Gouty arthritis patient pain management method and system based on MDT management mode

Abstract

The invention relates to a gouty arthritis patient pain management method and system based on an MDT management mode, and the method comprises the steps: automatically collecting the multivariate clinical data of a patient through integrating an existing information system of a hospital, carrying out the cleaning, standardization and structural processing through a special data preprocessing module, and carrying out the cleaning, standardization and structural processing. A machine learning prediction model trained by historical data is utilized to automatically generate a personalized MDT nursing plan, tasks are intelligently distributed to corresponding team members by means of a workflow engine, the execution condition is continuously monitored in the intervention process, and the nursing scheme is dynamically adjusted based on clinical data and prediction results which are periodically updated. And a complete'monitoring-prediction-intervention-reevaluation 'closed-loop management process is formed. The problems that a traditional MDT mode depends on manpower, efficiency is low, and decision-making lacks data support are effectively solved, and the accuracy, the collaboration and the continuity of pain management and wound surface nursing of the gouty calculus combined infection patient are improved.

Description

Technical Field

[0001] This application relates to the field of medical big data technology, and in particular to a pain management method and system for patients with gouty arthritis based on the MDT management model. Background Technology

[0002] In the traditional medical model, pain and wound management for patients with gouty arthritis, especially those who have developed tophi and infections, often relies on the experience of a single department. Conventional methods focus primarily on local dressing changes and basic drug interventions, lacking a systematic and comprehensive assessment of the patient's pain causes, nutritional status, psychological factors, and wound healing environment. This fragmented management model easily leads to ineffective nursing interventions, poor coordination between different treatment stages, resulting in poor pain control, slow wound healing, and negatively impacting patient experience and compliance.

[0003] Multidisciplinary team (MDT) collaboration is considered an effective way to improve the quality of management for such complex cases, aiming to integrate resources from multiple specialties such as rheumatology, wound care, and nutrition. However, it faces significant challenges in practice: communication and coordination among experts from different disciplines heavily rely on frequent, manually organized meetings, which are inefficient and difficult to sustain; treatment decisions are based more on experts' momentary experience rather than on objective, continuous, and dynamic data; and the development and subsequent adjustment of nursing plans lack quantitative basis and efficient tracking and implementation mechanisms, making it difficult to fully translate the theoretical advantages of the MDT model into stable and replicable clinical practice results.

[0004] In recent years, artificial intelligence (AI) technology has demonstrated tremendous potential in medical data mining and predictive analysis, offering new possibilities for optimizing clinical decision-making. However, how to seamlessly embed AI technology into the entire process of multidisciplinary team (MDT) collaboration, and build a technical solution that integrates automated data integration, intelligent predictive analysis, personalized plan generation, closed-loop task management, and dynamic feedback adjustment to truly achieve data-driven, precise, efficient, and personalized management, remains a pressing technical challenge in the field of smart healthcare. Summary of the Invention

[0005] To achieve the above objectives, this application provides the following technical solution: According to a first aspect of the present invention, the present invention claims protection for a pain management method for patients with gouty arthritis based on an MDT (Multidisciplinary Team) management model, comprising: S1. Collect clinical data of patients with gouty arthritis through an electronic health record system. The clinical data includes the patient's demographic information, laboratory test results, imaging test results, wound characteristics information, and pain scores. S2, The clinical data is cleaned and standardized using an artificial intelligence data preprocessing module to generate a standardized dataset; S3, The standardized dataset is input into a trained machine learning model to predict the progress of the patient's wound healing and the trend of pain changes. The machine learning model is trained based on historical data and includes a regression model and a classification model. S4. Based on the output of the machine learning model, a personalized MDT care plan is generated, which involves a multidisciplinary team including rheumatologists, wound and ostomy specialists, and caregivers, and specifies care measures and schedules. S5, push the personalized MDT care plan to the multidisciplinary team and patients through a mobile healthcare application, and monitor the implementation of the care plan; S6. Regularly collect subsequent clinical data, use the machine learning model to reassess wound healing progress and pain trends, and dynamically adjust the personalized MDT care plan.

[0006] Furthermore, the wound characteristic information in the clinical data in S1 includes wound size, granulation tissue coverage, infection status, and exudate volume. The pain score is collected using the Visual Analogue Scale (VAS) or the Numerical Rating Scale (NRS) and input into the electronic health record system via sensor devices or patient self-reporting.

[0007] Furthermore, in step S2, the steps of cleaning and standardizing the clinical data by the artificial intelligence data preprocessing module include: Detect missing values ​​and fill them using multiple imputation or k-nearest neighbor-based imputation methods; Perform one-hot encoding or label encoding on categorical variables, and normalize or standardize continuous variables to make the data meet the input requirements of machine learning models. Use anomaly detection algorithms, including isolated forest or local outlier algorithm, to identify and process outlier data points; The data is divided into a training set and a test set, where the training set is used for model training and the test set is used for model validation. Feature selection methods, including recursive feature elimination or tree-based feature importance assessment, are applied to select a subset of features relevant to wound healing and pain management.

[0008] Furthermore, the training steps of the machine learning model in S3 include: Collect historical clinical datasets, including clinical data, wound healing outcomes, and pain management effects from multiple patients with gouty arthritis; The historical clinical dataset was annotated, where wound healing results included wound shrinkage rate, granulation tissue coverage time, and healing effectiveness, and pain management results included changes in pain scores and patient satisfaction. An initial model is trained using a supervised learning algorithm, including random forest, support vector machine, or gradient boosting decision tree, to predict wound healing progress and pain change trends. Cross-validation was used to evaluate model performance, and hyperparameters were adjusted to optimize model accuracy and generalization ability. Deploy pre-trained machine learning models to cloud servers or edge computing devices for real-time or batch prediction; The machine learning model is periodically incrementally learned or retrained using newly collected clinical data to maintain its timeliness.

[0009] Furthermore, in step S4, the step of generating a personalized MDT care plan includes: Based on the predicted output of the machine learning model, the optimal care measures for the patient are determined, including moist care measures, debridement, dressing selection, dietary advice and health education content. Natural language processing technology is used to generate structured care plan documents, including care steps, frequency, and expected goals; The care plan is assigned to the appropriate members of the multidisciplinary team, and task execution is coordinated through a workflow management system.

[0010] Furthermore, the moist care measures include: In wound care, iodine disinfection, wet dressings, sodium bicarbonate irrigation, and silver ion dressings are used. The nursing care steps are dynamically adjusted according to the wound condition, and the adjustment of the nursing care steps is based on the results of convolutional neural network analysis of wound images by a machine learning model.

[0011] Furthermore, in S5, the mobile healthcare application includes a patient client and a doctor client: The patient-side app is used to collect patients' self-reported pain scores, wound photos, and dietary records, and to promote treatment adherence through push notifications. The physician's side is used to display care plans, patient data, and model prediction results, and provides decision support tools for adjusting care plans.

[0012] Furthermore, in S6, the step of periodically collecting subsequent clinical data includes: Using IoT devices to automatically collect wound images and physiological parameters, such IoT devices include smart dressings or portable cameras; The collected data is transmitted to a central database via the application programming interface (API) for model re-evaluation.

[0013] Furthermore, the method also includes a sample size optimization step: Generative adversarial networks (GANs) or synthetic minority oversampling techniques (SMOTE) are used to generate synthetic clinical data to augment the training dataset. The optimal sample size is determined using Bayesian optimization methods for model training and validation. The statistical analysis method is integrated into the artificial intelligence system: The collected data is processed using an automated statistical analysis module, which performs t-tests, χ² tests, or nonparametric tests. The analysis results are presented in a visual format, including charts and dashboards, to support decision-making by multidisciplinary teams.

[0014] According to a second aspect of the present invention, the present invention claims protection for a pain management system for gouty arthritis patients based on an MDT (Multidisciplinary Team) management model, comprising: One or more processors; A memory storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the aforementioned pain management method for gouty arthritis patients based on an MDT management model.

[0015] This invention relates to a method and system for pain management in patients with gouty arthritis based on a multidisciplinary team (MDT) management model. By integrating with existing hospital information systems, it automatically collects diverse clinical data from patients. This data is then cleaned, standardized, and structured using a dedicated data preprocessing module. A machine learning prediction model trained on historical data is used to automatically generate personalized MDT nursing plans. A workflow engine intelligently assigns tasks to appropriate team members, continuously monitoring implementation during intervention. Based on regularly updated clinical data and prediction results, the nursing plan is dynamically adjusted, forming a complete closed-loop management process of "monitoring-prediction-intervention-reassessment." This invention effectively solves the problems of reliance on manual labor, low efficiency, and lack of data support for decision-making in traditional MDT models, improving the accuracy, coordination, and continuity of pain management and wound care for patients with gouty tophi complicated by infection. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the pain management method for gouty arthritis patients based on an MDT management model, as claimed in this application. Figure 2 A second flowchart illustrating a pain management method for gouty arthritis patients based on an MDT management model, as claimed in this application. Figure 3The third flowchart is a method for pain management of gouty arthritis patients based on an MDT management model, which is claimed in the embodiments of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0018] The terms "first," "second," and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0019] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] According to a first embodiment of the present invention, the present invention claims protection for a pain management method for patients with gouty arthritis based on an MDT management model, referring to... Figure 1 ,include: S1. Collect clinical data of patients with gouty arthritis through an electronic health record system. The clinical data includes the patient's demographic information, laboratory test results, imaging test results, wound characteristics information, and pain scores. S2, The clinical data is cleaned and standardized using an artificial intelligence data preprocessing module to generate a standardized dataset; S3, The standardized dataset is input into a trained machine learning model to predict the progress of the patient's wound healing and the trend of pain changes. The machine learning model is trained based on historical data and includes a regression model and a classification model. S4. Based on the output of the machine learning model, a personalized MDT care plan is generated, which involves a multidisciplinary team including rheumatologists, wound and ostomy specialists, and caregivers, and specifies care measures and schedules. S5, push the personalized MDT care plan to the multidisciplinary team and patients through a mobile healthcare application, and monitor the implementation of the care plan; S6. Regularly collect subsequent clinical data, use the machine learning model to reassess wound healing progress and pain trends, and dynamically adjust the personalized MDT care plan.

[0021] In this embodiment, the data collection step involves automatically collecting clinical data of patients diagnosed with tophi complicated with infection through the hospital's Electronic Health Record (EHR) system application programming interface (API). The clinical data includes structured patient identification information, age, gender, and admission time, as well as unstructured laboratory test report text, imaging report text, wound description text, and pain self-rating scores recorded per instance. Data preprocessing steps: The collected clinical data is input into a central data processing server; the server runs a data preprocessing program, which first identifies and marks missing and outlier values ​​in the data, then transforms unstructured text reports into structured data points using named entity recognition technology, and merges them with existing structured data to finally generate a unified and standardized patient feature dataset. The feature vectors constructed by this system have a clear clinicopathological correlation with the prediction target. For example: Features used to predict wound healing progress mainly include: wound size, granulation tissue coverage, infection status (such as white blood cell count and C-reactive protein), exudate volume, serum uric acid level, and patient blood glucose level. Machine learning models learn the complex nonlinear relationships between these features and healing outcomes (such as wound shrinkage rate) through historical data. For example, the model learns that high serum uric acid levels and high CRP are generally negatively correlated with slower healing, while increased granulation tissue coverage is positively correlated with faster healing.

[0022] The features used to predict pain trends mainly include: current pain score (VAS / NRS), historical pain score sequence, degree of joint swelling, blood uric acid level, inflammatory markers, whether analgesics are used and the type of analgesics. By analyzing these features, the model learns to predict pain scores or pain levels (e.g., "relieved", "maintained", "worsened") at future time points (e.g., 3 days later, 7 days later).

[0023] The essence of model training is to use a large number of {feature vector, actual result} sample pairs from historical data to continuously adjust the internal parameters of the model in order to minimize the error between its prediction results and the actual results, so as to finally obtain a predictive model that can accurately infer the future healing and pain trends of new patients from their characteristics.

[0024] Predictive model application steps: Input the standardized patient feature dataset into a machine learning predictive model pre-deployed on a server; the model is trained based on historical gout patient data, and after executing the model, it outputs predictive indicators for specific patients, including wound healing speed scores and pain evolution trend curves; Nursing plan generation steps: A dedicated plan generation module receives the predictive indicators and automatically generates an initial, personalized multidisciplinary team (MDT) nursing plan based on a built-in clinical rule base. The plan clearly defines the tasks, execution time points, and expected nursing goals for different roles such as wound and ostomy therapists, rheumatologists, and specialist nurses. The nursing goals include debridement methods, dressing selection, irrigation solution composition, key health education content, and follow-up frequency. Plan distribution and monitoring steps: The generated personalized MDT care plan is distributed to the dedicated terminal device applications of each MDT member through the internal secure network; at the same time, a corresponding patient treatment timeline is created, and the actual execution time and executor identity of each care measure are automatically recorded to form an execution log; Dynamic adjustment steps: After the predetermined intervention time point, the data collection and data preprocessing steps are executed again to obtain the patient's subsequent clinical data; the new standardized dataset is input into the prediction model again to obtain updated predictive indicators; the MDT leader reviews the execution log of the initial plan and the updated predictive indicators through a terminal device, and based on this, manually adjusts or confirms the system-recommended adjustment plan in the plan generation module to generate an updated MDT care plan, which is then redistributed.

[0025] Furthermore, the wound characteristic information in the clinical data in S1 includes wound size, granulation tissue coverage, infection status, and exudate volume. The pain score is collected using the Visual Analogue Scale (VAS) or the Numerical Rating Scale (NRS) and input into the electronic health record system via sensor devices or patient self-reporting.

[0026] In this embodiment, during the data collection step, the patient's self-reported pain score is entered via an application interface on their smartphone, which provides a Visual Analogue Scale (VAS) slider. The submitted score, along with the patient's identification identifier and submission timestamp, is encrypted and transmitted to the EHR system. The wound description text is entered by the wound and ostomy technician using a voice-to-text function on their mobile device during dressing changes. The description includes the wound size, color, nature of exudate, and an estimated granulation tissue coverage.

[0027] Furthermore, referring to Figure 2 In step S2, the steps of cleaning and standardizing the clinical data by the artificial intelligence data preprocessing module include: Detect missing values ​​and fill them using multiple imputation or k-nearest neighbor-based imputation methods; Perform one-hot encoding or label encoding on categorical variables, and normalize or standardize continuous variables to make the data meet the input requirements of machine learning models. Use anomaly detection algorithms, including isolated forest or local outlier algorithm, to identify and process outlier data points; The data is divided into a training set and a test set, where the training set is used for model training and the test set is used for model validation. Feature selection methods, including recursive feature elimination or tree-based feature importance assessment, are applied to select a subset of features relevant to wound healing and pain management.

[0028] In this embodiment, data reception and caching: the central data processing server continuously receives clinical data streams pushed from the EHR system API through a message queue and temporarily stores them in a cache database; Data cleaning sub-steps: The preprocessing program periodically extracts batches of patient data from the cache; first, it performs integrity checks, filters out data records with missing required fields, and automatically sends retransmission requests to the data source system; for abnormal outliers in numerical fields, the program compares them with the historical data range of the same patient, and if they exceed the reasonable fluctuation range, they are marked as data to be reviewed, and their inclusion in subsequent analysis is suspended, while an alert is sent to the administrator. Data standardization sub-step: For laboratory test results, the program identifies their names and units, and converts all results into standard units and terms by referring to a unified set of medical standard terms; for imaging and wound description text, the program uses a pre-trained natural language processing model to extract key medical entities from the text, including "gouty tophi", "bone destruction", "exudate", "redness and swelling", etc., and converts these entities into standardized Boolean values ​​or classification values. Dataset construction sub-steps: Merge the cleaned structured data with the standardized entities extracted from the text, align them by patient identifiers and timestamps, and construct a feature vector dataset arranged in time series for each patient; Data output: The final processed standardized dataset is sent to a dedicated model input database table, waiting to be called by the prediction model.

[0029] Furthermore, referring to Figure 3 The training steps of the machine learning model in S3 include: Collect historical clinical datasets, including clinical data, wound healing outcomes, and pain management effects from multiple patients with gouty arthritis; The historical clinical dataset was annotated, where wound healing results included wound shrinkage rate, granulation tissue coverage time, and healing effectiveness, and pain management results included changes in pain scores and patient satisfaction. An initial model is trained using a supervised learning algorithm, including random forest, support vector machine, or gradient boosting decision tree, to predict wound healing progress and pain change trends. Cross-validation was used to evaluate model performance, and hyperparameters were adjusted to optimize model accuracy and generalization ability. Deploy pre-trained machine learning models to cloud servers or edge computing devices for real-time or batch prediction; The machine learning model is periodically incrementally learned or retrained using newly collected clinical data to maintain its timeliness.

[0030] In this embodiment, the detailed explanation of the training and prediction logic of the machine learning model includes: S3.1 Model Training Phase: After collecting historical clinical datasets, they are labeled. The labeling includes: Wound healing outcome labels: For example, the percentage reduction in wound area relative to baseline on day 14 can be used as a continuous value label (for regression models), or whether a 50% reduction in area is achieved within 28 days can be used as a binary label (for classification models).

[0031] Pain management effectiveness labels: For example, the change in pain score on day 7 relative to baseline can be used as a regression label, or whether pain decreased by ≥2 points within 5 days can be used as a classification label.

[0032] Subsequently, supervised learning algorithms are used for training. Taking Gradient Boosting Decision Tree (GBDT) as an example, its training process involves constructing an ensemble model composed of multiple decision trees. Each tree learns an approximation of the prediction residuals of the previous tree. By iteratively adding trees, the model gradually corrects the errors of previous models, and finally, the predictions of all trees are weighted and summed to obtain the final prediction value. Hyperparameters during the training process (such as maximum tree depth, learning rate, and number of subtrees) are optimized using cross-validation techniques to prevent overfitting while maintaining accuracy.

[0033] S3.2 Prediction and Application Stage: The preprocessed, standardized dataset of new patients is fed into the trained model. The model will then calculate one or more predicted outputs for that patient. For regression models, a specific numerical value is directly output, such as a predicted wound area reduction rate of 35%.

[0034] For classification models, the output is a probability distribution or category label, such as the patient having an 85% probability of belonging to the "rapid pain relief" category.

[0035] These quantitative predictions provide core, data-driven decision-making support for the subsequent generation of personalized care plans; Historical data preparation: Extract historical patient data from an independent model training database containing complete treatment cycle data of patients with tophi over the past few years, including their clinical characteristics, the nursing interventions received, and the final recorded healing and pain relief results; Feature engineering: Perform the same process as the data collection and preprocessing steps in claims 1 and 3 on the historical data to construct a standard historical dataset for training; and have clinical experts label each patient record with a final outcome label based on healing effectiveness, degree of pain score reduction, etc. Model training: The standard historical dataset is randomly divided into training and validation sets according to a preset ratio; the training set data is used to train various candidate machine learning model architectures, including decision tree ensemble models and multilayer perceptron neural networks; the training process adjusts the internal parameters of the model through iterative optimization algorithms to minimize the difference between the predicted results and the true outcome labels; Model Validation and Selection: The performance of each candidate model is evaluated using validation set data, and its prediction accuracy, recall, and other metrics are calculated. The evaluation results are reviewed by engineers and clinical experts, and the model with the best overall performance is selected as the prediction model for the production environment. Model Deployment: The selected production environment prediction model and all its required preprocessing parameters are encapsulated into a standalone software container; this container is deployed to the model runtime environment of the central data processing server and configured to read real-time data from the model input database table and write its output to the prediction results database.

[0036] Furthermore, in step S4, the step of generating a personalized MDT care plan includes: Based on the predicted output of the machine learning model, the optimal care measures for the patient are determined, including moist care measures, debridement, dressing selection, dietary advice and health education content. Natural language processing technology is used to generate structured care plan documents, including care steps, frequency, and expected goals; The care plan is assigned to the appropriate members of the multidisciplinary team, and task execution is coordinated through a workflow management system.

[0037] In this embodiment, the core of the plan generation module is a structured clinical rule base. This rule base is predefined by clinical experts based on medical guidelines and practical experience, and its form is a set of logical rules in the format of "IF<condition>THEN<action>". These rules dynamically associate the predicted output of the machine learning model with specific nursing interventions and implementing roles.

[0038] Rule base application examples: Rule 1 (Related Role: Wound and Ostomy Specialist): Conditions: IF the predicted wound healing progress is "slow" AND the infection status is "positive". Action: THEN Add the following task to the care plan: [Wound and Ostomy Practitioner] Perform "Silver Ion Dressing Change" daily and recommend [Microbial Culture].

[0039] Rule 2 (Responsible role: Rheumatologist): Conditions: IF predicts a persistently high-risk trend in pain and serum uric acid levels >480 μmol / L Action: THEN Add the following task to the care plan: [Rheumatologist] Assess and adjust the uric acid-lowering drug treatment regimen (e.g., increase the allopurinol dose or change the medication).

[0040] Rule 3 (Related roles: nutritionist, nurse): Condition: IF the predicted healing progress is "medium" or "slow". Action: THEN Add the following tasks to the care plan: [Dietitian] Conduct dietary assessments and develop a low-purine diet plan; [Nursing staff] Provide daily health education to patients, emphasizing the importance of dietary control.

[0041] The system automatically iterates through all relevant rules, summarizes all the "action" tasks generated after matching, removes duplicates, and forms a structured, personalized MDT (Multidisciplinary Team) care plan that includes specific measures, execution roles, and timelines. In this way, a closed loop is achieved from data prediction to intelligent multidisciplinary task allocation.

[0042] The clinical rule base in the nursing plan generation process is constructed by a group of senior experts in the MDT team, who jointly develop a series of logical rules based on clinical guidelines and experience. These rules exist in the form of "if-then" conditional statements, such as: if the predicted wound healing speed score is lower than the threshold X, then the debridement glue plus silver ion dressing is preferred, and the follow-up interval is shortened to Y days. Engineers translate and encode these rules into judgment logic that can be executed by the plan generation module.

[0043] Furthermore, the moist care measures include: In wound care, iodine disinfection, wet dressings, sodium bicarbonate irrigation, and silver ion dressings are used. The nursing care steps are dynamically adjusted according to the wound condition, and the adjustment of the nursing care steps is based on the results of convolutional neural network analysis of wound images by a machine learning model.

[0044] In this embodiment, the nursing plan generation step further includes: while outputting the text plan, the plan generation module also generates a structured task list; this list automatically creates task items and assigns them to the to-do lists of relevant MDT members through the hospital's workflow engine interface, and sets reminder notifications.

[0045] Furthermore, in S5, the mobile healthcare application includes a patient client and a doctor client: The patient-side app is used to collect patients' self-reported pain scores, wound photos, and dietary records, and to promote treatment adherence through push notifications. The physician's side is used to display care plans, patient data, and model prediction results, and provides decision support tools for adjusting care plans.

[0046] In this embodiment, during the planning, distribution, and monitoring steps, MDT members receive task reminders through their terminal device applications. After each nursing intervention is performed, members record the details of the intervention through the application's interface, including the batch number of the materials used, the start and end times of the operation, and observation records during the operation. This intervention data is uploaded in real time and integrated into the patient's intervention log.

[0047] Furthermore, in S6, the step of periodically collecting subsequent clinical data includes: Using IoT devices to automatically collect wound images and physiological parameters, such IoT devices include smart dressings or portable cameras; The collected data is transmitted to a central database via the application programming interface (API) for model re-evaluation.

[0048] In this embodiment, the system-recommended adjustment scheme is a revised MDT nursing plan generated by the plan generation module by comparing the differences between the initial predictive indicators and the updated predictive indicators, and combining the completion status of the execution log, and reapplying the clinical rule base.

[0049] Furthermore, the method also includes a sample size optimization step: Generative adversarial networks (GANs) or synthetic minority oversampling techniques (SMOTE) are used to generate synthetic clinical data to augment the training dataset. The optimal sample size is determined using Bayesian optimization methods for model training and validation. The statistical analysis method is integrated into the artificial intelligence system: The collected data is processed using an automated statistical analysis module, which performs t-tests, χ² tests, or nonparametric tests. The analysis results are presented in a visual format, including charts and dashboards, to support decision-making by multidisciplinary teams.

[0050] In this embodiment, when the system assesses that the amount of available training data is insufficient, a synthetic data generation technique is adopted. This technique generates synthetic patient data records that retain the key statistical features of the original data but have entirely new content based on the statistical distribution characteristics of the real patient feature dataset. These synthetic data are then labeled and added to the model training database to expand the training set.

[0051] The method also includes a statistical analysis step: at the end of the study period, the central data processing server automatically calls the built-in statistical analysis program; the program selects the final result data of the intervention group and the control group at the preset observation time points from the database, automatically selects the appropriate statistical test method, performs the significance test, and generates a standardized report containing statistical values, degrees of freedom, and p-values; the report is output in document form for researchers to download and review.

[0052] According to a second embodiment of the present invention, the present invention claims protection for a pain management system for gouty arthritis patients based on an MDT management model, comprising: One or more processors; A memory storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the aforementioned pain management method for gouty arthritis patients based on an MDT management model.

[0053] Before implementing this method, system deployment and configuration must be completed with the support of the hospital's information department. A secure data interface is established between the central data processing server and the hospital's existing Electronic Health Record (EHR) system, Laboratory Information System (LIS), and Picture Archiving System (PACS). Each interface undergoes strict access control authentication to ensure the security and privacy of data transmission.

[0054] The MDT team members, including rheumatologists, wound and ostomy specialists, specialist nurses, nutritionists, and psychotherapists, all have a dedicated client application installed on their office computers and mobile devices. This application maintains a real-time connection with a central server, ensuring instant synchronization of tasks and information.

[0055] Detailed implementation of the data collection process: The data interface module of the hospital's EHR system is set to run automatically on a scheduled basis. Every day at 2:00 AM, the system automatically starts the data collection process. This process first sends a query request to the EHR database to retrieve all patient records for which diagnostic information has been newly created or updated within the past 24 hours, and whose diagnostic codes contain a specific code for gouty tophi combined with infection.

[0056] For each qualified patient retrieved, the system will extract all available structured data fields, including patient ID, date of birth, gender, admission and discharge dates, diagnosis code, and surgical code. Simultaneously, the system will asynchronously initiate requests for unstructured data: requesting the patient's complete laboratory test reports from the LIS, relevant imaging reports from the PACS, and all nursing records and wound assessment notes from the nursing system.

[0057] All this data is collected by a data assembler component, which adds timestamps and data source tags to form a complete data packet. This packet is then placed in a secure, encrypted message queue, awaiting transmission to the central data processing server.

[0058] Detailed implementation of the data preprocessing steps: The central server's data ingestion service continuously listens to the message queue. When a new data packet is detected, it is retrieved into a temporary storage area.

[0059] The data cleaning subroutine then starts. It first verifies the integrity of the data, checking the existence of required fields. For missing fields, the system adopts different strategies based on the importance of the field: records missing key identifier fields (such as patient ID) are directly isolated and an alert is issued; records missing minor fields are marked as "partially missing" but continue to the next step, with the missing values ​​to be filled later.

[0060] For numerical data, the anomaly detection subroutine compares it to a predefined range of reasonable medical values. For extreme values ​​outside this range, the system marks the data as "suspicious" and compares it against the patient's historical data over time. If the value significantly deviates from the historical trend, it is temporarily excluded, and a verification task is generated and sent to the data administrator for manual confirmation.

[0061] Next, the Natural Language Processing (NLP) engine processes the text report. Loaded with a medical dictionary and grammatical model, the engine performs word segmentation, part-of-speech tagging, and semantic analysis on the report. It identifies and extracts key clinical entities and relationships, such as "Location: First metatarsophalangeal joint, Finding: Bone erosion, Severity: Moderate" from the imaging report, and transforms this information into structured data fields, entering them into standardized tables.

[0062] Finally, all the cleaned and transformed data is aligned and integrated according to patient identifiers and time points to generate standardized time-series data records for each patient, which are then stored in the "model-ready database".

[0063] The detailed implementation of the predictive model application steps: Each time a complete patient record is added to the "Model Ready Database", the predictive model service will be invoked.

[0064] The model service is a standalone software module that reads standardized feature vectors from a database. Internally, the model performs a series of complex mathematical calculations (the specific algorithms are trade secrets and will not be disclosed here), ultimately outputting a set of predictions. These predictions are returned in structured JSON format and typically include: the predicted category of wound healing speed (e.g., fast, medium, slow), predicted pain scores at multiple future time points, and the risk level of potential complications.

[0065] These predictions, along with patient identifiers, are written into a "prediction results database" and tagged with their version and generation time.

[0066] The plan generation module periodically scans the "prediction results database" for new records. For each new prediction result, it first reads its result.

[0067] The module then invokes its built-in clinical decision rule base, which consists of hundreds of condition-action rules. The module matches the patient's predicted outcomes and current clinical characteristics against the condition portions of these rules. For example, if a rule is matched: "If slow pain relief and purulent wound exudate are predicted, then the actions are: recommend microbial culture, enhanced anti-infection care, and increased dressing frequency," the module will add these actions to the initial draft plan.

[0068] The module summarizes all action suggestions from matching rules, removes duplicates, and resolves potential conflicts to form a preliminary, personalized MDT (Multidisciplinary Team) care plan draft. This draft details the suggested care measures, implementer roles, suggested implementation time windows, and expected goals.

[0069] Detailed implementation of the plan distribution and monitoring steps: The generated draft plan is first sent to the client application of the MDT team leader (usually a senior rheumatologist) for review. The leader can fully view the draft content, the predicted results and clinical data on which the draft was based, and has the authority to approve, modify, or reject it.

[0070] Once the plan is approved, the system sets its status to "effective." The workflow engine then intervenes, analyzing each task in the plan and creating specific to-do items, which are then assigned to the individual task lists of the appropriate role members. Each task includes a clear description, a deadline, and a link to the patient's information.

[0071] After logging into their application, MDT members can see all the tasks assigned to them in their "My To-Do" list. Once a task is completed, they need to mark it as "Completed" in the application and fill in the execution details as required, such as materials used, execution time, and observations made during the operation. The system automatically records all these operation logs, forming a complete and tamper-proof execution trail.

[0072] Detailed implementation of the dynamic adjustment process: The system sets a reassessment schedule for each patient (e.g., every seven days). Upon arrival at the scheduled time, the system automatically repeats the data collection and preprocessing steps to obtain the patient's latest clinical status.

[0073] The latest data is fed into the predictive model to generate updated forecasts. The system compares the new forecasts with the old ones, calculating the trends and differences in key indicators.

[0074] Based on the new prediction results and discrepancies, the plan generation module reapplies the rule base and may generate a suggestion for adjustment (e.g., "Pain relief is better than expected, it is recommended to appropriately reduce the frequency of analgesia assessment").

[0075] This adjustment proposal also requires review by the MDT (Multidisciplinary Team) leader. After comprehensive evaluation, the leader can approve the generation of a new care plan. The new plan will inherit any unfinished tasks from the original plan and update or add new task items. The system will send a notification of the plan update to relevant members, ensuring the entire team collaborates based on the latest consensus.

[0076] Patients open the pain reporting app on a smart terminal next to their bed or on their personal smartphone. The app interface displays an intuitive Visual Analogue Scale (VAS), typically a slider ranging from a smiley face icon for "no pain" to a crying face icon for "severe pain." Patients drag the slider to the position that best represents their pain level and then submit. The app collects the score, patient identification information, and the submission time accurate to the second, and uploads it.

[0077] When changing dressings, the wound and ostomy specialist uses a waterproof mobile handheld terminal. She opens the wound assessment application and selects the relevant patient. The assessment can be conducted in several ways: first, by filling out a form and selecting wound size, exudate type, etc., from drop-down menus and radio buttons; second, by voice input, directly describing what is seen, with the speech transcribed into text in real time; and third, with the patient's consent, taking and uploading photos of the wound. All these multimedia pieces of information are linked together to form a complete assessment record.

[0078] Data preprocessing is a multi-stage pipeline operation: Phase 1: Data Reception and Verification. After the data packet is received, it is first decrypted and parsed. The verifier checks the structural integrity and basic compliance of the data packet.

[0079] Phase Two: Data Cleaning. The cleaner processes data field by field. For missing values, the system first attempts to estimate an imputation value from the patient's historical data at other time points (e.g., previous weight). If an estimate cannot be estimated, categorical fields are filled with "unknown"; numeric fields are filled with the average value of the patient group. Once outliers are detected, they are not simply deleted, but a verification workflow is initiated, requiring confirmation from the data source system if necessary.

[0080] Phase Three: Data Transformation and Standardization. All date and time were converted to a unified time zone and format. All laboratory measurements were checked for units and converted to international standard units. When processing text using NLP technology, the engine not only identified entities but also attempted to determine their negative states (e.g., "no obvious bone destruction observed") and uncertainties (e.g., "suspected bone erosion"), accurately transforming them into structured data.

[0081] Phase Four: Data Integration and Output. Data from different systems regarding the same patient at the same point in time are merged. Ultimately, clean and standardized feature vectors are prepared for the machine learning model.

[0082] Building and maintaining a predictive model is an ongoing process. Historical data preparation: Extract patient data from a long historical time window from the data warehouse. This dataset needs to cover the complete cycle from admission to discharge outcome.

[0083] Feature engineering and annotation: A data preprocessing pipeline, identical to that used in the production environment, is applied to historical data to ensure fairness. Then, a review committee of clinical experts labels each record with a "gold standard" label (e.g., "good healing in 28 days") based on the final outcome recorded in the medical records.

[0084] Model Training and Validation: Data scientists build models on a training set using multiple machine learning algorithms. Each model is evaluated on a strictly isolated validation set, comparing their discrimination, calibration, and generalization ability. The best-performing model is selected.

[0085] Model Deployment and Monitoring: The selected model is encapsulated as a service accessible via a network API and deployed in a high-performance computing environment. The system continuously monitors the model's predictive performance, and if it detects a decline in predictive accuracy over time (model decay), it triggers a model retraining process.

[0086] The clinical rule base is the "brain" of the system. Rules are written in natural language by clinical experts, for example: "If a patient's 'malnutrition screening score' is above the threshold and their 'wound healing prediction' is slow, then 'trigger a nutritionist consultation'." Engineers translate these natural language rules into logical statements that the system can execute. After the plan generation module executes these rules, the output is not just a document, but a structured, actionable task list. This list is directly passed to the hospital's workflow engine via an application programming interface (API), which then creates specific task items in the system and assigns them to the appropriate personnel or role groups, achieving seamless automation from decision-making to execution.

[0087] MDT members receive task notifications on a mobile app. Upon completion of the task, they must open the app, locate the task, and tap the "Complete" button. The system prompts them to record necessary execution details, such as selecting the type of dressing used from a preset list or inputting brief operational notes via voice. This execution data is uploaded in real time and permanently linked to the corresponding care plan item and patient record. This makes the entire MDT process completely transparent and traceable, allowing managers to clearly see the plan's execution status and any potential delays.

[0088] The core of dynamic adjustment is "prediction-actual" difference analysis. The system compares the latest actual patient outcomes with previous predictions. The plan generation module contains specialized logic to analyze the meaning of these differences. For example, if the actual pain score decreases much less than the predicted decrease, the module identifies this as a "negative difference" and searches the rule base for rules to address it, generating adjustment suggestions such as "recommend reassessing the analgesia regimen." This demonstrates the system's advanced intelligence, moving from static prediction to dynamic response.

[0089] During the training phase of model development, if certain types of patients (e.g., "young patients with large wounds") are found to be very rare, data augmentation techniques are employed to ensure the model can accurately predict even these rare cases. This technique does not simply replicate data; instead, it analyzes the deep feature distributions of these rare samples and then generates entirely new, dummy sample data based on these distributions. This synthetic data is statistically similar to real rare samples, thus effectively expanding the training dataset and improving the model's robustness without introducing bias.

[0090] Towards the end of the study, researchers no longer need to manually collect data or operate statistical software. They can simply initiate the analysis task with a single click on the system interface. The system automatically extracts all preset outcome indicators for the treatment and control groups from the database. Then, it automatically performs tests for normality and homogeneity of variance, intelligently selecting the appropriate statistical method (parametric or nonparametric) based on the test results. Next, the system calls its built-in statistical calculation engine to execute these tests and automatically populates the calculation results (statistics, p-values, etc.) into a pre-prepared research report template, instantly generating a preliminary, academically compliant statistical analysis report, greatly improving research efficiency.

[0091] This embodiment, through the extremely detailed implementation described above, clearly demonstrates how artificial intelligence technology can be deeply embedded into every aspect of MDT management, constructing a complete closed-loop system from data perception to intelligent decision-making, and then to execution monitoring and dynamic optimization. This method not only significantly improves the efficiency and quality of medical services, but more importantly, it provides a data-driven, scientifically precise, and personalized medical management paradigm, exhibiting remarkable advancement and innovation.

[0092] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0093] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0094] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A pain management method for patients with gouty arthritis based on an MDT (Multidisciplinary Team) management model, characterized in that, include: S1. Collect clinical data of patients with gouty arthritis through an electronic health record system. The clinical data includes the patient's demographic information, laboratory test results, imaging test results, wound characteristics information, and pain scores. S2, The clinical data is cleaned and standardized using an artificial intelligence data preprocessing module to generate a standardized dataset; S3, The standardized dataset is input into a trained machine learning model to predict the progress of the patient's wound healing and the trend of pain changes. The machine learning model is trained based on historical data and includes a regression model and a classification model. S4. Based on the output of the machine learning model, a personalized MDT care plan is generated, which involves a multidisciplinary team including rheumatologists, wound and ostomy specialists, and caregivers, and specifies care measures and schedules. S5, push the personalized MDT care plan to the multidisciplinary team and patients through a mobile healthcare application, and monitor the implementation of the care plan; S6. Regularly collect subsequent clinical data, use the machine learning model to reassess wound healing progress and pain trends, and dynamically adjust the personalized MDT care plan.

2. The method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, The wound feature information in the clinical data in S1 includes wound size, granulation tissue coverage, infection status, and exudate volume. The pain score is collected using the Visual Analogue Scale (VAS) or the Numerical Rating Scale (NRS) and input into the electronic health record system via sensor devices or patient self-reporting.

3. The method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, In step S2, the steps of cleaning and standardizing the clinical data by the artificial intelligence data preprocessing module include: Detect missing values ​​and fill them using multiple imputation or k-nearest neighbor-based imputation methods; Perform one-hot encoding or label encoding on categorical variables, and normalize or standardize continuous variables to make the data meet the input requirements of machine learning models. Use anomaly detection algorithms, including isolated forest or local outlier algorithm, to identify and process outlier data points; The data is divided into a training set and a test set, where the training set is used for model training and the test set is used for model validation. Feature selection methods, including recursive feature elimination or tree-based feature importance assessment, are applied to select a subset of features relevant to wound healing and pain management.

4. The method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, The training steps for the machine learning model in S3 include: Collect historical clinical datasets, including clinical data, wound healing outcomes, and pain management effects from multiple patients with gouty arthritis; The historical clinical dataset was annotated, where wound healing results included wound shrinkage rate, granulation tissue coverage time, and healing effectiveness, and pain management results included changes in pain scores and patient satisfaction. An initial model is trained using a supervised learning algorithm, including random forest, support vector machine, or gradient boosting decision tree, to predict wound healing progress and pain change trends. Cross-validation was used to evaluate model performance, and hyperparameters were adjusted to optimize model accuracy and generalization ability. Deploy pre-trained machine learning models to cloud servers or edge computing devices for real-time or batch prediction; The machine learning model is periodically incrementally learned or retrained using newly collected clinical data to maintain its timeliness.

5. A method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, In step S4, the step of generating a personalized MDT care plan includes: Based on the predicted output of the machine learning model, the optimal care measures for the patient are determined, including moist care measures, debridement, dressing selection, dietary advice, and health education content. Natural language processing technology is used to generate structured care plan documents, including care steps, frequency, and expected goals; The care plan is assigned to the appropriate members of the multidisciplinary team, and task execution is coordinated through a workflow management system.

6. A method for pain management in patients with gouty arthritis based on an MDT management model according to claim 5, characterized in that, The moist care measures include: In wound care, iodine disinfection, wet dressings, sodium bicarbonate irrigation, and silver ion dressings are used. The nursing care steps are dynamically adjusted according to the wound condition, and the adjustment of the nursing care steps is based on the results of convolutional neural network analysis of wound images by a machine learning model.

7. A method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, In S5, the mobile medical application includes a patient client and a doctor client: The patient-side app is used to collect patients' self-reported pain scores, wound photos, and dietary records, and to promote treatment adherence through push notifications. The physician's side is used to display care plans, patient data, and model prediction results, and provides decision support tools for adjusting care plans.

8. A method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, In step S6, the step of periodically collecting subsequent clinical data includes: Using IoT devices to automatically collect wound images and physiological parameters, such IoT devices include smart dressings or portable cameras; The collected data is transmitted to a central database via the application programming interface (API) for model re-evaluation.

9. A method for pain management in patients with gouty arthritis based on an MDT management model according to claim 1, characterized in that, The method also includes a sample size optimization step: Generative adversarial networks (GANs) or synthetic minority oversampling techniques (SMOTE) are used to generate synthetic clinical data to augment the training dataset. The optimal sample size is determined using Bayesian optimization methods for model training and validation. The statistical analysis method is integrated into the artificial intelligence system: The collected data is processed using an automated statistical analysis module, which performs t-tests, χ² tests, or nonparametric tests. The analysis results are presented in a visual format, including charts and dashboards, to support decision-making by multidisciplinary teams.

10. A pain management system for patients with gouty arthritis based on an MDT (Multidisciplinary Team) management model, characterized in that, include: One or more processors; A memory having stored one or more programs that, when executed by one or more processors, cause the one or more processors to implement a pain management method for gouty arthritis patients based on an MDT management model according to any one of claims 1 to 9.

Images