Medical support devices and medical support programs

Abstract

The objective is to provide a clinical support device and program that allows for the identification of information gaps and inconsistencies based on a comparison of the patient's clinical information with the expected values ​​according to the clinical pathway, enabling additional interviews and other measures, and also enabling the presentation of priority confirmation items and treatment advice to physicians based on the uncertainty and medical validity of the information. [Solution] Subjective interview information from patients and their companions is converted into structured data, a clinical pathway model corresponding to the disease is identified, deficiencies and ambiguities in the interview information, and logical inconsistencies with expectations are detected, and questions are automatically generated to supplement information and resolve contradictions as needed. Then, disease risk, deviation from prediction, and uncertainty of information are integrated to derive the degree of deviation, and if the degree of deviation exceeds a threshold, an emergency warning is issued, and even if it is below the threshold and normal, the system notifies the physician that the information is highly consistent, thereby supporting the optimization of medical care.

Description

Technical Field

[0001] The present invention relates to a medical support device and a medical support program. Specifically, based on the comparison between the clinical information of a patient and the expected values according to a clinical pathway, it identifies deficiencies and contradictions in the information and conducts additional interviews, etc., and based on the uncertainty and medical validity of the information, it can present priority confirmation items and medical advice to doctors. The present invention relates to a medical support device and a medical support program capable of doing so.

Background Art

[0002] In the medical treatment process in a medical institution, "interviewing" to confirm a patient's symptoms, past history, lifestyle, etc. is an extremely important step in determining the direction of medical treatment. Through appropriate interviewing, a doctor can collect necessary information and select an appropriate clinical pathway (treatment plan) or formulate a treatment policy. Conventionally, these interviews have mainly been conducted through face-to-face hearings by doctors and nurses. In recent years, for the purpose of improving the efficiency of medical treatment and the comprehensiveness of information, automatic interviewing systems using computers and tablet terminals have been widely introduced.

[0003] Particularly, in pediatric medicine where it is difficult for a patient himself / herself to accurately convey detailed subjective symptoms, and in veterinary medicine where the subject (patient) cannot speak, interviews from companions such as guardians and owners become the main source of information. In such cases, the accuracy of the information is easily affected by the subjectivity of the companion and the ambiguity of memory, and there is a problem that objective information necessary for medical treatment may be lacking or contradictions in the factual relationship may easily occur. Therefore, there has been a demand for a technology to accurately collect truly necessary information with high precision in a limited medical treatment time in light of a specific disease or clinical pathway.

[0004] As related prior art, for example, a system is known that automatically generates dynamic questions that delve deeper into more detailed body parts and symptoms based on the answers selected by the patient (e.g., Patent Document 1). In addition, a technology has been proposed that supports medical treatment by identifying diseases that can be predicted from the results of the medical interview and presenting relevant guidelines and medical information to the doctor's terminal (e.g., Patent Document 2). Furthermore, with the recent advancements in natural language processing technology, a method has also been developed that uses artificial intelligence (AI) to automatically generate the next medical interview items to be checked (physical condition, tests, medication, etc.) from past progress information accumulated in electronic medical records, etc. (e.g., Patent Document 3). [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Patent No. 4228352 [Patent Document 2] Japanese Patent Publication No. 2022-142234 [Patent Document 3] Japanese Patent Publication No. 2025-144629 [Overview of the project] [Problems that the invention aims to solve]

[0006] However, while conventional medical interview support technologies, such as those described in Patent Documents 1 to 3, contribute to refining the input information and generating continuous questions based on past data, they still have challenges in terms of qualitatively evaluating information to ensure the quality of medical care. Specifically, a mechanism has not been established to maintain a medically "ideal state (expected state of clinical parameters)" as a standard (expected value) for a particular disease or phase of treatment, and to logically identify information deficiencies (insufficiencies) or medical inconsistencies (discrepancies) by comparing this with current clinical information in real time.

[0007] Furthermore, in the operation of clinical pathways that involve complex cases or require multifaceted observation, simply asking many questions is insufficient; a medically-based judgment is needed regarding "which indicators should be in what state at that point in time." Existing systems lack the dynamic logic to immediately detect deficiencies or inconsistencies in input data in the clinical setting and to lead to specific reconfirmations (such as additional interviews) to resolve them. As a result, physicians face the cognitive burden of having to reorganize and confirm information again during consultations.

[0008] In particular, in veterinary medicine, where patients (subjects) cannot directly express their subjective symptoms, clinical information is entered based on the subjective interpretations of accompanying persons such as pet owners. This often leads to discrepancies in interpretation between the "expected values ​​corresponding to evaluation items" necessary for objectively understanding the course of treatment and the actual "clinical parameters." Against this backdrop, there has been a strong desire for highly accurate clinical support technology that can immediately identify information deficiencies and inconsistencies based on criteria defined for each treatment process, and automatically suggest specific additional questions to resolve them before treatment begins.

[0009] This invention was conceived in view of the above points, and relates to a medical support device and a medical support program that can identify information deficiencies or inconsistencies and conduct additional interviews, etc., based on a comparison of the patient's clinical information with the expected value according to the clinical pathway. [Means for solving the problem]

[0010] To achieve the above objectives, the medical support device of the present invention comprises: a clinical parameter generation unit that generates clinical parameters including the patient's attribute information and chief complaint from clinical information obtained from past medical records or acquired interview information concerning the patient; a clinical management data storage unit that holds clinical management data defining the correspondence between a clinical path model that defines the expected medical state to be applied to the patient, one or more evaluation categories that indicate the clinical evaluation axis associated with the clinical path model, and one or more evaluation items necessary to determine the evaluation category; a clinical path identification unit that identifies the clinical path model based on the generated clinical parameters by referring to the clinical management data; an evaluation item identification unit that identifies one or more evaluation categories and one or more evaluation items necessary to determine the evaluation category associated with the identified clinical path model; and a clinical path determination unit that determines the suitability of the clinical information to the clinical path model based on a comparison of the clinical parameters and the expected values ​​corresponding to the evaluation items.

[0011] This system includes a clinical parameter generation unit that generates clinical parameters, including the patient's attribute information and chief complaint, from clinical information obtained from past medical records or acquired interview information. This allows for the extraction and structuring of indicators that reflect patient attributes and chief complaints from atypical clinical information such as past medical records and interview information. As a result, even information that is prone to subjectivity from accompanying persons such as pet owners can be consistently handled by the system as objective data (clinical parameters) that can be compared with medical standards.

[0012] Furthermore, by incorporating a clinical management data storage unit that holds clinical management data defining the correspondence between a clinical pathway model that defines the expected medical state to be applied to the patient, one or more evaluation categories that indicate the clinical evaluation axes associated with the clinical pathway model, and one or more evaluation items necessary to determine the evaluation categories, the "ideal medical state" according to the clinical pathway can be stored in a hierarchical structure linked to evaluation categories and items. As a result, the system can automatically recognize multifaceted evaluation axes based on the treatment plan and identify the evaluation items that truly need to be confirmed in accordance with each individual disease state without excess or deficiency. Consequently, it can provide a foundation for comprehensive and highly accurate treatment decisions based on medical evidence that does not rely on subjective judgment.

[0013] Furthermore, by incorporating a clinical pathway identification unit that references clinical management data and identifies clinical pathway models based on generated clinical parameters, it is possible to analyze the patient's current condition from clinical parameters and automatically narrow down the optimal clinical pathway model to be applied from the clinical management data. This allows for the immediate determination of a standard and appropriate treatment framework for each case, without relying on the physician's experience or memory, and supports the initiation of a rapid and high-quality treatment process.

[0014] Furthermore, by including an evaluation item identification unit that identifies one or more evaluation categories associated with the identified clinical pathway model and one or more evaluation items necessary to determine the evaluation categories, the system can automatically identify specific evaluation axes (categories) and items that should be medically confirmed at this time within the identified clinical pathway. This allows for the extraction of only the necessary information specific to the case from a vast number of examination items without omission, standardizing the items to be confirmed during medical treatment and preventing the inclusion of unnecessary information, thereby significantly improving the quality and efficiency of medical care.

[0015] Furthermore, by incorporating a clinical pathway determination unit that determines the fit of clinical information to the clinical pathway model based on a comparison of clinical parameters and corresponding expected values ​​for evaluation items, the collected clinical information can be automatically verified to meet the expected values ​​defined as the "ideal state" in the clinical pathway. This makes it possible to logically and in real time identify information deficiencies and medical inconsistencies as "fitting status." As a result, the progress and risks of the treatment process can be determined with high accuracy based on objective evidence, without relying on the physician's subjectivity or experience.

[0016] Furthermore, when the clinical pathway evaluation unit calculates at least one of the following indicators as the degree to which clinical information deviates from expected values, a first indicator indicating the degree to which clinical information deviates from predictions based on the clinical pathway model, and a third indicator indicating the degree of clinical information deficiency or lack of consistency among multiple pieces of information contained in the clinical information, the clinical pathway evaluation unit calculates the degree to which clinical information is deviated from expected values, a second indicator indicating the degree to which clinical information deviates from predictions based on the clinical pathway model, and a third indicator indicating the degree of clinical information deficiency or lack of consistency among multiple pieces of information contained in the clinical information. This allows for a quantitative understanding of the patient's condition from the perspective of medical validity and the reliability of the information. As a result, the urgency of treatment and the uncertainty of the data can be immediately visualized, contributing to improved accuracy in physicians' interpretation of the patient's condition and to more accurate judgments regarding the next steps in the medical interview.

[0017] Furthermore, when the clinical pathway determination unit calculates the fit status by multiplying the clinical pathway model or evaluation category defined for each clinical pathway model by at least one of the first, second, and third indicators in the clinical management data, dynamic optimization according to the characteristics of the disease and evaluation axis becomes possible. For example, by weighting the first indicator, which indicates the degree of risk, when it relates to vital activities, and the third indicator, which indicates uncertainty, in the early stages of the interview, it is possible to achieve a highly medically valid evaluation that cannot be captured by uniform threshold determination.

[0018] Furthermore, if the system includes an additional questionnaire generation unit that generates questions to obtain additional medical information based on the suitability status determined by the clinical pathway determination unit, it can dynamically generate the necessary questions to resolve any deficiencies or inconsistencies in the identified information based on the determination results. This allows for the re-collection of highly accurate information from the patient before the start of treatment and automatically supplements any deficiencies in clinical information.

[0019] Furthermore, the additional questionnaire generation unit generates questions to identify urgency or severity when the first indicator exceeds a predetermined threshold, questions to examine deviations from the predicted state based on the clinical pathway model when the second indicator exceeds a predetermined threshold, and questions to supplement missing information or resolve inconsistencies in information when the third indicator exceeds a predetermined threshold, enabling the system to ask the most appropriate and minimal questions tailored to the nature of the information deficiencies. This allows the system to improve the quality of information itself, enabling physicians to complete a comprehensive and reliable dataset essential for medical treatment with minimal effort before commencing treatment.

[0020] Furthermore, the system includes an output unit that outputs the clinical pathway determination results to a terminal accessible to physicians. The output unit presents clinical information and the determination results from the clinical pathway determination unit side-by-side. When it also presents advice on interpreting the clinical information or items to confirm with the patient, physicians can intuitively grasp the validity of the information, and the system can automatically recommend "points for interpretation" and "risks that should be immediately confirmed" from a vast amount of data. This helps organize the physician's thinking process during treatment, prevents judgment errors, and supports highly accurate and efficient decision-making in medical treatment.

[0021] Furthermore, if the system includes a medical information storage unit that defines a medical profile containing one or more clinical patterns suggesting a specific disease or condition, and a clinical support information generation unit that generates clinical support information to identify a candidate for the patient's condition or disease by calculating the degree of fit between the medical profile and at least one of the first, second, and third indicators calculated by the clinical pathway determination unit, then by comparing the first to third indicators calculated by the clinical pathway determination unit with the clinical patterns in the medical information storage unit, it becomes possible to identify disease candidates with high accuracy, taking into account the current progress of treatment and the amount of missing information. This not only detects deviations from the clinical pathway, but also enables the generation of clinical support information based on medical validity even under uncertain information, contributing to the acceleration of treatment and the reduction of judgment errors by physicians.

[0022] Furthermore, when the clinical pathway identification unit dynamically generates a temporary clinical pathway model by combining predefined clinical components according to the sufficiency status of the clinical pathway model identification based on clinical parameters, it dynamically generates a temporary model by combining predefined clinical components according to the sufficiency status of the clinical pathway model identification based on clinical parameters (information deficiency, degree of inconsistency, etc.), thereby enabling uninterrupted medically consistent clinical support even for cases for which no existing model exists or for unknown symptoms.

[0023] In order to achieve the above object, the medical support program of the present invention includes a clinical parameter generation step of generating clinical parameters including the patient's attribute information and chief complaint from past medical records regarding the patient or clinical information obtained from acquired interview information; a clinical path model defining the medical expected state to be applied to the patient, one or more evaluation categories indicating clinical evaluation axes associated with the clinical path model, and clinical management data defining the mutual correspondence relationship of one or more evaluation items necessary for determining the evaluation categories. Based on the generated clinical parameters, a clinical path identification step of identifying the clinical path model; an evaluation item identification step of identifying one or more of the evaluation categories associated with the identified clinical path model and one or more of the evaluation items necessary for determining the evaluation categories; and a clinical path determination step of determining the conformity state of the clinical information with respect to the clinical path model based on the comparison between the clinical parameters and the expected values corresponding to the evaluation items, for causing a computer to execute.

[0024] By executing the above medical support program, a series of processes from parameter generation from atypical clinical information, identification of clinical paths and evaluation items, and determination of the conformity state based on expected values can be realized as a highly automated process. As a result, it becomes possible to quickly provide objective and reproducible medical support that eliminates human bias by instantaneously utilizing clinical management data reflecting vast medical knowledge in a general-purpose computer environment.

Advantages of the Invention

[0025] The medical support device and medical support program according to the present invention can identify deficiencies and contradictions in information and conduct additional interviews, etc., based on the comparison between the patient's clinical information and the expected values corresponding to the clinical path, and can present priority confirmation items and medical advice to doctors based on the uncertainty and medical validity of the information.

Brief Description of the Drawings

[0026] [Figure 1]This is a system diagram showing the overall configuration of a medical support device according to an embodiment of the present invention. [Figure 2] This is a block diagram showing the configuration of a medical support device according to an embodiment of the present invention. [Figure 3] This figure shows the contents of the clinical path table stored in the clinical management data storage unit. [Figure 4] This figure shows the contents of the evaluation category table stored in the clinical management data storage unit. [Figure 5] This figure shows the contents of the evaluation item table stored in the clinical management data storage unit. [Figure 6] This diagram shows the contents of the judgment rule table stored in the clinical management data storage unit. [Figure 7] This diagram shows the contents of the treatment slot table stored in the clinical management data storage unit. [Figure 8] This figure shows an example of the output screen for the judgment result on the second terminal. [Figure 9] This figure shows the control flow of a medical support program according to an embodiment of the present invention. [Modes for carrying out the invention]

[0027] The embodiments of the present invention relating to the medical support device and medical support program will be described below with reference to the drawings to facilitate understanding of the present invention. First, Figure 1 is an overall configuration diagram of the medical support system including the medical support device in this embodiment. This system provides advanced support for physicians' medical practice in medical institutions (e.g., animal hospitals and pediatric clinics) by evaluating the quality of clinical information obtained from patients or their companions based on medical "expectations" and conducting additional interviews as necessary.

[0028] As shown in Figure 1, the medical support system mainly consists of a medical support device 1, which acts as a server device for analyzing and determining patient attributes and symptoms; a first terminal 2 for patients and accompanying persons to input medical history; and a second terminal 3 for doctors to view the determination results and advice information. Each device is connected to the others via a communication network 4 such as the Internet, LAN, or dedicated line, enabling data transmission and reception. In this embodiment, a server-client type system is illustrated, but it may also be configured as a standalone type where all functions are integrated into a single computer.

[0029] The first terminal 2 is a tablet, smartphone, or dedicated kiosk terminal, and accepts patient-side questionnaire input and voice input. The non-standard data collected here (text, choices, voice, etc.) is transmitted to the medical support device 1 via the communication network 4 and used to generate the clinical parameters described later.

[0030] The second terminal 3 is a desktop PC or mobile device, and displays the results of the medical support device 1, as well as related advice and priority confirmation items on its screen. This allows physicians to instantly identify any "medical inconsistencies" or "information gaps" in patient information before starting treatment, enabling them to provide efficient medical care.

[0031] Figure 2 is a block diagram of the medical support device 1, which is the core control unit of the present invention. The medical support device 1 is equipped with standard computer hardware such as a CPU or other processor, RAM, storage, and a communication interface, and comprises a clinical parameter generation unit 10, a clinical management data storage unit 20, a clinical path identification unit 30, an evaluation item identification unit 40, a clinical path determination unit 50, a medical information storage unit 60, a medical support information generation unit 70, an additional questionnaire generation unit 80, and an output unit 90 as physical or logical functional blocks. These units perform their respective functions when the medical support program stored in the storage is executed by the processor. The details of each component of the medical support device 1 will be described below.

[0032] [Clinical Parameter Generation Unit] The clinical parameter generation unit 10 analyzes the raw clinical information collected through interviews via the first terminal 2 and plays the role of converting and structuring it into clinical parameters that can be processed logically by the subsequent clinical path identification unit 30 and clinical path determination unit 50. Here, clinical parameters refer not only to numerical data but also to a collection of logical information elements such as keywords, flags, and categories that have specific medical meanings.

[0033] Furthermore, the clinical information input to the clinical parameter generation unit 10 includes standardized data such as multiple-choice questionnaire responses, text entered by the patient via the first terminal 2, as well as non-standardized data such as images, videos, or past medical history obtained from electronic medical records, and text entered by the physician via the second terminal 3, etc. The clinical parameter generation unit 10 utilizes morphological analysis, syntactic analysis, and semantic analysis technologies such as natural language processing (NLP) algorithms and large-scale language models (LLM) to extract medically significant information from the text obtained from both the patient and the physician.

[0034] Specifically, it first identifies "attribute information" such as the patient's type, breed, age, weight, whether they are spayed or neutered, and their medical history. Next, it extracts information about the "chief complaint," including the location, time of onset, frequency, duration, and nature of the symptoms (e.g., color of vomit, quality of cough sound). For example, in response to an atypical input such as "I vomited a yellow liquid about three times since last night," this component generates a series of logical clinical parameters such as "Symptoms: Vomiting, Frequency: 3 times / day, Contents: Bile-like (yellow), Onset time: The previous night."

[0035] Furthermore, the clinical parameter generation unit 10 may be configured to assign a confidence score to each generated parameter according to the uncertainty of the information and the degree of subjectivity of the observer. This makes it possible to process objective numerical data (such as weight and body temperature) and observational information that is prone to subjectivity from the observer (such as "the animal looks unwell") with different weights in the subsequent judgment process. In this way, by standardizing diverse forms of input into a common standard called "clinical parameters," it becomes possible to directly compare the condition of an animal (patient) that cannot speak with expected values, which are medical standards.

[0036] [Clinical Management Data Storage Unit] The clinical management data storage unit 20 holds clinical management data, which is the core structure of medical knowledge that forms the basis of the clinical support logic of the present invention. This data is not merely a list of information, but has a relational structure in which the clinical path model, evaluation categories, and evaluation items are interconnected, as shown in Figures 3 to 5. The definitions of each table and the technical roles they play will be described in detail below.

[0037] At the first level, clinical management data holds a "clinical path table" (Figure 3) that defines the treatment templates to be applied according to the attributes of each case. In order to identify each template without duplication within the system, this table includes a unique management number (ID) corresponding to the patient type, symptoms, and type of visit, as well as multiple attribute conditions used for matching with clinical parameters, such as "chief complaint," "patient type," "age / months," and "type of visit: first visit / follow-up." A "valid / invalid" flag controls whether the clinical path model is "available" at the moment, preventing physicians from selecting outdated clinical path models while maintaining links with past case data.

[0038] As a second layer, the clinical management data holds an "evaluation category table" (Figure 4) that retrieves one or more evaluation axes that should be prioritized in a given clinical pathway from the ID of the identified clinical pathway model. Each record holds an "identification code" to uniquely identify the category and a "category name" that is recognizable by users such as physicians.

[0039] Furthermore, this table defines a "medically expected state," which indicates the clinical course that the case should ideally follow, as a parameter that constitutes the judgment logic. In addition, this table includes weighting coefficients (w1, w2, w3) that determine the contribution to each judgment index (P1, P2, P3) of urgency, prediction deviation, and uncertainty described later, as well as threshold data (T1, T2) for determining mild or major deviations, and notification priority that defines the display order and emphasis of information provided to the physician when an abnormality is detected.

[0040] In this way, by linking information via IDs, abstract clinical pathway models can be translated into concrete evaluation categories. In particular, because independent weighting coefficients and thresholds are defined for each category, it is possible to realize a detailed judgment logic based on medical validity, such as giving a heavy rating to abnormalities in "general condition" and flexibly judging minor deviations in "medication status".

[0041] As the third layer, clinical management data holds an "evaluation item table" (Figure 5) that concretizes individual evaluation axes via the ID of each evaluation category. This table has an "ID" and an "identification code" to uniquely identify records, as well as a "medical item code," which is a normalized code to identify calculation processing within the system. Furthermore, each medical item is associated with "confirmation items" that interactively present the matters that the physician should confirm or the content of additional questions to be asked in that item, and "data attributes" that define the calculation properties, such as whether the input value is numerical, boolean, or categorical.

[0042] Furthermore, the evaluation item table includes logical parameters for calculating the three judgment indicators (first indicator P1, second indicator P2, and third indicator P3), which are detailed below. Specifically, an "importance" is defined, which specifies how much weight the detected abnormal values ​​or deviations from the expected state carry in clinical management. This "importance" is used as a calculation coefficient in the calculation logic of each judgment indicator, particularly when evaluating severity and urgency. In addition, each item has a "missing information coefficient" defined, which specifies the degree to which the absence of clinical information, i.e., the state of incomplete information, affects the judgment result. This "missing information coefficient" contributes particularly to the calculation of the judgment indicator (P3) that ensures comprehensiveness of information, and the system automatically adds a risk when necessary interviews have not been conducted. The details of each judgment indicator P1 to P3 are explained below.

[0043] <First indicator P1> The first indicator, P1, is an index that shows the degree of deviation between clinical parameters and the "medically expected state (normal range or desirable state)" defined in the clinical management data. This quantifies the "magnitude of risk"—how medically serious or urgent the patient's current condition is.

[0044] <Second indicator P2> The second indicator, P2, is an index that shows the extent to which the content of the clinical parameters "contradicts medical predictions" when viewed from a standard clinical pathway model (normal course and predicted progression of the disease). This quantifies events with a low statistical frequency or the occurrence of clinical conditions that are not normally observed in specific attributes (type, age, etc.).

[0045] <Third indicator P3> The third indicator, P3, is an index that shows the reliability and completeness of the clinical information that forms the basis of medical treatment. Specifically, it quantifies the state of "incompleteness," where clinical information corresponding to the necessary evaluation items is missing, the lack of consistency (logical contradiction) between multiple responses, or "ambiguity of definition" due to the subjective opinion of the accompanying person.

[0046] Thus, the clinical management data storage unit 20 logically links clinical pathways, evaluation categories, and evaluation items via IDs, and multiplies them by the "weight coefficients (w1~w3)" shown in the "evaluation category table," thereby functioning as a template for medical reasoning that derives the fit of clinical information to the clinical pathway model from fragmented information. This makes it possible for even inexperienced physicians to receive comprehensive and highly accurate evaluation support equivalent to that of veteran physicians. Here, the fit status in this application is an indicator that shows how well the clinical information matches or deviates from the expected value defined in the clinical pathway model, and in the embodiment, it includes a numerical value calculated as the deviation degree D.

[0047] Furthermore, the clinical pathway models stored in this table may include not only templates based on standard clinical guidelines, but also models uniquely set up by specific medical institutions, and individual models customized to suit the medical history and constitution of specific patients. In addition, the various parameters (weighting coefficients, thresholds, importance, missing coefficients, etc.) associated with each clinical pathway model, evaluation category, and evaluation item stored in the clinical management data storage unit 20 are configured to be added, deleted, or modified as needed through user operation via the administrator terminal. This makes it possible to update evaluation criteria based on the latest medical evidence and to fine-tune the judgment sensitivity to suit the actual conditions of specific clinical settings without having to rebuild the entire system.

[0048] Furthermore, the clinical management data storage unit 20 holds one or more "clinical component data" that have been pre-validated for medical validity and arranged in a common format, as resources for constructing an ad-hoc model when an existing clinical pathway model cannot be identified. Specifically, it includes a "judgment rule table" (Figure 6) and a "treatment slot table" (Figure 7).

[0049] Each of these tables functions as a standard library of materials for dynamically generated clinical pathway models. That is, when a temporary model is constructed in the clinical pathway identification unit 30 described later, the artificial intelligence engine does not create text from scratch, but rather selects definitions (decision mechanisms and treatment items) registered in each of these tables based on identification information (ID), and builds the model by linking them together.

[0050] In this way, by generating models within a range of predefined options managed by the system, the vocabulary and logical structure of the generated models are controlled to stay within a medically correct range. This technically prevents the generation of unfounded information by the artificial intelligence engine, while enabling consistent clinical support even for unknown cases.

[0051] Specifically, the artificial intelligence engine is prevented from dynamically generating unique medical treatment item names that lack medical basis or unique calculation formulas that lack validity. Instead, it is only permitted to "select and rearrange" the existing identification information (IDs) registered in the "Judgment Rule Table" (Figure 6) and the "Treatment Slot Table" (Figure 7) that best match the current clinical parameters. This technically eliminates the risk of inappropriate evaluation axes being introduced unintended by the system, even for unknown cases, thereby ensuring the objectivity and reliability of clinical support.

[0052] [Medical Information Storage Department] The medical information storage unit 60 is a database that defines medical profiles containing one or more clinical patterns suggesting a specific disease or pathological condition, and is used as a matching source for inferring disease candidates for individual cases in the clinical support information generation unit 70 described later. The medical information storage unit 60 is not limited to a local database built on the storage within the clinical support device 1, but broadly encompasses databases on external servers connected via the communication network 4, or knowledge bases accumulating external medical information such as medical papers and clinical guidelines. In other words, the medical information storage unit 60 in this application also includes a logical storage area where the clinical support information generation unit 70 searches and references external medical information in real time and temporarily or permanently stores the search results. While the clinical management data storage unit 20 holds a standardized clinical process, this storage unit holds statistical and medical disease characteristics such as combinations of symptoms and trends in indicators that are likely to appear in actual diseases.

[0053] The clinical patterns included in the medical profile include combinations of clinical parameters that are frequently observed in specific diseases, specific numerical ranges of the first indicator P1 to the third indicator P3 calculated by the clinical path determination unit 50, or definitions of critical abnormalities (red flags) that should not be overlooked. For example, in a medical profile related to heart failure, a clinical pattern is defined as a combination of conditions such as the first indicator P1, which indicates urgency, being above a threshold, and the clinical parameter "body weight" increasing by a predetermined rate or more in the past week. By incorporating this memory unit, the system can go beyond simply detecting deviations from the path and provide more in-depth clinical support by determining what diseases those deviations suggest.

[0054] [Clinical Pathway Specialization Department] The clinical pathway identification unit 30 is responsible for automatically selecting and identifying the optimal treatment template (clinical pathway model) to be applied to the case from the clinical pathway table (Figure 3) in the clinical management data storage unit 20, based on the individual patient information structured by the clinical parameter generation unit 10. This identification process is performed not by single keyword matching, but by multidimensional filtering combining multiple attribute parameters.

[0055] Specifically, the first step is to narrow down the population that will serve as the basis for evaluation based on "patient type" and "age / months." In veterinary medicine, even with the same symptoms, the diseases that need to be differentiated can differ greatly depending on the species and age, so at this stage, the narrowing down is done to reflect individual differences. Next, the "chief complaint" and "type of visit" are compared, and for example, if it is "a dog's first visit due to vomiting," a screening pathway specifically for the initial visit is identified, and if it is "a cat's follow-up visit with chronic renal failure," a follow-up pathway for monitoring the progress is identified.

[0056] Furthermore, if multiple matching paths exist, or if a clear path cannot be identified, the clinical pathway identification unit 30 may prioritize selecting a path with a high degree of satisfaction of clinical parameters, or it may identify multiple paths in parallel and pass them to the subsequent determination unit. This configuration instantly determines the medical evaluation axes that should be confirmed for the patient at the beginning of treatment. This has technical significance in that it eliminates the need for physicians to manually search for appropriate paths from a vast number of clinical guidelines, and presents a standardized treatment process that eliminates hesitation from the start of treatment.

[0057] Furthermore, the clinical pathway identification unit 30 has a function to dynamically construct a temporary clinical pathway model without using an existing model, depending on the satisfaction status in the identification process from the clinical pathway table (Figure 3). The satisfaction status here refers to, for example, a state in which there are no candidate existing models that match the input clinical parameters, or a state in which the degree of agreement (fit) with the extracted candidate models falls below a predetermined arbitrary threshold.

[0058] In the specific generation process, the clinical pathway identification unit 30 uses an artificial intelligence engine to extract elements necessary for evaluating the case from the current clinical parameters (chief complaint, attributes, etc.). At this time, instead of allowing the artificial intelligence engine to freely generate text or make judgments, it is made to select regular identification information defined in the aforementioned "judgment rule table" (Figure 6) and "medical treatment slot table" (Figure 7), and construct a model by logically combining them.

[0059] As an example, suppose the chief complaint "sudden vomiting of unknown cause" is extracted from clinical parameters, but there is no existing clinical pathway model that matches it in the clinical pathway table (Figure 3). In this case, the artificial intelligence engine first refers to the "clinical slot table" (Figure 7) and selects a slot ID (e.g., S_ID / S_001) to check the characteristics of the vomiting. This slot ID is pre-associated with a standard question such as "What color is the vomit?" and a choice data type.

[0060] Next, the artificial intelligence engine refers to the "judgment rule table" (Figure 6) and selects the rule ID (e.g., R_ID / R_001) of the medical judgment logic to be applied to the item in question. The predefined logic associated with this ID, "assign the maximum risk value if the input value is red," is incorporated as the judgment algorithm of the dynamically generated clinical path model.

[0061] The clinical pathway identification unit 30 logically combines these selected R_IDs and S_IDs to generate "temporary evaluation item data" that conforms to the record format of the evaluation item table shown in Figure 5. In this way, instead of having the artificial intelligence engine create text or perform calculations from scratch, the system selects regular IDs that it manages and maps (places) them into the existing data structure to construct a temporary clinical pathway model.

[0062] This technically eliminates the risk of inappropriate questions unintended by the system or the inclusion of proprietary calculation formulas without medical basis, even for unknown cases. Since the generated temporary clinical pathway model has the same data specifications as existing models, it can be used in the subsequent clinical pathway determination unit 50 to calculate objective determination indicators (P1 to P3) without requiring any special processing.

[0063] [Evaluation Item Identification Section] The evaluation item identification unit 40 is responsible for extracting specific evaluation categories and evaluation items that should be evaluated in the case from the clinical management data storage unit 20, based on the clinical pathway model identified by the clinical pathway identification unit 30. This unit performs a mapping process that links the current clinical parameters to the extent to which the selected pathway covers the ideal medical state required by the selected pathway.

[0064] Specifically, the system first uses the identified ID as a key to refer to the evaluation category table (Figure 4) and identify multiple evaluation axes (e.g., "medication status," "stool," "general condition," etc.) that should be examined in that path. Next, it extracts the individual evaluation items associated with each category from the "evaluation item table" (Figure 5), and by logically combining this information, it generates an "evaluation dataset" that covers all the items to be collected and evaluated in the clinical practice, as well as each judgment criterion (weight, threshold, expected value, etc.).

[0065] A key technical feature of this component is that it not only extracts a list of items, but also compares them with the generated clinical parameters and classifies items into those for which information already exists and those for which information is missing. For example, if the parameter "appetite: present" has already been obtained through a pre-examination questionnaire from a companion, the evaluation item identification unit 40 marks that item on the evaluation matrix as already entered and prepares it for judgment processing. On the other hand, if the input data does not contain information on "changes in water intake" required by the pathway, it identifies this as an "unfulfilled item."

[0066] Furthermore, the evaluation item identification unit 40 may be configured to dynamically adjust the priority and necessity of evaluation items according to the patient's attributes and past medical history. For example, if a specific pathway is selected for an elderly patient, the importance of evaluation items related to "cardiopulmonary function" is automatically increased compared to younger individuals, and these items are identified as items that should be checked with priority.

[0067] In this way, the evaluation item identification unit 40 distinguishes between "essential items required by the pathway" and "currently possessed information," enabling accurate calculation of the third indicator P3, which is an indicator of the degree of information adequacy in the subsequent judgment unit, and accurate processing of missing information in the additional questionnaire generation unit 80. This is an extremely important process in veterinary medicine, where information is often insufficient, in order to ensure comprehensiveness of treatment and prevent overlooking serious symptoms.

[0068] [Clinical Pathway Evaluation Department] The clinical pathway determination unit 50 is responsible for quantitatively calculating how well the clinical parameters obtained from the patient "fit" to the medically expected state, based on the evaluation dataset (combined data in Figures 3 to 5) generated by the evaluation item identification unit 40. This unit does not merely determine "normal" or "abnormal," but calculates the three independent determination indices (P1, P2, and P3) mentioned above, and further integrates these to derive the degree of deviation D, which is the degree of fit of the clinical information to the clinical pathway model.

[0069] First, in calculating the first indicator P1, this component compares clinical parameters with the expected values ​​of the clinical pathway model. If the clinical parameters deviate from the expected values, the magnitude of the medical risk is quantified by multiplying the deviation by an importance coefficient, etc. As a result, deviations of high-importance items such as "decreased level of consciousness" are calculated as a larger indicator than deviations of "slight increase in heart rate."

[0070] Furthermore, in calculating the second indicator P2, this component checks for deviations in the content of clinical parameters from the standard course and medical predictions defined by the identified clinical pathway model. Specifically, an unexpected score is added when a combination of symptoms with a statistically low incidence or a sudden change in condition (anomaly) that is not normally predicted given the individual's attributes (species, age, etc.) is observed. This quantifies the signs of "irregular pathological conditions" that do not fit into typical case patterns.

[0071] Furthermore, in calculating the third indicator P3, this component logically verifies the reliability and completeness of the information that forms the basis of treatment. Specifically, it checks for empty slots (unfulfilled items) among the evaluation items required by the identified clinical pathway model to which data was not assigned, and sums up the "missing penalty." In addition, it verifies the consistency (inconsistency) between multiple clinical parameters and the ambiguity of definitions due to the subjective opinions of accompanying personnel, and adds an uncertainty score based on a logical exclusion rule. This clearly demonstrates the risk that a lack of information or inconsistencies may lead to missed treatment decisions.

[0072] Finally, the clinical pathway determination unit 50 multiplies each of these P1, P2, and P3 indicators by weighting coefficients (w1, w2, w3) defined for each evaluation category to calculate the degree of deviation D, which indicates the degree to which the entire case deviates from the clinical pathway model. The calculation formula is as follows: D = (P1 × w1) + (P2 × w2) + (P3 × w3)

[0073] The calculated deviation score D is then compared to a predetermined threshold, and if it exceeds a specific value, subsequent actions such as "warning the physician" or "generating additional questions" are triggered. In this way, by using multifaceted indicators, it is possible to determine medical validity that goes beyond mere numerical evaluation in animal medicine, where the quality of information is often unstable.

[0074] In the processing performed by the clinical path determination unit 50, the direct object of determination is the clinical parameters structured by the clinical parameter generation unit 10, which are broadly included in clinical information. Therefore, the comparison of clinical information and expected values ​​broadly includes not only the act of directly comparing clinical information, but also the act of determining consistency with expected values ​​through intermediate data formats (parameters, scores, etc.) extracted and generated from clinical information.

[0075] [Medical assistance information generation department] The medical support information generation unit 70 is responsible for generating medical support information that identifies candidate conditions or diseases in the patient by calculating the degree of fit between the judgment index (at least one of the first index P1, second index P2, and third index P3) calculated by the clinical pathway determination unit 50 and the medical profile held in the medical information storage unit 60.

[0076] The method for calculating the degree of fit by the medical support information generation unit 70 is not limited to a specific algorithm, but can employ various calculation methods that can ensure medical validity. For example, one method is to calculate the proximity or similarity (cosine similarity, etc.) between an index vector in a multidimensional space whose components are the calculated indices P1 to P3 and a standard clinical pattern (reference vector) for each disease defined in the medical information storage unit 60.

[0077] Alternatively, as another determination method, the medical information storage unit 60 may maintain a table defining the degree of confidence in a disease according to the combination of values ​​of each indicator P1 to P3, and the degree of fit may be calculated by matching with this table. Furthermore, the calculated indicators P1 to P3 and clinical parameters may be input to an artificial intelligence engine such as a machine learning model or a large-scale language model, and the probability of belonging to each disease candidate may be estimated based on the knowledge data held in the medical information storage unit 60.

[0078] As described above, the medical support information generation unit 70 compares the findings of the medical information storage unit 60 with those of the clinical path determination unit 50 via a common data format: an objective index of suitability. This makes it possible to perform logical disease prediction based on medical evidence, going beyond simple keyword searches, even in initial medical treatment where information uncertainty is high.

[0079] [Additional interview generation department] The additional questionnaire generation unit 80 works in conjunction with the clinical pathway determination unit 50 to dynamically generate additional questions necessary to resolve information deficiencies (incomplete information) or medical inconsistencies. Specifically, if the calculated deviation degree D meets predetermined conditions (e.g., near the normal threshold T1, or exceeding the critical threshold T2), or if a particular indicator among the constituent indicators (P1, P2, P3) exceeds a certain contribution rate, the unit performs an additional questionnaire to determine or reduce the value of that indicator.

[0080] As for the specific generation process, if the first indicator P1 is the main factor contributing to the increase in deviation degree D (e.g., if a risk factor suggesting urgency is detected), this component prioritizes generating emergency questions for triage purposes. These include items to immediately identify life-threatening risk factors, such as the level of consciousness, mucous membrane color, or the presence or absence of co-occurring symptoms, thereby improving the accuracy of the assessment.

[0081] Next, if the second indicator P2 is influencing the degree of deviation D, this component generates a questionnaire to check the changes in the parameter that deviated from the predictions of the clinical pathway model. This helps to distinguish whether the statistically "unexpected event" is temporary or due to the progression of the disease.

[0082] Furthermore, if the third indicator P3 is driving up the deviance D, this component performs processes to fill in information gaps or resolve inconsistencies. Specifically, it generates direct questions to fill in required slots determined to be "unfulfilled" in the evaluation items, or questions that include options to reconfirm the logic between answers (such as quantitative constraint inconsistencies). This process reduces the uncertainty of the information and ensures the reliability of the deviance D.

[0083] The generated questions are presented directly to the accompanying person via an interactive interface, or as recommended confirmations for the physician. This configuration enables the system to move beyond passive data collection and proactively seek out the information necessary to enhance medical validity, thus providing strong support for physicians in situations with limited time, enabling them to collect highly accurate clinical information with minimal effort.

[0084] It should be noted that the additional questionnaire generation unit 80 described above is not an essential component of the present invention, and can be omitted or selectively executed as appropriate depending on the system's operating mode and the required clinical accuracy. For example, based on the suitability status calculated by the clinical pathway determination unit 50, the system may be configured to present clinical support information to the physician via the output unit 90 described below without collecting any additional information. In this case, the output unit 90 directly outputs the determination result based on the primary information of the questionnaire, that is, an evaluation of medical validity within the scope of the information obtained at this time, and visualization of areas where information is lacking (unfulfilled items).

[0085] In another embodiment, the questions generated by the additional questionnaire generation unit 80 may be presented to the physician as "recommended confirmation items (hearing guide during medical consultation)" rather than being used as direct questions to the accompanying person. In this way, the system plays a supplementary role in suggesting the next actions to the physician to further improve accuracy while presenting analysis results based on primary information.

[0086] Thus, the medical support device 1 of the present invention is centered on the clinical pathway determination unit 50 determining compatibility with the expected medical state and includes both a form that immediately outputs the determination result (result based on primary information) and a form that further delves into the information via the additional questionnaire generation unit 80.

[0087] [Output section] The output unit 90 outputs the suitability status calculated by the clinical pathway determination unit 50 to the display screen of the second terminal 3 used by the physician, presenting it in a format that the user can intuitively understand. The output unit 90 includes, for example, an upper area that displays the input primary information, namely the "questions and answers from the medical interview (Q&A)," and a lower area that displays the system's determination results based on this information in text format. In the lower area, in addition to a summary of the entire case, such as "urine volume and final urination time are undetermined," the specific details of any missing (unfulfilled) required items or internal inconsistencies are clearly indicated in text format.

[0088] Furthermore, the output unit 90 may be configured to present, in addition to or instead of these detailed textual comments, visual information to provide a multifaceted understanding of the suitability status. Specifically, it may generate radar charts based on the P1, P2, and P3 indicators to graphically visualize the risks of the case and the incompleteness of the information. In addition, dynamic display control is performed according to the nature of the information, such as "highlighting in red" for items with high P1, and adding a "warning icon" to items where P2 (prediction deviation) is detected or where necessary information is missing. This makes it possible for physicians to visually identify areas in the vast amount of clinical information that require immediate confirmation or additional questioning.

[0089] This allows physicians to intuitively identify which information is missing or questionable immediately, using visual indicators while reading the presented Q&A and judgment text. This configuration strongly supports rapid decision-making with high medical validity, even in the early stages of treatment when the quality of information is unstable.

[0090] The above is an overview of the medical support system including the medical support device 1 according to an embodiment of the present invention. Next, the medical support program executed in the medical support device 1 will be described. The medical support program according to this embodiment is executed by a processor installed in a computer (medical support device 1, or a second terminal 3 used by a physician, etc.), and realizes the determination of medical validity based on primary information through the processing of each step shown below. Figure 9 is a flowchart of the overall processing by this program.

[0091] [STEP 101: Generating Clinical Parameters] The program reads "clinical information" obtained from past medical records (electronic medical record data, etc.) of the patient, or from questionnaire information entered by the patient's companion, and generates "clinical parameters" that include the patient's attribute information (type, age, etc.) and chief complaint (the core complaint of the symptoms). This step includes processes to extract chief complaint codes from free-form text in natural language and to integrate attribute data and questionnaire responses as structured data.

[0092] [STEP 102: Identifying Clinical Pathway Models] The program references clinical management data stored in the clinical management data storage unit 20 based on the generated clinical parameters. This clinical management data defines the correspondence between clinical pathway models that define the expected medical state, the associated evaluation categories, and the evaluation items necessary for judgment (Figures 3 to 5). The program selects a specific clinical pathway model that matches the chief complaint and attribute information and identifies the criteria framework for evaluation in this medical practice.

[0093] [STEP 103: Identifying evaluation items] The program identifies one or more evaluation categories associated with the identified clinical pathway model, and one or more evaluation items necessary to determine those categories. Here, by traversing the hierarchical structure of clinical management data, all items that "medically need to be confirmed" in the current case are extracted. Furthermore, this step may include a process to compare the identified evaluation items with already acquired clinical information and determine whether the information is complete or not (identification of incomplete items).

[0094] [STEP 104: Clinical pathway evaluation] The program determines the fit of clinical information to an identified clinical pathway model based on a comparison of clinical parameters obtained from clinical information with expected values ​​(normal range or desirable state) defined for each evaluation item. Specifically, it calculates a first indicator P1 showing the degree of deviation from expected values, a second indicator P2 showing the specificity of the event, and a third indicator P3 showing the absence or inconsistency of information, and then calculates a deviation degree D by integrating these to quantify the medical validity and risk of the case.

[0095] [Additional processing] Based on the determined suitability status, the program executes "STEP 105: Generate additional questionnaire" and "STEP 106: Output results" as needed. In the results output, the primary information in Q&A format from the questionnaire and the assessment results from STEP 104 (text and visualization information) are displayed side by side on the screen of the terminal used by the physician.

[0096] Furthermore, the execution order of each step (STEP101 to STEP106) in the aforementioned medical support program can be changed as appropriate, as long as there is no logical dependency such as using the results of a previous step in a later step, given the nature of the processing.

[0097] For example, the generation of clinical parameters (STEP 101) and the identification of clinical pathway models (STEP 102) may be performed in parallel, allowing for the identification of chief complaints and the narrowing down of pathway candidates simultaneously. Furthermore, a dynamic and real-time processing configuration can be adopted, such as sequentially starting calculations for clinical pathway determination (STEP 104) in the background once specific categories are determined during the identification of evaluation items (STEP 103).

[0098] Furthermore, it is not essential to calculate all three indicators—the first indicator P1, the second indicator P2, and the third indicator P3—for the judgment criteria. Depending on the operational purpose of the system and the characteristics of the target disease, configurations that, for example, focus only on the third indicator P3 to evaluate the reliability of the information, or configurations that combine other medical indicators to comprehensively evaluate the fit, are also included within the technical scope of the present invention.

[0099] The operation and effects of the aforementioned medical support device and medical support program when applied to specific cases will be described in detail below using examples. [Examples]

[0100] 1. Case summary and acquisition of primary information In this embodiment, we assume a case where a patient (pet: cat, 6 years old, first visit) comes to the clinic with the chief complaint of "concern about the cat's litter box behavior." First, the owner answers a pre-consultation questionnaire via an interactive interface using terminal 12 at home or elsewhere. In this embodiment, we assume that answers A1 to A4 are obtained from the owner for the pre-consultation questionnaire Q1 to Q4 as follows, and this is acquired as primary information, known as "clinical information." Q1: "Are you peeing?" A1: "Yes (I am)." Q2: "Frequency of urination" A2: "I go to the toilet many times (frequent)" Q3: "Amount of urine per urination" A3: "Unknown" Q4: "Are you feeling well?" A4: "I'm fine."

[0101] 2. Generation of clinical parameters The clinical parameter generation unit 10 analyzes the acquired clinical information and generates "clinical parameters," which are structured data suitable for subsequent calculation processing. Specifically, it generates a dataset that integrates patient attribute information (type: cat, age: 6 years / 72 months, medical history: none) acquired from electronic medical records, etc., the chief complaint code normalized from the questionnaire responses (chief complaint: urinary abnormality), and the status values ​​corresponding to each evaluation item (urination flag: True, frequency: High, amount: Unknown, energy level: Normal). Through this process, subjective responses from pet owners that include ambiguous nuances such as "the cat is urinating" and "the cat is energetic" are converted into quantitative clinical parameters that can be calculated by medical algorithms.

[0102] 3. Identification of clinical pathways The clinical pathway identification unit 30 uses the chief complaint code and animal species from the generated clinical parameters as search keys to refer to the clinical management data stored in the clinical management data storage unit 20. As a result, it identifies a "clinical pathway model for feline lower urinary tract disease" to assess the risk of conditions such as feline urethral obstruction and cystitis as the expected medical condition to be applied to this case.

[0103] 4. Identification of evaluation items The evaluation item identification unit 40 identifies evaluation categories ("urinary status," "general condition," etc.) associated with the identified clinical pathway model, and specific evaluation items necessary to determine them (last urination time, presence or absence of hematuria, specific urine volume, presence or absence of vomiting, etc.). Here, the program compares the identified evaluation items with the clinical parameters generated by the clinical parameter generation unit 10. As a result, it internally flags any data corresponding to items of high importance in the pathway, such as "last urination time" or "presence or absence of vomiting," that are not included in the current parameter set (i.e., are not satisfied).

[0104] 5. Generation of additional questionnaires The additional questionnaire generation unit 80 identifies the flagged unfulfilled items (last urination time, specific urine volume) and inconsistencies between the initial responses and expected values. Based on this, it automatically generates additional questions optimized to supplement the lack of information and to verify the highest priority risk of "suspected urethral obstruction." The owner's responses are obtained as answers to these dynamically generated questions.

[0105] 6. Evaluation of clinical pathways The clinical pathway determination unit 50 calculates determination indices (P1, P2, P3) based on a comparison between clinical parameters obtained from clinical information and expected values ​​corresponding to evaluation items. Next, it integrates these calculated indices to derive the degree of deviation D from the clinical pathway for the entire case. The clinical pathway determination unit 50 finally determines the fit status to the clinical pathway model by comparing the derived degree of deviation D with a predetermined threshold. The following shows an example of the calculation of determination indices in this embodiment.

[0106] <Calculation of the first indicator P1> Currently, the "frequency" parameter deviates from the expected value (normal frequency), so it is multiplied by an importance coefficient to calculate the likelihood of the presence of medical risks such as lower urinary tract disease.

[0107] <Calculation of the second indicator P2> This study evaluates the specificity of events in which sudden frequent urination occurs in individuals with no prior medical history within a identified clinical pathway. By scoring outliers from standard physiological prediction models, the degree to which a statistically "unexpected event" has occurred is calculated.

[0108] <Calculation of the third indicator P3> The "lack of clarity" and "logical consistency" of the information are evaluated. Specifically, a penalty for omissions is calculated for the absence (incompleteness) of "last urination time" and "detailed urine volume," which are essential for determining urgency. In addition, the ambiguity (contradiction / uncertainty) of the definition is evaluated, as the owner responded that "urination is occurring," but the specific manner of urination is "unknown." Based on medical knowledge, the amount of urination should be measurable in normal urination, and the third indicator P3 is calculated high to reflect the inconsistency between the claim that "urination is occurring" and the fact that "the amount is unknown."

[0109] 7. Scoring The deviation degree D is calculated by multiplying each index (P1, P2, P3) calculated by the clinical pathway evaluation unit 50 by weight coefficients (w1, w2, w3) defined for each evaluation category. In the "urinary abnormalities" category of this embodiment, the following weight setting is used, which places a high weight on the third index P3, in order to emphasize the detection of discrepancies in recognition and gaps in information in observational information of animals that cannot speak. Note that the numerical values ​​are just examples and will be dynamically optimized according to the nature of the clinical pathway. • Weight of the first index P1 (w1): 0.35 • Weight of the second indicator P2 (w2): 0.20 • Weight of the third indicator P3 (w3): ​​0.45

[0110] The deviation degree D is then calculated using the following formula. D = (P1 × w1) + (P2 × w2) + (P3 × w3) By comparing the calculated deviation degree D with the normal threshold (T1: 0.4) and critical threshold (T2: 0.7) defined in the clinical management data, the system performs actions according to the following two cases (Case A and Case B).

[0111] (Case A: When the degree of deviation is less than the threshold T1) The pet owner provides the following additional answers to the additional questionnaire generated by the additional questionnaire generation unit 80, and the lack of clarity in the information is considered resolved. A5: "The last time a substantial amount came out was 30 minutes ago." A6 "The amount is about the same as usual." A7: "I didn't see any blood, and there was no vomiting."

[0112] These responses allow the evaluation item identification unit 40 to fill in the required slots, such as "time of last urination" and "specific amount of urine," which were flagged, with normal values. The clinical pathway determination unit 50 determines that the uncertain state of "frequent but unknown amount" in the initial response has been resolved and calculates the third indicator of uncertainty, P3, as 0.15. Furthermore, since the most recent normal urination has been confirmed, the risk of fatal consequences such as urethral obstruction is ruled out, and the first indicator of risk, P1, is also updated to 0.10. In addition, since these events are within the range of predictions based on the clinical pathway model, the second indicator of unexpectedness, P2, is also set to a low value of 0.15.

[0113] Based on the above results, the calculated deviation degree D is given by the following formula. D=(0.35×0.10)+(0.20×0.15)+(0.45×0.15)

[0114] The calculated deviation D (=0.1325) is below the threshold T1 (e.g., 0.40) which indicates the normal range, so the system determines the fit to be "good (normal course)". In response, the output unit 90 provides the physician with appropriate advice, ensuring the reliability of the information, such as "The urgency is low. We recommend standard examinations and follow-up based on the chief complaint."

[0115] (Case B: When the degree of deviation exceeds the threshold T2) On the other hand, consider the case where, in response to similar additional questions, the owner gives the following answers suggesting a worsening of the situation. A5' "The last time a significant amount was produced is unknown." A6' "The amount of urine coming out might only be a few drops." A7' "The patient has vomited and appears lethargic."

[0116] These responses are significant events that directly contradict the initial clinical parameters ("producing" and "feeling normal"). The clinical pathway determination unit 50 determines that the owner's perception of "producing" was likely merely a medically inadequate drip rate, and calculates the third uncertainty indicator P3 to the maximum level of 0.90, reflecting the inconsistency with the initial information and the ambiguity of the information definition.

[0117] Furthermore, given the prolonged inability to observe effective urination and the deterioration of "basic indicators of vital activity (energy and appetite)," the risk of urethral obstruction and associated uremia is assessed as extremely high, and the first indicator of risk, P1, is raised to 0.85. As this is a statistically dangerous course, the second indicator, P2, which indicates the degree of unexpectedness, is also raised to 0.40.

[0118] As a result, the calculated deviation degree D is as follows: D=(0.35×0.85)+(0.20×0.40)+(0.45×0.90)

[0119] The calculated deviation degree D (=0.7825) exceeds the threshold T2 (e.g., 0.7) indicating a serious abnormality, so the system determines it to be a "serious deviation." The output unit 90 immediately sends a warning to the physician's terminal, displaying the advice "Urgent: Suspected urethral obstruction. Significant discrepancy between initial response and actual event. Immediate treatment recommended" as a priority check item, supporting the optimization of medical treatment based on urgency. [Examples]

[0120] 1. Case summary and acquisition of primary information In this example, we consider a case where a patient (pet: dog, 10 years old, first visit) with a chronic disease (e.g., heart disease or chronic kidney disease) requiring continuous treatment comes to our clinic wishing to continue treatment from their previous veterinarian due to reasons such as relocation. In this case, the patient does not have a referral letter from the previous veterinarian, and the details of the prescribed medication (drug name, exact dosage) are unknown. Then, we assume that the owner has answered the following pre-consultation questions Q1 to Q3, A1 to A3, and this will be acquired as primary information, known as "clinical information". Q1: "Do you give your child medication?" A1: "I give it to them every day without fail." Q2: "When did you receive the medication?" A2: "I received it two weeks ago (14 days ago)." Q3: "How many tablets are left?" A3: "10 tablets"

[0121] 2. Generation of clinical parameters The clinical parameter generation unit 10 analyzes the acquired clinical information and generates "clinical parameters," which are structured data suitable for subsequent calculation processing. Specifically, it generates a dataset that integrates patient attribute information (type: dog, age: 10 years / 120 months, medical history: present) obtained from electronic medical records, etc., chief complaint codes normalized from interview responses (chief complaint: medication confirmation), and status values ​​corresponding to each evaluation item (medication frequency: 1 per day, elapsed days: 14 days, remaining medication: 10 pills). Through this process, subjective responses from pet owners that include ambiguous nuances such as "I give it to him every day" or "I got it two weeks ago" are converted into quantitative clinical parameters that can be calculated by medical algorithms and normalized into a format that allows for quantitative constraint verification.

[0122] 3. Identification of clinical pathways The clinical pathway identification unit 30 refers to the clinical management data stored in the clinical management data storage unit 20 based on the chief complaint keyword "medication confirmation" included in the clinical parameters and the patient type. As a result, it identifies a "medication management clinical pathway model" as the expected medical state to be applied to this case.

[0123] The "Clinical Pathway Model for Medication Management" incorporates a knowledge base of expected drug consumption rates for specific diseases, the risk associated with medication discontinuation, and statistically frequently used prescription units. This system uses this model as a baseline to dynamically evaluate the extent to which real-world clinical parameters deviate from the "ideal state."

[0124] 4. Identification of evaluation items The evaluation item identification unit 40 identifies evaluation categories (such as "medication status" and "general condition") associated with the identified "clinical pathway model for medication management," and specific evaluation items necessary to determine them (such as drug name, dose per administration, number of administrations per day, total number of prescriptions, current number of remaining medications, and response after administration). Here, the program compares the identified evaluation items with the clinical parameters generated by the clinical parameter generation unit 10. As a result, it internally flags any items that are not included in the current parameter set (i.e., are not satisfied), such as "drug name" and "initial total number of prescriptions," which are of high importance in verifying the quantitative consistency in the pathway.

[0125] 5. Generation of additional questionnaires The additional questionnaire generation unit 80 identifies the flagged unfulfilled items (drug name, total prescription quantity) and verification items to resolve any potential quantitative discrepancies between the initial response and the number of remaining medications. Based on this, it automatically generates questions to reconstruct the medication adherence status from the physical context. The owner's responses are obtained as answers to these dynamically generated questions.

[0126] 6. Evaluation of clinical pathways The clinical pathway determination unit 50 calculates determination indices (P1, P2, P3) based on a comparison between clinical parameters obtained from clinical information and expected values ​​corresponding to evaluation items. Next, it integrates these calculated indices to derive a deviation degree D, which indicates the degree of deviation from the pathway for the entire case. The clinical pathway determination unit 50 finally determines the fit to the clinical pathway model by comparing the derived deviation degree D with a predetermined threshold. The following shows an example of the calculation of determination indices in this embodiment.

[0127] <Calculation of the first indicator P1> This quantifies the medical risk of acute exacerbation of a pre-existing condition due to improper medication administration. For example, if diuretics or cardiac stimulants are prescribed for heart disease, missing a few days of medication could lead to a fatal situation (such as pulmonary edema). Therefore, if the inconsistencies described below are detected, a significant bonus will be added to P1.

[0128] <Calculation of the second indicator P2> The discrepancy between the expected number of remaining medications (theoretical value) and the reported number of remaining medications (measured value) is calculated. Normally, assuming one tablet is taken "every day" for 14 days, if the total prescribed number of tablets is 14, the number of remaining tablets should be "0". However, if the reported value is "10 tablets", P2 is calculated as an outlier (unexpected situation) from the expected number of remaining medications.

[0129] <Calculation of the third indicator P3> This method evaluates ambiguity and logical inconsistencies in the definition of information. It detects inconsistencies between the owner's claim that "the medication is given every day without fail" and the arithmetic fact that "10 tablets remain after 14 days" (Level 3: Quantitative Constraint Inconsistency), resulting in a high P3 score. The lack of transparency due to the unknown name of the medication also amplifies P3.

[0130] 7. Scoring The deviation degree D is calculated by multiplying each index (P1, P2, P3) calculated by the clinical pathway determination unit 50 by weight coefficients (w1, w2, w3) defined for each evaluation category. In the "medication management" category of this embodiment, emphasis is placed on verifying the logical consistency between the subjective declaration of the owner and the objective number of remaining medications. Therefore, the following weighting is used, with a high weight given to the third index P3, which evaluates uncertainty and missing information. • Weight of the first index P1 (w1): 0.20 • Weight of the second indicator P2 (w2): 0.30 • Weight of the third indicator P3 (w3): ​​0.50

[0131] And, as in Example 1, the deviation degree D is calculated by the following formula. D = (P1 × w1) + (P2 × w2) + (P3 × w3)

[0132] (Case A: When the degree of deviation is less than the threshold T1) This scenario assumes that the pet owner provides the following additional answers to the questions generated by the additional questionnaire generation unit 80, and that quantitative consistency is proven. A4: "Initially, I was prescribed a 24-day supply (24 tablets)." A5: "The last time I gave it to him was this morning." A6: "They don't vomit or show any signs of discomfort after taking the medicine."

[0133] In this case, the total number of tablets prescribed (24) minus the number of days elapsed (14) equals the expected remaining number (10 tablets), which perfectly matches the actual number of remaining tablets. The clinical pathway evaluation unit determines that some of the initial information uncertainties have been resolved (the drug name remains unknown) and sets the third indicator of uncertainty, P3, to 0.10. Furthermore, because the number of remaining tablets is consistent, the first indicator of risk, P1, is updated to 0.00. In addition, since these events match the predictions based on the clinical pathway model, the second indicator of unexpectedness, P2, is also set to 0.00.

[0134] Based on the above results, the calculated deviation degree D is given by the following formula. D=(0.00×0.20)+(0.00×0.30)+(0.10×0.50)

[0135] The calculated deviation D (=0.05) is below the threshold T1 (e.g., 0.40) which indicates the normal range, so the system determines the fit to be "good (normal course)". In response, the output unit 90 provides the physician with appropriate advice, ensuring the reliability of the information, such as, "The urgency is low. We recommend continuing with a standard prescription based on the chief complaint."

[0136] (Case B: When the degree of deviation exceeds the threshold T2) On the other hand, let's consider the case where, in response to additional questions, responses suggesting a worsening of the situation are obtained, as follows: A4' "Recently, my child has been refusing to take medicine, or has been spitting it out after I've given it to them." A5' "I feel like he's a little less energetic since I gave him the medicine." A6' "I don't really know the name of the prescribed medication or how to administer it." A7' "Someone in the family is giving it to him, but they're not properly managing it."

[0137] These additional responses are significant inconsistencies with the initial response of "daily administration." The clinical pathway evaluation unit 50 determines that the owner's perception of "administering daily" is likely to be a misconception due to concealing difficulties in administration or discontinuation due to side effects, or due to inadequate management. Reflecting the inconsistency with the initial information and the absence of essential information, it calculates the third indicator of uncertainty, P3, to the maximum level of 0.90.

[0138] Furthermore, the clinical pathway evaluation unit 50, while recognizing the risk of loss of therapeutic effect and the occurrence of unknown side effects, determined that it could not definitively conclude that the underlying disease itself was in an urgently fatal state at present, and set the first indicator (P1) to approximately 0.10. This is because the system identified that the most pressing issue in this case was "inadequate medication adherence" rather than rapid disease progression, and it reflects the long-term, potential risk of treatment failure due to poor medication adherence. In addition, the fact that 10 tablets remained after a 14-day prescription represents the greatest deviation from the expected clinical pathway, so the second indicator (P2), which indicates the degree of unexpectedness, is calculated to be 1.00.

[0139] As a result, the calculated deviation degree D is as follows: D=(0.20×0.10)+(0.30×1.00)+(0.50×0.90)

[0140] The calculated deviation degree D (=0.77) exceeds the threshold T2 (e.g., 0.7) indicating a serious abnormality, so the system determines it to be a "serious deviation." The output unit 90 immediately outputs a warning to the physician's terminal, displaying the following advice as a priority check: "Urgent: Serious inconsistency in medication management. Based on the discrepancy with the number of remaining prescriptions, approximately 70% of medications may not be administered. There is an unknown drug name and indication of adverse events. Avoid unnecessary increases in medication or additional tests, and prioritize improving medication adherence and providing guidance." This prevents confusion in medical practice and supports the optimization of clinical decision-making.

[0141] It should be noted that the calculation formulas for the judgment indices (P1-P3), the weighting coefficients (w1-w3) for each evaluation category, and the threshold values ​​(T1, T2) used in each of the above-described embodiments are merely examples to illustrate the technical concept of the present invention. These are dynamically set, adjusted, or optimized within the scope of the present invention, depending on the applicable animal species, target disease, clinical guidelines for each facility, or the condition of individual patients. Furthermore, the clinical pathway model itself may be configured to be updated as appropriate based on the clinical data accumulated at each facility.

[0142] As described above, the medical support device and medical support program according to the present invention can identify deficiencies or inconsistencies in information based on a comparison between the patient's clinical information and the expected value according to the clinical pathway, conduct additional interviews, and, based on the uncertainty and medical validity of the information, present physicians with priority items to confirm and advice on treatment. [Explanation of Symbols]

[0143] 1 Medical support equipment 10 Clinical parameter generation unit 20 Clinical Management Data Storage Unit 30 Clinical Pathway Specific Section 40. Evaluation Item Identification Section 50 Clinical Pathway Determination Unit 60 Medical Information Storage Department 70 Medical assistance information generation department 80 Additional interview generation section 90 Output section 2. First Terminal 3. Second Terminal 4. Communication Network

Claims

1. A clinical parameter generation unit generates clinical parameters, including the patient's attributes and chief complaint, from clinical information obtained from past medical records or acquired interview information concerning the patient. A clinical management data storage unit holds clinical management data that defines a clinical pathway model defining the expected medical state to be applied to the patient, one or more evaluation categories indicating clinical evaluation axes associated with the clinical pathway model, and the interrelationships between one or more evaluation items necessary to determine the evaluation category. A clinical path identification unit identifies the clinical path model based on the clinical parameters generated by referring to the clinical management data, An evaluation item identification unit that identifies one or more evaluation categories associated with the identified clinical pathway model and one or more evaluation items necessary to determine the evaluation category, The system includes a clinical path determination unit that determines the suitability of the clinical information to the clinical path model based on a comparison between the clinical parameters and the expected values ​​corresponding to the evaluation items, The clinical pathway determination unit calculates at least one of the following as the degree to which the clinical information is suited to the clinical pathway model: a first index indicating the degree to which the clinical information deviates from the expected value; a second index indicating the degree to which the clinical information deviates from the prediction based on the clinical pathway model; and a third index indicating the degree of deficiency of the clinical information or the degree of inconsistency among multiple pieces of information included in the clinical information. Medical support equipment.

2. The aforementioned clinical pathway determination unit is In the clinical management data, the fitting status is calculated by multiplying the weighting coefficient defined for each clinical pathway model or evaluation category by at least one of the first indicator, the second indicator, and the third indicator. The medical support device according to claim 1.

3. The system further includes an additional questionnaire generation unit that generates questions for obtaining additional questionnaire information based on the suitability status determined by the clinical pathway determination unit. A medical support device according to claim 1 or claim 2.

4. The aforementioned additional questionnaire generation unit, If the first indicator exceeds a predetermined threshold, questions are generated to identify the urgency or severity. If the second indicator exceeds a predetermined threshold, questions are generated to examine the deviation from the state predicted based on the clinical path model. If the third indicator exceeds a predetermined threshold, questions are generated to supplement the missing information or to resolve any inconsistencies in the clinical information. The medical support device according to claim 3.

5. The system further includes an output unit that outputs the determination results from the clinical pathway determination unit to a terminal that can be viewed by a physician. The output unit presents the clinical information and the determination result from the clinical pathway determination unit in parallel, and further presents advice information regarding the interpretation of the clinical information or confirmation items for the patient. A medical support device according to claim 1 or claim 2.

6. It further comprises a medical information storage unit in which a medical profile is defined, which includes one or more clinical patterns that suggest a specific disease or condition. The system includes a clinical support information generation unit that generates clinical support information to identify candidate conditions or diseases in the patient by calculating the degree of fit between the medical profile and at least one of the first, second, and third indicators calculated by the clinical pathway determination unit. A medical support device according to claim 1 or claim 2.

7. The clinical pathway identification unit dynamically generates a temporary clinical pathway model by combining predefined clinical components according to the satisfaction status in the identification of the clinical pathway model based on the clinical parameters. A medical support device according to claim 1 or claim 2.

8. A clinical parameter generation step that generates clinical parameters, including the patient's attributes and chief complaint, from clinical information obtained from past medical records or acquired interview information concerning the patient, A clinical path identification step involves identifying the clinical path model based on the generated clinical parameters, by referring to clinical management data that defines a clinical path model defining the expected medical state to be applied to the patient, one or more evaluation categories indicating clinical evaluation axes associated with the clinical path model, and one or more evaluation items necessary to determine the evaluation category. An evaluation item identification step that identifies one or more evaluation categories associated with the identified clinical pathway model and one or more evaluation items necessary to determine the evaluation category, A clinical support program for causing a computer to perform a clinical path determination step, which determines the suitability of the clinical information to the clinical path model based on a comparison of the clinical parameters and the expected values ​​corresponding to the evaluation items, The clinical pathway determination step includes a process of calculating at least one of the following as the degree to which the clinical information fits the clinical pathway model: a first indicator indicating the degree to which the clinical information deviates from the expected value; a second indicator indicating the degree to which the clinical information deviates from the prediction based on the clinical pathway model; and a third indicator indicating the degree of deficiency of the clinical information or the degree of inconsistency among multiple pieces of information included in the clinical information. Medical support program.

9. The aforementioned clinical pathway determination step is: The fitting status is calculated by multiplying the weighting coefficient defined for each clinical pathway model or evaluation category by at least one of the first index, the second index, and the third index, including the fitting status calculation step. The medical support program according to claim 8.

10. The system further includes an additional questionnaire generation step, which constitutes questions for obtaining additional medical information based on the suitability status determined by the clinical pathway determination step. A medical support program according to claim 8 or claim 9.

11. The aforementioned additional questionnaire generation step is: If the first indicator exceeds a predetermined threshold, the process includes generating questions to identify the urgency or severity. If the second indicator exceeds a predetermined threshold, the process includes generating questions to examine the deviation from the state predicted based on the clinical path model. If the third indicator exceeds a predetermined threshold, the process includes generating questions to supplement missing information or questions to resolve inconsistencies in the clinical information. The medical support program according to claim 10.

12. The system further includes an output step that outputs the determination result from the clinical pathway determination step to a terminal that can be viewed by a physician. The output step includes presenting the clinical information and the result of the suitability determination in parallel, and further including the process of presenting advice information regarding the interpretation of the clinical information or priority confirmation items for the patient. A medical support program according to claim 8 or claim 9.

13. The system further includes a step for generating clinical support information to identify the patient's condition or potential disease, The clinical support information generation step includes a process for calculating the degree of fit between at least one of the first indicator, the second indicator, and the third indicator calculated by the clinical path determination step and a medical profile that includes one or more clinical patterns suggesting a specific disease or condition. A medical support program according to claim 8 or claim 9.

14. The clinical pathway identification step includes a process of dynamically generating a temporary clinical pathway model by combining predefined clinical components according to the satisfaction status in the identification of the clinical pathway model based on the clinical parameters. A medical support program according to claim 8 or claim 9.

Images