning means, and said document creation means outputs advice output from said guidance means as a document. a document creation program that controls the computer of the document creation device to display on the output screen;
A records management program comprising: an approval request display means for displaying an approval request document including an approval column;
The approval means accepts the advice input corresponding to the keywords included in the displayed approval request document, and transfers the learning data set including the document contents including the keywords of the approval request document and the accepted advice to the approval input means. outputs in response to authorization input from
learning means learns the degree of conformity of the advice corresponding to the keyword based on the learning data set output from the approval means;
Document creation means outputs a document on the screen,
Guidance means outputs advice matching keywords contained in a document output from said document creation means based on learning results in said learning means, and said document creation means outputs advice output from said guidance means as a document. control the computer to display on the output screen,
Records management program. an approval request display means for displaying an approval request document including an approval column;
the approval means accepts advice input according to the displayed approval request document, outputs the content of the approval request document and the accepted advice in response to the approval input from the approval input means;
A keyword extraction means extracts a keyword included in the approval request document output from the approval means,
The learning means learns the degree of matching of the advice corresponding to the keyword based on the learning data set containing the keyword extracted by the keyword extracting means and the advice output from the approval means, and creates a document. means for outputting the created document displayed on the screen to the guidance means, and for displaying, as a document, advice matching the keywords included in the created document, which is output from the guidance means based on the learning result of the learning means. control the computer to display on the screen,
Records management program. The approval request means displays the approval request document including the approval column,
the advice input guidance means displays an advice input screen in response to the approval input from the approval input means;
teaching data output means accepts advice input on the advice input screen, outputs the accepted advice as teaching data for machine learning,
A keyword extracting means extracts a keyword included in an approval request document displayed on the approval request display means and for which approval input is received from the approval input means, and a learning means extracts a keyword from the teacher output from the teacher data output means. controlling the computer to learn the degree of matching of the advice corresponding to the keyword based on a learning data set containing the data and the keyword extracted by the keyword extracting means;
Document approval program. Obtained by performing machine learning based on a learning data set consisting of keywords contained in an approval request document to which approval has been given by the document approval device and advice as teacher data input when approving the approval request document. In addition, learned data representing the degree of suitability of advice corresponding to a keyword is stored in a storage device,
The document preparation program
Created document display means displays the created document on the screen,
Created document output means outputs the document displayed on the screen to the guidance means, and advice display means outputs from the guidance means based on the learned data stored in the storage device. control the computer to display on-screen advice that matches the keywords contained in the document;
A writing program. 21. A recording medium recording the program according to any one of claims 16 to 20 .

Images