Intelligent monitoring method and system for monitoring the health of a patient in a hospital ward

Abstract

The application relates to the field of biomedical technology, and particularly relates to a digital management method for hierarchical nursing in a monitoring ward based on the Internet of Things. The method comprises the following steps: collecting isomerism data of a patient in real time through a nursing monitoring device to obtain principle physiological data flow; comprehensively calculating physiological indexes of the principle physiological data flow to obtain comprehensive physiological indexes; predicting individual physiological indexes according to the comprehensive physiological indexes, and predicting a probability of illness deterioration of the patient to obtain illness deterioration probability data; predicting a nursing level requirement according to a prediction result of the individual physiological indexes to obtain nursing level requirement probability data; generating an illness trend prediction vector according to the illness deterioration probability data and the nursing level requirement probability data to obtain the illness trend prediction vector. The application can more accurately predict an illness trend and give a hierarchical early warning, thereby better assisting medical staff in hierarchical nursing of the patient and improving the management level of the monitoring ward.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of biomedical technology, and particularly relates to a digital management method for grading nursing in an intensive care unit based on Internet of Things. BACKGROUND

[0002] As the most intensive department for patient condition monitoring and nursing in a hospital, the intensive care unit has an urgent need for real-time, accurate and personalized nursing. The traditional management mode of the intensive care unit mainly relies on manual recording and observation, which is low in efficiency, prone to errors and difficult to meet the growing medical needs. Therefore, a digital management method for grading nursing in an intensive care unit based on Internet of Things emerges as the times require. The popularity of various physiological parameter sensors, wearable devices and bedside nursing monitoring devices makes it possible to collect multi-dimensional patient data in real time. The development of computing, big data and artificial intelligence technologies makes it possible to process and analyze massive patient data in real time, providing a technical basis for disease prediction and early warning.

[0003] However, the existing methods still have some deficiencies, such as inaccurate disease trend prediction and inaccurate grading early warning. Some methods only consider a single or a few physiological indicators, ignoring other factors that may affect the development of the disease, such as the patient's age, medical history, medication, etc. Lack of integration of multi-dimensional data leads to insufficient information of the prediction model, affecting the accuracy of the prediction. Some methods only provide early warning for a single indicator, ignoring the correlation between indicators. For example, a patient's heart rate and blood pressure are usually correlated, and if only one of them is considered, some important early warning signals may be missed. SUMMARY

[0004] Therefore, it is necessary to provide a digital management method for grading nursing in an intensive care unit based on Internet of Things to solve at least one of the above technical problems.

[0005] To achieve the above-mentioned purpose, a digital management method for grading nursing in an intensive care unit based on Internet of Things comprises the following steps:

[0006] Step S1: Real-time acquisition of heterogeneous data of a patient by a nursing monitoring device to obtain a principle physiological data stream; comprehensive physiological indicator calculation on the principle physiological data stream to obtain a comprehensive physiological indicator;

[0007] Step S2: Individualized physiological indicator prediction according to the comprehensive physiological indicator, and patient disease deterioration probability prediction to obtain disease deterioration probability data; nursing level demand prediction according to the individualized physiological indicator prediction result to obtain nursing level demand probability data; disease trend prediction vector generation on the disease deterioration probability data and the nursing level demand probability data to obtain a disease trend prediction vector;

[0008] Step S3: dynamically setting a warning threshold according to the disease trend prediction vector and a preset warning rule to obtain a dynamic threshold matrix; triggering a warning signal and determining a warning level according to the disease trend prediction vector and the dynamic threshold matrix to obtain warning level data;

[0009] Step S4: recommending a nursing resource according to the disease trend prediction vector and the warning level data to obtain a recommended nursing scheme;

[0010] Step S5: tracking a physiological index change according to the recommended nursing scheme to obtain a physiological index change trajectory; and evaluating a nursing effect according to the physiological index change trajectory to obtain a nursing effect index, so as to realize the digital management task of the graded nursing in the monitoring ward.

[0011] The present application realizes comprehensive and accurate quantitative evaluation of the physiological state of the patient by collecting multi-source heterogeneous physiological data in real time and calculating the comprehensive physiological index (IPI) by using the deep learning model, and lays a solid data foundation for subsequent personalized prediction and accurate warning. The future physiological index of the patient is predicted by the individualized physiological model, and the disease trend prediction vector (PTV) is generated by combining the disease deterioration probability and the nursing level demand prediction, so as to realize the individualized and forward-looking prediction of the patient's disease development trend, and provide key information for early intervention and resource allocation. The warning threshold is dynamically set based on the PTV and the preset rule, and the comprehensive risk score is calculated by combining the patient's historical data and the risk factor weight, so as to realize the intelligent and graded warning mechanism, and more accurately identify high-risk patients. The individualized nursing scheme is intelligently recommended by comprehensively considering the disease trend, individual characteristics and available resources of the patient, so as to improve the accuracy and feasibility of the nursing scheme, and ensure that the patient receives the most appropriate nursing. The nursing effect is evaluated in real time by continuously tracking the physiological index change and the nursing execution of the patient, and the nursing scheme is dynamically adjusted according to the evaluation result, so as to realize the closed-loop management of the nursing process, and continuously optimize the nursing strategy, and finally improve the nursing quality and patient safety. Therefore, the present application provides a monitoring ward graded nursing digital management method based on the Internet of Things, which realizes more accurate disease trend prediction and graded warning by the technologies of individualized modeling, multi-dimensional data fusion, deep learning and dynamic threshold setting, so as to better assist medical staff in grading nursing of patients, and improve the management level of the monitoring ward.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: collecting heterogeneous data of the patient in real time by the nursing monitoring device to obtain the original physiological data stream;

[0014] Step S12: Associate the patient identity with the principle physiological data stream to obtain patient identity associated data; perform physiological data synchronization processing based on the patient identity associated data to obtain synchronized patient physiological data;

[0015] Step S13: Extract physiological data features from the patient's synchronous physiological data to obtain a physiological feature vector;

[0016] Step S14: Calculate the comprehensive physiological index based on the physiological feature vector to obtain the comprehensive physiological index.

[0017] This invention utilizes various nursing monitoring devices and sensors to collect heterogeneous patient data in real time, comprehensively capturing changes in the patient's physiological state, avoiding information omissions, and providing a richer data foundation for subsequent analysis and decision-making. Real-time data acquisition also promptly reflects dynamic changes in the patient's condition, aiding in the early detection of potential risks. Patient identity association and time synchronization processing of the raw physiological data stream ensure data accuracy and consistency, avoiding data confusion and errors. Time synchronization aligns data from different sampling frequencies, facilitating subsequent feature extraction and analysis, and improving the efficiency and accuracy of data processing. Extracting clinically significant features from the synchronized physiological data transforms complex waveform and numerical data into more representative feature vectors, reducing data dimensionality, computational load, and improving the efficiency and accuracy of subsequent analysis. The extracted features more effectively reflect changes in the patient's physiological state, providing a more reliable basis for disease assessment and prediction. Calculating the Integrated Physiological Index (IPI) using a deep learning model integrates multi-dimensional physiological characteristics into a single index, more comprehensively reflecting the patient's overall physiological state. The calculation of IPI takes into account the correlation and weight of various indicators, enabling a more accurate assessment of the patient's condition and providing a more reliable basis for subsequent disease prediction and early warning. Using deep learning models can capture complex nonlinear relationships, reflecting the patient's physiological state more accurately than traditional linear models.

[0018] Preferably, step S2 includes the following steps:

[0019] Step S21: Construct a time series of comprehensive physiological indicators to obtain the time series of comprehensive physiological indicators;

[0020] Step S22: Perform individualized physiological pattern learning on the time series of comprehensive physiological indicators to obtain an individualized physiological model;

[0021] Step S23: Use individualized physiological models to predict individualized physiological indicators and obtain the prediction results of individualized physiological indicators;

[0022] Step S24: Based on the prediction results of individualized physiological indicators, predict the probability of the patient's condition worsening and obtain the probability data of the patient's condition worsening.

[0023] Step S25: Based on the individualized physiological indicator prediction results and the preset nursing level classification rules, predict the nursing level demand to obtain nursing level demand probability data;

[0024] Step S26: Generate a disease trend prediction vector from the probability data of disease deterioration and the probability data of nursing level requirements.

[0025] This invention constructs IPI data into a time series, preserving information about IPI changes over time and providing a foundation for subsequent time series analysis and prediction. Time series data reflects the dynamic trends of patients' physiological states, helping to more accurately predict disease progression. Using an LSTM model to learn individualized physiological patterns from IPI time series can capture the patient-specific physiological change patterns. Compared to general models, individualized models can more accurately reflect each patient's unique physiological patterns, thereby improving prediction accuracy and personalization. Predicting future IPI values ​​using individualized physiological models can more accurately predict future trends in patients' physiological states. By analyzing IPI prediction results, the probability of patient disease deterioration can be predicted, providing early warning of potential deterioration risks, enabling medical staff to take timely intervention measures and reduce the possibility of disease progression. Based on IPI prediction results and preset nursing level classification rules, the probability of patients needing different nursing levels can be predicted, helping to plan and allocate nursing resources in advance, optimize resource allocation, avoid resource waste, and ensure patients receive timely and effective care. By integrating the probability of disease deterioration and the probability of nursing level requirements, a disease trend prediction vector (PTV) is generated, which can provide a more comprehensive and integrated disease prediction result. This provides a more reliable basis for subsequent early warning triggering and nursing plan recommendations, and helps to improve the accuracy of nursing decisions.

[0026] Preferably, step S25 includes the following steps:

[0027] Step S251: Predict the trend of physiological indicator changes based on the individualized physiological indicator prediction results to obtain the nursing demand trend prediction sequence;

[0028] Step S252: Perform dynamic and static nursing level matching based on the nursing demand trend prediction sequence and the preset nursing level classification rules to obtain nursing demand matching results. The preset nursing level classification rules include static classification rules and dynamic classification rules.

[0029] Step S253: Construct a nursing demand probability model based on the nursing demand matching results to obtain the nursing demand probability model; calculate the nursing demand probability based on the nursing demand probability model and the nursing demand matching results to obtain the nursing demand probability matrix.

[0030] Step S254: Obtain nursing resource allocation data; optimize the nursing resource demand based on the nursing demand probability matrix according to the nursing resource allocation data to obtain nursing level demand probability data.

[0031] This invention, through trend analysis of individualized physiological indicator prediction results, obtains a nursing demand trend prediction sequence. This allows for a more accurate prediction of future changes in patients' nursing needs, rather than relying solely on current physiological indicator values. This provides more precise nursing management for advance preparation and adjustment of nursing plans. Combining static classification rules (based on absolute IPI values) and dynamic classification rules (based on IPI change rates) for nursing level matching allows for a more comprehensive consideration of patients' condition and trends, avoiding the limitations of a single rule and improving the accuracy and reliability of nursing level assessment. By constructing a nursing demand probability model (e.g., GMM) and calculating the nursing demand probability matrix, the predicted nursing demand results can be expressed in probabilistic form, more accurately quantifying the degree of patients' need for different nursing levels and providing more refined decision support for the optimal allocation of nursing resources. Optimizing the nursing demand probability matrix based on nursing resource allocation data can incorporate the availability of nursing resources into the demand prediction of nursing levels, making the prediction results more in line with the actual situation. This avoids situations where patients' nursing needs cannot be met due to insufficient resources, thereby improving the utilization efficiency of nursing resources and ensuring that patients receive the most appropriate level of care.

[0032] Preferably, step S252 specifically includes:

[0033] The nursing demand trend prediction sequence and static segmentation rules are matched point by point to obtain the point-by-point matching results; preliminary labeling is performed based on the point-by-point matching results to obtain the static matching results.

[0034] Trend features are extracted from the nursing demand trend prediction sequence to obtain trend features; dynamic rule matching is performed on the trend features and dynamic partitioning rules to obtain dynamic matching results;

[0035] The static and dynamic matching results are fused together, and matching conflict elimination is performed to obtain the fused matching result.

[0036] Nursing needs matching results are generated based on the fusion matching results.

[0037] This invention uses static segmentation rules to initially assess nursing needs at each time point, providing a baseline level of nursing needs and a reference for subsequent dynamic adjustments. By extracting trend features and applying dynamic segmentation rules, it can capture dynamic changes in nursing needs and make more refined adjustments to nursing levels, thus more accurately reflecting the actual development trend of the patient's condition. Integrating static and dynamic matching results and performing conflict resolution allows for a comprehensive consideration of the patient's baseline nursing needs and dynamic trends, avoiding the limitations of a single rule and obtaining more reliable nursing level assessment results. Prioritizing dynamic matching results highlights the importance of monitoring the trend of disease changes, facilitating timely adjustments to nursing strategies. The final nursing needs matching result integrates static and dynamic assessments, providing a more comprehensive and accurate prediction of nursing levels, offering a more reliable basis for subsequent nursing resource recommendations and nursing plan adjustments.

[0038] Preferably, step S3 includes the following steps:

[0039] Step S31: Set dynamic warning thresholds based on the disease trend prediction vector and preset warning rules to obtain a dynamic threshold matrix;

[0040] Step S32: Perform single-indicator comparison on the disease trend prediction vector and the dynamic threshold matrix to obtain the single-indicator comparison result; perform multi-indicator fusion on the single-indicator comparison result to obtain the multi-indicator fusion result.

[0041] Step S33: Based on the multi-indicator fusion results and the dynamic threshold matrix, determine the early warning trigger and obtain the early warning trigger signal;

[0042] Step S34: Calculate the risk factor weights based on the disease trend prediction vector, the patient's historical physiological data, and the preset early warning classification results to obtain the risk factor weight table;

[0043] Step S35: Calculate the risk score based on the risk factor weight table and the early warning trigger signal to obtain the comprehensive risk score;

[0044] Step S36: Divide the warning levels according to the comprehensive risk score and the preset warning classification results to obtain warning level data.

[0045] This invention dynamically adjusts the warning threshold based on the disease trend prediction vector, avoiding the limitations of fixed thresholds. It enables personalized warnings based on individual patient conditions, improving the sensitivity and accuracy of warnings and reducing false alarms and missed alarms. First, single-indicator comparison is performed, followed by multi-indicator fusion, allowing for a more comprehensive assessment of the patient's disease risk. Multi-indicator fusion considers the combined effects of multiple risk factors, avoiding the one-sidedness of single indicators, thus improving the reliability of warnings. Warning triggering is based on the multi-indicator fusion results and the dynamic threshold matrix, enabling more accurate determination of whether patient intervention is needed and timely triggering of warnings, providing timely risk alerts to medical staff so they can take early intervention measures. By combining the disease trend prediction vector, the patient's historical physiological data, and the preset warning classification results to calculate risk factor weights, the impact of different risk factors on the patient's condition can be more accurately assessed, providing a more scientific basis for risk scoring. Calculating a comprehensive risk score based on the risk factor weight table integrates the effects of multiple risk factors, quantifying the patient's overall risk level and providing a more precise basis for warning level classification. Classifying early warning levels based on comprehensive risk scores allows for the grading of patients' risk levels, enabling healthcare professionals to take appropriate intervention measures based on different early warning levels, thereby improving the targetedness and effectiveness of nursing care.

[0046] Preferably, step S31 includes the following steps:

[0047] Step S311: Obtain the patient's historical physiological data; construct an early warning Bayesian network based on the patient's historical physiological data and the disease trend prediction vector to obtain the early warning Bayesian network;

[0048] Step S312: Use the patient's historical physiological data to set the probability distribution of the early warning Bayesian network to obtain a probabilistic Bayesian network;

[0049] Step S313: Extract the patient's current physiological indicators from the principle physiological data stream to obtain the patient's current physiological indicators;

[0050] Step S314: Input the patient's current physiological indicators and disease trend prediction vector into a probabilistic Bayesian network to calculate the dynamic threshold and obtain the dynamic threshold matrix.

[0051] This invention constructs a Bayesian network using historical patient physiological data and disease trend prediction vectors. This network effectively expresses and learns the complex probabilistic relationships between different physiological indicators, risk factors, and disease trends, providing a more interpretable and reasoning-capable model for dynamic threshold setting. Setting the conditional probability distribution of the Bayesian network using historical patient physiological data incorporates individual patient differences and historical information into the model, making it more closely aligned with the patient's actual situation and improving the accuracy and personalization of dynamic threshold setting. Real-time extraction of the patient's current physiological indicators captures the latest changes in the patient's physiological state, providing real-time input data for dynamic threshold calculation and ensuring that the warning threshold is dynamically adjusted according to changes in the patient's condition. Inputting the patient's current physiological indicators and disease trend prediction vectors into a probabilistic Bayesian network for inference allows for dynamic calculation of the warning threshold based on the patient's real-time state and predicted trends, improving the sensitivity and accuracy of warnings and better adapting to dynamic changes in the patient's condition. Using a Bayesian network for inference effectively utilizes prior knowledge and real-time data, improving the reliability of threshold setting.

[0052] Preferably, step S311 specifically includes:

[0053] The patient's historical physiological data and disease trend prediction vector were preprocessed to obtain preprocessed data.

[0054] An initial node set is defined based on the preprocessed data to obtain the initial node set; node similarity is calculated based on the preprocessed data and the initial node set to obtain the node similarity matrix;

[0055] Perform node clustering analysis on the node similarity matrix to obtain the node clustering results; construct a hierarchical structure based on the node clustering results and the initial node set to obtain the node hierarchical structure;

[0056] By using preprocessed data to learn dependencies in the node hierarchy, a node hierarchy with dependencies is obtained.

[0057] A Bayesian network for early warning is constructed based on the hierarchical structure of nodes with dependencies.

[0058] This invention improves data quality and consistency through data preprocessing steps, including missing value handling, standardization, and one-hot encoding. This provides a more reliable data foundation for subsequent analysis and modeling, and enhances model performance and stability. A clearly defined initial node set ensures the Bayesian network includes all relevant variables, laying the foundation for building a model that comprehensively reflects the patient's physiological state and disease trends. Calculating the node similarity matrix quantifies the correlation between different variables, providing a basis for subsequent clustering analysis and hierarchical structure construction, and helping to discover potential relationships between variables. Node clustering analysis groups highly similar nodes together, simplifying the Bayesian network structure and improving model efficiency and interpretability, while also helping to discover intrinsic connections between variables. Constructing a node hierarchy based on the clustering results reflects the hierarchical relationships between variables, making the Bayesian network structure clearer, easier to understand and interpret, and helping to more effectively express dependencies between variables. Learning node dependencies through a data-driven approach avoids subjective human intervention and makes the Bayesian network structure more consistent with the actual data distribution, thereby improving the model's accuracy and reliability. The final early warning Bayesian network, based on a data-driven hierarchical structure and dependencies, can more accurately reflect the complex relationship between the patient's physiological state and disease trend, and provide a more reliable model foundation for subsequent dynamic threshold calculation and early warning.

[0059] Preferably, step S4 includes the following steps:

[0060] Step S41: Retrieve individualized patient information based on the disease trend prediction vector to obtain patient feature profiles;

[0061] Step S42: Obtain the status of nursing resources based on the warning level data to obtain a list of available resources;

[0062] Step S43: Generate an initial nursing care plan based on the warning level data, patient characteristic profile, and available resource list;

[0063] Step S44: Based on the patient's characteristic profile and the list of available resources, the initial nursing plan is adjusted to obtain a recommended nursing plan.

[0064] This invention, by retrieving individualized patient information, including age, gender, allergy history, and past medical history, can construct a more complete patient profile, providing more comprehensive information support for subsequent nursing plan development and facilitating the creation of more personalized care plans. Obtaining real-time nursing resource status information based on the alert level, such as available beds, on-duty medical staff, and nursing monitoring equipment, ensures the feasibility of recommended nursing plans in practice, avoiding situations where insufficient resources prevent the implementation of nursing plans and improving the efficiency of nursing resource utilization. Generating an initial nursing plan based on the alert level, patient profile, and available resource list can quickly provide a preliminary nursing plan framework, laying the foundation for subsequent personalized adjustments and accelerating the nursing plan development process. Personalizing the initial nursing plan based on the patient profile and available resource list ensures that the recommended nursing plan better meets the individual needs and actual situation of the patient, improving the effectiveness and safety of nursing care. Adjustments are made considering resource constraints to ensure the feasibility of the plan.

[0065] Preferably, step S5 includes the following steps:

[0066] Step S51: Record the nursing care execution status according to the recommended nursing care plan to obtain nursing care execution record data;

[0067] Step S52: Track changes in physiological indicators according to the recommended nursing plan to obtain the trajectory of physiological indicator changes;

[0068] Step S53: Conduct a preliminary assessment of the nursing effect based on the nursing execution record data and the trajectory of physiological indicator changes, and obtain preliminary effect assessment data;

[0069] Step S54: Based on the preliminary effect evaluation data and recommended nursing plan, dynamically adjust the nursing plan to obtain the adjusted nursing plan;

[0070] Step S55: Generate nursing effect indicators based on the preliminary effect evaluation data and the adjusted nursing plan.

[0071] This invention provides objective data support for nursing effectiveness evaluation by meticulously recording nursing procedures, including the content, time, executor, and patient response. This facilitates tracking the nursing process and improves the transparency and traceability of nursing care. Continuously tracking changes in patients' physiological indicators allows for dynamic monitoring of the effectiveness of nursing plans, providing real-time physiological data for evaluation and helping to promptly detect changes in the patient's condition. Preliminary assessments based on nursing execution records and physiological indicator changes quickly determine the effectiveness of the nursing plan, providing timely feedback for dynamic adjustments and helping to optimize nursing strategies. Dynamically adjusting the nursing plan based on the preliminary assessment results allows healthcare professionals to personalize the plan according to the patient's actual condition and nursing outcomes, improving the precision and effectiveness of nursing care and ensuring that patients always receive the most appropriate care. Generating Care Effectiveness Indicators (CEIs) quantifies nursing effectiveness, providing objective indicators for nursing quality evaluation and data support for continuous improvement of nursing plans, thus driving continuous improvement in nursing quality. Using CEIs allows for more effective comparison of the effects of different nursing plans and enables data-driven nursing decisions. Attached Figure Description

[0072] Figure 1 A flowchart illustrating the steps of a digital management method for graded nursing care in an intensive care unit based on the Internet of Things (IoT).

[0073] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0074] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0075] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0076] The technical method 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0077] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0078] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0079] To achieve the above objectives, please refer to Figures 1 to 3 A digital management method for graded nursing care in intensive care units based on the Internet of Things includes the following steps:

[0080] Step S1: Collect heterogeneous data from patients in real time using nursing monitoring equipment to obtain the underlying physiological data stream; calculate the comprehensive physiological indicators from the underlying physiological data stream to obtain the comprehensive physiological indicators;

[0081] Step S2: Based on comprehensive physiological indicators, perform individualized physiological indicator prediction and predict the probability of patient condition deterioration to obtain condition deterioration probability data; based on the individualized physiological indicator prediction results, predict nursing level needs to obtain nursing level need probability data; generate a condition trend prediction vector from the condition deterioration probability data and the nursing level need probability data.

[0082] Step S3: Set dynamic warning thresholds based on the disease trend prediction vector and preset warning rules to obtain a dynamic threshold matrix; trigger warning signals based on the disease trend prediction vector and dynamic threshold matrix, and determine the warning level to obtain warning level data.

[0083] Step S4: Recommend nursing resources based on the disease trend prediction vector and early warning level data to obtain a recommended nursing plan;

[0084] Step S5: Recommend nursing plans and track changes in physiological indicators to obtain the trajectory of physiological indicator changes; evaluate the nursing effect based on the trajectory of physiological indicator changes to obtain nursing effect indicators, so as to realize the task of digital management of hierarchical nursing in the intensive care unit.

[0085] In this embodiment of the invention, reference Figure 1 The diagram shown is a flowchart illustrating the steps of the IoT-based digital management method for graded nursing care in intensive care units according to the present invention. In this example, the IoT-based digital management method for graded nursing care in intensive care units includes the following steps:

[0086] Step S1: Collect heterogeneous data from patients in real time using nursing monitoring equipment to obtain the underlying physiological data stream; calculate the comprehensive physiological indicators from the underlying physiological data stream to obtain the comprehensive physiological indicators;

[0087] In this embodiment of the invention, various patient care monitoring devices and sensors (such as ECG sensors, blood pressure monitors, pulse oximeters, smart mattresses, bedside monitors, environmental sensors, etc.) are connected to collect multiple physiological and environmental data from the patient in real time, forming a raw physiological data stream (RPDS). Using the patient's unique RFID identifier, data from different devices are associated with the patient's identity and time-synchronized to generate patient synchronized physiological data (PSPD). Then, physiological features such as heart rate variability, blood pressure characteristics, blood oxygen saturation characteristics, and respiratory characteristics are extracted from the PSPD to form a physiological feature vector (PFV). Finally, a pre-trained deep learning model (e.g., an MLP-based model) is used to calculate the PFV to obtain an integrated physiological index (IPI), which is then stored in the time-series database Influ×DB.

[0088] Step S2: Based on comprehensive physiological indicators, perform individualized physiological indicator prediction and predict the probability of patient condition deterioration to obtain condition deterioration probability data; based on the individualized physiological indicator prediction results, predict nursing level needs to obtain nursing level need probability data; generate a condition trend prediction vector from the condition deterioration probability data and the nursing level need probability data.

[0089] In this embodiment of the invention, the patient's IPI data for the past 24 hours is read from Influ×DB to construct an IPI time series (IPITS). An LSTM model is used to train the IPITS to learn the patient's individualized physiological patterns, resulting in an individualized physiological model (IPM). The IPI value for the next 12 hours is predicted using the IPI prediction. Based on the IPI prediction results, the probability of disease deterioration (POD) is calculated, and the probability of needing different levels of care in the next 12 hours is predicted according to preset care level classification rules and the IPI change rate. Finally, the POD and the probability of needing different levels of care are integrated to generate a disease trend prediction vector (PTV).

[0090] Step S3: Set dynamic warning thresholds based on the disease trend prediction vector and preset warning rules to obtain a dynamic threshold matrix; trigger warning signals based on the disease trend prediction vector and dynamic threshold matrix, and determine the warning level to obtain warning level data.

[0091] In this embodiment of the invention, the warning threshold is dynamically adjusted based on the Patient Value Matrix (PTV) and preset warning rules to form a Dynamic Threshold Matrix (DTM). Each indicator in the PTV is compared with its corresponding threshold in the DTM, and the comparison results are logically ANDed to determine whether a warning is triggered. If a warning is triggered, the risk factor weights are calculated based on the PTV, the patient's historical physiological data, and the preset warning classification results. A comprehensive risk score is then calculated based on the risk factor weights and the warning trigger signal. Finally, the warning level is determined based on the comprehensive risk score and the preset warning classification results.

[0092] Step S4: Recommend nursing resources based on the disease trend prediction vector and early warning level data to obtain a recommended nursing plan;

[0093] In this embodiment of the invention, individualized patient information is retrieved from the electronic medical record system based on the Patient Feature Record (PTV) to form a patient feature profile. Available nursing resource status information is obtained from the hospital resource management system based on the alert level to form an available resource list. Based on the alert level, patient feature profile, and available resource list, a matching initial nursing plan is retrieved from a predefined nursing plan template library. Then, based on the patient feature profile and available resource list, medical staff personalize the initial nursing plan to ultimately obtain a recommended nursing plan.

[0094] Step S5: Recommend nursing plans and track changes in physiological indicators to obtain the trajectory of physiological indicator changes; evaluate the nursing effect based on the trajectory of physiological indicator changes to obtain nursing effect indicators, so as to realize the digital management of hierarchical nursing in the intensive care unit.

[0095] In this embodiment of the invention, medical staff record the implementation of recommended nursing plans using mobile nursing terminals or electronic nursing record systems, forming nursing execution record data. The system continuously collects patients' physiological indicator data and records the physiological indicator data before and after nursing execution according to the monitoring frequency specified in the recommended nursing plan, forming a physiological indicator change trajectory. Based on the nursing execution record data and the physiological indicator change trajectory, the nursing effect is initially assessed. Based on the initial assessment results and the recommended nursing plan, the nursing plan is dynamically adjusted. Finally, based on the initial assessment results and the adjusted nursing plan, a nursing effectiveness index (CEI) is generated.

[0096] Preferably, step S1 includes the following steps:

[0097] Step S11: Collect heterogeneous data from patients in real time using nursing monitoring equipment to obtain the underlying physiological data stream;

[0098] Step S12: Associate the patient identity with the principle physiological data stream to obtain patient identity associated data; perform physiological data synchronization processing based on the patient identity associated data to obtain synchronized patient physiological data;

[0099] Step S13: Extract physiological data features from the patient's synchronous physiological data to obtain a physiological feature vector;

[0100] Step S14: Calculate the comprehensive physiological index based on the physiological feature vector to obtain the comprehensive physiological index.

[0101] In this embodiment of the invention, real-time data collection from patients is achieved in the intensive care unit through various deployed nursing monitoring devices and sensors. For example, an ECG sensor attached to the patient's chest collects electrocardiogram data at a frequency of 250Hz; a non-invasive blood pressure monitor connected to the patient's arm collects blood pressure data every minute; a pulse oximeter collects blood oxygen saturation and pulse rate data every second; and a smart mattress sensor monitors the patient's position and respiratory rate in real time. Bedside monitors transmit comprehensive data including heart rate, respiratory rate, blood pressure, and blood oxygen saturation every 5 seconds via a standard interface (e.g., HL7). Environmental temperature and humidity sensors collect ambient temperature and humidity data every minute. All collected data is timestamped and transmitted in real-time to a central data processing server via the ward's wireless network, forming a Raw Physiological Data Stream (RPDS). The RPDS data stream is in JSON format and stored in a Kafka distributed message queue for subsequent processing.

[0102] RPDS data is read from a Kafka message queue. Each data packet contains a unique RFID identifier for the patient. The system uses these RFID identifiers to associate data from different sensors and nursing monitoring devices with the corresponding patient identity, generating patient identity association data. Considering the different sampling frequencies of different devices, the system uses a timestamp alignment algorithm to synchronize the patient identity association data. For example, using a 1-minute time window, all data collected from the same patient within that time window is aggregated. If data from a certain sensor is missing within a time window, it is filled using data from the previous time window. After synchronization, synchronized physiological data (PSPD) is generated, containing multi-dimensional physiological data for each patient within each time window. PSPD data is stored in Parquet format in the Hadoop Distributed File System (HDFS).

[0103] PSPD data is read from HDFS. Feature extraction is performed on the patient's synchronous physiological data within each time window. For example, wavelet transform is used to extract heart rate variability (HRV) features from ECG data; peak detection algorithm is used to extract systolic blood pressure, diastolic blood pressure, and mean arterial pressure from blood pressure data; statistical methods are used to extract the mean, standard deviation, and coefficient of variation of blood oxygen saturation from blood oxygen saturation data; and fast Fourier transform is used to extract respiratory rate and amplitude features from respiratory signals. All extracted features are combined into a numerical vector, namely the physiological feature vector (PFV). PFV data is stored in an HBase database, using patient ID and time window as primary keys.

[0104] PFV data is read from the HBase database. A pre-trained deep learning model (e.g., a multilayer perceptron (MLP) based model) is used to calculate the Integrated Physiological Index (IPI) of the PFV. The model takes the PFV as input and outputs a value between 0 and 1, representing the patient's overall physiological state. For example, if the PFV is input into a trained MLP model and the model outputs 0.85, it indicates that the patient's current physiological state is good. The IPI data, along with the corresponding timestamps, is stored in the time-series database Influ×DB for subsequent trend analysis.

[0105] Preferably, step S2 includes the following steps:

[0106] Step S21: Construct a time series of comprehensive physiological indicators to obtain the time series of comprehensive physiological indicators;

[0107] Step S22: Perform individualized physiological pattern learning on the time series of comprehensive physiological indicators to obtain an individualized physiological model;

[0108] Step S23: Use individualized physiological models to predict individualized physiological indicators and obtain the prediction results of individualized physiological indicators;

[0109] Step S24: Based on the prediction results of individualized physiological indicators, predict the probability of the patient's condition worsening and obtain the probability data of the patient's condition worsening.

[0110] Step S25: Based on the individualized physiological indicator prediction results and the preset nursing level classification rules, predict the nursing level demand to obtain nursing level demand probability data;

[0111] Step S26: Generate a disease trend prediction vector from the probability data of disease deterioration and the probability data of nursing level requirements.

[0112] As an example of the present invention, reference is made to... Figure 2 As shown, step S2 in this example includes:

[0113] Step S21: Construct a time series of comprehensive physiological indicators to obtain the time series of comprehensive physiological indicators;

[0114] In this embodiment of the invention, IPI data for a specified patient over the past 24 hours are read from the Influ×DB database and arranged in timestamp order to generate an IPI time series (IPITS) for that patient. The sampling frequency of the time series is once per minute. If data is missing, linear interpolation is used to fill in the gaps. For example, if the patient ID is "12345", the IPI values ​​for that patient per minute over the past 24 hours are queried from Influ×DB to generate an IPI time series containing 1440 data points, which is then stored as a NumPy array.

[0115] Step S22: Perform individualized physiological pattern learning on the time series of comprehensive physiological indicators to obtain an individualized physiological model;

[0116] In this embodiment of the invention, a Long Short-Term Memory (LSTM) network model is used to train the IPITS generated in step S21 to learn individualized physiological patterns of patients. This LSTM model consists of two LSTM layers, each containing 64 hidden units, followed by a fully connected layer, with an output dimension of 1. Training is performed using the Adam optimizer with a learning rate of 0.001, a batch size of 32, and 100 training epochs. The loss function is the mean squared error (MSE). An early stopping strategy is used during training to prevent overfitting. After training, the trained LSTM model is saved as an individualized physiological model (IPM).

[0117] Step S23: Use individualized physiological models to predict individualized physiological indicators and obtain the prediction results of individualized physiological indicators;

[0118] In this embodiment of the invention, the IPM trained in step S22 is used to predict the IPI for the next 12 hours. The IPI data of the past 12 hours is used as input to the model to predict the IPI value for the next 12 hours, generating a prediction sequence containing 720 data points. For example, the IPI data of patient "12345" for the past 12 hours is input into its corresponding IPM, and the model outputs the predicted IPI value for the next 12 hours.

[0119] Step S24: Based on the prediction results of individualized physiological indicators, predict the probability of the patient's condition worsening and obtain the probability data of the patient's condition worsening.

[0120] In this embodiment of the invention, the probability of a patient's condition worsening is calculated based on the IPI prediction results. An IPI threshold of 0.3 is preset. If any predicted IPI value within the next 12 hours is less than or equal to 0.3, the patient's condition is considered to have potentially worsened. The ratio of the number of time points within the next 12 hours where the IPI value is less than or equal to 0.3 to the total number of predicted time points is calculated as the probability of worsening condition (POD). For example, if patient "12345" has 120 predicted IPI values ​​less than or equal to 0.3 within the next 12 hours, then POD = 120 / 720 = 0.167.

[0121] Step S25: Based on the individualized physiological indicator prediction results and the preset nursing level classification rules, predict the nursing level demand to obtain nursing level demand probability data;

[0122] In this embodiment of the invention, based on the IPI prediction results and preset nursing level classification rules, the probability of a patient needing different nursing levels within the next 12 hours is predicted. The preset nursing level classification rules are as follows: IPI greater than or equal to 0.7 is Level 1 nursing care, IPI between 0.4 and 0.7 is Level 2 nursing care, and IPI less than 0.4 is Level 3 nursing care. The ratio of the number of time points belonging to each nursing level within the next 12 hours to the total predicted time points is calculated as the probability of needing that nursing level. For example, if patient "12345" has an IPI prediction value greater than or equal to 0.7 for 360 time points within the next 12 hours, then the probability of needing Level 1 nursing care is 360 / 720 = 0.5.

[0123] Step S26: Generate a disease trend prediction vector from the probability data of disease deterioration and the probability data of nursing level requirements;

[0124] In this embodiment of the invention, the probability of disease deterioration (POD) and the probability of nursing level requirement calculated in step S25 are integrated together to generate a disease trend prediction vector (PTV). PTV is a four-dimensional vector, representing the probability of disease deterioration, the probability of level 1 nursing requirement, the probability of level 2 nursing requirement, and the probability of level 3 nursing requirement, respectively. For example, the PTV of patient "12345" is [0.167, 0.5, 0.3, 0.033].

[0125] Preferably, step S25 includes the following steps:

[0126] Step S251: Predict the trend of physiological indicator changes based on the individualized physiological indicator prediction results to obtain the nursing demand trend prediction sequence;

[0127] Step S252: Perform static and dynamic nursing level matching based on the nursing demand trend prediction sequence and the preset nursing level division rules to obtain the nursing demand matching result. The preset nursing level division rules include static division rules and dynamic division rules;

[0128] Step S253: Construct a nursing demand probability model based on the nursing demand matching result to obtain the nursing demand probability model; calculate the nursing demand probability according to the nursing demand probability model and the nursing demand matching result to obtain the nursing demand probability matrix;

[0129] Step S254: Obtain the nursing resource allocation data; optimize the nursing demand probability matrix according to the nursing resource allocation data to obtain the nursing level demand probability data.

[0130] In the embodiment of the present invention, based on the individualized physiological index prediction result (IPI prediction sequence for the next 12 hours), the Savitzky-Golay filter is used to smooth the IPI prediction sequence to remove noise and highlight the trend. Then, calculate the change amount of the IPI value at adjacent time points to obtain the IPI change rate sequence. For example, if the IPI prediction value at the i-th time point is 0.6 and the IPI prediction value at the i+1-th time point is 0.55, then the IPI change rate at the i-th time point is 0.55 - 0.6 = -0.05. The IPI change rate sequence is used as the nursing demand trend prediction sequence. A positive value indicates a trend of improvement in the condition, a negative value indicates a trend of deterioration in the condition, and a zero value indicates a stable condition.

[0131] Static division rule: Divide the nursing level according to the absolute value of IPI. When IPI≥0.7, it is first-level nursing; when 0.4≤IPI<0.7, it is second-level nursing; when IPI<0.4, it is third-level nursing. Dynamic division rule: Divide the nursing level according to the IPI change rate. When the IPI change rate ≤ -0.05, it is third-level nursing; when -0.05 < IPI change rate < 0.02, it is second-level nursing; when the IPI change rate ≥ 0.02, it is first-level nursing. Match the nursing demand trend prediction sequence with the static division rule and the dynamic division rule respectively. For example, if the IPI prediction value at a certain time point is 0.5 and the IPI change rate is -0.06, then the static matching result is second-level nursing and the dynamic matching result is third-level nursing. Store the static matching results and dynamic matching results at all time points as two sequences respectively. Subsequently, conflict resolution is performed on the two sequences. Give priority to the dynamic matching result, that is, if the static matching result and the dynamic matching result are inconsistent, then adopt the dynamic matching result. The result after conflict resolution is used as the final nursing demand matching result.

[0132] A Gaussian Mixture Model (GMM) was used to model the nursing demand matching results, resulting in a nursing demand probability model. The GMM had three components, corresponding to the three nursing levels. The Expectation-Maximization (EM) algorithm was used to train the GMM to learn the probability distribution for each nursing level. After training, the nursing demand matching results were input into the trained GMM model to calculate the probability of belonging to each nursing level at each time point, resulting in a 720×3 nursing demand probability matrix. Each row of the matrix represents a time point, and each column represents a nursing level.

[0133] The system retrieves currently available nursing resource allocation data from the hospital information system, including the number of available beds and medical staff for each nursing level. For example, there are 10 available beds for Level 1 nursing, 20 for Level 2 nursing, and 5 for Level 3 nursing. Based on this data, the nursing demand probability matrix is ​​adjusted. For instance, if the probability of demand for Level 3 nursing is high at a certain point in time, but the available beds for Level 3 nursing are full, the probability of demand for Level 3 nursing at that point is reduced, and the excess probability is proportionally distributed to the demand probabilities for Level 1 and Level 2 nursing. The final optimized nursing level demand probability data is still a 720×3 matrix, but it takes into account the limitations of nursing resources.

[0134] Preferably, step S252 specifically includes:

[0135] The nursing demand trend prediction sequence and static segmentation rules are matched point by point to obtain the point-by-point matching results; preliminary labeling is performed based on the point-by-point matching results to obtain the static matching results.

[0136] Trend features are extracted from the nursing demand trend prediction sequence to obtain trend features; dynamic rule matching is performed on the trend features and dynamic partitioning rules to obtain dynamic matching results;

[0137] The static and dynamic matching results are fused together, and matching conflict elimination is performed to obtain the fused matching result.

[0138] Nursing needs matching results are generated based on the fusion matching results.

[0139] In this embodiment of the invention, a nursing demand trend prediction sequence (i.e., an IPI change rate sequence) and a static classification rule (the correspondence between absolute IPI values ​​and nursing levels) are obtained. The absolute IPI value at each time point is compared with the static classification rule to determine the corresponding nursing level. For example, an absolute IPI value of 0.8 matches Level 1 nursing, 0.5 matches Level 2 nursing, and 0.2 matches Level 3 nursing. The matching results at each time point are recorded, forming a sequence of length 720, i.e., the point-by-point matching results. These point-by-point matching results are used as the static matching results, with the numbers 1, 2, and 3 representing Level 1, Level 2, and Level 3 nursing, respectively.

[0140] Obtain the nursing demand trend prediction sequence (i.e., the IPI change rate sequence). Extract trend features using the sliding window method. Set the window size to 5, meaning the trend feature at the current time point is calculated using the IPI change rate over 5 consecutive time points. Calculate the mean and standard deviation of the IPI change rate within the window. Use the mean as the trend strength and the standard deviation as the trend volatility. For example, if the IPI change rate over 5 consecutive time points is [-0.02, -0.03, -0.04, -0.05, -0.06], then the trend strength is -0.04 and the trend volatility is 0.014. Combine the trend strength and volatility at each time point into a two-dimensional vector as the trend feature for that time point. Then, match the trend feature at each time point with dynamic classification rules. For example, if the trend strength is less than -0.05, match level 3 nursing care; if the trend strength is between -0.05 and 0.02, match level 2 nursing care; if the trend strength is greater than 0.02, match level 1 nursing care. The matching results at each time point are recorded to form a sequence of length 720, which is the dynamic matching result. The numbers 1, 2, and 3 are used to represent level 1, level 2, and level 3 nursing care, respectively.

[0141] The static and dynamic matching results are compared. If the two results are the same, the result is adopted directly. If the two results are different, the dynamic matching result is preferred. For example, if the static matching result at a certain time point is Level II nursing care and the dynamic matching result is Level III nursing care, the final result is Level III nursing care. The final matching result at each time point is recorded to form a sequence of length 720, which is the merged matching result. The numbers 1, 2, and 3 represent Level I, Level II, and Level III nursing care, respectively.

[0142] The merged matching results are used as the final care needs matching results. This sequence contains the predicted care level for each time point, serving as input for subsequent steps. For example, the final generated care needs matching result sequence might resemble [1, 1, 2, 2, 3, 3, 2, 1, ...], where each number represents the care level at the corresponding time point.

[0143] Preferably, step S3 includes the following steps:

[0144] Step S31: Set dynamic warning thresholds based on the disease trend prediction vector and preset warning rules to obtain a dynamic threshold matrix;

[0145] Step S32: Perform single-indicator comparison on the disease trend prediction vector and the dynamic threshold matrix to obtain the single-indicator comparison result; perform multi-indicator fusion on the single-indicator comparison result to obtain the multi-indicator fusion result.

[0146] Step S33: Based on the multi-indicator fusion results and the dynamic threshold matrix, determine the early warning trigger and obtain the early warning trigger signal;

[0147] Step S34: Calculate the risk factor weights based on the disease trend prediction vector, the patient's historical physiological data, and the preset early warning classification results to obtain the risk factor weight table;

[0148] Step S35: Calculate the risk score based on the risk factor weight table and the early warning trigger signal to obtain the comprehensive risk score;

[0149] Step S36: Divide the warning levels according to the comprehensive risk score and the preset warning classification results to obtain warning level data.

[0150] As an example of the present invention, reference is made to Figure 3 As shown, step S3 in this example includes:

[0151] Step S31: Set dynamic warning thresholds based on the disease trend prediction vector and preset warning rules to obtain a dynamic threshold matrix;

[0152] In this embodiment of the invention, a Predicted Disease Trend Vector (PTV) is obtained. Preset warning rules include a Probability of Disease Deterioration (POD) threshold, a Level 1 Nursing Care Requirement probability threshold, a Level 2 Nursing Care Requirement probability threshold, and a Level 3 Nursing Care Requirement probability threshold. The base thresholds are set as follows: POD > 0.2, Level 1 Nursing Care Requirement Probability < 0.6, Level 2 Nursing Care Requirement Probability > 0.3, and Level 3 Nursing Care Requirement Probability > 0.1. The base thresholds are dynamically adjusted based on the patient's current Individual Percentage Indicator (IPI) value. For example, if the patient's current IPI value is below 0.5, the POD threshold is lowered to 0.15, and the Level 3 Nursing Care Requirement Probability Threshold is increased to 0.15. The dynamically adjusted thresholds are combined to form a 4×1 Dynamic Threshold Matrix (DTM).

[0153] Step S32: Perform single-indicator comparison on the disease trend prediction vector and the dynamic threshold matrix to obtain the single-indicator comparison result; perform multi-indicator fusion on the single-indicator comparison result to obtain the multi-indicator fusion result.

[0154] In this embodiment of the invention, each indicator in the PTV is compared with the corresponding threshold in the DTM to obtain a single-indicator comparison result. For example, if the POD in the PTV is 0.25 and the POD threshold in the DTM is 0.15, then the single-indicator comparison result of POD is True. A logical AND operation is performed on the single-indicator comparison results of all indicators to obtain a multi-indicator fusion result. For example, if the single-indicator comparison results of all indicators are True, then the multi-indicator fusion result is True; if the single-indicator comparison result of any one indicator is False, then the multi-indicator fusion result is False.

[0155] Step S33: Based on the multi-indicator fusion results and the dynamic threshold matrix, determine the early warning trigger and obtain the early warning trigger signal;

[0156] In this embodiment of the invention, if the result of multi-indicator fusion is true, an early warning is triggered and the early warning trigger signal is 1; otherwise, no early warning is triggered and the early warning trigger signal is 0.

[0157] Step S34: Calculate the risk factor weights based on the disease trend prediction vector, the patient's historical physiological data, and the preset early warning classification results to obtain the risk factor weight table;

[0158] In this embodiment of the invention, historical physiological data of patients, such as IPI data for the past 7 days and past medical history, are obtained from a patient database. Preset warning classification results include low risk, medium risk, and high risk. A logistic regression model is used, with PTV, historical physiological data of patients, and warning classification results as input, to train a risk factor weight calculation model. The model outputs the weight of each risk factor, such as age, past medical history, current IPI value, and POD. These weights are stored in a table to form a risk factor weight table.

[0159] Step S35: Calculate the risk score based on the risk factor weight table and the early warning trigger signal to obtain the comprehensive risk score;

[0160] In this embodiment of the invention, if the warning trigger signal is 1, a comprehensive risk score is calculated based on the risk factor weight table. The value of each risk factor is multiplied by its corresponding weight, and then all results are summed to obtain the comprehensive risk score. For example, if the age weight is 0.2 and the patient's age is 60 years, then the age score is 0.2 * 60 = 12. The comprehensive risk score is obtained by summing all risk factor scores. If the warning trigger signal is 0, the comprehensive risk score is 0.

[0161] Step S36: Divide the warning levels according to the comprehensive risk score and the preset warning classification results to obtain the warning level data;

[0162] In this embodiment of the invention, the comprehensive risk score is divided into warning levels based on a preset warning classification result. For example, a comprehensive risk score below 5 is considered low risk, corresponding to warning level 1; a comprehensive risk score between 5 and 10 is considered medium risk, corresponding to warning level 2; and a comprehensive risk score above 10 is considered high risk, corresponding to warning level 3. The divided warning levels are then output as warning level data.

[0163] Preferably, step S31 includes the following steps:

[0164] Step S311: Obtain the patient's historical physiological data; construct an early warning Bayesian network based on the patient's historical physiological data and the disease trend prediction vector to obtain the early warning Bayesian network;

[0165] Step S312: Use the patient's historical physiological data to set the probability distribution of the early warning Bayesian network to obtain a probabilistic Bayesian network;

[0166] Step S313: Extract the patient's current physiological indicators from the principle physiological data stream to obtain the patient's current physiological indicators;

[0167] Step S314: Input the patient's current physiological indicators and disease trend prediction vector into a probabilistic Bayesian network to calculate the dynamic threshold and obtain the dynamic threshold matrix.

[0168] In this embodiment of the invention, the patient's IPI data for the past 7 days, age, gender, and disease-related diagnostic information, such as the presence of hypertension or diabetes, are obtained from a database. Simultaneously, a Disease Trend Prediction Vector (PTV) is obtained. Using Python's `pgmpy` library, a Bayesian network is constructed based on the patient's historical physiological data and PTV. Network nodes include: age, gender, hypertension, diabetes, average IPI value (past 7 days), IPI trend (from PTV), probability of disease worsening (from PTV), and level of care requirement (from PTV). A hill-climbing algorithm is used to learn the dependencies between network nodes and set the network structure. For example, age may affect the probability of hypertension and diabetes; historical average IPI and IPI trend may affect the probability of disease worsening and level of care requirement. The final generated Bayesian network is stored in `pgmpy`'s BayesianModel object format.

[0169] Using patients' historical physiological data, a conditional probability distribution (CPD) is set for each node in the Bayesian network. For discrete variables, such as gender, hypertension, and diabetes, the CPD is learned from historical data using maximum likelihood estimation. For continuous variables, such as IPI, it is assumed to follow a Gaussian distribution, and the parameters (mean and variance) of the Gaussian distribution are estimated using historical data. The learned CPD is added to the Bayesian network to obtain a probabilistic Bayesian network. For example, based on historical data, a CPD with a higher probability of disease progression is set for hypertensive patients. The final probabilistic Bayesian network is still stored in `pgmpy` BayesianModel object format, but includes the CPD information for each node.

[0170] Obtain the latest raw physiological data stream (RPDS). Extract the patient's current physiological indicators, including the latest IPI value, heart rate, blood pressure, blood oxygen saturation, etc. Store these indicators in a dictionary, for example, `{"IPI": 0.65, "HeartRate": 80, "BloodPressure": "120 / 80", "SpO2": 98}`.

[0171] The patient's current physiological indicators and PTV (Prognostic Threat) are input as evidence into a probabilistic Bayesian network. Using a Bayesian network inference algorithm, such as belief propagation, the probability distributions of the probabilities of disease deterioration, Level 1 care need, Level 2 care need, and Level 3 care need are calculated given the evidence. Dynamic thresholds are determined from the probability distributions based on a preset risk level (e.g., controlling the false alarm rate to within 5%). For example, if the probability distribution of disease deterioration shows that the false alarm rate is below 5% when the probability is greater than 0.18, then 0.18 is used as the dynamic threshold for POD. The four calculated dynamic thresholds are combined into a 4×1 dynamic threshold matrix (DTM).

[0172] Preferably, step S311 specifically includes:

[0173] The patient's historical physiological data and disease trend prediction vector were preprocessed to obtain preprocessed data.

[0174] An initial node set is defined based on the preprocessed data to obtain the initial node set; node similarity is calculated based on the preprocessed data and the initial node set to obtain the node similarity matrix;

[0175] Perform node clustering analysis on the node similarity matrix to obtain the node clustering results; construct a hierarchical structure based on the node clustering results and the initial node set to obtain the node hierarchical structure;

[0176] By using preprocessed data to learn dependencies in the node hierarchy, a node hierarchy with dependencies is obtained.

[0177] A Bayesian network for early warning is constructed based on the hierarchical structure of nodes with dependencies.

[0178] In this embodiment of the invention, the patient's IPI data, age, gender, hypertension, diabetes, and disease trend prediction vector (PTV) for the past 7 days are obtained. Missing values ​​in the IPI data are filled using linear interpolation. Continuous variables, such as age and IPI data, are standardized to convert them to data with a mean of 0 and a standard deviation of 1. Categorical variables, such as gender, hypertension, and diabetes, are encoded using one-hot encoding. The preprocessed data is stored in a PandasDataFrame as preprocessed data.

[0179] An initial node set is defined based on the variables contained in the preprocessed data. The node set includes: age, gender, hypertension, diabetes, mean IPI (past 7 days), IPI trend (from PTV), probability of disease worsening (from PTV), and level of care requirement (from PTV). Each node represents a variable.

[0180] The Pearson correlation coefficient is used to calculate the similarity between variables in the preprocessed data. For continuous variables, the Pearson correlation coefficient is calculated directly. For categorical variables, the one-hot encoded vectors are converted into numerical values ​​before calculating the Pearson correlation coefficient. The calculated correlation coefficients are stored in a matrix to form a node similarity matrix. The rows and columns of the matrix correspond to the nodes in the initial node set.

[0181] Hierarchical clustering algorithm is used to perform cluster analysis on the node similarity matrix. The number of clusters is set to 4. Based on the similarity values ​​in the node similarity matrix, nodes with high similarity are grouped into the same cluster. The clustering results are stored in a list, where each element of the list represents a cluster, and the cluster contains the nodes contained in that cluster.

[0182] Based on the node clustering results, a node hierarchy is constructed. Each cluster is treated as a parent node, and the nodes within that cluster are treated as child nodes. If a cluster contains only one node, that node is directly used as the parent node. For example, if the clustering results are [[Age, Hypertension], [Diabetes], [Average IPI, IPI Trend], [Probability of Disease Worsening, Nursing Level Requirements]], then the constructed hierarchy would be: four parent nodes representing four clusters, with the first two parent nodes each having two child nodes, and the last two parent nodes each having two child nodes.

[0183] Constraint-based structure learning algorithms, such as the PC algorithm, are used to learn the dependencies between nodes in a node hierarchy. Based on preprocessed data, it is determined whether dependencies exist between parent and child nodes, and between child nodes themselves. For example, data can be used to determine whether age affects hypertension. The learned dependencies are then added to the node hierarchy to obtain a node hierarchy with dependencies.

[0184] Based on the hierarchical node structure with dependencies, a Bayesian network is constructed using the `pgmpy` library. The dependencies between parent and child nodes are converted into directed edges in the Bayesian network. The final Bayesian network is stored as a `pgmpy` BayesianModel object.

[0185] Preferably, step S4 includes the following steps:

[0186] Step S41: Retrieve individualized patient information based on the disease trend prediction vector to obtain patient feature profiles;

[0187] Step S42: Obtain the status of nursing resources based on the warning level data to obtain a list of available resources;

[0188] Step S43: Generate an initial nursing care plan based on the warning level data, patient characteristic profile, and available resource list;

[0189] Step S44: Based on the patient's characteristic profile and the list of available resources, the initial nursing plan is adjusted to obtain a recommended nursing plan.

[0190] In this embodiment of the invention, a disease trend prediction vector (PTV) is obtained. Based on the patient ID, individualized patient information is retrieved from the electronic medical record system, including age, gender, allergy history, past medical history, current diagnosis, laboratory test results, imaging test results, and the most recent nursing record. The retrieved information is stored in a JSON-formatted data structure to form a patient feature profile. For example, the patient feature profile contains the following information: `{"Age": 65, "Gender": "Male", "Allergy History": "Penicillin", "Past Medical History": "Hypertension", "Current Diagnosis": "Pneumonia", ...}`.

[0191] Obtain early warning level data. Based on the early warning level, retrieve the status information of currently available nursing resources from the hospital resource management system. For example, retrieve the number of available beds, on-duty medical staff, ventilators, and patient monitors for Level 1, Level 2, and Level 3 nursing care. Store the retrieved resource status information in a JSON-formatted data structure to form an available resource list. For example, the available resource list might contain the following information: `{"Level 1 nursing beds": 10, "Level 2 nursing beds": 20, "Level 3 nursing beds": 5, "On-duty medical staff": 15, "Ventilators": 3, "Patient monitors": 10, ...}`.

[0192] Based on the alert level data, patient characteristic files, and available resource list, a matching initial nursing care plan is retrieved from a predefined nursing care plan template library. The nursing care plan template library contains templates for different diseases, alert levels, and resource configurations. For example, for a pneumonia patient with an alert level of 2 and sufficient level 2 nursing beds, the initial nursing care plan template might include: monitoring vital signs every 4 hours, providing nursing care every 6 hours, and performing a chest X-ray daily. The retrieved nursing care plan template is used as the initial nursing care plan.

[0193] Based on the patient's profile and available resources, healthcare professionals personalize the initial care plan. For example, if the patient is allergic to penicillin, the penicillin-type drug in the initial care plan is replaced with another antibiotic. If current ventilator resources are insufficient, the respiratory support plan is adjusted based on the patient's respiratory status and blood oxygen saturation. Medication dosages and frequencies are adjusted according to the patient's age and medical history. The adjusted care plan is output as the recommended care plan in JSON format, for example: `{"Monitoring Frequency": "Every 4 hours", "Nursing Plan": "Healthcare staff care, once daily", "Medication Plan": "Cefuroxime sodium, twice daily", ...}`.

[0194] Preferably, step S5 includes the following steps:

[0195] Step S51: Record the nursing care execution status according to the recommended nursing care plan to obtain nursing care execution record data;

[0196] Step S52: Track changes in physiological indicators according to the recommended nursing plan to obtain the trajectory of physiological indicator changes;

[0197] Step S53: Conduct a preliminary assessment of the nursing effect based on the nursing execution record data and the trajectory of physiological indicator changes, and obtain preliminary effect assessment data;

[0198] Step S54: Based on the preliminary effect assessment data and recommended nursing plan, dynamically adjust the nursing plan to obtain the adjusted nursing plan;

[0199] Step S55: Generate nursing effect indicators based on the preliminary effect evaluation data and the adjusted nursing plan to obtain nursing effect indicators.

[0200] In this embodiment of the invention, medical staff record the execution of recommended nursing plans through mobile nursing terminals or electronic nursing record systems. After each nursing operation, medical staff record the specific content of the operation, the execution time, the executor, and the patient's response in the system. For example, if medical staff perform "medical staff nursing", they record the operation time, medication name, dosage, changes in the patient's respiratory status, etc. The system automatically stores this information in the database, forming nursing execution record data, and associates it with the patient ID and recommended nursing plan. The data is stored in JSON format, for example: `{"Operation": "medical staff nursing", "Time": "2024-10-27 10:00:00", "Executor": "medical staff Zhang San", "Medication": "salbutamol", "Dosage": "2.5mg", "Patient Response": "Cough reduced", ...}`

[0201] The system continuously collects patients' physiological data, such as IPI, heart rate, blood pressure, blood oxygen saturation, and respiratory rate, from IoT sensors and nursing monitoring devices. Based on the monitoring frequency specified in the recommended nursing protocol, such as once per hour, the system records physiological data before and after nursing intervention. This data is then arranged chronologically to form a trajectory of physiological indicator changes. For example, if the recommended nursing protocol specifies hourly IPI monitoring, the system records the IPI value hourly and correlates these values ​​with the corresponding nursing intervention time to create an IPI change trajectory, which is stored as time-series data.

[0202] The system performs a preliminary assessment of nursing effectiveness based on nursing execution records and physiological indicator changes. For example, it compares changes in IPI values ​​before and after nursing execution to determine whether the nursing effect is positive, negative, or unchanged. An increase in IPI is considered a positive effect; a decrease is considered a negative effect; and little change is considered no significant effect. The assessment results are quantified, for example, using 1 to represent positive, -1 to represent negative, and 0 to represent no significant change. These quantified results are used as preliminary effectiveness assessment data.

[0203] The system dynamically adjusts the recommended nursing care plan based on preliminary effectiveness assessment data. If the preliminary effectiveness assessment data shows a negative nursing effect, the system will adjust the nursing care plan according to pre-set rules or in conjunction with an expert knowledge base. For example, if the patient's IPI value continues to decline, the system may suggest increasing the monitoring frequency, adjusting the medication dosage, or changing the nursing care plan. The adjusted nursing care plan will be used as the new recommended nursing care plan. If the preliminary effectiveness assessment data shows a positive nursing effect or no significant change, the original recommended nursing care plan will remain unchanged.

[0204] The system generates a Care Effectiveness Index (CEI) based on preliminary efficacy assessment data and the adjusted nursing plan. CEI can be a comprehensive indicator, such as a weighted sum of indicators like IPI change, patient comfort score, and nursing cost. It can also be a combination of multiple independent indicators, such as IPI change rate, complication rate, and length of hospital stay. CEI is used to quantify nursing effectiveness and provide data support for subsequent nursing quality assessment and improvement. For example, CEI can be defined as: `CEI = 0.5 * IPI change rate + 0.3 * patient comfort score - 0.2 * nursing cost`.

[0205] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0206] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. An Internet of Things-based digital management method for grading nursing in an intensive care unit, characterized by, The method comprises the following steps: Step S1: Real-time acquisition of heterogeneous data of a patient by a nursing monitoring device to obtain an original physiological data stream; comprehensive physiological indicators are calculated from the original physiological data stream to obtain comprehensive physiological indicators; Step S2: Individualized physiological indicator prediction is performed according to the comprehensive physiological indicators, and patient condition deterioration probability prediction is performed to obtain condition deterioration probability data; nursing level demand prediction is performed according to the individualized physiological indicator prediction result to obtain nursing level demand probability data; condition trend prediction vector generation is performed on the condition deterioration probability data and the nursing level demand probability data to obtain a condition trend prediction vector; Step S2 specifically comprises: Step S21: Time series construction is performed on the comprehensive physiological indicators to obtain a comprehensive physiological indicator time series; Step S22: Individualized physiological model learning is performed on the comprehensive physiological indicator time series to obtain an individualized physiological model; Step S23: Individualized physiological indicator prediction is performed using the individualized physiological model to obtain an individualized physiological indicator prediction result; Step S24: Patient condition deterioration probability prediction is performed according to the individualized physiological indicator prediction result to obtain condition deterioration probability data; Step S25: Nursing level demand prediction is performed according to the individualized physiological indicator prediction result and a preset nursing level division rule to obtain nursing level demand probability data; Step S26: Condition trend prediction vector generation is performed on the condition deterioration probability data and the nursing level demand probability data to obtain a condition trend prediction vector; Step S25 specifically comprises: Step S251: Physiological indicator change trend prediction is performed on the individualized physiological indicator prediction result to obtain a nursing demand trend prediction sequence; Step S252: Static and dynamic nursing level matching is performed according to the nursing demand trend prediction sequence and a preset nursing level division rule to obtain a nursing demand matching result, wherein the preset nursing level division rule comprises a static division rule and a dynamic division rule; Step S253: A nursing demand probability model is constructed according to the nursing demand matching result to obtain a nursing demand probability model; nursing demand probability calculation is performed according to the nursing demand probability model and the nursing demand matching result to obtain a nursing demand probability matrix; Step S254: Nursing resource configuration data is obtained; nursing resource demand optimization is performed on the nursing demand probability matrix according to the nursing resource configuration data to obtain nursing level demand probability data; Step S3: Dynamic early warning threshold setting is performed according to the condition trend prediction vector and a preset early warning rule to obtain a dynamic threshold matrix; early warning signal triggering and early warning level determination are performed according to the condition trend prediction vector and the dynamic threshold matrix to obtain early warning level data; Step S4: Nursing resource recommendation is performed according to the condition trend prediction vector and the early warning level data to obtain a recommended nursing scheme; Step S5: Physiological indicator change tracking is performed on the recommended nursing scheme to obtain a physiological indicator change trajectory; nursing effect evaluation is performed according to the physiological indicator change trajectory to obtain a nursing effect indicator, thereby achieving the task of digital management of graded nursing in a monitoring ward.

2. The intensive care unit grading nursing digital management method based on the Internet of Things according to claim 1, characterized in that, Step S1 comprises the following steps: Step S11: Real-time acquisition of heterogeneous data of a patient by a nursing monitoring device to obtain an original physiological data stream; Step S12: Patient identity association of the original physiological data stream to obtain patient identity association data; physiological data synchronization processing according to the patient identity association data to obtain patient synchronized physiological data; Step S13: Physiological data feature extraction of the patient synchronized physiological data to obtain a physiological feature vector; Step S14: Comprehensive physiological index calculation according to the physiological feature vector to obtain a comprehensive physiological index. 3.The Internet of Things-based intensive care unit hierarchical nursing digital management method according to claim 1, characterized in that, Step S252 specifically comprises: Point-by-point matching of the nursing demand trend prediction sequence and the static division rule to obtain a point-by-point matching result; preliminary labeling according to the point-by-point matching result to obtain a static matching result; Trend feature extraction of the nursing demand trend prediction sequence to obtain a trend feature; Dynamic rule matching of the trend feature and the dynamic division rule to obtain a dynamic matching result; Matching result fusion of the static matching result and the dynamic matching result, and matching conflict elimination processing to obtain a fused matching result; Nursing demand matching result generation according to the fused matching result to obtain a nursing demand matching result. 4.The Internet of Things-based intensive care unit hierarchical nursing digital management method according to claim 1, characterized in that, Step S3 comprises the following steps: Step S31: Dynamic threshold setting according to the disease trend prediction vector and a preset warning rule to obtain a dynamic threshold matrix; Step S32: Single-index comparison of the disease trend prediction vector and the dynamic threshold matrix to obtain a single-index comparison result; multi-index fusion of the single-index comparison result to obtain a multi-index fusion result; Step S33: Warning trigger judgment according to the multi-index fusion result and the dynamic threshold matrix to obtain a warning trigger signal; Step S34: Risk factor weight calculation according to the disease trend prediction vector, the patient historical physiological data, and a preset warning classification result to obtain a risk factor weight table; Step S35: Risk score calculation according to the risk factor weight table and the warning trigger signal to obtain a comprehensive risk score; Step S36: Warning level division according to the comprehensive risk score and a preset warning classification result to obtain warning level data.

5. The intensive care unit grading nursing digital management method based on the Internet of Things according to claim 4, characterized in that, Step S31 comprises the following steps: Step S311: Obtaining patient historical physiological data; warning Bayesian network construction according to the patient historical physiological data and the disease trend prediction vector to obtain a warning Bayesian network; Step S312: Conditional probability distribution setting of the warning Bayesian network by using the patient historical physiological data to obtain a probabilistic Bayesian network; Step S313: Patient current physiological index extraction of the original physiological data stream to obtain a patient current physiological index; Step S314: Inputting the patient current physiological index and the disease trend prediction vector into the probabilistic Bayesian network for dynamic threshold calculation to obtain a dynamic threshold matrix. 6.The intensive care unit grading nursing digital management method based on Internet of Things according to claim 5, characterized in that, Step S311 specifically comprises: Data preprocessing of the patient historical physiological data and the disease trend prediction vector to obtain preprocessed data; Initial node set definition according to the preprocessed data to obtain an initial node set; According to the pretreatment data and the initial node set, node similarity calculation is performed to obtain a node similarity matrix; Node clustering analysis is performed on the node similarity matrix to obtain a node clustering result; and according to the node clustering result and the initial node set, a hierarchical structure is constructed to obtain a node hierarchical structure; The node hierarchical structure is learned by using the pretreatment data to obtain a node hierarchical structure with dependency; According to the node hierarchical structure with dependency, a Bayesian network is constructed to obtain a warning Bayesian network. 7.The intensive care unit grading nursing digital management method based on Internet of Things according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: According to the disease trend prediction vector, patient individual information retrieval is performed to obtain a patient characteristic file; Step S42: According to the warning level data, nursing resource state acquisition is performed to obtain an available resource list; Step S43: According to the warning level data, the patient characteristic file and the available resource list, an initial nursing scheme is generated to obtain an initial nursing scheme; Step S44: According to the patient characteristic file and the available resource list, the initial nursing scheme is adjusted to obtain a recommended nursing scheme. 8.The intensive care unit nursing level digital management method based on Internet of Things according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: According to the recommended nursing scheme, nursing execution record is recorded to obtain nursing execution record data; Step S52: According to the recommended nursing scheme, physiological index change tracking is performed to obtain a physiological index change trajectory; Step S53: According to the nursing execution record data and the physiological index change trajectory, nursing effect preliminary evaluation is performed to obtain preliminary effect evaluation data; Step S54: According to the preliminary effect evaluation data and the recommended nursing scheme, nursing scheme dynamic adjustment is performed to obtain an adjusted nursing scheme; Step S55: According to the preliminary effect evaluation data and the adjusted nursing scheme, nursing effect index generation is performed to obtain a nursing effect index.

Images