Ai-based medical consent data analysis and risk prediction system

An AI-based system analyzes patient input behavior to predict medical risks, enhancing the reliability of consent procedures and reducing disputes by objectively evaluating patient understanding.

KR103046922B1Active Publication Date: 2026-11-11JOONGANG IT CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
KR1020260014790
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2026-01-26
Publication Date
2026-11-11
Estimated Expiration
2046-01-26

AI Technical Summary

Technical Problem

Conventional electronic medical consent systems fail to adequately analyze patient input behavior during the consent process, leading to subjective evaluations of patient understanding and increased risk of medical disputes, with limited utilization of generated data for risk management.

Method used

An AI-based system that collects and analyzes patient input behavior data, generating a feature vector to predict medical risks through machine learning, providing notifications and log data for medical staff to manage disputes.

Benefits of technology

Objective and quantitative risk assessment reduces the likelihood of medical disputes by identifying high-risk patients, optimizing resource allocation, and improving the reliability of consent procedures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112026010433069-PAT00002_ABST
    Figure 112026010433069-PAT00002_ABST
Patent Text Reader

Abstract

A system for predicting a patient's medical risk by analyzing data of a medical consent form provided as an electronic document based on artificial intelligence (AI) is disclosed. The AI-based medical consent form data analysis and risk prediction system according to the present invention includes a data collection unit that collects patient input behavior data generated during the process of creating a medical consent form, a feature vector generation unit that generates a feature vector based on the input behavior data, and an AI risk prediction unit that predicts the patient's medical risk from the feature vector.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to a system for predicting a patient's medical risk by analyzing data from a medical consent form provided as an electronic document based on artificial intelligence (AI). Background Technology

[0002] Generally, before performing medical procedures such as surgery, interventions, or examinations, medical institutions explain the details, risks, and side effects of the procedure to patients and require them to complete a medical consent form to secure their agreement. While traditionally paper-based medical consent forms were used, the use of electronic consent forms has been increasing recently.

[0003] The electronic document-based medical consent system has the advantage of increasing document management efficiency and simplifying the consent process by allowing patients to review the contents of medical consent forms provided via terminals and sign them electronically. In addition, the electronic document-based medical consent system contributes to improving the operational efficiency of medical institutions by enabling the computerized management of basic history, such as the time of consent creation and whether a signature was made.

[0004] However, conventional electronic document-based medical consent systems have limitations in that they fail to sufficiently analyze or utilize information regarding how well patients actually understood the contents of the consent form or what behaviors they exhibited during the completion process. For example, even when a patient repeatedly modifies specific items, takes an excessive amount of time to input data, or undergoes abrupt changes in input behavior immediately before signing, this behavioral information is often simply stored as a log or is not utilized at all.

[0005] Consequently, medical professionals are compelled to rely on subjective judgment or experience to assess the patient's completion or understanding of the medical consent form, and there is a problem in that it is difficult to objectively evaluate in advance the likelihood of medical disputes or the risk of subsequent objections for each individual patient. In particular, since it is difficult to prove the appropriateness of the consent procedure or the patient's understanding after a medical dispute has occurred, it can place a significant burden on medical institutions and medical staff.

[0006] Furthermore, in conventional technology, even when data related to medical consent forms is linked with electronic medical record (EMR) systems, it often remains at the level of simple document storage or verification; consequently, the various data generated during the consent process are not being fully utilized for medical risk management.

[0007] Related prior art documents include Korean Registered Patent Publication No. 10-1257286 (Title of Invention: EMR-based Smart Electronic Medical Consent System, Date of Publication: April 17, 2013), Korean Published Patent Publication No. 10-2025-0125863 (Title of Invention: Device and Method for Providing Electronic Document Consent Procedure Service, Date of Publication: August 22, 2025), Korean Published Patent Publication No. 10-2025-0062614 (Title of Invention: Method and System for Patient Authentication through Generation of Electronic Medical Consent Image, Date of Publication: May 8, 2025), and Korean Registered Patent Publication No. 10-2617442 (Title of Invention: Method and System for Managing Electronic Medical Consent, Date of Publication: December 27, 2023). Prior art literature

[0008] Korean Registered Patent Publication No. 10-1257286 (April 17, 2013) Korean Published Patent Publication No. 10-2025-0125863 (August 22, 2025) Korean Published Patent Publication No. 10-2025-0062614 (May 8, 2025) Korean Registered Patent Publication No. 10-2617442 (December 27, 2023) The problem to be solved

[0009] The objective of the present invention is to provide an AI-based medical consent data analysis and risk prediction system that can prevent medical disputes and improve the reliability of the medical consent procedure by effectively collecting and analyzing patient input behavior data generated during the process of drafting medical consent forms, and predicting the patient's medical risks based on this data and providing this information to medical staff. means of solving the problem

[0010] The above objective is achieved by an AI-based medical consent form data analysis and risk prediction system, characterized in that, according to one embodiment of the present invention, the system comprises a data collection unit that collects patient input behavior data generated during the process of creating a medical consent form, a feature vector generation unit that generates a feature vector based on the input behavior data, and an AI risk prediction unit that predicts the patient's medical risk from the feature vector.

[0011] Preferably, the input behavior data may include at least one of the input time per medical consent form item, the number of revisions to the medical consent form, the time taken to sign, the scroll movement pattern, and the comprehension evaluation result.

[0012] Preferably, the feature vector generation unit analyzes the input behavior data to derive a plurality of characteristic values ​​representing the temporal distribution, frequency of occurrence, or pattern of change of the input behavior data, and can generate the feature vector using the plurality of characteristic values.

[0013] Preferably, the feature vector generation unit can generate the feature vector to reflect the change pattern over time of each of the plurality of feature values ​​by considering the time sequence of the input behavior data throughout the medical consent form creation process.

[0014] Preferably, the feature vector generation unit may assign different weights to each of the plurality of feature values ​​based on at least one of the magnitude of change over time, the frequency of change, or the time at which the change occurs, and generate the feature vector using the feature values ​​with the weights reflected.

[0015] Preferably, the AI ​​risk prediction unit calculates a medical risk score of the patient based on the feature vector and classifies the patient into one of a plurality of risk grades based on the medical risk score.

[0016] Preferably, the plurality of risk grades may include a general group, a caution group, and a risk group.

[0017] Preferably, the AI-based medical consent data analysis and risk prediction system may further include a notification unit that provides a notification to medical staff when the medical risk score exceeds a preset threshold.

[0018] Preferably, the AI-based medical consent data analysis and risk prediction system further includes a log recorder that stores risk log data including patient identification information, medical consent identification information, medical risk score, and time of occurrence, and said risk log data can be provided in conjunction with a medical institution system. Effects of the invention

[0019] According to one embodiment of the present invention, a feature vector is generated based on input behavior data, such as input time per item of a medical consent form, number of revisions, time required for signing, scroll movement patterns, and comprehension evaluation results, and by analyzing this using AI, medical risk for each patient can be predicted in the form of a score or grade. Accordingly, unlike conventional methods that relied on the subjective judgment or experience of medical staff, medical risk for each patient can be evaluated objectively and quantitatively.

[0020] According to one embodiment of the present invention, when a medical risk score exceeds a preset threshold, a notification is provided to medical staff, allowing them to identify patients with a high likelihood of medical disputes in advance and perform procedures such as providing additional explanations, reconfirming, or supplementing. Through this, the likelihood of medical disputes can be lowered in advance, and the legal and administrative burden on medical institutions can be reduced.

[0021] According to one embodiment of the present invention, by analyzing and recording input behavior data throughout the entire process of drafting a medical consent form, evidence can be provided to objectively verify how a patient reviewed and signed the contents of the medical consent form. Accordingly, the appropriateness and reliability of the consent procedure are improved, and clear judgment data can be provided even in the event of a subsequent dispute.

[0022] According to one embodiment of the present invention, by generating a feature vector that reflects the temporal order, pattern of change, magnitude and frequency of change of input behavioral data, rather than merely aggregating input results, it is possible to analyze patients' medical consent signing behaviors in a time-series and contextual manner, and to enable advanced medical risk prediction that is distinct from simple statistical processing methods.

[0023] According to one embodiment of the present invention, since only patients with high medical risk can be selected and notified to medical staff, there is no need to perform the same level of additional explanation or verification procedures for all patients. Accordingly, medical staff can utilize limited time and resources more efficiently and improve the overall quality of medical services. Brief explanation of the drawing

[0024] FIG. 1 is a schematic diagram illustrating the overall configuration of an AI-based medical consent form data analysis and risk prediction system according to one embodiment of the present invention. Figure 2 is a block diagram showing the detailed configuration of the server of Figure 1. FIG. 3 is a flowchart illustrating the overall flow of an AI-based medical consent form data analysis and risk prediction method according to one embodiment of the present invention. FIG. 4 is a flowchart illustrated to explain in detail the process of generating a feature vector based on input behavior data collected according to one embodiment of the present invention. Specific details for implementing the invention

[0025] In order to fully understand the structure and operational advantages of the present invention and the objectives achieved through the implementation of the present invention, reference should be made to the accompanying drawings illustrating preferred embodiments of the present invention and the contents described therein.

[0026] Hereinafter, the present invention will be described in detail by explaining preferred embodiments of the invention with reference to the attached drawings. However, in describing the present invention, descriptions of already known configurations or functions will be omitted in order to clarify the gist of the present invention.

[0027] FIG. 1 is a schematic diagram illustrating the overall configuration of an AI-based medical consent form data analysis and risk prediction system according to one embodiment of the present invention.

[0028] Referring to FIG. 1, an AI-based medical consent data analysis and risk prediction system (100) according to one embodiment of the present invention may include a server (110), a medical institution system (120), an AI engine (130), and a patient terminal (140).

[0029] A patient terminal (140) is a terminal through which a patient views, completes, and signs a medical consent form provided by a medical institution, and may be, for example, a smartphone, tablet, kiosk, or other electronic terminal. During the process of the patient completing the medical consent form, input behavior data may be generated, such as the time taken to enter each item of the medical consent form, the act of modifying the input content, the time taken to sign, screen scrolling or page movement patterns, and the results of an understanding evaluation. This input behavior data is data that reflects the patient's medical consent form completion behavior and subsequently becomes a subject of analysis for medical risk prediction.

[0030] The server (110) may be configured to communicate with the patient terminal (140) and the medical institution system (120) to collect input behavior data generated from the patient terminal (140) and to perform medical risk prediction in conjunction with the AI ​​engine (130) described later based on the collected data. The server (110) may function as a central node for data collection, management, and AI integration in the system (100) of the present invention, and the specific internal configuration or detailed processing method of the server (110) may be explained in more detail later with reference to the drawings.

[0031] The medical institution system (120) is an internal system of a hospital, clinic, or medical service provider and may include an electronic medical record (EMR) system, an order of service (OCS), a picture archiving and transmission system (PACS), etc. The medical institution system (120) may be linked to the operating environment of the medical consent form, or linked to information related to patients and medical consent forms, and information related to medical risks calculated by the system (100) of the present invention.

[0032] The AI ​​engine (130) is linked with the server (110) and can perform the function of predicting medical risk for each patient using a feature vector generated from the patient's input behavior data as input. For example, the AI ​​engine (130) can calculate the patient's medical risk score based on the feature vector and, if necessary, classify the patient into one of a plurality of risk grades. This AI engine (130) may include a machine learning or deep learning-based model, but is not limited thereto.

[0033] In this way, an AI-based medical consent form data analysis and risk prediction system (100) according to one embodiment of the present invention collects input behavior data of the medical consent form creation process occurring at a patient terminal (140) through a server (110), generates a feature vector based on the input behavior data, and then predicts medical risks for each patient using an AI engine (130). Through this, medical staff can recognize medical risks based on the patient's medical consent form creation behavior in advance and utilize this for the prevention or management of medical disputes.

[0034] FIG. 2 is a block diagram showing the detailed configuration of the server (110) of FIG. 1.

[0035] Referring to FIGS. 1 and 2, the server (110) may be configured to include a data collection unit (210), a feature vector generation unit (220), an AI risk prediction unit (230), an alert unit (240), a log recording unit (250), and a control unit (260). Each component included in the server (110) may be implemented in hardware, software, or a combination thereof, and although they are shown functionally separated, some components may be integrated or separated depending on the implementation form.

[0036] The data collection unit (210) may be configured to collect patient input behavior data generated during the process of creating a medical consent form. Here, input behavior data refers to data that reflects temporal, behavioral, and reactive characteristics that appear during the series of processes in which a patient views, creates, modifies, and signs a medical consent form.

[0037] For example, the data collection unit (210) can collect at least one of the input time per medical consent form item occurring at the patient terminal (140), the input delay time for each item, the number of medical consent form modifications or the time of modification occurrence, the total time required until signature, the screen scroll movement speed and direction, the page transition frequency, and the comprehension evaluation result.

[0038] This input behavior data can reflect how carefully patients reviewed the contents of the medical consent form, whether they hesitated or repeatedly modified specific items, or whether they underwent a sufficient verification process prior to signing. In other words, input behavior data quantitatively reflects patients' behavior in completing medical consent forms and can be utilized as foundational data for predicting medical risks.

[0039] According to one embodiment, the data collection unit (210) may collect input behavior data in the form of a log that includes the time flow during the input process, rather than simply collecting input results. For example, by recording the time when a specific medical consent form item was first viewed, the time when input began, the time when modification occurred, and the time when moving to the next item along with timestamps, the process of a patient completing a medical consent form can be constructed as time-series data.

[0040] According to another embodiment, the data collection unit (210) can collect input behavior data in units of events. For example, whenever an event occurs, such as an item click, input start, input completion, modification occurrence, or signature button selection, the type of the event, the time of occurrence, and related item information can be recorded. This event-based data can be used to analyze the patient's behavioral patterns more precisely when generating feature vectors later.

[0041] According to another embodiment, the data collection unit (210) may additionally collect the time taken for the understanding evaluation or the response patterns for each evaluation item, along with the results of the understanding evaluation. For example, the response delay time for the understanding evaluation items or whether an incorrect answer occurred can be used as an indirect indicator of whether the patient has fully understood the contents of the medical consent form.

[0042] The data collection unit (210) may operate by receiving input behavior data directly from the patient terminal (140), or by collecting input behavior data by monitoring input events occurring through the medical consent form creation interface on the server (110). Additionally, depending on the implementation type, input behavior data may be collected in real time, or collected collectively after the medical consent form is completed.

[0043] In this way, the data collection unit (210) can provide basic data that quantitatively and objectively reflects the patient's medical consent form completion behavior by widely collecting various input behavior data generated during the medical consent form completion process. This input behavior data is subsequently analyzed and processed by the feature vector generation unit (220) to generate feature vectors, and furthermore, can be used as core input data for predicting medical risks for each patient by the AI ​​risk prediction unit (230).

[0044] The feature vector generation unit (220) may be configured to generate a feature vector that quantitatively represents the patient's medical consent form completion behavior by analyzing the patient's input behavior data collected by the data collection unit (210). Here, the feature vector is a data structure that represents the patient's input behavior in multiple dimensions, and can subsequently be input into the AI ​​list prediction unit (230) to be used for medical risk prediction.

[0045] According to one embodiment, the feature vector generation unit (220) can derive a plurality of characteristic values ​​representing the temporal distribution, frequency of occurrence, or pattern of change of the input from the input behavior data. For example, values ​​such as the average input time for a specific medical consent form item, deviation of the input time, number of modifications, interval between modifications, or scroll movement frequency may be derived as characteristic values. These characteristic values ​​may be the result of individually quantifying the behavioral characteristics that appear during the process of a patient writing a medical consent form.

[0046] In addition, the feature vector generation unit (220) can generate a feature vector by considering the time sequence of input behavior data throughout the entire process of writing the medical consent form, rather than merely calculating individual feature values. For example, changes over time, such as changes in input patterns at the beginning and end of writing the medical consent form, whether there is a rapid increase in the frequency of modifications after a specific point in time, or input delay phenomena at the stage immediately before signing, can be reflected in the feature vector. Through this, the patient's behavior of writing the medical consent form can be expressed as a time-series behavior pattern rather than a value at a single point in time.

[0047] According to another embodiment, the feature vector generation unit (220) may generate a feature vector using the feature values ​​with weights after assigning different weights to each of the derived multiple feature values. For example, relatively high weights may be assigned to feature values ​​with large magnitudes of change, feature values ​​where changes occur frequently, or feature values ​​where a sudden change occurs at a specific point in time. Such weight assignment may be intended to emphasize characteristics that are more closely associated with medical risks among the patient's medical consent form signing behaviors.

[0048] For example, repetitive modification actions or input delays occurring immediately before signing may have significant meaning in predicting medical risks, so higher weights may be assigned to characteristic values ​​associated with that point in time. On the other hand, temporary input delays occurring during the initial stage of drafting the medical consent form may be assigned relatively lower weights. In this way, the feature vector generation unit (220) can generate a feature vector by considering both the temporal context and the behavioral importance of the characteristic values.

[0049] The feature vector generation unit (220) may include a rule-based method, a statistical processing method, or preprocessing logic designed to be suitable for AI model training, depending on the implementation form, but is not limited thereto. In addition, the number of dimensions, components, and weighting method of the feature vector may be set in various ways depending on the operating environment of the medical institution system (120) or the characteristics of the medical procedure to which it is applied.

[0050] In this way, the feature vector generated by the feature vector generation unit (220) is not merely a set of simple input data, but a behavior-based feature representation that comprehensively reflects the patient's medical consent form completion behavior, thereby contributing to the AI ​​risk prediction unit (230) predicting medical risks for each patient more accurately and reliably. Through this, the present invention enables advanced behavior analysis-based medical risk prediction that is distinguished from simple statistical aggregation methods.

[0051] The AI ​​risk prediction unit (230) may be configured to predict a patient's medical risk by using a feature vector generated by the feature vector generation unit (220) as input. Here, medical risk may be a concept that includes the possibility of a medical dispute arising in light of the patient's medical consent form completion behavior, the possibility of the patient's lack of understanding, or the possibility of raising an objection to medical treatment afterward.

[0052] According to one embodiment, the AI ​​risk prediction unit (230) can calculate the patient's medical risk as a quantified medical risk score based on a feature vector. For example, the medical risk score may be expressed as a continuous value within a predefined range, and a higher value may indicate that there is a relatively high medical risk for the patient. This medical risk score may be the result of comprehensively reflecting the patient's overall behavior regarding the signing of medical consent forms.

[0053] Additionally, the AI ​​risk prediction unit (230) can classify the patient into one of a plurality of risk grades based on the calculated medical risk score. For example, the patient may be classified into a risk grade such as a general group, a caution group, or a risk group depending on whether the medical risk score falls within a specific range. This risk grade classification can be used by medical staff to intuitively understand the status of the patient's medical consent form and to decide whether to provide additional explanations or perform verification procedures.

[0054] According to another embodiment, the AI ​​risk prediction unit (230) may derive not only the medical risk score but also some elements of key characteristic values ​​or feature vectors that influenced the calculation of the risk score. For example, if repeated modifications to specific medical consent form items or input delays immediately before signing contributed to an increase in the medical risk score, the relevant factors may be internally identified and subsequently utilized in the process of providing notifications or logging. This can improve the interpretability of the medical risk prediction results.

[0055] The AI ​​risk prediction unit (230) may include a machine learning or deep learning-based artificial intelligence model suitable for time series data processing in conjunction with the AI ​​engine (130) to learn temporal change characteristics included in the patient's input behavior data or feature vector.

[0056] For example, the AI ​​engine (130) may include a Recurrent Neural Network (RNN), a Long Short-Term Memory (LSTM) based model, or a similar neural network model capable of time series learning configured to reflect the time sequence, magnitude of change, and timing of change of input behavioral data.

[0057] Additionally, the AI ​​engine (130) may include a multi-layer perceptron (MLP), a tree-based learning model, or a combination thereof that takes a feature vector as input and calculates a medical risk score.

[0058] According to one embodiment, the AI ​​engine (130) illustrated in FIG. 1 may be configured to perform the function of the AI ​​risk prediction unit (230) by being linked with the server (110), or the AI ​​risk prediction unit (230) may call the AI ​​engine (130) to perform medical risk prediction processing. That is, the AI ​​risk prediction unit (230) may be logically or physically connected to the AI ​​engine (130) as a functional component within the server (110) to transmit feature vectors to the AI ​​engine (130) and receive medical risk prediction results from the AI ​​engine (130).

[0059] Additionally, the prediction model included in the AI ​​risk prediction unit (230) or executed in the AI ​​engine (130) may be provided in a trained state using pre-built training data, or it may be continuously updated or retrained by reflecting input behavior data and risk log data accumulated in the operating environment or actual usage process of the medical institution system (120). For example, to reflect differences in patient characteristics or medical consent form writing patterns by medical institution, the AI ​​engine (130) may adjust the parameters of the prediction model or update the model based on data linked with the medical institution system (120).

[0060] According to one embodiment, the AI ​​engine (130) may include a plurality of prediction models, and the AI ​​risk prediction unit (230) may selectively apply at least one of the plurality of prediction models according to the type of medical act, the type of medical consent form, or the policy of the medical institution. Through this, even with the same input behavioral data, more suitable medical risk prediction results can be derived depending on the medical environment or characteristics of the medical consent form to which it is applied.

[0061] In this way, the AI ​​risk prediction unit (230) performs medical risk prediction based on feature vectors through linkage with the AI ​​engine (130), and enables the learning, updating, and selective application of the prediction model, thereby providing a flexible and scalable AI-based medical risk prediction structure suitable for the actual operating environment of a medical institution.

[0062] The notification unit (240) may be configured to provide a notification to medical staff when the medical risk score calculated by the AI ​​risk prediction unit (230) exceeds a preset threshold. Here, the threshold may be preset according to the medical institution's operating policy, the risk level of the medical procedure, or the type of medical consent form, or may be changed as needed.

[0063] According to one embodiment, the notification unit (240) may generate notification information for a patient whose medical risk score exceeds a threshold, including at least one of the patient's identification information, medical consent identification information, medical risk score, or risk grade. This notification information may be provided in a summarized form so that medical staff can quickly grasp the patient's condition. In particular, the medical risk score or risk grade may be provided in a visual form including a graph, color display, or icon, thereby enabling medical staff to intuitively grasp the patient's medical risk level and assist in selecting follow-up actions such as providing additional explanations, requesting re-consent, or adjusting the treatment procedure.

[0064] For example, the notification unit (240) may provide medical risk-related information to medical staff using mobile terminals in the form of mobile push notifications, and in this case, the notification may include messages such as "detection of high-risk medical consent form completion" or "occurrence of patient requiring attention." As another example, the notification unit (240) may visually display the patient at risk on the dashboard screen of the medical institution system (120), or provide different colors, icons, or highlights according to the risk level.

[0065] According to another embodiment, the notification unit (240) may apply different notification methods depending on the level of medical risk. For example, for patients in the caution group, only a dashboard display may be provided, and for patients in the risk group, a pop-up message may be additionally provided along with a mobile push notification. This allows for the reduction of unnecessary notification burdens on medical staff while enabling rapid recognition of critical risk situations.

[0066] The notification unit (240) can provide information on when the medical risk score exceeds a threshold. For example, the notification may include information on which stage of the medical consent form completion process the risk increased, or whether the risk was detected at the stage immediately before signing. Through this, the medical staff can determine when to provide additional explanations or perform a verification procedure.

[0067] Additionally, the notification unit (240) may receive or record the results of the medical staff's response after the notification is provided. For example, information regarding whether the medical staff provided additional explanations to the patient, requested re-consent, or performed separate measures may be linked to be stored in the log record unit (250). This information may be used as reference material in the event of a future medical dispute.

[0068] In this way, the notification unit (240) can support real-time or near-real-time monitoring of the patient's medical consent form completion status by providing medical risk-related information to medical staff in a timely manner when the medical risk score exceeds a threshold. Through this, medical staff can perform appropriate responses tailored to the situation of each patient.

[0069] The log recorder (250) may be configured to store risk log data including patient identification information, medical consent form identification information, medical risk score, time of risk occurrence and related processing results. Here, the risk log data is data that systematically records the process of drafting the medical consent form and the resulting medical risk prediction results, and can be used as evidence for preventing medical disputes or for subsequent review.

[0070] According to one embodiment, the log recording unit (250) may record the time at which the medical risk score was calculated, whether the score exceeded a threshold, the risk grade to which the patient was classified, and whether the notification unit (240) provided a notification regarding the risk. Through this, the log can be configured so that the medical risk prediction results and subsequent actions taken therefrom can be tracked in chronological order.

[0071] Additionally, the log record (250) can store information including the results of the medical staff's response. For example, information regarding whether the medical staff provided additional explanations to the patient, requested re-consent, or performed a separate verification procedure can be stored linked to the risk log data. This information on the results of the response can be used as data to clearly show what measures were actually taken after the medical risk prediction.

[0072] According to another embodiment, the log recorder (250) may also store some information of key characteristic values ​​or feature vectors that are directly used for medical risk prediction among the input behavioral data. For example, factors that have influenced the increase in the medical risk score, such as repetitive modification behavior, input delay in a specific interval, or abrupt behavioral changes immediately before the signature stage, may be included in the log. Through this, the basis for the medical risk prediction result can be presented more clearly.

[0073] The log record book (250) can be linked with a medical institution system (120), namely an electronic medical record system (EMR), an order of services (OCS), or a hospital information system (HIS) to provide risk log data. For example, it can be linked so that the patient's medical consent history and medical risk prediction results can be viewed together on the patient's electronic medical record screen. Through this, medical staff can utilize the patient's medical risk information more intuitively during the treatment process.

[0074] The log record book (250) can manage the retention period, access rights, or viewing history of risk log data. For example, risk log data can be stored for a certain period or longer in preparation for potential medical disputes, or access rights can be granted only to specific users, such as medical staff or administrators. Additionally, the log record book (250) can increase the reliability of medical information management by recording the viewing history of risk log data. Here, the risk log data can be utilized by the AI ​​engine (130) to improve the performance or retrain the medical risk prediction model, thereby enabling risk prediction that continuously reflects the actual operating environment of the medical institution.

[0075] The control unit (260) may be configured to control the operation of the data collection unit (210), the feature vector generation unit (220), the AI ​​risk prediction unit (230), the notification unit (240), and the log recording unit (250), and to manage the data flow between these components. For example, the control unit (260) may comprehensively control the timing of input behavior data collection, the timing of feature vector generation, whether to perform medical risk prediction, the conditions for providing notifications, and the timing of log recording.

[0076] FIG. 3 is a flowchart illustrating the overall flow of an AI-based medical consent form data analysis and risk prediction method according to one embodiment of the present invention.

[0077] Referring to FIGS. 1 and FIGS. 3, in step (310), the server (110) can collect data on the patient's input behavior that occurs during the process of creating a medical consent form. For example, the server (110) can collect data such as the input time per item of the medical consent form, the number of modifications, the time required for signing, scroll movement patterns, and the results of an understanding evaluation that occur on the patient terminal (140).

[0078] Next, in step (320), the server (110) can generate a feature vector based on the collected input behavior data. That is, the server (110) can generate a feature vector by reflecting the temporal distribution, frequency of occurrence, or pattern of change of the collected input behavior data. A more specific operation regarding this will be explained in detail with reference to FIG. 4, which will be described later.

[0079] Next, in step (330), the server (110) can calculate the patient's medical risk score from the generated feature vector. At this time, the server (110) can calculate the patient's risk score through linkage with the AI ​​engine (130).

[0080] Next, in step (340), the server (110) can predict the patient's medical risk based on the calculated medical risk score. That is, the server (110) can classify the patient into one of a plurality of risk grades (general group, caution group, risk group, etc.).

[0081] Next, if the medical risk score exceeds a preset threshold in step (350), the server (110) may provide a notification to the medical staff. The notification may be provided via mobile push notification, dashboard display, or other electronic notification methods.

[0082] Next, in step (360), the server (110) can store risk log data including a medical risk score, risk grade, whether a notification is provided, and the time of occurrence. The stored risk log data can be provided in conjunction with an electronic medical record system.

[0083] FIG. 4 is a flowchart illustrated to explain in detail the process of generating a feature vector based on input behavior data collected according to one embodiment of the present invention.

[0084] Referring to FIGS. 1 and FIGS. 4, in step (410), the server (110) can analyze input behavior data collected during the process of creating a medical consent form. At this time, the server (110) can analyze the time information, occurrence order, and event type of the input behavior data.

[0085] Next, in step (420), the server (110) can derive multiple characteristic values ​​representing the temporal distribution, frequency of occurrence, or pattern of change of the input behavior data based on the analysis results of the input behavior data. For example, the server (110) can derive the average and deviation of the input time, the frequency of modification occurrence, the interval between modifications, and changes in the scroll movement pattern as characteristic values.

[0086] Next, in step (430), the server (110) may assign different weights to each of a plurality of characteristic values ​​based on at least one of the magnitude of the change over time, the frequency of the change, or the time at which the change occurred. For example, the server (110) may assign a relatively high weight to characteristic values ​​corresponding to input delays or repetitive modification actions that occurred in the step immediately preceding the signature.

[0087] Next, in step (440), the server (110) can generate a feature vector using weighted feature values. The generated feature vector can be used as input data for calculating a medical risk score performed in step (330) of FIG. 3.

[0088] As such, according to the feature vector generation process illustrated in Fig. 4, rather than simply aggregating input behavioral data, a behavior-based feature vector can be generated that reflects the chronological order, patterns of change, and the relative importance of feature values. Through this, medical risk prediction that more precisely reflects the patient's medical consent form signing behavior can be performed.

[0089] Although the present invention has been described in detail through preferred embodiments, it is obvious to those skilled in the art that the present invention is not limited to the aforementioned preferred embodiments and can be modified and varied in various ways without departing from the spirit and scope of the invention. Accordingly, such modifications or variations should be deemed to fall within the scope of the claims of the present invention. Explanation of the symbols

[0090] 110: Server 120: Medical institution system 130: AI Engine 140: Patient terminal 210: Data Collection Unit 220: Feature Vector Generation Unit 230: AI Risk Prediction Department 240: Notification Department 250: Logbook 260: Control unit

Claims

Claim 1 An AI-based medical consent form data analysis and risk prediction system comprising: a data collection unit for collecting patient input behavior data generated during the process of creating a medical consent form; a feature vector generation unit for generating a feature vector based on the input behavior data; and an AI risk prediction unit for predicting the patient's medical risk from the feature vector, wherein the feature vector generation unit analyzes the input behavior data to derive a plurality of characteristic values ​​representing the temporal distribution, frequency of occurrence, or pattern of change of the input behavior data, and generates the feature vector using the plurality of characteristic values. Claim 2 An AI-based medical consent data analysis and risk prediction system according to claim 1, characterized in that the input behavior data includes at least one of the input time per medical consent form item, the number of revisions to the medical consent form, the time required for signing, scroll movement patterns, and comprehension evaluation results. Claim 3 delete Claim 4 The AI-based medical consent form data analysis and risk prediction system according to claim 1, wherein the feature vector generation unit generates the feature vector to reflect the change pattern over time of each of the plurality of feature values ​​by considering the time sequence of the input behavioral data throughout the entire process of creating the medical consent form. Claim 5 The AI-based medical consent data analysis and risk prediction system according to claim 1, wherein the AI ​​risk prediction unit calculates a medical risk score of the patient based on the feature vector and classifies the patient into one of a plurality of risk grades based on the medical risk score. Claim 6 An AI-based medical consent data analysis and risk prediction system according to claim 5, characterized in that the plurality of risk grades include a general group, a caution group, and a risk group. Claim 7 An AI-based medical consent data analysis and risk prediction system according to claim 5, further comprising a notification unit that provides a notification to medical staff when the medical risk score exceeds a preset threshold. Claim 8 An AI-based medical consent form data analysis and risk prediction system according to claim 1, further comprising a log recorder that stores risk log data including patient identification information, medical consent form identification information, medical risk score, and time of occurrence, wherein the risk log data is linked to a medical institution system.

Citation Information

Patent Citations

  • Electronic contract generation method and device, computer equipment and storage medium

    CN112818300A

  • Apparatus and Method for Providing an Electronic Document Consent Procedure Service

    KR1020250125863A

  • Method and apparatus for authenticating handwritten signature using multiple authentication algorithms

    KR102144289B1

  • Document-based fraud detection

    US20240303664A1