Nursing behavior pattern recognition platform

Abstract

The invention relates to the technical field of data processing, in particular to a nursing behavior pattern recognition platform. The platform carries out data quantification on a patient responsible for each nursing task and specific nursing content to obtain a real-time scene association vector. The space states of all the patients in the nursing responsibility group are quantified, a nursing opportunity cost matrix is constructed to determine the nursing path space economy loss, the scene similarity is obtained and used for representing the coherence degree of the nurse workload, the scene similarity and the nursing path space economy loss visually reflect the nurse working mode, and the nurse working mode is more accurate. By constructing the two-dimensional parameter space, the cognitive mode of the nurse in the current nurse responsibility group can be determined, and an effective alarm is given. The behavior time sequence data of the nurse and the state of the patient are effectively integrated through methods of vector fusion, matrix construction and the like, an effective alarm signal is fed back by analyzing the cognitive state mode of the nurse, and the nursing risk is avoided.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a nursing behavior pattern recognition platform. Background Technology

[0002] In the daily operations of hospitals, the working condition of nursing staff directly affects patient safety and the quality of medical care. Current technologies primarily rely on hospital information systems to record nursing procedures or on monitoring equipment to trigger threshold alarms for individual vital signs. These methods have significant technical limitations. They cannot objectively and continuously quantify changes in individual nurses' behavioral patterns due to fatigue or stress, such as repetitive and inefficient fixed patterns. These changes are key human factors leading to medical errors, and current technologies lack the ability to identify such hidden risks. Secondly, existing risk identification mechanisms are mostly passive, typically triggering alarms only after patients' physiological indicators have significantly deteriorated, thus missing the early intervention window. More importantly, current technologies cannot identify the associated risks among a group of patients cared for by the same nurse due to shared nursing deficiencies or a decline in the nurse's cognitive state; that is, they cannot provide early warning of potential group adverse events. Theoretically, a nurse's workflow involves not only physical movement but also the cognitive processing of multiple dynamically changing tasks. Furthermore, healthy physiological systems exhibit high complexity across multiple time scales, while illness or stress can reduce this complexity. Existing technologies have failed to effectively integrate nurses' behavioral time-series data, patients' physiological time-series data, and spatial information of the work environment, and have failed to delve into the dynamic interaction between nurses' cognitive state and patient group risk contained in these data. Summary of the Invention

[0003] To address the technical problem that existing technologies fail to effectively integrate nurses' behavioral time-series data and patients' physiological time-series data, and fail to deeply explore the dynamic interaction between nurses' cognitive states and patient group risks hidden behind these data, the present invention aims to provide a nursing behavior pattern recognition platform. The specific technical solution adopted is as follows: This invention proposes a nursing behavior pattern recognition platform, the platform comprising: The scenario state quantification module is used to obtain the multi-scale physiological complexity of each patient at the beginning of each nursing task based on the fluctuation characteristics of each patient's vital sign sequence at different time scales. The multi-scale physiological complexity and the patient's diagnosis result are encoded to form the patient's state vector. The state vector is fused with the nursing task content to obtain the real-time scenario association vector of each nursing task. The nursing pathway analysis module is used to construct a nursing opportunity cost matrix for all patients based on the multi-scale physiological complexity of each patient and the distance between patients; perform path optimization on the nursing opportunity cost matrix to obtain the theoretically optimal nursing pathway; and obtain the spatial economic loss of the nursing pathway based on the cost difference between the nurse's actual nursing pathway and the theoretically optimal nursing pathway. The scenario similarity acquisition module is used to obtain the scenario similarity of consecutive nursing tasks based on the vector similarity between the real-time scenario association vectors of adjacent nursing tasks in the actual nursing path. The behavior recognition and alarm module is used to construct a two-dimensional parameter space based on the scenario similarity and the economic loss of the nursing path space, obtain the two-dimensional risk coordinates of all nurses in the nursing responsibility group in the two-dimensional parameter space, identify the nurses' behavioral cognitive patterns based on the changing characteristics of the two-dimensional risk coordinates, and issue an alarm.

[0004] Furthermore, the method for obtaining the multi-scale physiological complexity includes: The vital sign sequence is processed using a multi-scale sample entropy algorithm to obtain a sample entropy-scale curve; the area enclosed by the sample entropy-scale curve and the coordinate system is taken as the multi-scale physiological complexity.

[0005] Furthermore, the method for obtaining the state vector includes: The state vector is obtained by encoding and merging the multi-scale physiological complexity with the static diagnostic information of the patient's basic pathology.

[0006] Furthermore, the method for obtaining the real-time scenario association vector includes: The nursing task text is processed using a natural language processing model to obtain a nursing task encoding vector; the nursing task encoding vector and the state vector are input into a pre-trained attention mechanism neural network to output the real-time context association vector.

[0007] Furthermore, the method for constructing the nursing opportunity cost matrix includes: The nursing opportunity cost matrix is ​​N×N, where N is the number of patients. Each element in the nursing opportunity cost matrix represents the cost of moving from one patient to another, and the cost is obtained based on the distance between the two patients and the multi-scale physiological complexity of the second patient.

[0008] Furthermore, a Traveling Salesman Problem (TSP) solver based on a genetic algorithm is used for path optimization.

[0009] Furthermore, the method for obtaining the spatial economic loss of the nursing pathway includes: The theoretical cost is obtained by summing the costs of the theoretically optimal nursing pathway in the nursing opportunity cost matrix, and the actual cost is obtained by summing the costs of the actual nursing pathway in the nursing opportunity cost matrix. The difference between the actual cost and the theoretical cost is used as the numerator, and the theoretical cost is used as the denominator to obtain the spatial economic loss of the nursing pathway.

[0010] Furthermore, the method for obtaining the scenario similarity includes: The cosine similarity between real-time contextual association vectors of adjacent nursing tasks is used as the similarity, and the average of all vector similarities in the actual nursing path is used as the contextual similarity.

[0011] Furthermore, based on the changing characteristics of the two-dimensional risk coordinates, the nurses' behavioral and cognitive patterns are identified and alarms are triggered, including: The average scenario similarity and average nursing path spatial economic loss of all nurses in the nursing responsibility group are mapped to a two-dimensional parameter space to obtain the two-dimensional risk coordinates of the nursing responsibility group. Two-dimensional risk coordinates are used as observation vectors for the HMM algorithm. The time series of the observation vectors are input into the trained HMM model. The Viterbi algorithm is used to determine the nurse's hidden working mode and feed back an alarm signal. The forward algorithm is used to determine the probability of transitioning from the hidden working mode to the cognitive failure mode at a future time. The alarm signal is fed back based on the probability.

[0012] Furthermore, if the hidden working mode is a behavioral rigidity mode, a first-level warning signal is fed back; if the probability is greater than a preset probability threshold, or if the hidden working mode is a cognitive failure mode, a second-level warning signal is fed back.

[0013] The present invention has the following beneficial effects: This invention first quantifies the data of the patients and specific nursing content for each nursing task to obtain a real-time scenario-related vector. To avoid the lag caused by detecting anomalies in a single indicator, this invention analyzes changes in patients' vital signs at different time scales, thus obtaining multi-scale physiological complexity. Compared to a single indicator threshold, this can serve as an earlier risk warning signal. Furthermore, it integrates the patient's diagnostic results with the determined state vector and nursing task content, ensuring that the real-time scenario-related vector simultaneously includes both the nursing task and the patient's specific state, avoiding the limitations of existing technologies that only differentiate nurses' work content based on the similarities and differences of nursing tasks. This invention further quantifies the spatial state of all patients in the nursing responsibility group, constructing a nursing opportunity cost matrix to determine the spatial economic loss of the nursing path. By evaluating the nurse's workflow on two orthogonal dimensions—the economy of spatial movement and the continuity of task logic—it quantifies the deviation in overall work efficiency and its potential impact on the group of patients. Furthermore, scenario similarity is used to characterize the coherence of nurses' workload. Scenario similarity and the spatial economic loss of nursing pathways can be used to intuitively reflect nurses' work patterns. By constructing a two-dimensional parameter space and observing the position and changes of coordinates in the space, the cognitive patterns of nurses in the current nurse responsibility group can be determined, and effective alarms can be triggered. Attached Figure Description

[0014] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a structural block diagram of a nursing behavior pattern recognition platform provided in one embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a nursing behavior pattern recognition platform proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The specific solution of the nursing behavior pattern recognition platform provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Please see Figure 1 The diagram illustrates a structural block diagram of a nursing behavior pattern recognition platform provided by an embodiment of the present invention. The platform includes: a scenario state quantification module 101, a nursing path analysis module 102, a scenario similarity acquisition module 103, and a behavior recognition and alarm module 104.

[0020] The scenario state quantification module is used to obtain the multi-scale physiological complexity of each patient at the beginning of each nursing task based on the fluctuation characteristics of each patient's vital sign sequence at different time scales. The multi-scale physiological complexity and the patient's diagnosis result are encoded to form the patient's state vector. The state vector is fused with the nursing task content to obtain the real-time scenario association vector for each nursing task.

[0021] In this embodiment of the invention, the vital signs sequence is a parameter sequence over an 8-hour period. Heart rate is selected as the vital signs parameter, but other parameters such as respiratory rate can also be selected for analysis. The specific selection can be made according to the patient's case, and will not be elaborated here. It should be noted that all patient information and nurses' nursing tasks can be directly accessed from the hospital's data system, and will not be elaborated further.

[0022] In the scenario state quantification module, to overcome the problems of traditional single-feature threshold alarm lag and inability to assess patient health reserves, this invention aims to obtain a more sensitive indicator of disease trend, namely multi-scale physiological complexity, by analyzing the inherent dynamic characteristics of vital sign signals. Healthy physiological systems typically exhibit high complexity and adaptability across multiple time scales, while disease or stress can lead to a reduction in this complexity. Therefore, using multi-scale physiological complexity can effectively characterize a patient's physiological features from the dynamic characteristics across multiple time scales.

[0023] Preferably, in this embodiment of the invention, the method for obtaining multi-scale physiological complexity includes: The vital sign sequence was processed using a multi-scale sample entropy algorithm to obtain a sample entropy-scale curve. The area enclosed by the sample entropy-scale curve and the coordinate system was used as the multi-scale physiological complexity. The multi-scale sample entropy algorithm can measure the irregularity of the signal at different time scales. The area obtained is a positive real number; the higher the value, the higher the dynamic complexity of the patient's physiological system at multiple time scales, indicating relatively stronger health reserves and resistance to interference. A continuous decrease in this characteristic can serve as an earlier risk warning signal than an abnormality in a single indicator.

[0024] The scenario state quantification module integrates nursing task content with patient state vectors, ensuring that each nursing task is not identified as having the same task characteristics, thus preventing the loss of a significant amount of clinical scenario information. In this embodiment, multi-scale physiological complexity and patient diagnostic results are first encoded to form the patient's state vector. Then, the state vector is integrated with the nursing task content to obtain a real-time scenario-related vector for each nursing task.

[0025] Preferably, in this embodiment of the invention, the method for obtaining the state vector includes: The state vector is obtained by encoding and merging the multi-scale physiological complexity with the static diagnostic information of the patient's basic pathology. The static diagnostic information is the result of the patient's clinical diagnosis, which can be converted into an encoded vector through natural language processing. The state vector is obtained by placing the multi-scale physiological complexity at the beginning or end of this vector.

[0026] Preferably, in this embodiment of the invention, the method for obtaining the real-time scenario association vector includes: Nursing task text is processed using a natural language processing model to obtain a nursing task encoding vector. The nursing task encoding vector and state vector are then input into a pre-trained attention mechanism neural network to output the real-time context association vector. In this embodiment, the natural language processing model used is the Word2Vec algorithm. Through processing by the natural language processing model, each nursing task (e.g., intravenous infusion, blood pressure measurement, etc.) can obtain a fixed high-dimensional vector representing its general functions and attributes.

[0027] It should be noted that the attention mechanism neural network dynamically adjusts the weights of each dimension in the standard nursing task vector based on the patient risk information carried by the real-time patient state vector. For example, for a critically ill patient with low multi-scale physiological complexity, when performing the task of "measuring vital signs," dimensions related to "circulatory system monitoring" and "respiratory system monitoring" will be given higher weights. This results in the output of a special real-time context-related vector. The specific model structure and training methods are well-known to those skilled in the art and will not be elaborated here.

[0028] The nursing pathway analysis module constructs a nursing opportunity cost matrix for all patients based on their multi-scale physiological complexity and the distance between patients. It then performs path optimization on this matrix to obtain the theoretically optimal nursing pathway. Finally, it calculates the spatial economic loss of the nursing pathway based on the cost difference between the nurse's actual nursing pathway and the theoretically optimal pathway. The module further analyzes the spatial aspects of all patients within the nursing responsibility group. By constructing a matrix and analyzing patient movement costs, it evaluates the nurse's workflow across two orthogonal dimensions: "economic efficiency of spatial movement" and "continuity of task logic." This quantifies the deviation in overall work efficiency and its potential impact on the group of patients.

[0029] To effectively construct a nursing opportunity cost matrix, we consider both the actual movement path in physical space and the patient's risk level. Based on the multi-scale physiological complexity of each patient and the distance between patients, we construct a nursing opportunity cost matrix for all patients.

[0030] Preferably, in one embodiment of the present invention, the method for constructing the nursing opportunity cost matrix includes: The nursing opportunity cost matrix is ​​N×N, where N is the number of patients. Each element in the matrix represents the cost of moving from one patient to another, calculated based on the distance between the two patients and the multi-scale physiological complexity of the next patient. This cost calculation is not solely based on physical distance but also takes into account the patient's immediate risk. As a specific example, in this embodiment of the invention, the cost is expressed by the formula: ;in Costs between patient i and patient j Let i be the distance between patient i and patient j. Let be the multi-scale physiological complexity of patient j. In this embodiment of the invention, the smaller the multi-scale physiological complexity, the higher the patient's risk, and the lower the final cost. This relationship ensures that in subsequent optimal path planning, the system will naturally tend to prioritize accessing patients with higher risk and closer proximity, which is completely consistent with the priority principle of clinical care.

[0031] The nursing pathway analysis module further optimizes the pathway based on the nursing opportunity cost matrix to obtain the theoretically optimal nursing pathway. It then calculates the spatial economic loss of the nursing pathway based on the cost difference between the nurse's actual nursing pathway and the theoretically optimal pathway. A higher value indicates a greater deviation from the rational principle of "prioritizing high-risk patients" in the nurse's actual movement decisions, representing a direct and objective quantitative indicator of a decline in their physical space planning ability.

[0032] Preferably, in this embodiment of the invention, a Traveling Salesman Problem (TSP) solver based on a genetic algorithm is used for path optimization. The genetic algorithm is chosen because it achieves a good balance between computational efficiency and solution accuracy when solving NP-hard problems of moderate size (the number of patients in a single nursing unit in a hospital is typically within this range). This solver takes the nursing opportunity cost matrix as input, and through iterative calculation, outputs a patient visit sequence that theoretically minimizes the total opportunity cost, i.e., the theoretically optimal nursing path. The specific optimization algorithm is a well-known technique to those skilled in the art and will not be elaborated upon here.

[0033] Preferably, in this embodiment of the invention, the method for obtaining the spatial economic loss of the nursing pathway includes: The theoretical cost is obtained by summing the costs of the theoretically optimal nursing pathway in the nursing opportunity cost matrix, and the actual cost is obtained by summing the costs of the actual nursing pathway in the nursing opportunity cost matrix. The difference between the actual cost and the theoretical cost is used as the numerator, and the theoretical cost is used as the denominator to obtain the spatial economic loss of the nursing pathway.

[0034] The scenario similarity acquisition module is used to obtain the scenario similarity of consecutive nursing tasks based on the vector similarity between the real-time scenario association vectors of adjacent nursing tasks in the actual nursing path.

[0035] Contextual similarity characterizes the coherence of switching between nursing tasks within a nurse's workflow. A higher similarity value indicates that the nurse is handling a series of similar nursing events in succession (e.g., performing the same procedure on several patients with similar conditions), resulting in lower cognitive costs for "mental context switching." Conversely, a low similarity value signifies frequent and drastic jumps in the nurse's thinking between tasks with vastly different contexts. Each jump consumes cognitive resources, contributing to cognitive fatigue and inducing distraction.

[0036] The method for obtaining the scenario similarity includes: The cosine similarity between real-time contextual association vectors of adjacent nursing tasks is used as the contextual similarity, and the average of the similarities of all vectors in the actual nursing path is used as the contextual similarity. Cosine similarity was chosen because it effectively measures the consistency of two high-dimensional vectors in direction, without being affected by their absolute magnitude, which precisely meets the need to assess the logical correlation of tasks.

[0037] The behavior recognition and alarm module is used to construct a two-dimensional parameter space based on the scenario similarity and the economic loss of the nursing path space, obtain the two-dimensional risk coordinates of all nurses in the nursing responsibility group in the two-dimensional parameter space, identify the nurses' behavioral cognitive patterns based on the changing characteristics of the two-dimensional risk coordinates, and issue an alarm.

[0038] Based on the scenario similarity and nursing pathway spatial economic loss obtained in the above modules, the work status of each nurse within the time frame was evaluated from two dimensions: job switching coherence and workflow execution. Therefore, a probabilistic inference model based on a two-dimensional risk state space can be constructed to transform continuous numerical indicators into hierarchical and interpretable early warning signals that meet the needs of clinical management decision-making.

[0039] In this embodiment of the invention, a two-dimensional parameter space is constructed using scenario similarity as the Y-axis and the nursing path spatial economic loss as the X-axis. The two-dimensional parameter space contains four quadrants, i.e., four patterns: (1) The first quadrant represents high spatial economic loss in nursing pathways and high scenario similarity. This region is defined as the expert workaround mode in this embodiment of the invention. In order to maintain the logical coherence of task processing (high Y-axis value), nurses intentionally choose physically suboptimal paths (high X-axis value). This is a behavior that trades spatial cost for cognitive efficiency.

[0040] (2) The second quadrant represents low spatial economic loss in nursing pathways and high scenario similarity. This region is defined as an efficient flow pattern in this embodiment of the invention. The nurse's physical movement path is economical (low X-axis value), and the tasks handled consecutively are highly similar in scenario (high Y-axis value), indicating that the nurse is handling batches of homogeneous nursing tasks in an orderly manner with low cognitive cost.

[0041] (3) The third quadrant represents low spatial economic loss in nursing pathways and low situational similarity. In this embodiment of the invention, it is defined as a rigid behavioral pattern. The nurse's movement path appears economical (low X-axis value), but the tasks performed are situationally unrelated and involve frequent jumps (low Y-axis value). This reflects that nurses may abandon clinical logical thinking due to fatigue and instead adopt inefficient patterns such as mechanical, fixed-order rounds.

[0042] (4) The fourth quadrant represents high spatial economic loss in nursing pathways and low scenario similarity. This region is defined as a cognitive failure mode in this embodiment of the invention. The nurse's behavior is spatially aimless (high X-axis value), while the nurse's thinking jumps chaotically between different scenarios (low Y-axis value). This state indicates that the nurse's cognition and behavior are out of sync.

[0043] Based on the regional representation of coordinate points in two-dimensional space, the cognitive patterns of nurses in the current nursing responsibility group can be diagnosed and alarms can be triggered. Therefore, the two-dimensional risk coordinates of all nurses in the nursing responsibility group are obtained in the two-dimensional parameter space. The behavioral cognitive patterns of nurses are identified and alarms are triggered based on the changing characteristics of the two-dimensional risk coordinates.

[0044] Preferably, in this embodiment of the invention, a Hidden Markov Model (HMM) is introduced for dynamic inference to capture the evolution trend of working modes over time. First, the structure of the HMM is defined, and the hidden states of the model are defined as the four typical working modes identified above: {efficient flow, expert flexibility, behavioral rigidity, and cognitive failure}. The model's emission probability learns the two-dimensional Gaussian mixture distribution of the observed two-dimensional risk coordinates in each hidden state. For example, the model learns that when in the "cognitive failure" state, the probability of the observed coordinate point falling in the third quadrant is the highest. The model's state transition probabilities are obtained by learning from historical data to obtain the transition probability matrix between different working modes.

[0045] Then, the average scenario similarity and average nursing path spatial economic loss of all nurses in the nursing responsibility group are mapped to the two-dimensional parameter space to obtain the two-dimensional risk coordinates of the nursing responsibility group; Two-dimensional risk coordinates are used as observation vectors for the HMM algorithm. The time series of the observation vectors are input into the trained HMM model. The Viterbi algorithm is used to determine the nurse's hidden working mode and feed back an alarm signal. The forward algorithm is used to determine the probability of transitioning from the hidden working mode to the cognitive failure mode at a future time. The alarm signal is fed back based on the probability.

[0046] Furthermore, in this embodiment of the invention, a two-level early warning mode is set. If the hidden working mode is a behavioral rigidity mode, a first-level early warning signal is fed back. The content of the first-level early warning signal may include: "First-level early warning: [Nurse N] responsibility group has entered the [behavioral rigidity] state. Risk coordinate analysis: The spatial economic loss of nursing pathways is low, but the scenario similarity of continuous nursing tasks is consistently low, suggesting a tendency to patrol according to fixed routes while ignoring task logic." This early warning provides managers with an early and interpretable basis for intervention, allowing them to take measures such as conversations and adjusting task allocation before the system function is severely impaired.

[0047] If the probability exceeds a preset probability threshold, or if the hidden working mode is a cognitive failure mode, a secondary warning signal is issued. The secondary warning signal may include: "[Nurse N]'s responsible group is about to enter or has already entered a [cognitive failure] state (transfer probability: [P%]). Risk coordinate analysis: Both spatial economic loss and task scenario similarity are in the disadvantage range, indicating that behavior and cognition are out of sync, and the patient group under their responsibility has a high risk of experiencing a group of adverse events; immediate intervention is recommended." This warning signifies that the system's internal regulatory mechanism is on the verge of collapse, providing managers with a decision-making basis for taking mandatory measures (such as assigning personnel to take over or forcibly reorganizing tasks) within the window before adverse events occur.

[0048] In summary, this invention quantifies data on the patients and specific nursing content for each nursing task to obtain real-time scenario-related vectors. The spatial states of all patients in the nursing responsibility group are quantified, and a nursing opportunity cost matrix is ​​constructed to determine the spatial economic loss of the nursing path. Scenario similarity is obtained to characterize the coherence of nurses' workload. Scenario similarity and spatial economic loss of the nursing path intuitively reflect the nurses' work patterns. By constructing a two-dimensional parameter space and observing the position and changes of coordinates in the space, the cognitive patterns of nurses in the current nursing responsibility group can be determined, and effective alarms can be issued. This invention effectively integrates nurses' behavioral time-series data and patients' own states through vector fusion, matrix construction, and other methods. By analyzing the nurses' cognitive state patterns, effective alarm signals are fed back to avoid nursing risks.

[0049] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A nursing behavior pattern recognition platform, characterized in that, The platform includes: The scenario state quantification module is used to obtain the multi-scale physiological complexity of each patient at the beginning of each nursing task based on the fluctuation characteristics of each patient's vital sign sequence at different time scales. The multi-scale physiological complexity and the patient's diagnosis result are encoded to form the patient's state vector. The state vector is fused with the nursing task content to obtain the real-time scenario association vector of each nursing task. The nursing pathway analysis module is used to construct a nursing opportunity cost matrix for all patients based on the multi-scale physiological complexity of each patient and the distance between patients; perform path optimization on the nursing opportunity cost matrix to obtain the theoretically optimal nursing pathway; and obtain the spatial economic loss of the nursing pathway based on the cost difference between the nurse's actual nursing pathway and the theoretically optimal nursing pathway. The scenario similarity acquisition module is used to obtain the scenario similarity of consecutive nursing tasks based on the vector similarity between the real-time scenario association vectors of adjacent nursing tasks in the actual nursing path. The behavior recognition and alarm module is used to construct a two-dimensional parameter space based on the scenario similarity and the economic loss of the nursing path space, obtain the two-dimensional risk coordinates of all nurses in the nursing responsibility group in the two-dimensional parameter space, identify the nurses' behavioral cognitive patterns based on the changing characteristics of the two-dimensional risk coordinates, and issue an alarm.

2. The nursing behavior pattern recognition platform according to claim 1, characterized in that, The method for obtaining the multi-scale physiological complexity includes: The vital sign sequence is processed using a multi-scale sample entropy algorithm to obtain a sample entropy-scale curve; the area enclosed by the sample entropy-scale curve and the coordinate system is taken as the multi-scale physiological complexity.

3. The nursing behavior pattern recognition platform according to claim 1, characterized in that, The method for obtaining the state vector includes: The state vector is obtained by encoding and merging the multi-scale physiological complexity with the static diagnostic information of the patient's basic pathology.

4. The nursing behavior pattern recognition platform according to claim 1, characterized in that, The method for obtaining the real-time scenario association vector includes: The nursing task text is processed using a natural language processing model to obtain a nursing task encoding vector; the nursing task encoding vector and the state vector are input into a pre-trained attention mechanism neural network to output the real-time context association vector.

5. A nursing behavior pattern recognition platform according to claim 1, characterized in that, The method for constructing the nursing opportunity cost matrix includes: The nursing opportunity cost matrix is ​​N×N, where N is the number of patients. Each element in the nursing opportunity cost matrix represents the cost of moving from one patient to another, and the cost is obtained based on the distance between the two patients and the multi-scale physiological complexity of the second patient.

6. The nursing behavior pattern recognition platform according to claim 1, characterized in that, A genetic algorithm-based Traveling Salesman Problem (TSP) solver is used for path optimization.

7. The nursing behavior pattern recognition platform according to claim 1, characterized in that, The methods for obtaining the spatial economic loss of the nursing pathway include: The theoretical cost is obtained by summing the costs of the theoretically optimal nursing pathway in the nursing opportunity cost matrix, and the actual cost is obtained by summing the costs of the actual nursing pathway in the nursing opportunity cost matrix. The difference between the actual cost and the theoretical cost is used as the numerator, and the theoretical cost is used as the denominator to obtain the spatial economic loss of the nursing pathway.

8. The nursing behavior pattern recognition platform according to claim 1, characterized in that, The method for obtaining the scenario similarity includes: The cosine similarity between real-time contextual association vectors of adjacent nursing tasks is used as the similarity, and the average of all vector similarities in the actual nursing path is used as the contextual similarity.

9. A nursing behavior pattern recognition platform according to claim 1, characterized in that, Based on the changing characteristics of two-dimensional risk coordinates, the system identifies nurses' behavioral and cognitive patterns and issues alerts, including: The average scenario similarity and average nursing path spatial economic loss of all nurses in the nursing responsibility group are mapped to a two-dimensional parameter space to obtain the two-dimensional risk coordinates of the nursing responsibility group. Two-dimensional risk coordinates are used as observation vectors for the HMM algorithm. The time series of the observation vectors are input into the trained HMM model. The Viterbi algorithm is used to determine the nurse's hidden working mode and feed back an alarm signal. The forward algorithm is used to determine the probability of transitioning from the hidden working mode to the cognitive failure mode at a future time. The alarm signal is fed back based on the probability.

10. A nursing behavior pattern recognition platform according to claim 9, characterized in that, If the hidden working mode is a behavioral rigidity mode, a first-level warning signal is fed back; if the probability is greater than a preset probability threshold, or if the hidden working mode is a cognitive failure mode, a second-level warning signal is fed back.

Images