Clinical quality control rule-based nursing document intelligent error correction and specification generation system

Abstract

The invention belongs to the technical field of medical informatization, and particularly relates to a nursing document intelligent error correction and specification generation system based on clinical quality control rules, which comprises a data access and processing layer, an AI capability support layer, a core intelligent engine layer and a business application layer, the data access and processing layer, the AI capability support layer, the core intelligent engine layer and the business application layer are connected in sequence, and data interaction and control instruction transmission are carried out among the layers through service interfaces and message middleware; through deep integration of a rule execution engine and a streaming data processing engine, real-time intelligent error correction and standard content generation of nursing documents are achieved, clinical quality control rules are deeply fused, and meanwhile efficient integration with an existing medical system is guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology, specifically referring to an intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules. Background Technology

[0002] In current medical and nursing practice, nursing documentation is the core tool for recording patient care processes and assessing nursing quality. However, existing nursing documentation management has many shortcomings: 1. Low efficiency of manual quality control: Traditional nursing document quality control relies on manual inspection, which is time-consuming, labor-intensive, and easily affected by subjective factors, making it difficult to achieve full and real-time quality control.

[0003] 2. Insufficient intelligence in existing systems: Some nursing information systems only have basic format verification functions and lack in-depth understanding and automated verification of clinical professional rules (such as nursing operation time limits, content completeness, terminology standardization, etc.).

[0004] 3. Lack of standardization in document generation: Nursing document generation relies heavily on manual input by nurses, which can easily lead to problems such as non-standard terminology, missing content, and non-standard format. Furthermore, it lacks intelligent generation capabilities based on clinical scenarios.

[0005] 4. Poor system integration: It lacks efficient data synchronization and service integration with existing HIS (Hospital Information System) and EMR (Electronic Medical Record System), making it difficult to form a closed-loop quality control process. Summary of the Invention

[0006] To address the problems of untimely quality control, non-standardized generation, insufficient integration of clinical rules, and poor system compatibility in existing nursing documentation, this invention provides an intelligent error correction and standardized generation system for nursing documentation based on clinical quality control rules. This system enables real-time intelligent error correction and standardized content generation of nursing documentation, deeply integrates clinical quality control rules, and ensures efficient integration with existing medical systems.

[0007] To achieve the above functions, the technical solution adopted by the present invention is as follows: a nursing document intelligent error correction and standardization generation system based on clinical quality control rules, including a data access and processing layer, an AI capability support layer, a core intelligent engine layer and a business application layer, wherein the data access and processing layer, the AI ​​capability support layer, the core intelligent engine layer and the business application layer are connected in sequence, and the data interaction and control command transmission between each layer are carried out through service interfaces and message middleware. The data access and processing layer includes a unified data interface gateway and a streaming data processing engine, which are used to access multi-source data in real time from various heterogeneous information systems in the hospital, and perform cleaning, standardization and time-series alignment processing to provide high-quality and standardized patient clinical data streams for the upper layer. The AI ​​capability support layer includes a medical NLP engine and a generative LLM engine. The medical NLP engine is used to parse unstructured nursing text and identify clinical entities and relationships. The generative LLM engine is used to generate standardized nursing document content based on structured data and contextual information, providing the system with natural language processing and generative large language model capabilities. The core intelligent engine layer includes a rule execution engine and a dynamic routing decision-maker. The rule execution engine loads and executes formally expressed clinical quality control rules to achieve millisecond-level real-time error correction and alarms. The dynamic routing decision-maker dynamically selects the optimal content generation strategy based on the characteristics of the document to be generated. The business application layer directly faces nurse users, providing real-time error correction and document generation services, and interacts with users through a front-end interface.

[0008] Furthermore, the rule execution engine includes a patient context manager, working memory, and a rule matcher. The patient context manager is used to maintain an independent state vector for each patient. The state vector includes at least the patient's clinical state and key time points. The working memory is used to store real-time factual data from the data access and processing layer. The rule matcher is based on the Rete algorithm.

[0009] Furthermore, the dynamic routing decision-maker is configured with a routing strategy, which is calculated through a decision function based on multiple features including document type, content complexity, context richness, and user preferences, to select between template-based generation, LLM engine generation, and hybrid generation strategies.

[0010] Furthermore, the document generation service is connected to the dynamic routing decision-maker in the core intelligent engine layer; when a document generation request is received, the document generation service calls the dynamic routing decision-maker to obtain the generation strategy, and calls the corresponding generation component according to the generation strategy: if the strategy is template-based generation, the template engine is called; if the strategy is LLM engine generation, the generative LLM engine in the AI ​​capability support layer is called.

[0011] Furthermore, the real-time error correction service is connected to the rule execution engine via a message middleware; when the rule execution engine triggers an execution action to generate an alarm, the alarm message is pushed to the real-time error correction service via the message middleware, and then pushed to the front-end user interface in real time via a WebSocket connection.

[0012] Furthermore, the data access and processing layer includes a unified data interface gateway and a streaming data processing engine. The unified data interface gateway supports HL7 and FHIR standard protocols and is equipped with a protocol adapter, a RESTful API interface, and a CDC-based database connector. The streaming data processing engine is used to clean, standardize, and align the accessed data.

[0013] Furthermore, the medical NLP engine employs a neural network model including BERT, BiLSTM, and CRF for named entity recognition and relation extraction of nursing text; the generative LLM engine is a large language model that has been further pre-trained and fine-tuned based on nursing domain data.

[0014] Furthermore, the system also includes a clinical quality control rule base. The rules in the clinical quality control rule base are expressed using a formal model of "rule identifier, triggering condition, inspection condition, execution action, priority, time constraint, and metadata" and are managed through a versioning mechanism.

[0015] Furthermore, the rule matcher is used to perform pattern matching between the facts in the working memory and the preloaded clinical quality control rule base. When the triggering condition is met, the rule instance is activated, and when the inspection condition is verified to be true, the execution action is performed.

[0016] The beneficial effects achieved by adopting the above-described solution in this invention are as follows: 1. Through the deep integration of the rule execution engine and the streaming data processing engine, a quality control model transformation from "post-event spot checks" to "real-time intervention during the event" has been achieved. The rule matching latency is less than 100 milliseconds, which can immediately remind nurses to correct errors. 2. The system achieves a fully automated closed loop from error detection, alarm generation, nurse processing to status update, ensuring the integrity and effectiveness of the quality control process; 3. Through a hybrid generation algorithm and a dynamic routing decision-maker, the system can intelligently select the optimal generation method based on the complexity and context of the document, ensuring both standardization and flexibility and personalization, thus significantly improving the efficiency and quality of document generation. 4. By transforming abstract nursing standards into formalized, structured, and executable rules and incorporating them into version control, the quality control standards become transparent, unified, and easy to maintain. After new rules are issued, they can be quickly implemented throughout the hospital by updating the rule base. The system can continuously optimize the rule base and algorithm model by monitoring indicators such as the trigger frequency, false alarm rate, and clinical adoption rate of the rules, and by combining feedback from nurses. Attached Figure Description

[0017] Figure 1This is a block diagram of the overall architecture of the intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules, as described in this invention. Figure 2 This is a flowchart of the rule execution engine of the intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules, as described in this invention. Figure 3 This is an architectural diagram of the clinical quality control rule base of the present invention, which is a nursing document intelligent error correction and standardization generation system based on clinical quality control rules. Figure 4 This is a flowchart illustrating Embodiment 1 of the present invention, which is an intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules. Detailed Implementation

[0018] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0019] This invention relates to an intelligent error correction and standardization generation system for nursing documentation based on clinical quality control rules. It comprises a data access and processing layer, an AI capability support layer, a core intelligent engine layer, and a business application layer. These layers are sequentially connected, with data interaction and control command transmission occurring between them via service interfaces and message middleware. The data access and processing layer includes a unified data interface gateway and a streaming data processing engine, used to access multi-source data in real time from various heterogeneous information systems within the hospital, performing cleaning, standardization, and time-series alignment to provide high-quality, standardized patient clinical data streams to the upper layers. The AI ​​capability support layer includes a medical NLP engine and a generative... The LLM engine and medical NLP engine are used to parse unstructured nursing text and identify clinical entities and relationships. The generative LLM engine is used to generate standardized nursing documentation based on structured data and contextual information, providing the system with natural language processing and generative large language model capabilities. The core intelligent engine layer includes a rule execution engine and a dynamic routing decision-maker. The rule execution engine loads and executes formally expressed clinical quality control rules, achieving millisecond-level real-time error correction and alerts. The dynamic routing decision-maker dynamically selects the optimal content generation strategy based on the characteristics of the document to be generated. The business application layer directly faces nurse users, providing real-time error correction and document generation services, and interacts with users through a front-end interface.

[0020] The rule execution engine includes a patient context manager, working memory, and a rule matcher. The patient context manager maintains an independent state vector for each patient, which includes at least the patient's clinical state and key time points. The working memory stores real-time fact data from the data access and processing layer. The rule matcher is based on the Rete algorithm. The system also includes a clinical quality control rule base. The rules in the clinical quality control rule base are expressed using a formal model of "rule identifier, trigger condition, check condition, execution action, priority, time constraint, and metadata," and are managed through a versioning mechanism. The rule matcher performs pattern matching between the facts in the working memory and the pre-loaded clinical quality control rule base. When the trigger condition is met, the rule instance is activated, and when the check condition is verified to be true, the execution action is executed.

[0021] The dynamic routing decision-maker is configured with routing strategies, which are calculated through a decision function based on multiple features, including document type, content complexity, context richness, and user preferences, to select between template-based generation, LLM engine generation, and hybrid generation strategies.

[0022] The document generation service is connected to the dynamic routing decision-maker in the core intelligent engine layer. When a document generation request is received, the document generation service calls the dynamic routing decision-maker to obtain the generation strategy, and calls the corresponding generation component according to the generation strategy: if the strategy is template-based generation, the template engine is called; if the strategy is LLM engine generation, the generative LLM engine in the AI ​​capability support layer is called.

[0023] The real-time error correction service is connected to the rule execution engine through a message middleware. When the rule execution engine triggers an action to generate an alarm, the alarm message is pushed to the real-time error correction service through the message middleware, and then pushed to the front-end user interface in real time through a WebSocket connection.

[0024] The data access and processing layer includes a unified data interface gateway and a streaming data processing engine. The unified data interface gateway supports HL7 and FHIR standard protocols and is equipped with a protocol adapter, a RESTful API interface, and a CDC-based database connector. The streaming data processing engine is used to clean, standardize, and align the accessed data.

[0025] The medical NLP engine uses neural network models including BERT, BiLSTM, and CRF to perform named entity recognition and relation extraction on nursing text; the generative LLM engine is a large language model that has been further pre-trained and fine-tuned based on nursing domain data. Example 1

[0026] I. Data Access and Processing Layer Interface with Hospital System: The unified data interface gateway adopts the FHIR R4 standard as the core data model. Data is accessed through protocol adapters, RESTful APIs, and database connectors. The accessed data enters the streaming data processing engine and goes through three stages: data cleaning, terminology standardization, and time-series alignment. The final output is a standardized, time-consistent patient data stream, which is continuously delivered to the core intelligent engine layer through a message queue.

[0027] II. The AI ​​capability support layer provides intelligent processing capabilities. The medical NLP engine adopts a model architecture of BERT+BiLSTM+CRF, which is responsible for named entity recognition and relation extraction of free text in nursing records, transforming unstructured text into structured information for use by the rule engine. The generative LLM engine is based on an open-source large model for domain adaptation and instruction fine-tuning, enabling it to understand and generate text that conforms to nursing standards. This engine is optimized for inference through TensorRT and provides API services.

[0028] III. The core intelligent engine layer is the control core: The rule execution engine's core components include: Patient Context Manager: Maintains an independent state vector for each patient, for example: Patient P001: {Clinical Status: "Postoperative Return", Return Time: "14:00", Key Time Points: {First Record Deadline: "16:00"}}. This manager receives real-time data streams from the data access and processing layer and dynamically updates the corresponding patient context. Working memory: Stores real-time events and data facts currently entering the engine, such as a "patient returns to ward" event; Rule matcher: Based on the Rete algorithm, it continuously matches facts in the working memory with tens of thousands of quality control rules preloaded in the rule base; Rule enforcement principle: Taking the "completeness check of the first postoperative nursing record" rule as an example; 1. Trigger: The data access layer pushes the "Patient P001 returns to the ward" event. After receiving it, the rule engine updates the context of patient P001 and stores the event in the working memory. 2. Matching: The rule matcher finds that the event meets the "triggering condition" of the rule COMPLETEENESS_POSTOP_001, so it activates the rule instance and sets a 2-hour timer; 3. Check: 2 hours later, the rule engine executes the "check conditions". Based on the patient context, it queries the data storage to "whether there is a nursing record of type 'first postoperative' for patient P001 between 14:00 and 16:00 that contains necessary fields such as consciousness status and wound condition". 4. Execution: If the query result is "record does not exist or fields are incomplete", the rule engine will immediately trigger the "execution action": generate a high-risk alert message; 5. Output: The alarm message is immediately sent to the business application layer.

[0029] The dynamic routing decision-maker is activated when the system needs to generate nursing documentation. It first analyzes several features of the document to be generated: F_structure (Document structure characteristics) F_complexity (content complexity feature) F_context (contextual feature) Then, these features are weighted and fused using the decision function Φ, and a routing decision is made based on the preset routing policy threshold. If the document is highly structured and has low complexity, it will be routed to the template generation module to quickly populate the data; If a document requires a high degree of creativity, reasoning, and natural language description, it is routed to the LLM engine for generation; If it falls between the two, a hybrid strategy is adopted; The decision will control the subsequent process and determine which generation component to invoke.

[0030] IV. Interaction between the business application layer and nurses: The real-time error correction service maintains communication with the nurse workstation front-end via a WebSocket long connection. When the core layer's rule engine generates an alarm, the alarm message is pushed to the connection manager through a message queue, and then pushed to the corresponding nurse workstation in real time by the connection manager to realize pop-up or sound reminders with a latency of less than 100 milliseconds. The document generation service provides an API to receive generation requests and parameters, calls the corresponding generation components based on the results of the dynamic routing decision-maker, and returns the generated document content, while performing quality assessment and caching.

[0031] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A nursing documentation intelligent error correction and standardization generation system based on clinical quality control rules, characterized in that, It includes a data access and processing layer, an AI capability support layer, a core intelligent engine layer, and a business application layer. The data access and processing layer, the AI ​​capability support layer, the core intelligent engine layer, and the business application layer are connected in sequence, and the layers interact with each other and transmit control commands through service interfaces and message middleware. The data access and processing layer includes a unified data interface gateway and a streaming data processing engine, which are used to access multi-source data in real time from various heterogeneous information systems in the hospital, and perform cleaning, standardization and time-series alignment processing to provide high-quality and standardized patient clinical data streams for the upper layer. The AI ​​capability support layer includes a medical NLP engine and a generative LLM engine. The medical NLP engine is used to parse unstructured nursing text and identify clinical entities and relationships. The generative LLM engine is used to generate standardized nursing document content based on structured data and contextual information, providing the system with natural language processing and generative large language model capabilities. The core intelligent engine layer includes a rule execution engine and a dynamic routing decision-maker. The rule execution engine loads and executes formally expressed clinical quality control rules to achieve millisecond-level real-time error correction and alarms. The dynamic routing decision-maker dynamically selects the optimal content generation strategy based on the characteristics of the document to be generated. The business application layer directly faces nurse users, providing real-time error correction and document generation services, and interacts with users through a front-end interface.

2. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 1, characterized in that, The rule execution engine includes a patient context manager, working memory, and a rule matcher. The patient context manager is used to maintain an independent state vector for each patient. The state vector includes at least the patient's clinical status and key time points. The working memory is used to store real-time factual data from the data access and processing layer. The rule matcher is based on the Rete algorithm.

3. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 2, characterized in that, The dynamic routing decision maker is configured with a routing strategy, which is calculated through a decision function based on multiple features including document type, content complexity, context richness, and user preferences, to select between template-based generation, LLM engine generation, and hybrid generation strategies.

4. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 3, characterized in that, The document generation service is connected to the dynamic routing decision-maker in the core intelligent engine layer. When a document generation request is received, the document generation service calls the dynamic routing decision-maker to obtain the generation strategy and calls the corresponding generation component according to the generation strategy: if the strategy is template-based generation, then the template engine is called. If the strategy is generated by the LLM engine, then the generative LLM engine in the AI ​​capability support layer is invoked.

5. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 2, characterized in that, The real-time error correction service is connected to the rule execution engine via a message middleware. When the rule execution engine triggers an execution action to generate an alarm, the alarm message is pushed to the real-time error correction service via the message middleware, and then pushed to the front-end user interface in real time via a WebSocket connection.

6. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 1, characterized in that, The unified data interface gateway supports HL7 and FHIR standard protocols and is equipped with a protocol adapter, RESTful API interface and CDC-based database connector. The streaming data processing engine is used to clean, standardize terms and align time sequence of the incoming data.

7. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 1, characterized in that, The medical NLP engine employs a neural network model including BERT, BiLSTM, and CRF for named entity recognition and relation extraction from nursing text; the generative LLM engine is a large language model that has been further pre-trained and fine-tuned based on nursing domain data.

8. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 2, characterized in that, It also includes a clinical quality control rule base, in which the rules are expressed using a formal model of "rule identifier, triggering condition, inspection condition, execution action, priority, time constraint, and metadata" and are managed through a versioning mechanism.

9. The intelligent error correction and standardization generation system for nursing documents based on clinical quality control rules according to claim 8, characterized in that, The rule matcher is used to perform pattern matching between facts in the working memory and the preloaded clinical quality control rule base. When the trigger condition is met, the rule instance is activated, and the execution action is performed when the check condition is verified to be true.

Images