WCBA-based nurse occupational burnout early warning and toughness improving system and method thereof

Abstract

The invention relates to the technical field of medical health information and digital psychological intervention, in particular to a WCBA-based nurse occupational burnout early warning and toughness improving system and method, and the system comprises a data collection and anonymization module which is used for collecting information from multi-source data and generating an anonymous data set; the risk prediction and portrait generation module is used for training a tiredness risk prediction model by utilizing a federated learning technology and generating an individual risk portrait containing a multi-dimensional risk level based on the tiredness risk prediction model; the WCBA intervention resource library management module is used for constructing a digital intervention resource library with a preset tiredness dimension label; and the personalized scheme generating and pushing module is used for matching the individual risk portrait with the resource library label, and generating and pushing a personalized WCBA intervention scheme to a nurse terminal. According to the method and the system, the problems of one-step intervention measure and contradiction between data utilization and privacy protection in the prior art are solved, and accurate early warning and personalized toughness improvement are realized.

Description

Technical Field

[0001] This invention relates to the fields of medical and health information technology and digital psychological intervention technology, and in particular to a nurse burnout early warning and resilience enhancement system and method based on WCBA. Background Technology

[0002] Nursing burnout has become a global public health problem affecting the quality of healthcare and the stability of the nursing workforce. It is mainly manifested as emotional exhaustion, depersonalization, and decreased personal achievement, severely damaging nurses' physical and mental health and directly leading to a decline in nursing quality, an increase in medical error rates, and high staff turnover. To address this challenge, the profession is actively exploring a shift from passive psychological assistance to proactive resilience building and early warning systems. Among these, mindfulness-based cognitive behavioral therapy and other psychological interventions have been proven to effectively improve individuals' psychological resilience and stress coping abilities.

[0003] However, most existing solutions adopt a "one-size-fits-all" approach, failing to accurately assess nurses' individual risk status in real time and across multiple dimensions, and also struggling to match them with resilience-enhancing resources that truly meet their current psychological needs. This results in weak targeting of interventions, low compliance, and ultimately affects the overall effectiveness. Therefore, how to build an intelligent system capable of accurate early warning and personalized intervention while ensuring nurses' privacy and security has become a key technical problem that needs to be solved in this field. Summary of the Invention

[0004] To address the above shortcomings, this invention provides a nurse burnout early warning and resilience enhancement system and method based on WCBA, aiming to improve the existing technology's inability to accurately assess the real-time, multi-dimensional risk status of individual nurses, and its difficulty in matching them with resilience enhancement resources that truly meet their current psychological needs.

[0005] In a first aspect, the present invention provides the following technical solution: a nurse burnout early warning and resilience enhancement system based on WCBA, comprising the following modules: The data acquisition and anonymization module collects occupation-related data of nurses from multiple heterogeneous data sources, anonymizes all collected data, and generates an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information. The risk prediction and profile generation module uses the anonymous dataset as training samples and employs federated learning technology to train a preset burnout risk prediction model. Using the trained burnout risk prediction model, it calculates the risk profile of each anonymous individual in the anonymous dataset and outputs the corresponding individual risk profile. The individual risk profile includes the risk level of the individual in multiple preset burnout dimensions. The WCBA intervention resource library management module pre-builds a digital WCBA intervention resource library, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention target. The personalized solution generation and push module matches the individual risk profile with the preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, it generates a personalized WCBA intervention plan for each anonymous individual and pushes the specific WCBA intervention modules contained in the plan to the nurse terminal corresponding to the anonymous individual.

[0006] Preferably, the process for generating the anonymous dataset includes: Nurses' shift scheduling and seniority data are extracted from the human resources management system; nurses' workload data for patients under their care is extracted from the hospital information system; and nurses' physiological signal data are collected from wearable devices. The extracted data are cleaned to remove invalid and outlier values, and then standardized according to a unified time window and data format. Generate a globally unique, meaningless anonymized identifier for each nurse, and use this anonymized identifier to replace all direct or indirect personal identification information in all data; The standardized scheduling data is associated and aggregated with seniority data, nursing workload data, and physiological signal data to form an anonymous, structured dataset. Each record in this dataset corresponds to an anonymization identifier, and the records only retain department, seniority attributes, and various occupation-related data, without containing any identity information that can directly locate a specific individual.

[0007] Preferably, the training process for the burnout risk prediction model includes: Based on the preset model type and structure, a global burnout risk prediction model is initialized on the central server. The current parameters of the burnout risk prediction model are distributed to the local clients of each data source participating in federated learning; Each local client uses its local anonymous dataset to train the received model parameters and calculate the updated values ​​of the local model parameters. Each local client encrypts the calculated local model parameter update value and uploads it to the central server. The central server decrypts and aggregates the received local update values, and uses the aggregation result to update the parameters of the burnout risk prediction model. Repeat the steps of parameter distribution, value calculation, and parameter update of the burnout risk prediction model until the performance index of the global burnout risk prediction model reaches the preset convergence threshold, and the training is completed.

[0008] Preferably, the calculation process for the anonymous individual includes: The trained burnout risk prediction model is deployed on a pre-defined model inference server. From the anonymous dataset, multi-dimensional feature data belonging to the same anonymous individual and spanning different data sources are extracted according to the anonymization identifier and assembled into the complete feature vector of the anonymous individual; The complete feature vector of each anonymous individual after assembly is input into the burnout risk prediction model for batch calculation; The system receives the calculation results output by the burnout risk prediction model for each anonymous individual. The calculation results include predicted values ​​for different preset burnout dimensions. Based on the predicted values, the system generates an individual risk profile for each anonymous individual, which includes a risk level assessment.

[0009] Preferably, the process for generating the individual risk profile includes: For each preset burnout dimension, the original predicted value output by the burnout risk prediction model is mapped to a unified preset risk assessment scale to obtain the corresponding scale score. The scale score is compared with a preset risk level threshold range for different burnout dimensions; Based on the comparison results, the risk level of the anonymous individual in each dimension of burnout was determined; The anonymized identifier of the anonymous individual, its risk level on each preset burnout dimension, and the corresponding scaling score are collectively encapsulated to generate a structured individual risk profile.

[0010] Preferably, the configuration process for the preset burnout dimension tags includes: Based on the core elements of mindfulness-based cognitive behavioral therapy, psychological intervention content is designed for each of the pre-defined multiple dimensions of burnout, and the psychological intervention content is produced in digital media format. The psychological intervention content targeting specific burnout dimensions or specific intervention goals is packaged into an independent, executable WCBA intervention module; Each WCBA intervention module is labeled with its specific preset burnout dimension tag; All WCBA intervention modules with dimension labels are stored in a pre-built digital WCBA intervention resource library, and an index is created for all modules for subsequent retrieval and retrieval.

[0011] Preferably, the process of matching the individual risk profile with the preset burnout dimension tags in the WCBA intervention resource library includes: Read the individual risk profile and analyze the risk level to which the anonymous individual belongs in each preset burnout dimension; According to the preset matching rules, the parsed risk level is matched with the preset burnout dimension label of the WCBA intervention module in the WCBA intervention resource library; Based on the matching results, all WCBA intervention modules that match the individual risk profile are selected from the WCBA intervention resource library to form a candidate intervention module set; Based on the severity of the risk levels of different burnout dimensions in the individual risk profile, the modules in the candidate intervention module set are prioritized to generate an ordered sequence of intervention modules.

[0012] Preferably, the process for generating the personalized WCBA intervention plan includes: From the priority-sorted sequence of intervention modules, a preset number of WCBA intervention modules are selected sequentially as the core intervention content; Based on the core intervention content and combined with the preset push frequency, an intervention schedule containing the execution order and push time points is generated for the anonymous individual. The core intervention content and the intervention schedule are encapsulated together to generate a structured, personalized WCBA intervention plan; Based on the preset mapping relationship between the anonymized identifier of the anonymous individual and the nurse terminal, the target nurse terminal is determined, and the modules in the personalized WCBA intervention plan are pushed to the target nurse terminal according to its schedule.

[0013] Secondly, this invention provides the following technical solution: a method for early warning and resilience enhancement of nurses' occupational burnout based on WCBA, the method comprising the following steps: We collected occupation-related data from multiple heterogeneous data sources for nurses, and anonymized all the collected data to generate an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information. Using the anonymized dataset as training samples, a pre-defined burnout risk prediction model is trained using federated learning technology. The trained burnout risk prediction model is then used to calculate the risk profile of each anonymous individual in the anonymized dataset, which includes the individual's risk level across multiple pre-defined burnout dimensions. A digital WCBA intervention resource library is pre-built, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention goal. The individual risk profile is matched with the preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, a personalized WCBA intervention plan is generated for each anonymous individual, and the specific WCBA intervention modules contained in the plan are pushed to the nurse terminal corresponding to the anonymous individual.

[0014] The present invention has the following beneficial effects: 1. In this invention, by adopting federated learning technology and the anonymization process of the system, it is possible to comprehensively utilize multi-dimensional heterogeneous data such as human resources, clinical work and physiological signals without centrally holding or transmitting any nurses' personal identity information. This fundamentally eliminates the risk of sensitive information leakage, resolves the inherent contradiction between medical data utilization and privacy protection, and provides a foundation for building a trustworthy intelligent management system.

[0015] 2. In this invention, by training a preset burnout risk prediction model and generating a structured individual risk profile, this solution transforms the traditional subjective and static assessment into an objective and quantitative dynamic monitoring. The system can accurately identify the risk level of each nurse in different burnout dimensions, thereby achieving early warning from group generalization to individual precision and realizing early intervention.

[0016] 3. In this invention, by intelligently matching and prioritizing individual risk profiles with the WCBA intervention resource library with clear dimensional labels, a customized intervention plan can be automatically generated for each nurse. This ensures that psychological intervention resources are highly aligned with the individual's real-time psychological needs, significantly improving the pertinence, compliance, and final effect of the intervention, and effectively avoiding resource waste.

[0017] 4. In this invention, data perception, intelligent analysis, decision-making and action execution are organically integrated to form a complete automated closed loop of assessment-early warning-intervention. This system can not only discover problems, but also automatically provide solutions and deliver them, greatly reducing the reliance on manual judgment, improving the efficiency and reliability of large-scale applications, and providing medical institutions with a systematic tool for the mental health management of nurses. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the architecture of the nurse burnout early warning and resilience enhancement system based on WCBA proposed in this invention; Figure 2 This is a flowchart illustrating the method for early warning and resilience enhancement of nurses based on WCBA proposed in this invention. Detailed Implementation

[0019] The technical solutions in 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: In a first embodiment of the present invention, the present invention provides a nurse burnout early warning and resilience enhancement system based on WCBA, such as Figure 1 As shown, it includes the following modules: The data acquisition and anonymization module collects occupation-related data from multiple heterogeneous data sources, anonymizes all collected data, and generates an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information.

[0021] Furthermore, the process for generating anonymized datasets includes: Nurses' shift scheduling and seniority data are extracted from the human resources management system; nurses' workload data for patients under their care is extracted from the hospital information system; and nurses' physiological signal data are collected from wearable devices. The extracted data are cleaned to remove invalid and outlier values, and then standardized according to a unified time window and data format. Generate a globally unique, meaningless anonymized identifier for each nurse, and use this anonymized identifier to replace all direct or indirect personal identification information in all data; The standardized scheduling data is linked and aggregated with seniority data, nursing workload data, and physiological signal data to form an anonymous, structured dataset. Each record in this dataset corresponds to an anonymous identifier, and the records only retain department, seniority attributes, and various occupation-related data, without containing any identity information that can directly locate a specific individual.

[0022] Specifically, by calling the API interface provided by the HR system or directly accessing its database view, the system extracts each nurse's shift scheduling data (such as shift E (day shift) / N (night shift), working hours, overtime hours) and seniority data. Through the hospital information integration platform or by directly connecting to the HIS database, it extracts nursing workload data related to the nurses' workload. This may include the number of patients cared for within a specific time period (such as a shift), patient severity scores (such as using the APACHE II score), and the frequency of nursing procedures performed. During data extraction, only the nurse's work ID is associated; sensitive information such as patient names is not obtained. Nurses voluntarily wear smart bracelets or other devices. The raw physiological signal data collected by the devices (such as photoplethysmography (PPG) signals) is first preliminarily calculated on the device or mobile app to generate derived indicators characterizing stress and recovery levels, such as time-domain indicators of heart rate variability (HRV) (such as SDNN) and sleep stage data (deep sleep, light sleep, REM sleep, wakefulness duration). These derived indicators are then uploaded to a designated secure receiving endpoint via a wireless network. Due to the diverse sources of data, cleaning and standardization are necessary to ensure data quality. The raw data is inspected, and procedural rules are used to remove invalid values ​​(such as signal loss segments in HRV data) and outliers (such as physiological data or work duration data that deviate significantly from the normal range and are identified by the 3σ principle or box plot method). The cleaned data is then aggregated according to a unified time window (e.g., based on shifts or calendar days). At the same time, data with different dimensions are standardized (e.g., using Z-Score standardization or Min-Max normalization) to transform them into a unified numerical range, eliminating the influence of dimensions and preparing for subsequent model training. For each nurse recorded in the HR system, the system generates a globally unique, meaningless random string as an anonymization identifier (for example, using the UUIDv4 algorithm to generate an identifier like f47ac10b-58cc-4372-a567-0e02b2c3d479). A temporary identity-anonymous ID mapping table with strict access control is established in memory. Then, using this mapping table, all direct personal identification information (such as name, employee number, ID card number) and indirect personal identification information (such as mobile phone MAC address, device ID) in all standardized data records are replaced with the corresponding anonymization identifier. After this process is completed, the temporary mapping table is destroyed to ensure that it is impossible to trace individuals back from the data. Using an anonymized identifier as the unique association key, scheduling data, seniority data, nursing workload data, and physiological signal data from different data sources belonging to the same anonymous individual are associated and spliced ​​to form a complete feature record spanning multiple dimensions. The feature records of all anonymous individuals are collected to form a structured anonymous dataset (e.g., stored as a Parquet file or in a dedicated table in a database). Each record in this dataset corresponds to an anonymized identifier, and the records only retain the two necessary group attributes of department and seniority, as well as various desensitized occupation-related data fields, completely excluding any identity information that can directly locate a specific individual. By following the steps above, we can improve the accuracy of early warning by fully integrating multi-source data, while fundamentally eliminating the risk of leakage of nurses' personal privacy information.

[0023] The risk prediction and profile generation module uses the anonymous dataset as training samples and employs federated learning technology to train a pre-defined burnout risk prediction model. Using the trained burnout risk prediction model, it calculates the risk profile of each anonymous individual in the anonymous dataset and outputs the corresponding individual risk profile. The individual risk profile includes the risk level of the individual in multiple pre-defined burnout dimensions.

[0024] Furthermore, the training process for the burnout risk prediction model includes: Based on the preset model type and structure, a global burnout risk prediction model is initialized on the central server. The current parameters of the burnout risk prediction model are distributed to the local clients of each data source participating in federated learning. Each local client uses its local anonymous dataset to train the received model parameters and calculate the updated values ​​of the local model parameters. Each local client encrypts the calculated local model parameter update values ​​and uploads them to the central server. The central server decrypts and aggregates the received local update values ​​and uses the aggregation results to update the parameters of the burnout risk prediction model. Repeat the steps of parameter distribution, value calculation, and burnout risk prediction model parameter update until the performance index of the global burnout risk prediction model reaches the preset convergence threshold, and the training is completed.

[0025] Furthermore, the calculation process for anonymous individuals includes: The trained burnout risk prediction model is deployed on a pre-defined model inference server. From the anonymous dataset, extract multi-dimensional feature data belonging to the same anonymous individual across different data sources according to the anonymization identifier, and assemble them into the complete feature vector of that anonymous individual; The complete feature vector of each anonymous individual after assembly is input into the burnout risk prediction model for batch calculation; The system receives the calculation results from the burnout risk prediction model for each anonymous individual. The calculation results include predicted values ​​for different preset burnout dimensions. Based on the predicted values, an individual risk profile including a risk level assessment is generated for each anonymous individual.

[0026] Furthermore, the process for generating an individual risk profile includes: For each preset burnout dimension, the original predicted value output by the burnout risk prediction model is mapped to a unified preset risk assessment scale to obtain the corresponding scale score. The scaled scores are compared with preset risk level threshold ranges for different burnout dimensions; Based on the comparison results, the risk level of the anonymous individual in each dimension of burnout was determined; The anonymized identifier of the anonymous individual, its risk level on each preset burnout dimension, and the corresponding scaling score are combined to generate a structured individual risk profile.

[0027] Specifically, on the central server, a global model is initialized according to a preset model type and structure. In a preferred embodiment, this model is a logistic regression model with an initial weight vector. and bias terms Randomly initialized or set to zero, the central server will set the current global model parameters ( , The data is distributed to local clients of each data source participating in federated learning (i.e., data processing nodes of HR, HIS, and wearable devices). Each local client uses its local anonymized dataset to train the received model. Taking logistic regression as an example, the goal is to minimize the cross-entropy loss function. The loss function of the local client is... It can be represented as: ; in, It is the number of local samples. It is the true label of sample i (e.g., whether it is highly emotionally exhausted). These are the model predictions, and σ is the sigmoid function. It is the L2 regularization coefficient, which the client calculates to obtain the updated values ​​(gradients) of the local model parameters using gradient descent. and ; Each local client will calculate the updated value ( , The data is encrypted using homomorphic encryption and other techniques before being uploaded to the central server. The server decrypts the received multiple local update values ​​and aggregates them using a federated averaging algorithm to update the global model, the expression of which is: ; ; in, It's the learning rate. This is the total number of clients participating in this round of training. This indicates an assignment operation, that is, updating the variable on the left with the result of the calculation on the right; Repeat the above steps until the performance metrics of the global model (such as the loss function value decreasing to a stable level, or the AUC area reaching a preset convergence threshold, such as 0.85) are achieved. Training is then complete, and the final burnout risk prediction model is obtained. The trained final model is deployed on a high-performance model inference server. Then, multi-dimensional feature data belonging to the same individual are extracted from the anonymized dataset according to the anonymization identifier, and assembled into the complete feature vector of that individual. The feature vectors of all the assembled anonymous individuals are input in batches into the deployed model for computation. For each individual, the model outputs raw predicted values ​​in one or more dimensions. (For example, scores on the emotional exhaustion dimension) Depersonalization dimension score ); For each preset burnout dimension, the original predicted values ​​are... Mapping to a uniform 0-100 point pre-defined risk assessment scale, for example, using a linear transformation: ; in, and The minimum and maximum values ​​of the predicted values ​​for this dimension on the validation set; Scale fractions Compare with a preset risk level threshold range, for example: Low risk Medium risk High risk is ; Based on the comparison results, the risk level of the individual in each dimension is determined. Finally, the individual's anonymized identifier, risk level in each burnout dimension (e.g., {emotional exhaustion: medium risk, depersonalization: low risk}) and corresponding scaling score are encapsulated into a structured individual risk profile in JSON format. Through the above steps, anonymized multi-source data can be transformed into quantifiable and interpretable individual risk profiles, providing a direct basis for decision-making to achieve precise intervention.

[0028] The WCBA intervention resource library management module pre-builds a digital WCBA intervention resource library, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention goals.

[0029] Furthermore, the configuration process for preset burnout dimension tags includes: Based on the core elements of mindfulness-based cognitive behavioral therapy, psychological intervention content is designed for each of the pre-defined multiple dimensions of burnout, and the psychological intervention content is produced in digital media format. The psychological intervention content targeting specific burnout dimensions or specific intervention goals is packaged into independent, executable WCBA intervention modules; Each WCBA intervention module is labeled with its specific pre-defined burnout dimension tags; All WCBA intervention modules with dimension labels are stored in a pre-built digital WCBA intervention resource library, and an index is created for all modules for subsequent retrieval and retrieval.

[0030] Specifically, firstly, led by a team of psychology experts, content was created based on the core elements of Mindfulness-Based Cognitive Behavioral Therapy (WCBA) and targeting pre-defined dimensions of burnout (e.g., based on the three classic dimensions of the Masler Burnout Scale: emotional exhaustion, depersonalization, and low personal achievement). Specifically, for emotional exhaustion, content focused on body scans, mindful breathing, and loving-kindness meditation was designed to help nurses become aware of and soothe excessive emotional and physical exhaustion. For depersonalization, content focused on cognitive restructuring, gratitude exercises, and shared humanistic perspectives was designed to help nurses regain their sense of well-being. To establish new emotional connections with patients and work, and reduce feelings of indifference and alienation, the program addresses low personal achievement by focusing on strengths discovery, achievement recording, and positive recall. It aims to help nurses identify and reinforce positive experiences and self-worth at work. All psychological intervention content is produced in digital media formats, including but not limited to audio files (5-15 minute guided meditations and breathing exercises), video files (animations or short videos of expert lectures on stress management techniques), and graphic materials (interactive e-manuals or infographics introducing cognitive behavioral adjustment methods). The software development team encapsulates the designed digital content into independent, executable WCBA intervention modules. Each module is a self-contained functional unit, which can be implemented as a microservice interface or a content package containing metadata. Each module contains core media files (such as the audio, video, or text content mentioned above) and metadata description files. The metadata description files specifically describe the module's title, introduction, expected completion time (e.g., 10 minutes), and intervention type (e.g., mindfulness breathing and cognitive exercises). Each packaged WCBA intervention module is labeled with its specific preset burnout dimension tags. In the resource library's backend management interface, each module is provided with a multi-select box that is linked to the preset burnout dimension list. For example, module A (body scan audio) is labeled as "emotional exhaustion," module B (cognitive reconstruction graphics and text) is labeled as "depersonalization," and module C (strengths discovery practice) is labeled as "low personal achievement." A complex module may be labeled with multiple tags simultaneously, such as a comprehensive meditation module that can be labeled with both "emotional exhaustion" and "depersonalization." Finally, all WCBA intervention modules labeled with dimension tags are stored in a pre-built digital WCBA intervention resource library. This resource library can technically be a relational database or a NoSQL document database. Each module is stored in the database as a record, and its tags are queryable fields. The system will create database indexes for key fields such as burnout dimension tags, intervention type, and duration. This ensures that subsequent modules can quickly retrieve and filter matching intervention module sets at millisecond speeds based on the risk dimensions parsed from the individual risk profile. Through the above steps, professional WCBA psychological intervention resources can be digitally packaged and standardized, providing a resource foundation for accurate and efficient matching with individual risk profiles.

[0031] The personalized plan generation and push module matches the individual risk profile with the preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, it generates a personalized WCBA intervention plan for each anonymous individual and pushes the specific WCBA intervention modules contained in the plan to the nurse terminal corresponding to the anonymous individual.

[0032] Furthermore, the process of matching individual risk profiles with the pre-defined burnout dimension labels in the WCBA intervention resource library includes: Read the individual risk profile and analyze the risk level of the anonymous individual in each preset burnout dimension; According to the preset matching rules, the parsed risk level is matched with the preset burnout dimension label in the WCBA intervention module of the WCBA intervention resource library; Based on the matching results, all WCBA intervention modules that match the individual risk profile are selected from the WCBA intervention resource library to form a set of candidate intervention modules; Based on the severity of the risk levels of different burnout dimensions in the individual risk profile, the modules in the candidate intervention module set are prioritized to generate an ordered sequence of intervention modules.

[0033] Furthermore, the process for generating personalized WCBA intervention plans includes: From the priority-sorted sequence of intervention modules, a predetermined number of WCBA intervention modules are selected sequentially as the core intervention content; Based on the core intervention content and combined with the preset push frequency, an intervention schedule containing the execution order and push time points is generated for the anonymous individual. The core intervention content and intervention schedule are packaged together to generate a structured and personalized WCBA intervention plan; Based on the pre-defined mapping relationship between the anonymized identifier of the anonymous individual and the nurse terminal, the target nurse terminal is determined, and the modules in the personalized WCBA intervention plan are pushed to the target nurse terminal according to its schedule.

[0034] Specifically, the personalized solution generation and push module reads the structured individual risk profile (e.g., a JSON object) output by the risk prediction and profile generation module. By parsing the object, it obtains the risk level (e.g., high risk, medium risk, and low risk) of the anonymous individual in each preset burnout dimension (e.g., emotional exhaustion, depersonalization). The system matches the parsed risk level with the dimension labels of modules in the WCBA intervention resource library according to preset matching rules. In a preferred embodiment, the matching rules are a set of IF-THEN logic, where rule 1 is: if the risk level of a certain dimension is high, then THEN prioritize matching modules with the label of that dimension and the type of reinforcement intervention; rule 2 is: if the risk level of a certain dimension is medium, then THEN match modules with the label of that dimension and the type of basic intervention; rule 3 is: if the risk level of a certain dimension is low, then THEN can match modules with the label of that dimension and the type of preventive intervention, or temporarily not match if resources are sufficient. Based on the above matching results, the module initiates a query to the WCBA intervention resource library, filters out all matching WCBA intervention modules, and forms a candidate intervention module set. The system sorts the modules in the candidate set according to the severity of the risk level of different burnout dimensions in the individual risk profile. The sorting logic can be based on a preset weight table. For example, the module corresponding to the high-risk dimension has a weight of 3, the module corresponding to the medium-risk dimension has a weight of 2, and the module corresponding to the low-risk dimension has a weight of 1. The system calculates the comprehensive priority score of each module (for example, if a module targets two dimensions at the same time, the score is the sum of the weights of the two dimensions), and then sorts them from high to low scores to finally generate an ordered sequence of intervention modules. From the priority-sorted sequence, a preset number of WCBA intervention modules (e.g., the first 3 in the first week) are selected sequentially as the core intervention content of this personalized plan. Based on the selected core intervention content and combined with the preset push frequency (e.g., a new module is pushed every Monday, Wednesday, and Friday), a structured intervention schedule is generated for the anonymous individual. This schedule clarifies the push time and execution order of each module in the future. The core intervention content (module ID list) and the intervention schedule are encapsulated together to generate a structured personalized WCBA intervention plan. Technically, this plan can be a JSON object containing a complete execution plan. The system queries a secure and independent mapping table based on an anonymous identifier to determine the target nurse terminal (such as the device identification code of a bound smartphone app) corresponding to the identifier. This mapping table is physically isolated from the main business database containing sensitive information to further protect privacy. The system (or integrated push service) accurately pushes the corresponding WCBA intervention module (or its access link) to the target nurse terminal at a specified time according to the schedule in the plan. The nurse will receive a prompt on their device, such as "You have a new mindfulness practice to check", thus completing the entire closed loop. Through the above steps, the risk analysis results at the front end can be automatically transformed into executable and personalized action plans and accurately pushed out, forming a complete intelligent closed loop of assessment-early warning-intervention.

[0035] Example 2: Existing solutions mostly adopt a "one-size-fits-all" approach, failing to accurately assess nurses' individual risk status in real time and across multiple dimensions, and also struggling to match them with resilience enhancement resources that truly meet their current psychological needs. This results in weak intervention targeting, low compliance, and ultimately affects the overall effectiveness. Therefore, how to build an intelligent system capable of accurate early warning and personalized intervention while ensuring nurses' privacy and security has become a key technical problem that needs to be solved in this field. To address the above problems, this invention provides a method for early warning and resilience enhancement of nurses based on WCBA, the structure of which is as follows: Figure 2 As shown. The specific implementation process of this method is as follows: We collected occupation-related data from multiple heterogeneous data sources for nurses, and anonymized all the collected data to generate an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information. Using an anonymous dataset as training samples, a pre-defined burnout risk prediction model is trained using federated learning techniques. The trained burnout risk prediction model is then used to calculate the risk profile of each anonymous individual in the anonymous dataset, which includes the risk level of the individual across multiple pre-defined burnout dimensions. A digital WCBA intervention resource library is pre-built, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention goal. Individual risk profiles are matched with preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, a personalized WCBA intervention plan is generated for each anonymous individual, and the specific WCBA intervention modules contained in the plan are pushed to the nurse terminal corresponding to the anonymous individual.

[0036] Specifically, after system initialization, the data collection and anonymization process is initiated. The system asynchronously collects occupation-related data such as shift schedules, years of service, nursing workload, and physiological signals from multiple heterogeneous data sources, including the human resource management system, hospital information system, and wearable devices voluntarily worn by nurses, through a secure interface. All data is processed in a pipeline in memory. Specifically, the data is first cleaned and standardized. Then, the system generates a globally unique anonymized identifier (such as a UUID) for each nurse in real time, and uses this identifier to replace all direct or indirect personal identity information in all data records. After processing, an anonymous dataset that retains only the department, years of service attributes, and various desensitized occupational data is persistently stored, while the temporary identity mapping table is immediately destroyed, thus building a privacy protection barrier at the data source. Based on the aforementioned anonymized dataset, the system enters the risk prediction and profile generation stage. A pre-set machine learning model (such as logistic regression or a lightweight neural network) is trained using federated learning technology. During this process, the model's parameters, rather than the original data, are encrypted and aggregated between the central server and the local clients of each data source. After multiple iterations until the model performance converges, the model is deployed on a dedicated inference server. The system then extracts cross-data source features for each anonymous individual in the anonymized dataset and assembles them into feature vectors for batch computation. The original predicted values ​​output by the model (such as 0.8 representing a high risk probability) are post-processed and mapped to a unified pre-set risk assessment scale (such as 0-100 points). Finally, by comparing with a pre-set threshold, the risk level is transformed into a risk level on pre-set dimensions such as emotional exhaustion, depersonalization, and low personal achievement, and encapsulated into a structured individual risk profile. Based on the core elements of Mindfulness-Based Cognitive Behavioral Therapy (WCBA), psychology experts designed and produced digital media intervention content (such as audio, video, and text) targeting the aforementioned pre-defined burnout dimensions. The development team packaged this content into independent, executable WCBA intervention modules, and the system labeled each module with its specific pre-defined burnout dimension. All labeled modules were stored in a digital resource library and an efficient index was created for quick retrieval. Finally, the system generates and pushes personalized intervention plans. It reads the generated individual risk profile, analyzes its risk level in different dimensions, and intelligently filters all matching WCBA intervention modules from the resource library according to preset matching rules (such as matching enhanced intervention modules for high-risk dimensions), forming a candidate set. The system then prioritizes this set according to the severity of the risk level and selects a preset number of modules as core intervention content. Next, the system generates an intervention schedule containing the execution order and time points for each anonymous individual based on the preset push frequency, and encapsulates the content and schedule together into a personalized WCBA intervention plan. Finally, the system determines the target nurse's terminal through a secure mapping relationship between an anonymous identifier and the terminal device, and accurately delivers the specific modules in the plan to the nurse's mobile terminal according to the schedule via message push service.

[0037] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A nurse burnout early warning and resilience enhancement system based on WCBA, characterized in that, Includes the following modules: The data acquisition and anonymization module collects occupation-related data of nurses from multiple heterogeneous data sources, anonymizes all collected data, and generates an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information. The risk prediction and profile generation module uses the anonymous dataset as training samples and employs federated learning technology to train a preset burnout risk prediction model. Using the trained burnout risk prediction model, it calculates the risk profile of each anonymous individual in the anonymous dataset and outputs the corresponding individual risk profile. The individual risk profile includes the risk level of the individual in multiple preset burnout dimensions. The WCBA intervention resource library management module pre-builds a digital WCBA intervention resource library, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention target. The personalized solution generation and push module matches the individual risk profile with the preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, it generates a personalized WCBA intervention plan for each anonymous individual and pushes the specific WCBA intervention modules contained in the plan to the nurse terminal corresponding to the anonymous individual.

2. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 1, characterized in that, The process for generating the anonymous dataset includes: Nurses' shift scheduling and seniority data are extracted from the human resources management system; nurses' workload data for patients under their care is extracted from the hospital information system; and nurses' physiological signal data are collected from wearable devices. The extracted data are cleaned to remove invalid and outlier values, and then standardized according to a unified time window and data format. Generate a globally unique, meaningless anonymized identifier for each nurse, and use this anonymized identifier to replace all direct or indirect personal identification information in all data; The standardized scheduling data is associated and aggregated with seniority data, nursing workload data, and physiological signal data to form an anonymous, structured dataset. Each record in this dataset corresponds to an anonymization identifier, and the records only retain department, seniority attributes, and various occupation-related data, without containing any identity information that can directly locate a specific individual.

3. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 1, characterized in that, The training process for the burnout risk prediction model includes: Based on the preset model type and structure, a global burnout risk prediction model is initialized on the central server. The current parameters of the burnout risk prediction model are distributed to the local clients of each data source participating in federated learning; Each local client uses its local anonymous dataset to train the received model parameters and calculate the updated values ​​of the local model parameters. Each local client encrypts the calculated local model parameter update value and uploads it to the central server. The central server decrypts and aggregates the received local update values, and uses the aggregation result to update the parameters of the burnout risk prediction model. Repeat the steps of parameter distribution, value calculation, and parameter update of the burnout risk prediction model until the performance index of the global burnout risk prediction model reaches the preset convergence threshold, and the training is completed.

4. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 3, characterized in that, The calculation process for the anonymous individual includes: The trained burnout risk prediction model is deployed on a pre-defined model inference server. From the anonymous dataset, multi-dimensional feature data belonging to the same anonymous individual and spanning different data sources are extracted according to the anonymization identifier and assembled into the complete feature vector of the anonymous individual; The complete feature vector of each anonymous individual after assembly is input into the burnout risk prediction model for batch calculation; The system receives the calculation results output by the burnout risk prediction model for each anonymous individual. The calculation results include predicted values ​​for different preset burnout dimensions. Based on the predicted values, the system generates an individual risk profile for each anonymous individual, which includes a risk level assessment.

5. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 4, characterized in that, The process for generating the individual risk profile includes: For each preset burnout dimension, the original predicted value output by the burnout risk prediction model is mapped to a unified preset risk assessment scale to obtain the corresponding scale score. The scale score is compared with a preset risk level threshold range for different burnout dimensions; Based on the comparison results, the risk level of the anonymous individual in each dimension of burnout was determined; The anonymized identifier of the anonymous individual, its risk level on each preset burnout dimension, and the corresponding scaling score are collectively encapsulated to generate a structured individual risk profile.

6. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 1, characterized in that, The configuration process for the preset burnout dimension tags includes: Based on the core elements of mindfulness-based cognitive behavioral therapy, psychological intervention content is designed for each of the pre-defined multiple dimensions of burnout, and the psychological intervention content is produced in digital media format. The psychological intervention content targeting specific burnout dimensions or specific intervention goals is packaged into an independent, executable WCBA intervention module; Each WCBA intervention module is labeled with its specific preset burnout dimension tag; All WCBA intervention modules with dimension labels are stored in a pre-built digital WCBA intervention resource library, and an index is created for all modules for subsequent retrieval and retrieval.

7. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 1, characterized in that, The process of matching the individual risk profile with the preset burnout dimension tags in the WCBA intervention resource library includes: Read the individual risk profile and analyze the risk level to which the anonymous individual belongs in each preset burnout dimension; According to the preset matching rules, the parsed risk level is matched with the preset burnout dimension label of the WCBA intervention module in the WCBA intervention resource library; Based on the matching results, all WCBA intervention modules that match the individual risk profile are selected from the WCBA intervention resource library to form a candidate intervention module set; Based on the severity of the risk levels of different burnout dimensions in the individual risk profile, the modules in the candidate intervention module set are prioritized to generate an ordered sequence of intervention modules.

8. The nurse burnout early warning and resilience enhancement system based on WCBA according to claim 7, characterized in that, The process for generating the personalized WCBA intervention plan includes: From the priority-sorted sequence of intervention modules, a preset number of WCBA intervention modules are selected sequentially as the core intervention content; Based on the core intervention content and combined with the preset push frequency, an intervention schedule containing the execution order and push time points is generated for the anonymous individual. The core intervention content and the intervention schedule are encapsulated together to generate a structured, personalized WCBA intervention plan; Based on the preset mapping relationship between the anonymized identifier of the anonymous individual and the nurse terminal, the target nurse terminal is determined, and the modules in the personalized WCBA intervention plan are pushed to the target nurse terminal according to its schedule.

9. A method for early warning and resilience enhancement of nurses' occupational burnout based on WCBA, characterized in that, The method for the WCBA-based nurse burnout early warning and resilience enhancement system according to any one of claims 1-8 includes the following steps: We collected occupation-related data from multiple heterogeneous data sources for nurses, and anonymized all the collected data to generate an anonymous dataset that only contains department and length of service attributes and does not contain personal identity information. Using the anonymized dataset as training samples, a pre-defined burnout risk prediction model is trained using federated learning technology. The trained burnout risk prediction model is then used to calculate the risk profile of each anonymous individual in the anonymized dataset, which includes the individual's risk level across multiple pre-defined burnout dimensions. A digital WCBA intervention resource library is pre-built, which stores multiple independent WCBA intervention modules. Each WCBA intervention module is configured with a preset burnout dimension label to identify its intervention goal. The individual risk profile is matched with the preset burnout dimension tags in the WCBA intervention resource library. Based on the matching results, a personalized WCBA intervention plan is generated for each anonymous individual, and the specific WCBA intervention modules contained in the plan are pushed to the nurse terminal corresponding to the anonymous individual.

Images