A patient-oriented digital staff post-discharge follow-up management system and method

Abstract

The application relates to the technical field of follow-up management systems, in particular to a patient-oriented digital employee post-diagnosis follow-up management system and method, which comprises a patient data access and hierarchical label module, a digital employee interaction and follow-up task distribution module, a follow-up data collection and cleaning module, an AI rehabilitation risk assessment and early warning module, a personalized intervention scheme generation and execution module and a follow-up closed-loop management and effect review module; the system is collaboratively operated through the six modules. In the application, the collaborative linkage and closed-loop operation of the modules solve various problems existing in traditional post-diagnosis follow-up and existing systems.

Description

Technical Field

[0001] This invention relates to the field of follow-up management system technology, and in particular to a digital employee post-diagnosis follow-up management system and method for patients. Background Technology

[0002] As the medical service model shifts towards full-cycle health management, post-diagnosis follow-up has become a core component of postoperative rehabilitation, chronic disease management, and specialist treatment, directly impacting patient recovery outcomes and medical safety.

[0003] Traditional post-diagnosis follow-up relies on medical staff to make phone calls, text messages, and offline visits. This has problems such as high labor costs, follow-up coverage of less than 60%, high missed follow-up rate, low standardization of follow-up content, and fragmented patient information records, making it difficult to meet the needs of refined management of a large number of patients. Summary of the Invention

[0004] To address the technical problems of high labor costs and low follow-up coverage caused by traditional post-diagnosis follow-up relying on manual telephone calls by medical staff, this invention provides a digital employee post-diagnosis follow-up management system and method for patients.

[0005] The technical solution adopted in this invention is: a digital employee post-diagnosis follow-up management system for patients, including a patient data access and hierarchical tagging module, a digital employee interaction and follow-up task distribution module, a follow-up data collection and cleaning module, an AI rehabilitation risk assessment and early warning module, a personalized intervention plan generation and execution module, and a follow-up closed-loop management and effect review module; the system operates through the coordinated operation of these six modules. The system comprises the following modules: Patient Data Access and Hierarchical Tagging Module for accessing full-dimensional patient data, structured parsing, and automated generation of hierarchical tags; Digital Employee Interaction and Follow-up Task Distribution Module for driving digital employees to achieve multimodal interaction adaptation and automatically distributing follow-up tasks based on patient tags; Follow-up Data Collection and Cleaning Module for collecting full follow-up data in real time and forming structured data after standardized cleaning; AI Rehabilitation Risk Assessment and Early Warning Module for assessing patient rehabilitation risks in real time and providing graded early warnings based on standardized follow-up data; Personalized Intervention Plan Generation and Execution Module for automatically generating personalized intervention plans based on risk warning results and patient tags, with digital employees completing the entire execution process; and Follow-up Closed-Loop Management and Effect Review Module for summarizing follow-up data throughout the entire process, quantitatively evaluating follow-up effects, and completing process optimization and review.

[0006] In one embodiment, the patient data access and hierarchical labeling module connects to the hospital's HIS system, LIS laboratory system, EMR electronic medical record system, PACS imaging system, patient-side mini-program, and IoT health devices through standard interfaces, accessing four categories of comprehensive data: basic patient information, core diagnosis and treatment information, basic health information, and rehabilitation risk information. The patient data access and hierarchical labeling module is equipped with a medical-specific natural language processing model, which can perform structured parsing of unstructured discharge summaries, medical records, and medical orders, automatically extract key feature items, and automatically generate three categories of three-level hierarchical labels—core labels, risk labels, and follow-up labels—based on a specialist follow-up knowledge base and clinical guidelines. The labels can be dynamically updated according to changes in the patient's condition.

[0007] In one embodiment, the digital employee interaction and follow-up task distribution module uses a multimodal virtual medical assistant that integrates speech recognition, speech synthesis, text interaction, and virtual digital human video interaction. It is equipped with a medical-specific ASR recognition model and TTS synthesis engine, supports dialect recognition, emotion perception, and multi-turn contextual dialogue, and has an interaction latency of ≤300ms. The digital employee interaction and follow-up task distribution module can automatically adapt the optimal interaction method according to the patient's hierarchical tags. At the same time, through the task scheduling engine, it automatically generates a personalized follow-up task list based on the patient tags, specifying the follow-up time, follow-up content, response rules, and timeout handling strategies. The digital employee automatically pushes follow-up tasks according to task priority, supports automatic re-issuance of missed visits, priority push of urgent tasks, and silent delay during non-working hours.

[0008] In one embodiment, the follow-up data collection and cleaning module collects follow-up data in real time: data automatically recorded by digital employee interaction, data actively reported by patients, and data automatically synchronized by IoT devices; the follow-up data collection and cleaning module adopts a four-layer data cleaning algorithm to sequentially complete deduplication, missing value filling, outlier removal, and standardization conversion, converting unstructured voice and text data into a structured follow-up database in a unified format and storing it in the system data platform.

[0009] In one embodiment, the follow-up data collection and cleaning module introduces a patient follow-up priority quantification formula: P = α × S + β × R + γ × D; where P is the comprehensive score of patient follow-up priority, with higher scores indicating higher priority; α is the risk weight coefficient, ranging from 0.4 ≤ α ≤ 0.6, preset by the clinical follow-up standards of each specialty; S is the patient risk level score, with low risk = 1, medium risk = 2, high risk = 3, and very high risk = 4, automatically assigned by patient stratification labels; β is the risk weight coefficient. The weighting coefficient for each recovery stage ranges from 0.2 to 0.3; R is the patient's recovery stage score, with 3 for the acute phase, 2 for the recovery phase, and 1 for the stable phase, assigned by the patient's stratification label; γ is the follow-up response weighting coefficient, with 0.1 to 0.3; D is the patient's historical follow-up response score, with 0 for no response, 1 for partial response, and 2 for complete response, automatically calculated from historical follow-up data; the system classifies patients into four follow-up levels—Special, Level 1, Level 2, and Level 3—based on the P-value and matches corresponding follow-up resources accordingly.

[0010] In one embodiment, the AI ​​rehabilitation risk assessment and early warning module has a built-in specialized risk assessment model trained on clinical cases, covering common postoperative diseases. It can conduct 24-hour uninterrupted intelligent assessment of patient rehabilitation risks based on standardized follow-up data. The AI ​​rehabilitation risk assessment and early warning module divides the early warning level into four levels: red, orange, yellow, and blue. When the corresponding early warning level is reached, a dual early warning mechanism is triggered: the digital staff pushes early warning reminders and temporary emergency guidance to the patient, and at the same time sends early warning notifications to the responsible medical staff through system pop-ups, text messages, and telephone calls. The red early warning reaches the patient within 10 seconds.

[0011] In one embodiment, the personalized intervention plan generation and execution module has a built-in specialist intervention knowledge base covering six major categories: medication adjustment, rehabilitation training, dietary guidance, psychological counseling, follow-up appointment, and emergency treatment. The personalized intervention plan generation and execution module can automatically generate a unique personalized intervention plan based on the patient's risk level, disease type, rehabilitation stage, follow-up data, and early warning results, through rule matching and intelligent recommendation algorithms, specifying the intervention content, intervention frequency, execution method, expected goals, and precautions.

[0012] In one embodiment, the follow-up closed-loop management and effect review module summarizes core indicators such as follow-up coverage rate, task completion rate, risk warning accuracy rate, intervention effectiveness rate, patient compliance, improvement of rehabilitation indicators, and medical staff response time, and introduces a quantitative evaluation formula for follow-up effect: E=δ×C+ε×A+ζ×Rc; where E is the comprehensive score of follow-up effect, and the higher the score, the better the follow-up management effect; δ is the compliance weight coefficient, with a value range of 0.3≤δ≤0.4; and C is the patient follow-up compliance rate. The score is calculated by comprehensively considering the follow-up task completion rate, medication adherence, rehabilitation training execution rate, and follow-up examination attendance rate; ε is the effective weighting coefficient for early warning, with a value range of 0.3≤ε≤0.4; A is the risk warning accuracy score, calculated from the true positive rate, missed judgment rate, and false judgment rate of early warning; ζ is the rehabilitation improvement weighting coefficient, with a value range of 0.2≤ζ≤0.3; Rc is the patient's rehabilitation indicator improvement score, comprehensively assessed by the degree of symptom relief, the rate of achievement of vital signs, the improvement of follow-up examination results, and the incidence of adverse events.

[0013] In one embodiment, the follow-up closed-loop management and effect review module can automatically generate three-level follow-up review reports for single patients, departments, and the entire hospital based on the comprehensive follow-up effect score E; automatically adjust follow-up strategies and intervention plans for patients with low effects; automatically optimize algorithm models and knowledge bases for system process shortcomings; and automatically generate performance data for medical and nursing work to achieve continuous system iteration and upgrades; at the same time, the review data is synchronized to the hospital management platform.

[0014] In one embodiment, a digital employee post-diagnosis follow-up management method for patients specifically includes the following steps: Data access and tag generation: Access patients' full-dimensional diagnosis and treatment and health data, parse unstructured text through NLP models, and automatically generate three types of hierarchical tags: core, risk, and follow-up. Interaction adaptation and task distribution: Digital staff adapts the optimal multimodal interaction method based on patient stratification tags, automatically generates personalized follow-up tasks, pushes them out according to priority, and handles missed visits, urgent tasks, and special circumstances. Data collection and cleaning: Real-time collection of follow-up data from digital employee interactions, patient self-reporting, and IoT device synchronization, followed by deduplication, completion, anomaly removal, and standardization to form a structured database; Risk assessment and early warning: Based on standardized follow-up data, the risk of patient recovery is assessed in real time through a specialized risk assessment model, and four levels of early warning are defined and a dual early warning mechanism is triggered. Intervention plan generation and execution: Based on risk warning results and patient tags, personalized intervention plans are automatically generated and executed by digital staff throughout the process. For patients with red alerts, offline medical staff are mobilized for collaborative intervention. Closed-loop management and debriefing optimization: Summarize the core indicators of the entire follow-up process, evaluate the follow-up effect through quantitative formulas, generate debriefing reports, and automatically adjust follow-up strategies and optimize the system.

[0015] The beneficial effects of this invention are as follows: Compared with the prior art, this invention solves various problems existing in traditional post-diagnosis follow-up and current systems through the coordinated linkage and closed-loop operation of modules. Firstly, it replaces the traditional manual follow-up mode, reducing the need for manual intervention by medical staff and solving the problems of low efficiency, high cost, and non-standardized follow-up. Secondly, through fully automated management, it breaks down the barriers of independent operation of each module, solving the problems of fragmented follow-up work and the inability to iterate processes. Thirdly, the digital staff achieves multimodal interaction adaptation, meeting the needs of different patient groups and solving the problems of single interaction methods, poor patient adaptability, and low compliance of existing digital staff. Finally, through automated patient stratification, personalized intervention plan generation, and real-time risk warning, it solves the problems of insufficient personalization of follow-up plans and delayed risk warnings. Attached Figure Description

[0016] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the management method steps in this invention; Figure 3 This is a flowchart of the patient data access and hierarchical tagging module in this invention; Figure 4 This is a flowchart of the digital employee interaction and follow-up task distribution module in this invention. Detailed Implementation

[0017] In the description of this invention, it should be noted that the terms "front", "up", "down", "left", "right", "vertical", "horizontal", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0018] To address the problems existing in the background technology, this application proposes the following technical solution: a digital employee post-diagnosis follow-up management system for patients, including a patient data access and hierarchical tagging module, a digital employee interaction and follow-up task distribution module, a follow-up data collection and cleaning module, an AI rehabilitation risk assessment and early warning module, a personalized intervention plan generation and execution module, and a follow-up closed-loop management and effect review module; the system operates through the coordinated operation of these six modules. The system comprises the following modules: Patient Data Access and Hierarchical Tagging Module for accessing full-dimensional patient data, structured parsing, and automated generation of hierarchical tags; Digital Employee Interaction and Follow-up Task Distribution Module for driving digital employees to achieve multimodal interaction adaptation and automatically distributing follow-up tasks based on patient tags; Follow-up Data Collection and Cleaning Module for collecting full follow-up data in real time and forming structured data after standardized cleaning; AI Rehabilitation Risk Assessment and Early Warning Module for assessing patient rehabilitation risks in real time and providing graded early warnings based on standardized follow-up data; Personalized Intervention Plan Generation and Execution Module for automatically generating personalized intervention plans based on risk warning results and patient tags, with digital employees completing the entire execution process; and Follow-up Closed-Loop Management and Effect Review Module for summarizing follow-up data throughout the entire process, quantitatively evaluating follow-up effects, and completing process optimization and review.

[0019] In this embodiment, the patient data access and hierarchical labeling module connects with the hospital's HIS system, LIS laboratory system, EMR electronic medical record system, PACS imaging system, patient-side mini-program, and IoT health devices through standard interfaces. It accesses four categories of comprehensive data: basic patient information, core diagnosis and treatment information, basic health information, and rehabilitation risk information. The patient data access and hierarchical labeling module is equipped with a medical-specific natural language processing model, which can perform structured parsing of unstructured discharge summaries, medical records, and medical orders, automatically extract key feature items, and automatically generate three categories of three-level hierarchical labels: core labels, risk labels, and follow-up labels, based on a specialist follow-up knowledge base and clinical guidelines. The labels can be dynamically updated according to changes in the patient's condition.

[0020] First, the system seamlessly integrates with hospital HIS systems, LIS laboratory systems, EMR electronic medical record systems, PACS imaging systems, patient-side mini-programs, and IoT health devices through standard interfaces. The accessed data is categorized into four main types: basic patient information (age, gender, contact information, residential address, emergency contact person), core treatment information (disease diagnosis, type of surgery, medication regimen, discharge date, follow-up examination items), basic health information (past medical history, allergy history, family medical history, self-care ability), and rehabilitation risk information (complication risk level, chronic disease control level, follow-up visit requirements). Second, the system employs a medical-specific Natural Language Processing (NLP) model to perform structured parsing of unstructured discharge summaries, medical records, and medical orders, automatically extracting 28 key feature items such as rehabilitation cycle, risk factors, medication frequency, dietary and exercise restrictions, and symptom observation points, eliminating the need for manual data entry. Finally, based on its built-in specialist follow-up knowledge base and clinical guidelines, the system automatically generates three-tiered labels according to extracted features: core labels (disease type, recovery stage), risk labels (low / medium / high / very high risk), and follow-up labels (follow-up frequency, follow-up method, follow-up content, follow-up duration). The label generation process is fully automated and supports dynamic updates. When a patient's condition changes, the system automatically re-matches labels, solving the technical problems of existing systems that rely on manual patient classification, are inefficient, have inconsistent standards, and are outdated. This lays a data foundation for subsequent personalized follow-up task allocation.

[0021] This module achieves seamless integration with multiple systems through standard interfaces, accurately accessing four categories of comprehensive patient data, breaking down medical data silos and eliminating the need for manual data entry. It employs a medical-specific NLP model to perform structured parsing of unstructured medical text, automatically extracting 28 key feature items to ensure accurate and efficient data extraction. Based on a built-in specialist follow-up knowledge base and clinical guidelines, it automatically generates three-tiered labels (core, risk, and follow-up) through feature item matching. The label generation process is fully automated and dynamically updated according to changes in the patient's condition, achieving standardized and precise patient classification. This completely solves the problems of low efficiency and inconsistent standards in traditional manual classification, providing accurate and real-time data support for subsequent follow-up task distribution, risk assessment, and other modules, ensuring the orderly operation of the entire system.

[0022] In this embodiment, the digital employee interaction and follow-up task distribution module uses a multimodal virtual medical assistant that integrates speech recognition, speech synthesis, text interaction, and virtual digital human video interaction. It is equipped with a medical-specific ASR recognition model and TTS synthesis engine, supporting dialect recognition, emotion perception, and multi-turn contextual dialogue, with an interaction latency of ≤300ms. The digital employee interaction and follow-up task distribution module can automatically adapt the optimal interaction method according to the patient's hierarchical tags. At the same time, through the task scheduling engine, it automatically generates a personalized follow-up task list based on the patient tags, specifying the follow-up time, follow-up content, response rules, and timeout handling strategies. The digital employee automatically pushes follow-up tasks according to task priority, supporting automatic re-issuance of missed visits, priority push of urgent tasks, and silent delay during non-working hours.

[0023] Based on hierarchical tags, this invention drives a digital employee to complete multimodal interaction adaptation and automated distribution of follow-up tasks, distinguishing it from existing single-function digital employees. The digital employee of this invention is a multimodal virtual healthcare assistant integrating speech recognition, speech synthesis, text interaction, and virtual digital human video interaction. It is equipped with a medical-specific ASR recognition model and TTS synthesis engine, supporting 12 dialects, emotion perception, and multi-turn contextual dialogue, with interaction latency controlled within 300ms. The system automatically adapts the optimal interaction method based on the patient's hierarchical tags: high-definition voice interaction is prioritized for elderly patients over 60 years old and visually impaired patients; text and mini-program interaction are used for middle-aged and young patients and patients with high health literacy; and virtual digital human video interaction is used for patients with high post-operative rehabilitation guidance needs. Simultaneously, the system uses a task scheduling engine to automatically generate personalized follow-up task lists based on patient tags, specifying follow-up time, follow-up content (symptom inquiry, medication reminders, rehabilitation guidance, and follow-up notifications), response rules, and timeout handling strategies. Digital staff automatically push follow-up tasks to patients according to task priority, supporting automatic re-issuance of missed visits, priority push of urgent tasks, and silent delay during non-working hours. It realizes autonomous distribution of follow-up tasks and initial interactive response without the need for manual intervention by medical staff, solving the problems of the existing digital staff's single interaction method, low efficiency of manual allocation of follow-up tasks, and poor patient adaptability.

[0024] This module utilizes a medical-grade ASR recognition model and a TTS synthesis engine to construct a multimodal digital employee. It supports three interaction methods: voice, text, and virtual digital human video, and is compatible with 12 dialects for recognition and emotion perception. Interaction latency is controlled within 300ms, improving the interactive experience for different patient groups. Based on generated patient stratification tags, an intelligent adaptation algorithm matches the optimal interaction method for patients of different ages, health literacy levels, and rehabilitation needs, solving the problem of poor interaction adaptability. Through a task scheduling engine, personalized follow-up task lists are automatically generated based on patient tags, clearly defining the details of each task. The digital employee automatically pushes tasks according to priority, with functions such as missed follow-up reschedules, priority for urgent tasks, and silent delays during non-working hours. This achieves fully automated distribution of follow-up tasks and initial interactive responses, completely replacing manual task allocation, significantly improving follow-up efficiency, and reducing medical staff labor costs.

[0025] In this embodiment, the follow-up data collection and cleaning module collects follow-up data in real time: data automatically recorded by digital staff interaction, data actively reported by patients, and data automatically synchronized by IoT devices; the follow-up data collection and cleaning module adopts a four-layer data cleaning algorithm to complete deduplication, missing value filling, outlier removal, and standardization conversion in sequence, converting unstructured voice and text data into a structured follow-up database in a unified format and storing it in the system data platform.

[0026] In the follow-up data collection and cleaning module, a patient follow-up priority quantification formula is introduced: P=α×S+β×R+γ×D; where P is the comprehensive score of patient follow-up priority, with higher scores indicating higher priority; α is the risk weight coefficient, ranging from 0.4≤α≤0.6, preset by the clinical follow-up standards of each specialty; S is the patient risk level score, with low risk=1, medium risk=2, high risk=3, and very high risk=4, automatically assigned by patient stratification labels; β is the weight of the recovery stage. The coefficient ranges from 0.2 to 0.3; R is the patient's recovery stage score, with acute phase = 3, recovery phase = 2, and stable phase = 1, assigned by the patient stratification label; γ is the follow-up response weight coefficient, with a range of 0.1 to 0.3; D is the patient's historical follow-up response score, with no response = 0, partial response = 1, and complete response = 2, automatically calculated from historical follow-up data; the system classifies patients into four follow-up levels—special, level one, level two, and level three—based on the P-value and matches corresponding follow-up resources.

[0027] This module is responsible for collecting all follow-up data generated from interactions between digital staff and patients, and performing standardized cleaning and structured storage to provide high-quality data support for subsequent risk assessment. The system collects follow-up data in real time through three methods: first, automatic recording of data from digital staff interactions (patient responses, follow-up response time, task completion status); second, data proactively reported by patients (symptom feedback, vital signs, medication adherence); and third, automatic synchronization of data from IoT devices (real-time health data from devices such as smart bracelets, blood pressure monitors, and blood glucose meters). After collection, the system employs a four-layer data cleaning algorithm: first, deduplication to remove redundant data; second, missing value imputation to fill in reasonable missing items based on historical patient data and the average of similar patients; third, outlier removal to identify and remove erroneous data using medical threshold rules; and fourth, standardization conversion to transform unstructured voice and text data into a unified format structured follow-up database, which is then stored in the system's data platform.

[0028] This module achieves comprehensive follow-up data collection through three complementary methods, encompassing digital employee interaction records, patient-initiated reports, and IoT device synchronization, ensuring the comprehensiveness and real-time nature of data collection and eliminating any key follow-up information. A four-layer progressive data cleaning algorithm is employed, sequentially performing deduplication, missing value imputation, outlier removal, and standardization transformation. Missing value imputation combines historical patient data with the average of similar patients, while outlier removal follows medical threshold rules, ensuring the accuracy and standardization of the cleaned data. Unstructured data is converted into a structured database in a unified format and stored in the data platform. Simultaneously, a patient follow-up priority quantification formula is introduced. Based on tagged data and historical response data collected by this module, patient follow-up priorities are accurately calculated, automatically classifying patients into four follow-up levels and matching corresponding resources. This achieves rational allocation of follow-up resources, providing high-quality, directly usable data support for the subsequent risk assessment module and solving the problems of fragmented data collection and non-standardized cleaning in traditional methods.

[0029] In this embodiment, the AI ​​rehabilitation risk assessment and early warning module has a built-in specialized risk assessment model trained on clinical cases, covering common postoperative diseases. It can conduct 24-hour uninterrupted intelligent assessment of patient rehabilitation risks based on standardized follow-up data. The AI ​​rehabilitation risk assessment and early warning module divides the early warning level into four levels: red, orange, yellow, and blue. When the corresponding early warning level is reached, a dual early warning mechanism is triggered: the digital staff pushes early warning reminders and temporary emergency guidance to the patient, and at the same time sends early warning notifications to the responsible medical staff through system pop-ups, SMS, and telephone. The red early warning reaches the patient within 10 seconds.

[0030] Based on follow-up data, a deep learning risk assessment model is used to perform real-time assessment and graded early warning of patient rehabilitation risks, solving the problems of lagging risk warnings, high false negative rates, and reactive responses in existing systems. The system incorporates a specialized risk assessment model trained on millions of clinical cases, covering more than 50 common diseases such as postoperative rehabilitation, hypertension, diabetes, and cardiovascular and cerebrovascular diseases. The model combines real-time patient symptoms, vital signs, medication adherence, and behavioral data to provide 24 / 7 intelligent assessment of patient rehabilitation risks. The system classifies warning levels into four levels: Red Alert (extremely high risk, life-threatening, requiring immediate medical intervention), Orange Alert (high risk, possible complications, requiring enhanced follow-up), Yellow Alert (medium risk, requiring routine attention and adjustment of the treatment plan), and Blue Alert (low risk, normal follow-up as planned). When the assessment result reaches the corresponding warning level, the system immediately triggers a dual warning mechanism: first, digital staff immediately push warning reminders and temporary emergency guidance to the patient; second, warning notifications are sent to responsible medical staff through system pop-ups, SMS, and telephone, with red warnings reaching patients within 10 seconds, ensuring rapid transmission of warning information. This module differs from the manual post-event screening in existing systems, enabling real-time risk perception, automatic assessment, and proactive early warning, significantly reducing the incidence of adverse events in patients.

[0031] This module uses standardized follow-up data as its core input and incorporates a specialized risk assessment model trained on millions of clinical cases, covering over 50 common diseases to ensure the professionalism and accuracy of risk assessment. Through deep learning algorithms, the model continuously analyzes patients' real-time symptoms, vital signs, and medication adherence data 24 / 7, achieving real-time assessment of rehabilitation risks and overcoming the limitations of traditional post-event manual screening. Warning levels are divided into four categories: red, orange, yellow, and blue, clearly defining the risk level and handling requirements for each level, triggering a dual warning mechanism: firstly, digital staff push warning reminders and temporary emergency guidance to patients, enabling initial intervention at the patient's end; secondly, warning notifications are pushed to responsible medical staff through three methods, with red warnings reaching patients within 10 seconds, ensuring timely intervention in the management of high-risk patients. The entire module achieves real-time risk perception, automatic assessment, tiered warning, and rapid transmission, effectively solving the problems of delayed warnings and high missed detection rates, providing a guarantee for patient medical safety. In this embodiment, the personalized intervention plan generation and execution module has a built-in specialist intervention knowledge base covering six major categories: medication adjustment, rehabilitation training, dietary guidance, psychological counseling, follow-up appointment, and emergency treatment. The personalized intervention plan generation and execution module can automatically generate a unique personalized intervention plan based on the patient's risk level, disease type, rehabilitation stage, follow-up data, and early warning results, through rule matching and intelligent recommendation algorithms, specifying the intervention content, intervention frequency, execution method, expected goals, and precautions.

[0032] Based on risk warning results and patient stratification tags, the system automatically generates personalized post-diagnosis intervention plans and drives digital employees to complete the entire execution process, solving the problems of manual intervention plan formulation, insufficient personalization, and inadequate execution in existing systems. The system has a built-in specialist intervention knowledge base covering six major categories of intervention content: medication adjustment, rehabilitation training, dietary guidance, psychological counseling, follow-up appointment scheduling, and emergency treatment. Each category is matched with standardized operating procedures and medical guidelines. Based on the patient's risk level, disease type, rehabilitation stage, follow-up data, and warning results, the system automatically generates a unique personalized intervention plan through rule matching and intelligent recommendation algorithms. The plan clearly defines the intervention content, frequency, execution method, expected goals, and precautions. The digital employee, as the sole executor of the intervention plan, accurately pushes medication reminders, rehabilitation training videos, dietary restrictions advice, and psychological counseling content to the patient according to the plan requirements, guiding the patient to complete the intervention tasks and recording the intervention execution status, patient feedback, and effects in real time. For patients under red alert, the system automatically links with the offline medical team to initiate a collaborative process of initial online digital staff intervention and emergency offline medical staff intervention, achieving integrated online and offline intervention without the need for medical staff to manually develop intervention plans, thus improving the timeliness, accuracy, and personalization of intervention.

[0033] This module includes a built-in specialist intervention knowledge base covering six major categories. All intervention content is aligned with standardized medical guidelines, ensuring the professionalism and compliance of intervention plans. Through rule matching and intelligent recommendation algorithms, combined with patient stratification tags and risk warning results, and integrating multi-dimensional information such as patient disease type, recovery stage, and follow-up data, it automatically generates personalized intervention plans, clearly defining the details of each intervention stage. This eliminates the need for manual planning by medical staff, addressing the issue of insufficient personalization. A digital employee serves as the sole executor, accurately pushing various intervention content according to the plan, guiding patients to complete intervention tasks, and recording execution status and feedback in real time to ensure effective intervention. For extremely high-risk patients with red alerts, it automatically connects with offline medical teams, constructing a collaborative model of online initial intervention + offline emergency intervention. This achieves integrated online and offline intervention, significantly improving the timeliness and accuracy of intervention, effectively improving patient recovery outcomes, and solving the problems of inadequate and inefficient implementation of traditional interventions.

[0034] In this embodiment, the follow-up closed-loop management and effect review module summarizes core indicators such as follow-up coverage rate, task completion rate, risk warning accuracy rate, intervention effectiveness rate, patient compliance, improvement of rehabilitation indicators, and medical staff response time. It also introduces a quantitative evaluation formula for follow-up effectiveness: E=δ×C+ε×A+ζ×Rc; where E represents the overall follow-up effectiveness score (higher scores indicate better follow-up management); δ is the compliance weighting coefficient (range 0.3≤δ≤0.4); and C is the patient's follow-up compliance score (0-100 points, determined by the follow-up...). The system calculates the following metrics: completion rate of follow-up tasks, medication adherence, rehabilitation training execution rate, and follow-up examination attendance rate. ε is the effective weighting coefficient for early warning (range 0.3 ≤ ε ≤ 0.4); A is the accuracy score for risk early warning (0-100 points, calculated from the true positive rate, missed judgment rate, and false judgment rate); ζ is the weighting coefficient for rehabilitation improvement (range 0.2 ≤ ζ ≤ 0.3); Rc is the score for improvement in patient rehabilitation indicators (0-100 points, comprehensively assessed from the degree of symptom relief, the rate of achievement of vital signs, the improvement of follow-up examination results, and the incidence of adverse events). Based on the E value, the system automatically generates three-level follow-up review reports for individual patients, departments, and the entire hospital. It automatically adjusts follow-up strategies and intervention plans for patients with low effectiveness, automatically optimizes the algorithm model and knowledge base to address system process shortcomings, and automatically generates performance data for medical and nursing work, enabling continuous iterative upgrades of the system. At the same time, the system will synchronize the review data to the hospital management platform, providing data support for medical institutions to optimize post-diagnosis management processes, improve the quality of medical services, and reduce medical risks, thus completing the closed loop of the entire post-diagnosis follow-up management process.

[0035] This module first summarizes the core indicators of the entire follow-up process, covering multiple dimensions such as follow-up coverage, task completion, accurate early warning, and effective intervention, ensuring the comprehensiveness of the review data. It introduces a quantitative evaluation formula for follow-up effectiveness, achieving quantitative evaluation of follow-up results through weighted allocation and comprehensive calculation of multiple indicators. This provides objective data for review and solves the problems of traditional follow-up lacking quantitative evaluation and directionless optimization. Based on the quantitative scores, it automatically generates three levels of review reports: single patient, department, and hospital-wide, accurately identifying low-performing patients, system process shortcomings, and key areas of medical and nursing work. It automatically adjusts follow-up strategies, optimizes algorithm models and knowledge bases for different problems, and automatically generates medical and nursing performance evaluation data, enabling continuous iteration of the system and follow-up work. Simultaneously, the review data is synchronized to the hospital management platform, providing data support for post-diagnosis management decisions in medical institutions, completely solving the problems of fragmented and uniterable traditional follow-up management, and promoting continuous optimization of follow-up management.

[0036] The follow-up closed-loop management and effect review module in this embodiment can automatically generate three-level follow-up review reports for individual patients, departments, and the entire hospital based on the comprehensive follow-up effect score E; automatically adjust follow-up strategies and intervention plans for patients with low effects; automatically optimize algorithm models and knowledge bases for system process shortcomings; and automatically generate performance data for medical and nursing work, enabling continuous iteration and upgrading of the system; at the same time, the review data is synchronized to the hospital management platform.

[0037] One method for managing post-diagnosis follow-up of patients using digital employee data includes the following steps: Data access and tag generation: Access patients' full-dimensional diagnosis and treatment and health data, parse unstructured text through NLP models, and automatically generate three types of hierarchical tags: core, risk, and follow-up. Interaction adaptation and task distribution: Digital staff adapts the optimal multimodal interaction method based on patient stratification tags, automatically generates personalized follow-up tasks, pushes them out according to priority, and handles missed visits, urgent tasks, and special circumstances. Data collection and cleaning: Real-time collection of follow-up data from digital employee interactions, patient self-reporting, and IoT device synchronization, followed by deduplication, completion, anomaly removal, and standardization to form a structured database; Risk assessment and early warning: Based on standardized follow-up data, the risk of patient recovery is assessed in real time through a specialized risk assessment model, and four levels of early warning are defined and a dual early warning mechanism is triggered. Intervention plan generation and execution: Based on risk warning results and patient tags, personalized intervention plans are automatically generated and executed by digital staff throughout the process. For patients with red alerts, offline medical staff are mobilized for collaborative intervention. Closed-loop management and debriefing optimization: Summarize the core indicators of the entire follow-up process, evaluate the follow-up effect through quantitative formulas, generate debriefing reports, and automatically adjust follow-up strategies and optimize the system.

[0038] This invention constructs a complete digital employee post-diagnosis follow-up management system. It employs a dual-quantification formula to accurately assess follow-up priority and effectiveness, integrating multimodal digital employees, automated task distribution, real-time risk warning, personalized intervention, and closed-loop review technologies. Unlike existing manual follow-up and single-function digital employee systems, it offers the following advantages: automated patient stratification and task distribution replace traditional manual follow-up, increasing follow-up coverage, reducing missed visits, lowering healthcare labor costs, and significantly improving follow-up efficiency; multimodal digital employees adapt to patient groups of different ages, health literacy levels, and disease types, significantly improving patient engagement and interaction experience; the patient follow-up priority quantification formula enables precise allocation of follow-up resources, improving the management quality of key patients; and AI-based real-time risk assessment... The system's assessment and tiered early warning capabilities can improve the early detection rate of adverse events in patients, effectively reduce the risk of postoperative complications and acute exacerbations of chronic diseases, and ensure patient safety. Personalized intervention plans are automatically generated and executed autonomously by digital staff, eliminating the need for medical personnel to manually develop plans. This significantly improves the timeliness and accuracy of interventions, leading to marked improvements in patient recovery. The quantitative formula for follow-up results and closed-loop management throughout the entire process enable continuous system optimization, breaking down data silos and achieving standardization, visualization, and traceability of follow-up data. Furthermore, the system is easy to deploy, adaptable to medical institutions at all levels and various disease types, and can be widely applied to post-diagnosis scenarios such as chronic disease management, postoperative rehabilitation, and specialist treatment. Overall, it promotes the transformation of post-diagnosis medical services towards refinement, intelligence, efficiency, and personalization, improving the service quality of medical institutions and the level of patient health management.

[0039] Although embodiments of the invention have been shown and described, the scope of the invention will be defined by the appended claims and their equivalents by those skilled in the art.

Claims

1. A digital employee post-diagnosis follow-up management system for patients, characterized in that, It includes modules for patient data access and hierarchical labeling, digital employee interaction and follow-up task distribution, follow-up data collection and cleaning, AI rehabilitation risk assessment and early warning, personalized intervention plan generation and execution, and follow-up closed-loop management and effect review. The system comprises the following modules: Patient Data Access and Hierarchical Tagging Module for accessing full-dimensional patient data, structured parsing, and automated generation of hierarchical tags; Digital Employee Interaction and Follow-up Task Distribution Module for driving digital employees to achieve multimodal interaction adaptation and automatically distributing follow-up tasks based on patient tags; Follow-up Data Collection and Cleaning Module for collecting full follow-up data in real time and forming structured data after standardized cleaning; AI Rehabilitation Risk Assessment and Early Warning Module for assessing patient rehabilitation risks in real time and providing graded early warnings based on standardized follow-up data; Personalized Intervention Plan Generation and Execution Module for automatically generating personalized intervention plans based on risk warning results and patient tags, with digital employees completing the entire execution process; and Follow-up Closed-Loop Management and Effect Review Module for summarizing follow-up data throughout the entire process, quantitatively evaluating follow-up effects, and completing process optimization and review.

2. The digital employee post-diagnosis follow-up management system for patients according to claim 1, characterized in that, The patient data access and hierarchical labeling module connects with hospital HIS systems, LIS laboratory systems, EMR electronic medical record systems, PACS imaging systems, patient-side mini-programs, and IoT health devices through standard interfaces. It accesses four categories of comprehensive data: basic patient information, core diagnosis and treatment information, basic health information, and rehabilitation risk information. The patient data access and hierarchical labeling module is equipped with a medical-specific natural language processing model to perform structured parsing of unstructured discharge summaries, medical records, and medical orders, automatically extracting key feature items, and automatically generating three categories of three-level hierarchical labels: core labels, risk labels, and follow-up labels, based on a specialist follow-up knowledge base and clinical guidelines. The labels can be dynamically updated according to changes in the patient's condition.

3. The digital employee post-diagnosis follow-up management system for patients according to claim 2, characterized in that, The digital employee interaction and follow-up task distribution module features a multimodal virtual medical assistant that integrates speech recognition, speech synthesis, text interaction, and virtual digital human video interaction. It is equipped with a medical-specific ASR recognition model and TTS synthesis engine, and supports dialect recognition, emotion perception, and multi-turn contextual dialogue. The digital employee interaction and follow-up task distribution module can automatically adapt the optimal interaction method based on the patient's stratified tags. At the same time, through the task scheduling engine, it can automatically generate a personalized follow-up task list based on the patient tags, specifying the follow-up time, follow-up content, response rules, and timeout handling strategy. The digital employee can automatically push follow-up tasks according to task priority, supporting automatic re-issuance of missed visits, priority push of urgent tasks, and silent delay during non-working hours.

4. The digital employee post-diagnosis follow-up management system for patients according to claim 3, characterized in that, The follow-up data collection and cleaning module collects follow-up data in real time: data automatically recorded by digital staff interaction, data actively reported by patients, and data automatically synchronized by IoT devices; the follow-up data collection and cleaning module adopts a four-layer data cleaning algorithm to complete deduplication, missing value filling, outlier removal, and standardization transformation in sequence, converting unstructured voice and text data into a structured follow-up database in a unified format and storing it in the system data platform.

5. A digital employee post-diagnosis follow-up management system for patients according to claim 4, characterized in that, In the follow-up data collection and cleaning module, a patient follow-up priority quantification formula is introduced: P=α×S+β×R+γ×D; where P is the comprehensive score of patient follow-up priority, with higher scores indicating higher priority; α is the risk weight coefficient, ranging from 0.4≤α≤0.6, preset by the clinical follow-up standards of each specialty; S is the patient risk level score, with low risk=1, medium risk=2, high risk=3, and very high risk=4, automatically assigned by patient stratification labels; β is the weight of the recovery stage. The coefficient ranges from 0.2 to 0.3; R is the patient's recovery stage score, with acute phase = 3, recovery phase = 2, and stable phase = 1, assigned by the patient stratification label; γ is the follow-up response weight coefficient, with a range of 0.1 to 0.3; D is the patient's historical follow-up response score, with no response = 0, partial response = 1, and complete response = 2, automatically calculated from historical follow-up data; the system classifies patients into four follow-up levels—special, level one, level two, and level three—based on the P-value and matches corresponding follow-up resources.

6. A digital employee post-diagnosis follow-up management system for patients according to claim 5, characterized in that, The AI-powered rehabilitation risk assessment and early warning module incorporates a specialized risk assessment model trained on clinical cases, covering common postoperative diseases. Based on standardized follow-up data, it can conduct 24 / 7 intelligent assessment of patient rehabilitation risks. The module categorizes warning levels into four levels: red, orange, yellow, and blue. When a corresponding warning level is reached, a dual warning mechanism is triggered: a digital employee pushes warning reminders and temporary emergency guidance to the patient, while simultaneously sending warning notifications to responsible medical staff via system pop-ups, SMS, and telephone. The red warning response time is ≤10 seconds.

7. A digital employee post-diagnosis follow-up management system for patients according to claim 6, characterized in that, The personalized intervention plan generation and execution module has a built-in specialist intervention knowledge base covering six major categories: medication adjustment, rehabilitation training, dietary guidance, psychological counseling, follow-up appointment, and emergency treatment. The personalized intervention plan generation and execution module can automatically generate exclusive personalized intervention plans based on the patient's risk level, disease type, rehabilitation stage, follow-up data, and early warning results, through rule matching and intelligent recommendation algorithms, specifying the intervention content, intervention frequency, execution method, expected goals, and precautions.

8. A digital employee post-diagnosis follow-up management system for patients according to claim 7, characterized in that, The follow-up closed-loop management and effect review module summarizes core indicators such as follow-up coverage rate, task completion rate, risk warning accuracy rate, intervention effectiveness rate, patient compliance, improvement of rehabilitation indicators, and medical staff response time. It also introduces a quantitative evaluation formula for follow-up effectiveness: E=δ×C+ε×A+ζ×Rc; where E is the comprehensive follow-up effectiveness score, with higher scores indicating better follow-up management; δ is the compliance weighting coefficient, ranging from 0.3≤δ≤0.4; and C is the patient's follow-up compliance score, calculated by... The following parameters are considered: completion rate of visit tasks, medication adherence, rehabilitation training execution rate, and follow-up examination rate; ε is the effective weighting coefficient for early warning, with a value range of 0.3 ≤ ε ≤ 0.4; A is the accuracy score of risk early warning, calculated from the true positive rate, missed judgment rate, and false judgment rate of early warning; ζ is the weighting coefficient for rehabilitation improvement, with a value range of 0.2 ≤ ζ ≤ 0.3; Rc is the score for improvement of patient rehabilitation indicators, comprehensively assessed from the degree of symptom relief, the rate of achievement of vital signs, the improvement of follow-up examination results, and the incidence of adverse events.

9. A digital employee post-diagnosis follow-up management system for patients according to claim 8, characterized in that, The follow-up closed-loop management and effect review module can automatically generate three-level follow-up review reports for individual patients, departments, and the entire hospital based on the comprehensive follow-up effect score E; it can automatically adjust follow-up strategies and intervention plans for patients with low effects, automatically optimize algorithm models and knowledge bases for system process shortcomings, and automatically generate performance data for medical and nursing work, so as to realize continuous iterative upgrades of the system; at the same time, the review data is synchronized to the hospital management platform.

10. A digital employee post-diagnosis follow-up management method for patients, characterized in that, Specifically, the following steps are included: Data access and tag generation: Access patients' full-dimensional diagnosis and treatment and health data, parse unstructured text through NLP models, and automatically generate three types of hierarchical tags: core, risk, and follow-up. Interaction adaptation and task distribution: Digital staff adapts the optimal multimodal interaction method based on patient stratification tags, automatically generates personalized follow-up tasks, pushes them out according to priority, and handles missed visits, urgent tasks, and special circumstances. Data collection and cleaning: Real-time collection of follow-up data from digital employee interactions, patient self-reporting, and IoT device synchronization, followed by deduplication, completion, anomaly removal, and standardization to form a structured database; Risk assessment and early warning: Based on standardized follow-up data, the risk of patient recovery is assessed in real time through a specialized risk assessment model, and four levels of early warning are defined and a dual early warning mechanism is triggered. Intervention plan generation and execution: Based on risk warning results and patient tags, personalized intervention plans are automatically generated and executed by digital staff throughout the process. For patients with red alerts, offline medical staff are mobilized for collaborative intervention. Closed-loop management and debriefing optimization: Summarize the core indicators of the entire follow-up process, evaluate the follow-up effect through quantitative formulas, and generate a debriefing report.

Images