Nursing plan creation support system

Abstract

By providing a nursing care plan creation support system that enables even new nurses to easily create complete nursing care plans, searches for nursing care plans for similar cases from accumulated nursing care plan records, and enables automatic or semi-automatic creation of nursing care plans using additional learning in a large-scale language model, it becomes easier to create high-quality nursing care plans in a short amount of time. [Solution] The nursing care plan creation support system comprises a nursing diagnosis classification recording means, a nursing outcome classification recording means, a nursing intervention classification recording means, a nursing outcome classification feature vector recording means, a nursing intervention classification feature vector recording means, a nursing diagnosis item selecting means, a nursing outcome item selecting means, and a nursing intervention item selecting means. The nursing diagnosis classification feature vector recording means converts any portion of each classification record document into a feature vector and records it, and the nursing diagnosis item selecting means inputs the patient's symptoms and findings and a group of items of the content of each selected classification into a large-scale language model and selects items appropriate to the patient's symptoms and findings.

Description

[Technical Field]

[0001] The present invention relates to a nursing care plan creation support system that uses a large-scale language model and search expansion and generation in combination. [Background technology]

[0002] In the medical field, doctors use the symptoms reported by patients visiting the hospital and the physical findings (symptoms and findings) obtained during the examination to plan tests such as blood tests and radiological tests, or to supplement missing symptoms and physical findings to arrive at an accurate diagnosis and formulate effective treatment. Depending on the severity of the illness, hospitalization may be necessary. Nurses investigate the doctor's diagnosis, the patient's complaints and difficulties, and identify nursing issues. They then create nursing care plans that include intervention methods to resolve the issues, evaluate the effectiveness of the intervention results, and aim for further resolution.

[0003] The North American Nursing Diagnosis Association (NANDA) has been working to standardize and internationalize nursing plans, and since 2002 it has been called NANDA-International (NANDA-I), which has been utilized as a guide for clarifying patients' health issues and providing appropriate nursing interventions. Nursing diagnoses are problems that nursing must address from a different perspective than a doctor's disease name diagnosis, and are defined as 123 nursing diagnoses across 13 areas. Individual nursing diagnoses are expressed in three ways: "existential nursing diagnoses," "risk nursing diagnoses," and "wellness nursing diagnoses." Each nursing diagnosis is further comprised of five elements: (1) diagnostic label, (2) definition, (3) diagnostic indicators, (4) associated factors, and (5) risk factors.

[0004] Based on NANDA nursing diagnoses, there are NOC (Nursing Outcomes Classification) and NIC (Nursing Interventions Classification) that classify how nursing interventions should be performed and what kind of care should be provided. Because there is a linkage (interlocking) between making a nursing diagnosis using NANDA nursing diagnoses, determining nursing outcomes with NOC, and providing nursing interventions with NIC, creating a series of processes such as "NANDA-NOC-NIC" allows for more effective and appropriate fulfillment of patient needs.

[0005] Machine learning has made remarkable progress in recent years, with large-scale language models (LLMs) becoming increasingly widespread. Neural network parameters, numbering in the tens of billions to trillions, are trained on massive amounts of data, amounting to terabytes, to handle tasks such as translation, speech and image recognition, and text summarization. Generative AI, which generates images, music, and documents based on queries, questions, or prompts, is also being put to practical use. It has become known that performance improves by scaling models (scaling laws), and the scale of models is rapidly increasing. By additionally training large amounts of clinical data into large-scale language models and quantifying the relationships between items using attention mechanisms, it is now possible to make diagnoses based on actual clinical symptoms, findings, and test data (Patent Document 2).

[0006] As the range of applications of large-scale language models expands, knowledge data from various fields is required. Furthermore, with social conditions and technological advances in each field, there is a need to constantly incorporate the latest knowledge data. However, training large amounts of data requires large-scale computational resources, enormous amounts of electricity, and costs. For this reason, large-scale language models cannot be reconstructed frequently. When large-scale language models are queried, they can generate answers that are not based on facts, even if they do not have the necessary knowledge stored internally. This phenomenon, known as hallucination, has limited their widespread use in fields such as medicine, where errors can lead to serious accidents.

[0007] In order to utilize large-scale language models based on the latest information in the field, it is necessary to obtain additional up-to-date information, and two approaches are currently being used. The first is called additional training (fine tuning) or transfer learning, in which additional information is used to train part of the output layer of an existing large-scale language model (fine tuning) or only the final layer (transfer learning) to create a specialized large-scale language model. This is highly useful as it results in a domain-specific large-scale language model. However, while it is not as costly as training a large-scale language model from scratch, it does require a certain amount of cost and technical skill for training.

[0008] Another approach is the RAG Retrieval-Augmented Generation (RAG) approach. The large-scale language model itself is not modified. Additional information is stored in a separate database (RAG database), and the information necessary to solve a question is searched for in the RAG database. The obtained information is added to the question, and an answer is obtained from the large-scale language model. Although there is a large amount of potentially additional data, the additional information required to answer a given question is limited. Therefore, only this limited additional information is searched for in the RAG database, and the extracted additional information is added to the question in the question input box of the large-scale language model (context learning). This approach does not involve any learning changes to the large-scale language model itself; it simply adds context information to the question input box, reducing costs and technical hurdles. Furthermore, it offers excellent responsiveness, allowing for the immediate reflection of urgent information, such as information on emergency drug side effects (Non-Patent Document 1, Non-Patent Document 2).

[0009] The RAG database is a database created by cutting document information to be added into small fragments (chunks) as needed, converting each chunk into a feature vector (chunk vector), and storing it in a database. When asking a question, the question itself is converted into a feature vector, and chunk vectors with a high similarity to the vector of the question are searched for in the RAG database, and the contents of the obtained chunk group are added to the question. This makes it possible to respond to specific fields and the latest information without the need for expensive additional training of large-scale language models, and by limiting the basis for responses to questions to the RAG database, hallucinations can be prevented (Patent Document 3). Prior art documents relevant to this application include the following: [Prior art documents] [Patent documents]

[0010] [Patent Document 1] Patent No. 6792750 "Diagnosis Support System" [Patent Document 2] Patent No. 7592232 "Diagnosis and prescription support system using large-scale language models" [Patent Document 3] Patent application no. 2024-226197 "Large-scale language model using search expansion generation" [Patent Document 4] Patent No. 7441391 "Electronic medical record system using large-scale language models" [Patent Document 5] Patent No. 6902693 "Document Creation System"

[0011] [Non-Patent Document 1] https: / / ja.wikipedia.org / wiki / %E5%A4%A7%E8%A6%8F%E6%A8%A1%E8%A8%80%E8%AA%9E%E3%83%A2%E3%83%87%E3%83%AB [Non-patent document 2] https: / / www.idnet.co.jp / column / page_308.html Summary of the Invention [Problem to be solved by the invention]

[0012] The aforementioned "NANDA-NOC-NIC" is a massive document, running to hundreds of pages each. Selecting the appropriate nursing diagnoses, nursing outcomes, and nursing interventions for each hospitalized patient can be a significant burden, especially for novice nurses. Leaving pages of thick nursing diagnoses, nursing outcomes, and nursing interventions can take hours. Experienced nurses understand the relationship between a patient's condition and the corresponding nursing diagnoses, nursing outcomes, and nursing interventions, so even when a patient's clinical symptoms become slightly more complex, it still takes a considerable amount of time. Naturally, even experienced nurses are unsure whether they have covered all the necessary nursing diagnoses, nursing outcomes, and nursing interventions for a given patient. As a result, creating and recording nursing plans requires a great deal of time and effort, which takes away from essential nursing care time, leading to overtime and sometimes even medical accidents.

[0013] The present invention has been made to solve these conventional problems, and its purpose is to (1) enable even new nurses to easily create complete nursing care plans, (2) enable nursing care plans for similar cases to be searched for from a collection of accumulated nursing care plan records, and (3) furthermore, to provide a nursing care plan creation support system that enables automatic or semi-automatic creation of nursing care plans using additional learning on a large-scale language model, thereby facilitating the creation of high-quality nursing care plans in a short amount of time. [Means for solving the problem]

[0014] As a means for achieving the above object, In the nursing care plan creation support system according to claim 1, (1) Classification of nursing diagnoses (NANDA), nursing outcomes (NOC), and nursing interventions (NIC) and its contents Nursing diagnosis classification recording means, nursing outcome classification recording means, and nursing intervention classification recording means are provided, Documents recorded by each classification and recording means consists of classification labels, classification definitions, and classification contents. Documents recorded by each classification and recording meansand a nursing diagnosis classification feature vector recording means (nursing diagnosis RAG database) which converts any part of the above into a feature vector and records it, a nursing outcome classification feature vector recording means (nursing outcome RAG database), and a nursing intervention classification feature vector recording means (nursing intervention RAG database), and a feature vector corresponding to the patient's symptom and findings is created (symptom and finding feature vector creating means), and a nursing diagnosis classification feature vector searching means, a nursing outcome classification feature vector searching means, and a nursing intervention classification feature vector searching means are provided, and the extracted similar nursing diagnoses are The system is provided with nursing diagnosis classification selection means, nursing outcome classification selection means, and nursing intervention classification selection means for selecting an appropriate classification for the patient from classification groups, similar nursing diagnosis classification groups, and similar nursing outcome classification groups, and is provided with symptom finding corresponding nursing diagnosis item group extraction means, symptom finding corresponding nursing outcome item group extraction means, and symptom finding corresponding nursing intervention item group extraction means for inputting the patient's symptom findings and a group of items of the content of each selected classification into a large-scale language model and extracting an item group corresponding to the patient's symptom findings from the group of items of the content of each classification, and is provided with nursing diagnosis item selection means, nursing outcome item selection means, and nursing intervention item selection means for selecting items appropriate for the patient's symptom findings from the extracted item groups. In the nursing care plan creation support system according to claim 2, in the nursing care plan creation support system according to claim 1, (2) It is provided with a symptom finding feature vector recording means (symptom finding RAG database) that records the symptom finding and its feature vector for each patient whose treatment has been completed, a nursing plan recording means (nursing plan database) that records the nursing plan for each patient, a similar patient nursing plan extraction means that searches the symptom finding RAG data database using the symptom finding feature vector of a new patient and extracts a list of nursing plans of patients with high similarity, and a patient-matched nursing plan selection means that selects a nursing plan that is suitable for the patient. In the nursing care plan creation support system according to claim 3, in the nursing care plan creation support system according to claim 1 or 2, (3) Supervised additional learning is performed on the LLM using the accumulated patient symptoms and nursing plans (symptoms / nursing plan additional learning method), and the symptoms of new patients are Based on prompt input , and the corresponding nursing plan, By outputting from the LLM where the additional learning was performed, It is equipped with a nursing care plan creation support tool that assists in creating nursing care plans. [Effects of the Invention]

[0015] In the nursing care plan creation support system according to claim 1 of the present invention, (1) Classification of nursing diagnoses (NANDA), nursing outcomes (NOC), and nursing interventions (NIC) and its contents Nursing diagnosis classification recording means, nursing outcome classification recording means, and nursing intervention classification recording means are provided, Documents recorded by each classification and recording means consists of classification labels, classification definitions, and classification contents. Documents recorded by each classification and recording means and a nursing diagnosis classification feature vector recording means (nursing diagnosis RAG database) which converts any part of the above into a feature vector and records it, a nursing outcome classification feature vector recording means (nursing outcome RAG database), and a nursing intervention classification feature vector recording means (nursing intervention RAG database), and a feature vector corresponding to the patient's symptom findings is created (symptom findings feature vector creating means), and a nursing diagnosis classification feature vector searching means, nursing outcome classification feature vector searching means, and nursing intervention classification feature vector searching means are provided, and the extracted similar nursing diagnosis classification groups and similar nursing diagnosis classification groups are The system is provided with a nursing diagnosis classification selection means, a nursing outcome classification selection means, and a nursing intervention classification selection means that select an appropriate classification for the patient from similar nursing outcome classification groups, and is provided with a symptom finding corresponding nursing diagnosis item group extraction means, a symptom finding corresponding nursing outcome item group extraction means, and a symptom finding corresponding nursing intervention item group extraction means that input the patient's symptom findings and a group of items of the content for each selected classification into a large-scale language model and extract an item group that corresponds to the patient's symptom findings from the group of items of the content for each classification, and is provided with a nursing diagnosis item selection means, a nursing outcome item selection means, and a nursing intervention item selection means that select items appropriate for the patient's symptom findings from the extracted group of items, so that even a new nurse can easily create a nursing care plan without any omissions. In the nursing care plan creation support system according to claim 2 of the present invention,(2) It is provided with a symptom finding feature vector recording means (symptom finding RAG database) that records the symptom finding and feature vectors of each patient whose treatment has been completed, and a nursing plan recording means (nursing plan database) that records the nursing plan for each patient, and is provided with a similar patient nursing plan extraction means that searches the symptom finding RAG data database using the feature vector of the symptom finding of a new patient and extracts a list of nursing plans of patients with high similarity, and is provided with a patient-matched nursing plan selection means that selects a nursing plan that is suitable for the patient, so that nursing plans for similar cases can be searched for from the accumulated nursing plans. In the nursing care plan creation support system according to claim 3 of the present invention, (3) Supervised additional learning is performed on the LLM using the accumulated patient symptoms and nursing plans (symptoms / nursing plan additional learning method), and the symptoms of new patients are Based on prompt input , and the corresponding nursing plan, By outputting from the LLM where the additional learning was performed, The system is equipped with a nursing care plan creation support means that supports the creation of nursing care plans, allowing for additional learning of large-scale language models to enable automatic or semi-automatic planning of nursing care plans. [Brief explanation of the drawings]

[0016] [Figure 1] 1 illustrates an embodiment of a network configuration according to the present invention. [Figure 2] This is an example of a user interface for a large-scale language model (LLM). [Figure 3] We show the general relationship between large-scale language models and RAG databases. [Figure 4] This shows the 13 areas that make up nursing diagnoses and the system of nursing diagnoses for each area. [Figure 5] An example of a nursing diagnosis is shown below. [Figure 6] This is an example of nursing outcomes. [Figure 7] This is an example of a nursing intervention. [Figure 8] We show that the contents of each classification, NANDA, NOC, and NIC, are converted into feature vectors and recorded in the RAG database. [Figure 9]This shows the process flow for determining a nursing diagnosis based on the patient's symptoms and initial appearance, using both RAG and LLM nursing diagnoses. [Figure 10] This is an example of a selected and confirmed nursing diagnosis. [Figure 11] This shows the process flow for determining nursing outcomes using the RAG and LLM together based on the patient's symptoms, findings, and confirmed nursing diagnoses. [Figure 12] This is an example of a selected and confirmed nursing outcome. [Figure 13] This shows the process flow for determining nursing interventions based on the patient's symptoms, findings, and confirmed nursing outcomes, using both RAG and LLM. [Figure 14] This is an example of a selected and confirmed nursing intervention. [Figure 15] This shows the process of creating a nursing care plan using the selected and confirmed nursing diagnoses, nursing outcomes, and nursing interventions in the LLM. [Figure 16] (a) An explanatory diagram for constructing a RAG data database of accumulated patient symptoms and findings and a corresponding nursing care plan database. (b) A flow chart for searching for a corresponding nursing care plan from the symptoms and findings of a new patient. [Figure 17] (a) An explanatory diagram of supervised additional learning of accumulated patient symptoms and corresponding nursing care plans for the LLM. (b) A flow chart of outputting a corresponding nursing care plan from the additionally trained LLM based on prompt input of new patient symptoms. DETAILED DESCRIPTION OF THE INVENTION

[0017] FIG. 1 shows a typical system configuration of the present invention. Large-scale language models consist of huge amounts of data, numerous CPUs (Central Processing Units), GPUs (Graphics Processing Units), and high-speed networks connecting them, so they are built on huge servers such as cloud data centers and provided via the web. In hospitals and other facilities, many PC terminals are connected via a LAN (Local Area Network) connected to the web. There are also servers within the hospital that operate electronic medical records. In recent years, there has been an increasing number of cases where people use mobile devices such as smartphones and tablets to access cloud services such as large-scale language models and electronic medical records while moving around the hospital or from outside the hospital.

[0018] Servers, terminals, and mobile devices all consist of memory for recording programs and data, recording media such as hard disks that permanently record said programs and data as needed, a CPU that reads and processes said programs and data, a GPU that performs high-speed parallel processing as needed, and a communication module. As cloud services become more stable and cheaper, there are increasing cases of people migrating some or all of their electronic medical records to the cloud. Conversely, there is also a movement underway to move some or all of the large-scale language models to servers or terminals installed within hospitals with increased processing capabilities and storage capacity (edge ​​computing). Furthermore, the development of small-scale language models, which limit the number of parameters in the language model, is also progressing. Note that, although they are small-scale, they are still sufficiently large compared to the time before the advent of large-scale language models, and any embodiment, including this form, is included in the present invention.

[0019] Figure 2 shows an example of a user interface for a large-scale language model (LLM). Development of LLM is currently progressing rapidly, with numerous models being developed, including ChatG PT (a registered trademark of OpenAI), Bard, LaMDA (a registered trademark of Google), and LLaMA (a registered trademark of Meta). The use of any LLM is within the scope of the present invention. Naturally, the user interface will differ, but typically, as shown in Figure 2, it consists of a box (prompt input box) for inputting prompts to instruct and inquire of the LLM, a box (answer display box) for displaying the answers to those prompts, and a box (usage history box) for displaying the history of prompts and answers as a usage log.

[0020] Recently, in addition to using LLM alone as described above, there has been an increase in cases where LLM itself has an API (Application Programming Interface) and LLM functions are used from external software. In this case, prompts, answers, and history are input and output to and from external software via API, so the display format is not limited to that shown in Figure 2 and is controlled by the external software.

[0021] In large-scale language models, to represent a certain vocabulary, a one-hot vector is used, which is a long vector consisting of zeros with the same number of dimensions as the number of types of vocabulary used, with a single 1 placed in the position corresponding to the vocabulary in question. All the vocabulary in a large amount of literature is replaced with vectors of this format, and deep learning is used to determine the associations (attention) between each vocabulary vector.In response to a query (prompt), vocabulary that is likely to appear next to the query and the answer sentences that have already been generated is generated and added one by one to create an answer sentence. If the information from which answers are generated is contained in the large amount of literature, it is expected that correct sentences that are rational and highly useful will be generated. However, if the information on which the answer is generated is not contained in the large amount of literature, the answer generation process will proceed by mechanically adopting vocabulary with a high probability, which is known to result in the generation of false answers that are not based on any evidence (hallucination). If this hallucination occurs in medical settings or other settings, it could pose a risk to the patient's life, which is one of the reasons why the application of large-scale language models to core business processes has not progressed.

[0022] Training large-scale language models requires large-scale servers equipped with a large number of parallel computing units (GPUs), as well as a large amount of electricity and money, in order to read a large amount of documents and use deep learning to find relationships between vocabulary words. New documents are generated every day, and it is not realistic to reflect all of them in a large-scale language model without any time delay. Furthermore, the large amount of documents that are scanned are mainly public documents that are available on the web, etc., but they do not include sensitive information such as internal company documents or hospital electronic medical records, so it is said that the documents that can be collected are only a small portion of the total documents that exist on Earth.

[0023] In order to utilize large-scale language models based on the latest information in the field, it is necessary to obtain additional up-to-date information, and two approaches are currently being taken. One is called additional learning (fine tuning) or transfer learning, which uses additional information to train part of the output layer of an existing large-scale language model (fine tuning) or only the final layer (transfer learning), resulting in a specialized large-scale language model. It is highly useful as it is a domain-specific large-scale language model, but it still requires a certain amount of cost and technical skill to learn, although not as much as training a large-scale language model from scratch. In addition, since the additional information contains sensitive information such as personal information and descriptions of medical conditions, it may be used for learning purposes, and there is a risk that it may be viewed outside the organization. To prevent this, it may be necessary to build and operate a large-scale language model with additional training exclusively within your company or hospital.

[0024] Another approach is the search expansion generation (RAG) approach (Figure 3). The large-scale language model itself is not changed. The text of the additional information is cut into small fragments (chunks) as needed, and feature vectors are calculated for each chunk and stored in a separate database (RAG database). The system calculates the feature vector of the question, searches for the information needed to solve the question based on the similarity between the feature vector of the question and the feature vector of the chunks recorded in the RAG database, adds the text information of the chunks found to be highly similar to the question, and searches for an answer in a large-scale language model (Figure 3).

[0025] The degree of similarity between feature vectors is often expressed by the magnitude of the inner product between the vectors (cosine similarity), but the Manhattan distance, which is the sum of the absolute values ​​of the differences between the elements of both vectors, may also be used. The threshold value of the degree of similarity that serves as the extraction standard is variable depending on the situation. If the extracted list is too large, you can narrow it down by raising the criteria, or if the list is too small, you can lower the criteria to make the list larger. In some cases, you can set the top 10, etc., and automatically adjust the criteria so that the number of extracted items matches the set value.

[0026] Even if there is a large amount of data in a related field, the additional information required to answer a certain question is limited. Therefore, only this limited additional information is searched for in the RAG database, and the additional information contained in the extracted chunks is added to the question (context) in the question input box of the large-scale language model. This approach does not involve any changes to the large-scale language model itself through learning, but simply adds contextual information to the question input box, which keeps costs and technical hurdles low. Furthermore, it is highly responsive, being able to immediately reflect information of high urgency, such as information on emergency side effects of drugs.

[0027] Figure 4 shows the 13 areas that make up nursing diagnoses and the system of nursing diagnoses for each area. Each of these has been further subdivided, and currently a total of 123 nursing diagnoses have been defined. The number of diagnoses is increasing in response to the demands of the field. Nursing diagnoses can be expressed in three ways: "existential nursing diagnoses" based on the symptoms and findings that the patient already has, "risk nursing diagnoses" that do not currently recognize symptoms or findings but aim to prevent symptoms and findings that are at high risk of appearing in the future, and "wellness nursing diagnoses" that aim to improve current symptoms and findings.

[0028] Figure 5 shows an example of a nursing diagnosis. Each nursing diagnosis consists of five elements: (1) diagnostic label, (2) definition, (3) diagnostic indicators, (4) associated factors, and (5) risk factors. The diagnostic label is the name of the nursing diagnosis, and the definition indicates the definition of the nursing diagnosis. A diagnostic index is a list of symptoms and findings that support the nursing diagnosis. Related factors are a list of factors that are not directly related to the nursing diagnosis but may indirectly affect it. Risk factors are a list of factors that may worsen the current symptoms.

[0029] Figure 6 is an example of nursing outcomes. First, there is the label and definition of nursing outcomes. These are semi-quantitative evaluations of a group of indicators on a scale of 1 to 5 to assess whether nursing has improved or worsened a patient's symptoms. The degree of improvement is evaluated comprehensively by adding up the scores of each evaluation, which allows for an objective evaluation of the results of nursing practice.

[0030] Figure 7 is an example of a protective intervention. First, there are labels and definitions of nursing interventions. "Actions" is a list of nursing intervention practices to improve the nursing outcomes. The aim is to improve the patient's symptoms (improve nursing outcomes) by taking the necessary "actions" for the patient.

[0031] Figure 8 shows that the contents of each classification, NANDA, NOC, and NIC, are converted into feature vectors and recorded in the RAG database (nursing diagnosis classification feature vector recording means, nursing outcome classification feature vector recording means, and nursing intervention classification feature vector recording means). As mentioned above, each diagnostic document for a nursing diagnosis consists of five elements: (1) diagnostic label, (2) definition, (3) diagnostic indicators, (4) related factors, and (5) risk factors. The entire document containing all of these can be converted into a feature vector, or only the diagnostic labels and definitions that are part of the document can be converted into feature vectors, and the rest (contents by classification) can be recorded in a separate document database that can be referenced from the feature vector. When building a RAG database, documents are typically cut into chunks of a certain length and recorded as feature vectors. In this invention, because the entire NANDA, NOC, and NIC classified document, or each group of content items for each classification, has semantic unity, fragmenting it into smaller fragments could actually hinder semantic searches. For this reason, in this application, the RAG database is built using the entire classified document, classification labels and definitions, or each group of content items for each classification, which are semantic units, as chunks. The same applies to NOCs and NICs.

[0032] Figure 9 shows the process flow for determining a nursing diagnosis based on the patient's symptoms and findings, using the nursing diagnosis RAG and LLM in combination. The patient's symptoms and findings are converted into feature vectors (symptom and finding feature vector creation means), the nursing diagnosis RAG is searched, and a list of nursing diagnoses with high similarity, i.e., high possibility of application, is extracted (nursing diagnosis classification feature vector search means). The nurse in charge selects at least one nursing diagnosis that is thought to be suitable for the patient from the extracted nursing diagnosis list (nursing diagnosis classification selection means). Along with the list of symptoms and findings of the patient, lists of diagnostic indicators, related factors, high risks, and related conditions included in the selected nursing diagnoses, if necessary, are input into the LLM context, and a group of items for diagnostic indicators, related factors, high risks, and related conditions that are likely to apply to the patient are extracted (means for extracting nursing diagnosis items corresponding to symptoms and findings). The nurse selects only the items that are suitable for the patient from the extracted group of diagnostic indicators, related factors, high risks, and related conditions that are likely to apply to the patient (nursing diagnosis item selection means). Items that have not been extracted may also be added as necessary. In this way, a group of items including nursing diagnosis labels, corresponding diagnostic indicators, related factors, high risks, and related conditions is determined.

[0033] NANDA, NOC, and NIC are being developed with the aim of covering and standardizing the vocabulary used in nursing care plans. Because comprehensiveness is a priority, there is considerable overlap in content between diagnostic labels. A unique choice is difficult. For this reason, after listing the likely diagnostic labels, as well as diagnostic indicators, associated factors, high-risk, and related conditions, it is necessary to select items that seem appropriate to the patient's symptoms and findings, and in some cases add items that were not included in the list.

[0034] Figure 10 is an example of a selected and confirmed nursing diagnosis. From among the nursing diagnosis labels extracted as being highly likely, the nurse examines the content and selects the nursing diagnosis label that seems appropriate. The diagnostic indicators, related factors, high risk factors, and related conditions included in the selected nursing diagnosis are also selected at the same time.

[0035] Figure 11 shows the process flow for determining nursing outcomes from the patient's symptoms, findings, and determined nursing diagnoses using both the nursing outcome RAG and LLM. At least the patient's symptoms and findings are included, and if necessary, the diagnostic label determined in Figure 9, as well as the diagnostic indicators, related factors, high risk, and related condition items are also converted into feature vectors, and the nursing outcome RAG is searched for (nursing outcome classification feature vector search means). A list of nursing outcome labels with high similarity to the feature vector is extracted. The nurse selects a nursing outcome label that is deemed appropriate for the patient (nursing outcome classification selection means). The complete list of indicators included in the selected nursing outcome, along with the list of symptom findings for the patient, and, if necessary, the confirmed nursing diagnostic label, diagnostic indicators, related factors, high risk, and related condition items, are entered into the LLM context. This allows for the extraction of a group of nursing outcome indicators that are likely to apply to the patient (a symptom-finding-corresponding nursing outcome item group extraction means). The nurse selects only nursing outcome indicators that are suitable for the patient from the extracted group of nursing outcome indicators that are likely to apply to the patient (nursing outcome item selection means). Nursing outcome indicators that have not been extracted may also be added as necessary. In this way, a group of nursing outcome indicators for the patient is determined.

[0036] Figure 12 is an example of a selected and confirmed nursing outcome. From among the nursing outcome labels extracted as being highly likely, the nurse examines the content and selects the nursing outcome label that seems appropriate. Each list of nursing outcome indicators included in the selected nursing outcome is also selected at the same time.

[0037] Figure 13 shows the flow of the process for determining nursing intervention activity items based on the patient's symptoms and findings and the determined nursing outcome indicators, using nursing intervention RAG and LLM in combination. At least the patient's symptoms and findings are included, and if necessary, the list of nursing outcome indicators determined in FIG. 11 is also converted into a feature vector, and a nursing intervention RAG is searched for (nursing intervention classification feature vector search means). A list of nursing intervention labels with high similarity in feature vectors is extracted. The nurse selects a nursing intervention label that is deemed appropriate for the patient (nursing intervention classification selection means). A complete list of activity items included in the selected nursing intervention label, along with a list of symptoms and findings of the patient, and, if necessary, the determined nursing outcome indicators, are entered into the context of the LLM. This allows extraction of a group of nursing intervention activities that are likely to be applicable to the patient (a symptom and finding corresponding nursing intervention item group extraction means). The nurse selects only nursing intervention activity groups that are suitable for the patient from the extracted nursing intervention activity groups that are likely to be applicable to the patient (nursing intervention item selection means). Nursing intervention activities that have not been extracted may also be added as necessary. In this way, the nursing intervention activity group for the patient is determined.

[0038] Figure 14 is an example of a selected and confirmed nursing intervention. From among the nursing intervention labels extracted as likely possibilities, the nurse examines the content and selects the nursing intervention label that seems appropriate. Each list of nursing intervention activity items included in the selected nursing intervention label is also selected at the same time.

[0039] Figure 15 shows the process of creating a nursing care plan in LLM using the selected and confirmed nursing diagnoses, nursing outcomes, and nursing interventions. The selected and confirmed nursing diagnoses, nursing outcomes, and nursing interventions are entered as the context of the LLM, and a nursing care plan document is created and recorded in the electronic medical record, etc.

[0040] 16A is an explanatory diagram for constructing a RAG data database of accumulated patient symptoms and findings and a corresponding nursing care plan database, and FIG. 16B is a flowchart for searching for a corresponding nursing care plan based on the symptoms and findings of a new patient. By completing the steps up to Figure 15, even a new nurse can create a new, complete nursing care plan. Depending on the medical institution, patients' disease profiles are often concentrated into a relatively small number of categories. In such cases, it is more efficient to reuse nursing care plans that have been accumulated in the past rather than creating new nursing care plans from scratch for each patient. (a) For each patient, symptom observation data is converted into a feature vector and recorded in the RAG database (symptom observation feature vector recording means, or symptom observation RAG database), and the patient's nursing care plan is recorded in the nursing care plan database (nursing care plan recording means) so that it can be referenced from the symptom observation RAG database. (b) For new patients for whom a nursing care plan needs to be developed, the symptoms and findings of the patient are converted into feature vectors and searched through the symptom and findings RAG database. Patients with high similarity are extracted, and the nursing plans of these patients are listed from the nursing plan database (similar patient nursing plan extraction means). The nurse selects a nursing plan that is deemed appropriate for the patient (patient-matched nursing plan selection means). Depending on the patient's symptoms and findings, some modifications are made to the nursing care plan for the new patient. This makes it possible to create an efficient nursing care plan in a short amount of time.

[0041] Figure 17 shows the flow of creating a nursing care plan using additional learning. (a) The accumulated patient symptom findings and corresponding nursing plans are applied to the LLM as a corpus for supervised incremental learning. Additional learning is performed on the LLM using the patient's symptoms and findings as inquiry data and the nursing plan as response teaching data (symptoms and findings / nursing plan additional learning method). (b) The symptoms and findings of a new patient are entered into the prompt, and the corresponding nursing care plan is output from the additionally trained LLM. The nurse appropriately modifies the output nursing care plan in accordance with the symptoms and findings of the patient, and finalizes the nursing care plan.

[0042] Furthermore, since there is a risk of information leakage if the content entered into the prompt is used to train a large-scale language model, it is useful to explicitly declare in the prompt that learning is prohibited, or to use a paid version of a large-scale language model that is guaranteed not to be used for learning, or even to use a large-scale language model that has been independently constructed within the medical institution.

[0043] Although the embodiments have been described above, the specific configuration of the present invention is not limited to the above-described embodiments, and the present invention also includes design changes and the like that do not deviate from the gist of the invention. In this application, a nursing care plan creation support system using NANDA, NOC, and NIC is described, but attempts are being made to create nursing care plans other than NANDA, NOC, and NIC. In these cases, a nursing care plan creation support system similar to this application is also conceivable. Similar support systems are also conceivable for creating nursing care plans for care workers, and are included in this application.

Claims

1. A nursing care plan creation support system characterized by comprising the following nursing care plan creation support means for creating nursing care plans. (1) A nursing diagnosis classification recording means, a nursing outcome classification recording means, and a nursing intervention classification recording means are provided, which record the classifications and contents of nursing diagnoses (NANDA), nursing outcomes (NOC), and nursing interventions (NIC), and the documents recorded by each of the classification recording means consist of classification labels, classification definitions, and contents of each classification. A nursing diagnosis classification feature vector recording means (nursing diagnosis RAG database), a nursing outcome classification feature vector recording means (nursing outcome RAG database), and a nursing intervention classification feature vector recording means (nursing intervention RAG database) are provided, which convert any part of the document recorded by each of the classification recording means into a feature vector and record it. A feature vector corresponding to the symptom findings of a patient is created (symptom finding feature vector creation means), and similar nursing diagnosis classification feature vector recording means, nursing outcome classification feature vector recording means, and nursing intervention classification feature vector recording means are searched for. and a nursing diagnosis item selection means, a nursing outcome item selection means, and a nursing intervention item selection means for selecting an appropriate classification for the patient from the extracted similar nursing diagnosis classification groups, similar nursing diagnosis classification groups, and similar nursing outcome classification groups. The nursing plan creation support means is characterized by comprising: a nursing diagnosis item selection means, a nursing outcome item selection means, and a nursing intervention item selection means for inputting the patient's symptom findings and a group of items of the contents of each selected classification into a large-scale language model and extracting an item group corresponding to the patient's symptom findings from the group of items of the contents of each classification.

2. A nursing plan creation support system as described in claim 1, characterized in that it is equipped with the following nursing plan creation support means. (2) A nursing plan creation support means characterized by comprising: a symptom finding feature vector recording means (symptom finding RAG database) that records the symptom finding and its feature vector for each patient whose treatment has been completed; a nursing plan recording means (nursing plan database) that records the nursing plan for each patient; a similar patient nursing plan extraction means that searches the symptom finding RAG data database using the symptom finding feature vector of a new patient and extracts a list of nursing plans of patients with high similarity; and a patient-matched nursing plan selection means that selects a nursing plan that is suitable for the patient.

3. A nursing plan creation support system as described in claim 1 or 2, characterized in that it is equipped with the following nursing plan creation support means. (3) A nursing plan creation support means for performing supervised additional learning on the LLM using accumulated patient symptom findings and corresponding nursing plans (symptom findings / nursing plan additional learning means), and supporting the creation of nursing plans by outputting a corresponding nursing plan from the LLM where the additional learning was performed based on prompt input of a new patient's symptom findings.

Images