Liver transplantation recipient intelligent interaction follow-up visit management system based on compliance behavior portrait

Abstract

The invention belongs to the technical field of medical information, and relates to a liver transplantation recipient intelligent interaction follow-up visit management system based on a compliance behavior portrait, which comprises a portrait generation module used for generating a dynamic compliance behavior portrait representing the current state of a liver transplantation recipient; the risk prediction module is used for generating a compliance risk prediction result including a risk level and a risk type; the interactive intervention module is used for generating a personalized interactive intervention strategy matched with the risk level, the risk type and the current state according to the compliance risk prediction result and the dynamic compliance behavior portrait; the risk early warning module determines to trigger the interaction intervention module to execute a personalized interaction intervention strategy according to a comparison result of a risk level in the compliance risk prediction result and a preset risk level threshold value, or sends a high-risk early warning notification to the medical care terminal platform; the system solves the problem that the system cannot know the non-compliance behavior of the patient only after the patient does not record on time, and is lack of predictability.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology and relates to an intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiles. Background Technology

[0002] Liver transplantation is an effective treatment for end-stage liver disease. Recipients require long-term comprehensive health management, including the use of immunosuppressants, to ensure the long-term survival of the transplanted liver and a good quality of life. With the increasing application of information and communication technologies in the healthcare field, various digital tools for post-operative patient management have emerged, such as mobile health applications and remote follow-up platforms. These tools aim to assist patients in self-management through information technology and enhance communication between doctors and patients.

[0003] In existing technologies, postoperative management systems for liver transplant recipients typically offer some basic functions. These systems or mobile applications mostly include medication reminders, manual entry of health data such as blood pressure and weight, and online text and image communication with medical staff. Patients can receive medication notifications according to a preset schedule and periodically upload some basic physiological indicators. Medical staff can then view these scattered data records in the backend and respond and provide guidance via messages or online chat when needed.

[0004] However, existing technologies have significant shortcomings in practical applications. First, the clinical data collected by existing systems is disconnected from patient behavioral data. For example, the system cannot correlate an abnormal lab report with a patient's recent online consultation behavior for comprehensive analysis. Furthermore, management methods are often static and one-way. For instance, medication reminders are based on fixed schedules, and the system only becomes aware of non-compliance after the patient fails to record on time, lacking predictability. The health education or intervention measures provided by the system are often standardized, failing to fully consider the individualized needs of patients due to changes in their condition, psychological fluctuations, etc., resulting in suboptimal intervention outcomes. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiles.

[0006] A liver transplant recipient intelligent interactive follow-up management system based on compliance behavior profiles includes: The profile generation module is used to acquire objective clinical data and user interaction behavior data of liver transplant recipients, integrate the objective clinical data and user interaction behavior data, and generate a dynamic compliance behavior profile that represents the current state of liver transplant recipients. The risk prediction module, based on the dynamic compliance behavior profile, uses a preset compliance risk prediction model to analyze and generate compliance risk prediction results including risk level and risk type. It uses historical training data to fit the model parameters and calculates risk scores and classifies risk levels based on the dimensional scores of the dynamic compliance behavior profile. The interactive intervention module generates personalized interactive intervention strategies that match the risk level, risk type, and current status based on the compliance risk prediction results and dynamic compliance behavior profiles. The risk warning module determines whether to trigger the interactive intervention module to execute a personalized interactive intervention strategy or send a high-risk warning notification to the healthcare platform based on the comparison between the risk level in the compliance risk prediction results and the preset risk level threshold. The preset risk level threshold is set based on the clinical expert experience or historical data and is used to distinguish between low risk, medium risk and high risk.

[0007] A further aspect of this invention involves obtaining objective clinical data and user interaction behavior data from liver transplant recipients, including the following steps: Obtain medication records and examination reports uploaded by liver transplant recipients to obtain objective clinical data; The system captures in real time the keywords asked by liver transplant recipients in the follow-up management platform, the reading time of health knowledge articles, and their type preferences. It then performs structured analysis on the keywords asked, the reading time of health knowledge articles, and their type preferences to obtain user interaction behavior data. The structured analysis includes extracting the features of the keywords asked and quantifying the reading time of health knowledge articles and their type preferences to obtain standardized user interaction behavior data. This involves correlation analysis between physiological indicators revealed by objective clinical data and psychological and cognitive states reflected by user interaction behavior data. The correlation analysis is to link physiological indicators in objective clinical data with psychological and cognitive characteristics in user interaction behavior data.

[0008] A further aspect of this invention generates a dynamic compliance behavior profile, comprising the following steps: Establish an event-triggered profile update mechanism, which uses received objective clinical data or user interaction behavior data as trigger events. When the event-triggered profile update mechanism detects the event, it adjusts the dimension labels and their weights in the dynamic compliance behavior profile in real time to generate an updated dynamic compliance behavior profile.

[0009] A further aspect of this invention involves adjusting the dimension labels and their weights in the dynamic compliance behavior profile in real time, including: dynamically adjusting the scores of each dimension using an update formula based on the amount of change in new event data.

[0010] A further aspect of the present invention includes the following steps: Personalized interactive intervention strategies are implemented on the user terminals of liver transplant recipients to achieve interactive intervention; Monitor liver transplant recipients' response to interactive interventions to obtain intervention feedback data; Intervention feedback data is used as new user interaction behavior data, and an event-triggered profile update mechanism is used to drive the dynamic compliance behavior profile to be updated.

[0011] A further aspect of the present invention generates personalized interactive intervention strategies, including: From the pre-set intervention strategy library, based on the risk level and risk type in the compliance risk prediction results, the initial intervention content, intervention form and intervention timing are matched; By utilizing the current state of liver transplant recipients as represented by dynamic compliance behavior profiles, the initial intervention content, form, and timing can be adjusted to generate personalized interactive intervention strategies.

[0012] A further aspect of the present invention is that the preset intervention strategy library is a structured knowledge base that stores standard intervention templates for different risk levels and risk types. The standard intervention templates include intervention content, intervention form, and intervention timing.

[0013] A further aspect of the present invention includes the following steps: We continuously collect personalized interactive intervention strategies and intervention feedback data from multiple liver transplant recipients to form strategy-effect data pairs. Based on the strategy-effect data, statistical analysis was conducted to obtain the results of the intervention effectiveness evaluation of different intervention strategies on different dynamic compliance behavior profiles; Based on the results of the intervention effectiveness assessment, optimize the pre-set intervention strategy library.

[0014] In a further embodiment of the present invention, the risk warning module further includes the following steps: Compare the risk level in the risk prediction results with the preset risk level threshold; If the risk level does not exceed the risk level threshold, the interactive intervention module will be triggered to execute a personalized interactive intervention strategy. If the risk level exceeds or equals the risk level threshold, a high-risk warning notification will be automatically generated. This high-risk warning notification will include at least the complete compliance risk prediction result that triggered the warning and the corresponding dynamic compliance behavior profile. The high-risk warning notification will be sent to the medical staff platform to prompt medical staff to make manual intervention decisions.

[0015] A further aspect of this invention involves receiving and storing the basic information, medical history, and surgical information of a liver transplant recipient to form a basic information file; and automatically generating a follow-up plan for the liver transplant recipient based on the basic information file and a preset follow-up template, and storing it in the follow-up management platform.

[0016] In summary, the present invention has the following beneficial technical effects: 1. By integrating objective clinical data and user interaction data of liver transplant recipients, a dynamic compliance profile that reflects their comprehensive physiological and psychological state in real time was constructed. Based on this profile, a forward-looking risk prediction was made, transforming follow-up management from a traditional passive response and delayed treatment to a proactive and predictive intervention model. This model can identify risks and intervene before non-compliance occurs, improving the effectiveness and foresight of management. 2. A closed-loop management process has been established, from interactive intervention to behavioral feedback to profile updates, making every doctor-patient interaction a diagnostic and learning process. The system can dynamically adjust and optimize its management strategies based on real-time feedback from patients on intervention measures, enabling intervention measures to be highly tailored to individual patient differences and dynamic needs, thereby improving patient acceptance and satisfaction. 3. By setting risk level thresholds, the system can efficiently handle routine risk events, while accurately identifying and promptly warning of high-risk situations requiring medical expert intervention. The hierarchical management model not only ensures the safety and reliability of follow-up management, but also optimizes the allocation of medical resources, enabling professional medical staff to focus their efforts on the most critical cases, thus achieving simultaneous improvement in management efficiency and medical quality. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This discloses a first flowchart of an embodiment of this application.

[0019] Figure 2 This discloses a schematic diagram of the framework in the embodiments of this application.

[0020] Figure 3 This discloses a second process diagram in the embodiments of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The following is in conjunction with the appendix Figure 1 - Figure 3 A preferred description of the present invention is provided below.

[0023] See attached document Figure 1 - Figure 3 This invention proposes an intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling, comprising the following modules: The profile generation module is used to acquire objective clinical data and user interaction behavior data of liver transplant recipients, integrate the objective clinical data and user interaction behavior data, and generate a dynamic compliance behavior profile that represents the current state of liver transplant recipients. The risk prediction module, based on the dynamic compliance behavior profile, uses a preset compliance risk prediction model to analyze and generate compliance risk prediction results including risk level and risk type. It uses historical training data to fit the model parameters and calculates risk scores and classifies risk levels based on the dimensional scores of the dynamic compliance behavior profile. The interactive intervention module generates personalized interactive intervention strategies that match the risk level, risk type, and current status based on the compliance risk prediction results and dynamic compliance behavior profiles. The risk warning module determines whether to trigger the interactive intervention module to execute a personalized interactive intervention strategy or send a high-risk warning notification to the medical care platform based on the comparison between the risk level in the compliance risk prediction results and the preset risk level threshold. The preset risk level threshold is set based on the clinical expert experience or historical data and is used to distinguish between low-risk, medium-risk and high-risk levels.

[0024] In one embodiment of the present invention, obtaining objective clinical data and user interaction behavior data of liver transplant recipients includes the following steps: Obtain medication records and examination reports uploaded by liver transplant recipients to obtain objective clinical data; The system captures in real time the keywords of questions asked by liver transplant recipients, the reading time of health knowledge articles, and their type preferences within the follow-up management platform. It then performs structured analysis on these keywords, reading time, and type preferences to obtain user interaction behavior data. By correlating the physiological state revealed by objective clinical data with the psychological and cognitive state reflected by user interaction behavior data, we can enhance the accuracy of dynamic compliance behavior profiles in representing the potential compliance motivation of liver transplant recipients.

[0025] To achieve a comprehensive and accurate profile of liver transplant recipients' compliance behavior, the first step is to guide liver transplant recipients to proactively submit their medical information through the user terminal interface of the follow-up management platform.

[0026] Specifically, the system receives medication records and periodic examination reports. After receiving this information in the system backend, it uses image recognition or data extraction technologies to analyze and structure key indicators, such as immunosuppressant blood drug concentrations and liver function values, thereby forming objective clinical data reflecting the recipient's physiological health status. Simultaneously, the system continuously captures all interaction traces of liver transplant recipients within the follow-up management platform in real time. For example, when a user enters a question in the platform's health consultation module, the system extracts keywords from the question using text analysis algorithms; when a user browses health education articles, the system records the reading time and the type of articles they prefer. These raw interaction logs are cleaned, denoised, and feature-extracted, transforming them into user interaction behavior data that can be quantitatively analyzed.

[0027] The scores for each portrait dimension are calculated using a weighted fusion algorithm, as shown in the following formula: in, This represents the comprehensive score of the i-th profile dimension, such as physiological risk, psychological state, and cognitive level. The dimension definition is based on the liver transplant recipient follow-up guidelines. For example, the physiological risk dimension includes liver function indicators (such as ALT and AST) and blood concentration of immunosuppressants; the psychological state dimension includes anxiety level and confidence in treatment. The standardized value of the clinical objective value in the i-th dimension is obtained by using Min-Max standardization: Where X is the original clinical value, such as blood drug concentration. and Based on historical data or clinical safety ranges, for example, tacrolimus blood concentration range of 5–15 ng / mL; For the standardized values ​​of user interaction behavior data in the i-th dimension, for example, for the psychological state dimension, It can be calculated based on the sentiment analysis score of the question keywords (such as the frequency of negative words) and the reading time deviation (the ratio of actual reading time to expected reading time); and The initial values ​​are weights, set by liver transplant experts based on the importance of the dimensions or their impact on adherence behavior, for example: Physiological risk dimension: =0.8, =0.2, emphasizing the dominance of clinical data; Psychological state dimension: =0.3, =0.7, emphasizing the dominance of behavioral data; Sum of weights: + =1, the score range is .

[0028] In this process, the system does not process these two types of data in isolation, but performs a correlation analysis process. This process integrates the physiological indicators revealed by objective clinical data with the psychological and cognitive states reflected by user interaction data. For example, the system can associate a newly uploaded test report with abnormal transaminase levels with the user's recent frequent searches for keywords such as "drug-induced liver injury" and "rejection reaction". Through this correlation analysis, the system can generate a dynamic compliance behavior profile that is far more accurate than that of a single data source. This dynamic compliance behavior profile not only includes physiological facts, but also represents potential compliance motivations that may affect future behavior.

[0029] In one embodiment of the present invention, generating a dynamic compliance behavior profile includes the following steps: Establish an event-triggered profile update mechanism, which uses received objective clinical data or user interaction behavior data as trigger events. When the event-triggered profile update mechanism detects the event, it adjusts the dimension labels and their weights in the dynamic compliance behavior profile in real time to generate an updated dynamic compliance behavior profile.

[0030] The steps for real-time adjustment of the dimension labels and their weights in the dynamic compliance behavior profile include: dynamically adjusting the scores of each dimension using an update formula based on the amount of change in new event data.

[0031] Specifically, in order to ensure that the compliance profile of liver transplant recipients can reflect their latest status in real time, an event-triggered profile update mechanism was added after the initial dynamic compliance profile was generated.

[0032] Generally, the mechanism runs continuously in the background of the system, acting as a data listener to specifically monitor the data flow associated with a specific liver transplant recipient and define newly received information as a trigger event. For example, when a liver transplant recipient uploads a new lab report via a terminal, and the system successfully parses it to generate new objective clinical data, this constitutes a triggering event. Similarly, when the system detects that a user has completed a new keyword search or article reading within the platform and processes it into new user interaction data, this also constitutes a triggering event. Once the event-triggered profile update mechanism detects either type of triggering event, it immediately activates a preset update algorithm. This update algorithm adjusts the relevant dimension labels and their weights in the user's current dynamic compliance behavior profile in real time, based on the type and content of the triggering event. For instance, new objective clinical data indicating worsening liver function will directly increase the weight of the "physiological risk" dimension label in the profile; while new user interaction data regarding "drug side effects" will increase the weight of the "anxiety level" dimension label. After this real-time, targeted adjustment, the system immediately generates an updated dynamic compliance behavior profile, which replaces the old version and becomes the user's latest status representation in the system for subsequent steps.

[0033] Real-time adjustments are achieved through the following image update formula: in, This represents the updated score for the i-th dimension after the t-th time trigger. The change in new event data is calculated as follows: ,in It is a new data value. It is the historical average. This refers to the normal range of data, such as the normal range of liver function indicators; and These are the decay factor and the update factor, respectively, satisfying... The initial value is set to =0.7, =0.3, emphasizing historical continuity, generally The larger the value, the greater the impact of new event data on the profile. This "trust level" or "influence weight" is usually referred to as the learning rate in machine learning. Usually The complementary parameters, generally, in standard designs, are set to ensure that the updated image score is not infinitely amplified or reduced, and are usually set to... and satisfy =1- In this way, updating the portrait becomes a weighted average process, ensuring numerical stability. Therefore, when When dynamically adjusted, It will change automatically and complementaryly. If Increase The corresponding reduction means that the weight of historical data decreases and the weight of new data increases.

[0034] Learning rate Example of dynamic adjustment: Adjustments are made based on the type and urgency of the event, such as emergency clinical events. Temporarily raise the level to 0.6 for rapid response, such as when blood drug concentration exceeds the limit; routine behavioral events: It maintains a value of 0.2 and updates smoothly, similar to reading an article; it is implemented through an event classifier, whose rules are defined by experts, such as marking an emergency if the rate of change in blood drug concentration is >10%. To avoid excessive fluctuations; a lower limit of 0.1 ensures the system is sensitive to minor changes, while an upper limit of 0.9 prevents historical data from being ignored; Specifically: Routine behavioral events, such as reading health articles and asking routine questions: use a lower learning rate and set... =0.2, then =0.8; This indicates that the profile mainly relies on historical data and reacts mildly to new behavioral data, avoiding drastic fluctuations in the profile due to a single accidental behavior.

[0035] Significant clinical events, such as uploaded test reports showing key indicators exceeding the normal range: Use a moderate learning rate and set... =0.5, then =0.5; This means that historical information and current new information are considered equally, and the profile will be significantly adjusted to the new state.

[0036] In emergency clinical events, such as undetectable blood drug concentrations or reports suggesting a possible acute rejection reaction: use a higher learning rate and set... =0.8, then =0.2; This indicates that the system highly trusts new data, and the profile will respond quickly, almost entirely driven by the latest events, in order to immediately trigger high-risk warnings.

[0037] In one embodiment of the present invention, the risk types include the risk of forgetful missed doses, the risk of intentional dose reduction, and the risk of missed follow-up appointments. Specifically, in adherence risk prediction, the system's precise definition of risk types is the foundation for achieving accurate intervention. When the pre-set adherence risk prediction model receives the updated dynamic adherence behavior profile, it identifies and outputs the specific risk type based on the combination of multi-dimensional features in the profile.

[0038] For example, if the profile shows that a liver transplant recipient's attendance record is irregular and their lifestyle is disrupted, but their user interaction data does not show concerns about drug side effects or questions about the treatment plan, the model will determine that their primary risk is the risk of missed doses due to forgetfulness. Conversely, if the profile reveals that after uploading a report of poor liver function indicators, the user interaction data shows frequent keyword searches related to immunosuppressant side effects and drug toxicity, accompanied by emotionally charged questions, the model will determine that they face the risk of intentionally reducing medication. Furthermore, when the profile shows that the user's response to follow-up reminders is delayed, and they inquire about transportation difficulties or examination costs, the model will categorize this as the risk of missed follow-up appointments. Ultimately, these precisely categorized risk types, along with risk levels, constitute a complete adherence risk prediction result, providing a clear direction for subsequent intervention strategy development.

[0039] The compliance risk prediction model uses a logistic regression algorithm to calculate the risk score using the following formula: in, To comply with risk probability, range For example, risk levels are categorized as: low risk (0–0.3), medium risk (0.3–0.7), and high risk (0.7–1); P1, P2, ..., P n The scores for the portrait dimensions are used as feature inputs; θ0, θ1, ..., θ n The model parameters (feature weights) are fitted using historical training data; Feature weight training method: Collect profile dimensions scores and actual adherence outcomes of historical liver transplant recipients, such as missed medication doses and missed follow-up appointments, and label them as binary categories, such as 1 indicating a risk event and 0 indicating no event.

[0040] Optimize parameters using maximum likelihood estimation via logistic regression or gradient descent. Regularization (such as L2 regularization) prevents overfitting.

[0041] Since the features have been standardized, the weights typically fall within the range of... Within the range. For example: θ i >0 indicates that the risk increases as the score of this dimension increases (e.g., the weight θ1 of the physiological risk dimension can be set to 1.5). θ i <0, the risk decreases when the score of this dimension increases (e.g., the weight θ2 of the treatment confidence dimension can be set to -0.8). When fitting the weights, medical prior knowledge is introduced, such as prioritizing physiological risk over behavioral risk, to ensure that the model conforms to clinical logic.

[0042] The formula for classifying risk types is as follows: in, For the profile dimension score vector (e.g.) (corresponding to physiological, psychological, and cognitive dimensions) This is a weight matrix, where each row corresponds to a risk type, such as... Risk of forgetting to take a dose Risks associated with intentional medication reduction Risk of missed follow-up appointments; The bias vector is used to adjust the classification threshold; the training data uses multi-class labeled data, where each instance contains profile features and the true risk type, labeled by healthcare professionals; the weight matrix... The fitting method is Softmax regression, which minimizes the cross-entropy loss. The feature weights have no fixed range but are standardized.

[0043] Example of weight distribution: Forgetfulness risk profile: high weighting for behavioral pattern dimension (e.g., W1=0.9), low weighting for clinical dimension; Intentional risk behavior: high weighting for the psychological anxiety dimension (e.g., W2=0.9), and medium weighting for the interactive behavior dimension; Its design is based on the behavioral patterns of liver transplant recipients. For example, intentional risk is often related to psychological factors, while amnesia risk is related to daily routines.

[0044] In one embodiment of the present invention, after generating the personalized interactive intervention strategy, the following steps are further included: Personalized interactive intervention strategies are implemented on the user terminals of liver transplant recipients to achieve interactive intervention; Monitor liver transplant recipients' response to interactive interventions to obtain intervention feedback data; Intervention feedback data is used as new user interaction behavior data, and an event-triggered profile update mechanism is used to drive the dynamic compliance behavior profile to be updated.

[0045] Specifically, after generating a personalized interactive intervention strategy, the system translates this strategy into actual instructions. For example, it might push a customized reminder message to the liver transplant recipient's user terminal, such as a smartphone, or display a health education article addressing their current concerns within an application, thus completing the interactive intervention. Following this, the system enters a continuous monitoring state, actively tracking and recording all user responses to the intervention. For instance, does the user immediately click on the reminder, or choose to delay or ignore it? Does the user read the recommended article completely, or quickly swipe past it? These raw response behaviors are captured by the system and transformed into structured intervention feedback data. The core of this process is that the system does not consider this interaction as the end point, but rather as a new user signal with a clear intent. Subsequently, this intervention feedback data is characterized by new user interaction behavior data and immediately sent to the aforementioned event-triggered profile update mechanism. Upon receiving this latest data, the profile update mechanism immediately drives the user's dynamic compliance behavior profile to update and adjust, ensuring that the profile can reflect the user's latest state after receiving the intervention, forming a complete closed-loop management process.

[0046] In one embodiment of the present invention, generating a personalized interactive intervention strategy includes the following steps: From the pre-set intervention strategy library, based on the risk level and risk type in the compliance risk prediction results, the initial intervention content, intervention form and intervention timing are matched; the pre-set intervention strategy library is a structured knowledge base that stores standard intervention templates for different risk levels and risk types. The standard intervention templates include intervention content, intervention form and intervention timing.

[0047] By utilizing the current state of liver transplant recipients as represented by dynamic compliance behavior profiles, the initial intervention content, form, and timing can be adjusted to generate personalized interactive intervention strategies. The adjustments include enabling personalized interactive intervention strategies to address the problems indicated by the risk type while adapting to individual preferences and acceptance levels revealed by dynamic compliance behavior profiles.

[0048] Specifically, when generating personalized interactive intervention strategies, the system first queries and matches a pre-defined intervention strategy library based on the risk level and type from the adherence risk prediction results generated in the previous step. Generally, this intervention strategy library is a structured knowledge base that stores standard intervention templates for different risk scenarios. For example, when the adherence risk prediction result is "moderate level, intentional medication reduction risk," the system will match an initial intervention content, which might be a popular science article about balancing drug efficacy and side effects, an initial intervention format (in-app push notification), and an initial intervention timing (the user's typical active period).

[0049] Then, the system uses the user's real-time dynamic compliance behavior profile to adjust its intervention. For example, if the user's dynamic compliance behavior profile shows a high level of knowledge but also high anxiety, the system will determine that a simple popular science article may not be in-depth enough, and thus adjust the initial intervention content to an in-depth interpretation or research summary written by an authoritative expert; if the profile shows that the user prefers video content to text, the system will change the intervention form from article push to sending a short video link; if the profile shows that the user has recently shown signs of nighttime activity, the system may adjust the intervention timing to late at night to better align with their current sleep schedule. In this way, the system ultimately generates a highly customized personalized interactive intervention strategy that not only addresses the core issues pointed to by the compliance risk prediction results, but also fully adapts its expression and timing to the individual preferences and acceptance levels revealed by the dynamic compliance behavior profile.

[0050] Taking the adjustment of intervention timing as an example, the system uses the following formula for personalized calculation: in, The timing of initial intervention is set based on the risk level, for example, high risk corresponds to immediate intervention, medium risk corresponds to 2 hours later, and low risk corresponds to 1 day later. User time preference scores are calculated by analyzing user activity history, such as nighttime activity level = number of nighttime activities / total number of activities, and then standardized. ; This is an adjustment factor, in hours, with a range of values. ∈[-3,3], representing the maximum adjustment range; The settings are based on optimization through A / B testing, for example, the test results show... The highest user response rate is achieved during early intervention; negative values ​​indicate early intervention, while positive values ​​indicate delayed intervention. Dynamic optimization is performed based on intervention feedback. For example, if users respond positively to the timing of adjustments, then increase .

[0051] Content fit calculation formula: in, To determine the feature value of the intervention content in the i-th dimension, such as the correlation between the content topic and physiological risks, keyword matching degree is extracted through text analysis. The score of the user profile on the i-th dimension; sim is ,scope ; The weights are initially set by experts, such as physiological risk weights. Psychological weight Then optimize based on feedback data; like .

[0052] In one embodiment of the present invention, establishing a self-optimizing learning loop includes the following steps: We continuously collect personalized interactive intervention strategies and intervention feedback data from multiple liver transplant recipients to form strategy-effect data pairs. Statistical analysis was conducted based on strategy-effect data to obtain the results of evaluating the intervention effectiveness of different intervention strategies on different dynamic compliance behavior profiles. Based on the results of the intervention effectiveness assessment, the pre-set intervention strategy library is optimized to improve the accuracy of subsequent generation of personalized interactive intervention strategies.

[0053] Specifically, the system continuously collects and records the personalized interactive intervention strategies implemented on all liver transplant recipients and the resulting intervention feedback data. Each set of data is structured and stored as a strategy-effect data pair. This pair includes not only the specific strategy content and the user's feedback behavior but also a snapshot of the user's dynamic compliance behavior profile at the time of intervention. Once a sufficient number of strategy-effect data pairs have been accumulated, the system automatically launches a statistical analysis engine. This engine analyzes massive amounts of data, for example, by comparing the performance of different strategies in user groups with similar dynamic compliance behavior profiles, to calculate the intervention effectiveness evaluation results of different intervention strategies for different user profiles. This result may be expressed as a set of confidence scores or association rules, such as "For the high anxiety profile group, the success rate of video-based intervention strategies is 40% higher than that of article-based strategies." Finally, based on this objective intervention effectiveness evaluation result, the system automatically adjusts and optimizes the preset intervention strategy library. Such optimization may include increasing the matching priority of efficient strategies, reducing or eliminating inefficient strategies, or even combining existing strategies to generate new candidate strategies, thereby directly improving the accuracy and success rate of generating personalized interactive intervention strategies for other users.

[0054] The optimization process is achieved through the following priority update formula: Effectiveness is the intervention effectiveness score, ranging from [0,1], calculated based on user feedback (such as whether the intervention was followed within 24 hours after the intervention). , These are strategic features (such as content type, format, and timing) and user profile features (such as risk level and psychological state). for The training data includes policy-effect pairs, with effect labels being binary (1: effective, 0: ineffective); training uses logistic regression with a regularization parameter C=1.0; due to feature standardization, the weights... The numerical range is usually in ,For example: γ a =0.8, where 'a' represents the video format, meaning the video format is more effective; γ b =1.2, b represents video format, meaning intervention is more effective for high-risk users; Typically, the model is retrained monthly, incorporating new data to maintain timely evaluation.

[0055] Strategy library optimization formula: in, The original strategy priority is used, with initial values ​​evenly distributed. The latest evaluation validity score, The learning rate controls the strength of the influence of historical priorities; for example... Range of values , At this initial stage, it quickly absorbs new feedback; The market is in a stable phase, with smooth fluctuations. Adjustments are made based on the frequency of strategy usage: high-frequency strategies Higher (e.g., 0.85), low-frequency strategy Lower (e.g., 0.65); For example, intervention is carried out every 100 times as an optimization cycle, and the priority is updated in batches after the intervention, with high priority strategies being selected first.

[0056] In one embodiment of the present invention, after generating a compliance risk prediction result including risk level and risk type based on a dynamic compliance behavior profile, an automated hierarchical response process is initiated. After generating the compliance risk prediction result, the risk level in the compliance risk prediction result is compared with a preset risk level threshold to determine whether the risk level in the compliance risk prediction result exceeds the preset risk level threshold. If the risk level does not exceed the risk level threshold, the interactive intervention module is triggered to execute a personalized interactive intervention strategy. If the risk level exceeds or equals the risk level threshold, a high-risk warning notification will be automatically generated. This high-risk warning notification will include at least the complete compliance risk prediction result that triggered the warning and the corresponding dynamic compliance behavior profile. The high-risk warning notification will be sent to the medical staff platform to prompt medical staff to make manual intervention decisions.

[0057] Specifically, the system performs a judgment step, comparing the risk level in the adherence risk prediction result with a risk level threshold pre-set in the system backend by medical experts or administrators. This risk level threshold is a specific value or level, representing the upper limit of risk that the system can automatically handle. If the judgment result shows that the risk level does not exceed the preset risk level threshold, the system will continue to execute automated personalized interactive intervention. Once the judgment result shows that the risk level has exceeded or equaled this risk level threshold, the system considers the situation to be beyond the safe scope of automated intervention and requires professional intervention. At this time, the system will automatically generate a high-risk warning notification. When generating this notification, the system will package the complete adherence risk prediction result that triggered the warning, as well as the user's latest dynamic adherence behavior profile, together as the core content of the notification. Subsequently, this high-risk warning notification, including the decision-making basis, will be sent and presented to the healthcare platform in real time, such as by popping up a warning window on the healthcare worker's workstation interface or adding it to the emergency to-do list, to ensure that healthcare workers can notice the high-risk event as soon as possible and make timely and accurate human intervention decisions based on the comprehensive information provided in the notification.

[0058] In one embodiment of the present invention, before the intelligent interactive follow-up management system is put into operation, basic filing and planning steps are first performed, including the following steps: Receive and store basic information, medical history, and surgical information of liver transplant recipients to form a basic information file; Based on basic information files and preset follow-up templates, follow-up plans are automatically generated for liver transplant recipients and stored in the follow-up management platform.

[0059] Specifically, when a liver transplant recipient is first included in the system's management, medical staff will enter the patient's basic information, past medical history, and key surgical information recorded during their hospital stay into the system through the medical staff platform. Alternatively, the system can automatically receive this data from the hospital's information system via an interface, process it in a structured manner, and then store it, thus establishing a unique basic information file for each patient. Subsequently, the system will access this newly generated basic information file and access pre-set follow-up templates within the platform, developed by clinical experts based on the latest treatment guidelines. These follow-up templates define the standard follow-up frequency and items for patients at different stages and risk levels. The system algorithm will automatically select the most suitable pre-set follow-up template based on key information such as the surgery date and risk stratification in the patient's basic information file, and automatically generate a detailed personalized follow-up plan including specific dates and items. This personalized follow-up plan will then be stored in the follow-up management platform and linked to the patient's basic information file.

[0060] It should be noted that the formulas described above, through the principle of dimensional consistency and mathematical standardization methods (such as normalization, dimensionless parameter conversion, or unit system unification), can translate physical quantities with different properties into unitless standard values ​​or superimposed parameters of the same dimension. This eliminates the interference of different dimensions on the computational logic, allowing the formulas to retain the original data distribution characteristics while possessing mathematical rationality and adaptability to objective laws. The descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention.

[0061] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0062] It should be noted that the human information (including but not limited to human device information and personal information) and data (including but not limited to data used for analysis, data stored and data displayed) involved in this invention are all information and data authorized by the human body or fully authorized by all parties. The collection, use and processing of related data require relevant legal standards.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A liver transplant recipient intelligent interactive follow-up management system based on compliance behavior profiling, characterized in that, include: The profile generation module is used to acquire objective clinical data and user interaction behavior data of liver transplant recipients, integrate the objective clinical data and user interaction behavior data, and generate a dynamic compliance behavior profile that represents the current state of liver transplant recipients. The risk prediction module, based on the dynamic compliance behavior profile, uses a preset compliance risk prediction model to analyze and generate compliance risk prediction results including risk level and risk type. It uses historical training data to fit the model parameters and calculates risk scores and classifies risk levels based on the dimensional scores of the dynamic compliance behavior profile. The interactive intervention module generates personalized interactive intervention strategies that match the risk level, risk type, and current status based on the compliance risk prediction results and dynamic compliance behavior profiles. The risk warning module determines whether to trigger the interactive intervention module to execute a personalized interactive intervention strategy or send a high-risk warning notification to the medical care platform based on the comparison between the risk level in the compliance risk prediction results and the preset risk level threshold. The preset risk level threshold is set based on the clinical expert experience or historical data and is used to distinguish between low-risk, medium-risk and high-risk levels.

2. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 1, characterized in that, Obtaining objective clinical data and user interaction behavior data from liver transplant recipients includes the following steps: Obtain medication records and examination reports uploaded by liver transplant recipients to obtain objective clinical data; The system captures in real time the keywords asked by liver transplant recipients in the follow-up management platform, the reading time of health knowledge articles, and their type preferences. It then performs structured analysis on the keywords asked, the reading time of health knowledge articles, and their type preferences to obtain user interaction behavior data. The structured analysis includes extracting the features of the keywords asked and quantifying the reading time of health knowledge articles and their type preferences to obtain standardized user interaction behavior data. This involves correlation analysis between physiological indicators revealed by objective clinical data and psychological and cognitive states reflected by user interaction behavior data. The correlation analysis is to link physiological indicators in objective clinical data with psychological and cognitive characteristics in user interaction behavior data.

3. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 1, characterized in that, Generating a dynamic compliance behavior profile includes the following steps: Establish an event-triggered profile update mechanism, which uses received objective clinical data or user interaction behavior data as trigger events. When the event-triggered profile update mechanism detects the event, it adjusts the dimension labels and their weights in the dynamic compliance behavior profile in real time to generate an updated dynamic compliance behavior profile.

4. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 3, characterized in that, The steps for real-time adjustment of the dimension labels and their weights in the dynamic compliance behavior profile include: dynamically adjusting the scores of each dimension using an update formula based on the amount of change in new event data.

5. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 3, characterized in that: Personalized interactive intervention strategies are implemented on the user terminals of liver transplant recipients to achieve interactive intervention; Monitor liver transplant recipients' response to interactive interventions to obtain intervention feedback data; Intervention feedback data is used as new user interaction behavior data, and an event-triggered profile update mechanism is used to drive the dynamic compliance behavior profile to be updated.

6. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 5, characterized in that, Generate personalized interactive intervention strategies, including: From the pre-set intervention strategy library, based on the risk level and risk type in the compliance risk prediction results, the initial intervention content, intervention form and intervention timing are matched; By utilizing the current state of liver transplant recipients as represented by dynamic compliance behavior profiles, the initial intervention content, form, and timing can be adjusted to generate personalized interactive intervention strategies.

7. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 6, characterized in that, The pre-defined intervention strategy library is a structured knowledge base that stores standard intervention templates for different risk levels and risk types. The standard intervention templates include intervention content, intervention form, and intervention timing.

8. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 5, characterized in that: We continuously collect personalized interactive intervention strategies and intervention feedback data from multiple liver transplant recipients to form strategy-effect data pairs. Statistical analysis based on strategy-effect data pairs yielded evaluation results of the intervention effectiveness of different intervention strategies on different dynamic compliance behavior profiles. Based on the results of the intervention effectiveness assessment, optimize the pre-set intervention strategy library.

9. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 1, characterized in that, The risk warning module also includes the following steps: Compare the risk level in the risk prediction results with the preset risk level threshold; If the risk level does not exceed the risk level threshold, the interactive intervention module will be triggered to execute a personalized interactive intervention strategy. If the risk level exceeds or equals the risk level threshold, a high-risk warning notification will be automatically generated. This high-risk warning notification will include at least the complete compliance risk prediction result that triggered the warning and the corresponding dynamic compliance behavior profile. The high-risk warning notification will be sent to the medical staff platform to prompt medical staff to make manual intervention decisions.

10. The intelligent interactive follow-up management system for liver transplant recipients based on compliance behavior profiling as described in claim 1, characterized in that, The system receives and stores the basic information, medical history, and surgical information of liver transplant recipients to form a basic information file. Based on the basic information file and the preset follow-up template, it automatically generates a follow-up plan for the liver transplant recipient and stores it in the follow-up management platform.

Images