Risk identification method for pediatric intensive care unit nursing

Abstract

The invention discloses a risk identification method for pediatric intensive care unit nursing, which relates to the technical field of medical early warning, and comprises the following steps: collecting multi-modal data, and carrying out time synchronization on the collected data; data features are extracted based on the data after time synchronization, fusion processing is carried out through linear weighting, an optimization target is determined, effective fusion of multi-modal data is achieved, and fused high-quality data is obtained; constructing a risk assessment model, calculating a risk value for each time point in combination with historical data, obtaining an overall risk value through weighted average processing, judging a risk level, and optimizing prediction accuracy through a self-correction algorithm; the state of a patient is monitored in real time, the early warning level is set, abnormal conditions are early warned according to the corrected risk value variation, feedback of medical staff is collected, and the system performance is continuously optimized. The method has remarkable clinical application value, and can effectively improve early warning accuracy, optimize medical resource allocation, reduce medical risks and improve overall medical quality.

Description

Technical Field

[0001] The present invention relates to the technical field of medical early warning, and in particular to a risk identification method for pediatric intensive care unit nursing. Background Art

[0002] Currently, clinical assessments of patient conditions primarily rely on static indicators such as vital signs and laboratory parameters. While these can reflect a patient's current condition, they are unable to fully assess the progression of the disease. Traditional early warning scoring systems, primarily based on the assessment of indicators at a single point in time, are subject to lags and fail to fully consider the crucial factor of how quickly the patient's condition changes. Furthermore, existing early warning systems often employ coarse grading methods, with inconsistent response standards, making it difficult to meet the demands of refined management. Furthermore, clinical practice faces challenges such as high false alarm rates and inefficient allocation of medical resources. This is particularly true for critically ill patients, whose conditions often evolve rapidly and in complex ways. Existing early warning systems struggle to detect potential risks in a timely manner, leading to delays in treatment.

[0003] In actual applications, medical staff often need to rely on personal experience to make judgments and lack unified quantitative standards. This not only increases the workload but also may affect the accuracy of medical decisions. At the same time, the rational allocation of medical resources is also facing challenges, with over-monitoring and under-monitoring coexisting. Therefore, it is of great significance to develop a new system that can dynamically assess patient risks and achieve accurate early warning. Existing technologies urgently need an early warning method that combines static risk assessment with dynamic change trends to improve early warning accuracy, optimize medical resource allocation, and enhance patient monitoring quality. Summary of the Invention

[0004] In view of the problems existing in the existing risk identification methods for pediatric intensive care unit nursing, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that the disease changes rapidly and is complex. The existing early warning system is difficult to detect potential risks in a timely manner, and lacks a unified quantitative standard, which affects the accuracy of medical decision-making.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, the present invention provides a risk identification method for pediatric intensive care unit nursing, which includes: collecting multimodal data, including continuous physiological indicator data, intermittent physiological indicator data and environmental data, and time-synchronizing the collected data; extracting data features based on the time-synchronized data, performing multimodal data feature fusion processing through linear weighting, and based on a preset optimization goal, performing data fusion processing on the multimodal data and medical imaging data to obtain high-quality fused data; constructing a risk assessment model based on the fused data, combining historical data, calculating the risk value for each time point, obtaining the overall risk value through weighted average processing, determining the risk level, and optimizing the prediction accuracy through a self-correction algorithm; monitoring the patient status in real time, setting the warning level, warning of abnormal conditions based on the corrected risk value change, collecting feedback from medical staff, and continuously optimizing system performance.

[0007] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing described in the present invention, the continuous physiological indicator data includes heart rate indicator data, blood pressure indicator data, body temperature indicator data, blood oxygen saturation indicator data and respiratory rate indicator data; the intermittent physiological indicator data includes blood glucose level indicator data, electrolyte level indicator data and coagulation function indicator data; the environmental data includes indoor temperature, relative humidity, environmental noise level, light intensity, CO2 concentration, oxygen concentration and particulate matter concentration.

[0008] As a preferred solution of the risk identification method for pediatric intensive care unit nursing described in the present invention, the time synchronization includes: for continuous physiological indicator data, a hardware timestamp is used, and each data packet carries precise time information; for intermittent physiological indicator data, the timestamp of the actual sampling moment is recorded; for environmental data, the timestamp is loaded regularly according to the preset sampling cycle; the timestamp of the continuous physiological indicator data is used as the main reference, the intermittent physiological indicator data is associated through the timestamp, and the environmental data is matched according to the principle of the latest timestamp; the time alignment rules are determined, including: the continuous physiological indicator data is aligned according to the highest sampling frequency, the intermittent physiological indicator data maintains the original timestamp, the latest continuous physiological indicator data is associated, and the environmental data is interpolated and aligned to the time point of the continuous physiological indicator data.

[0009] As a preferred solution of the risk identification method for pediatric intensive care unit nursing described in the present invention, the data fusion processing is a fusion processing of multimodal data and medical imaging data, including: extracting data features of the multimodal data, including time domain features, frequency domain features, statistical features and morphological features; extracting data features of the medical imaging data including texture features, shape features, intensity features and spatial features; fusing the multimodal data and medical imaging data in a linear weighted manner, wherein the feature weight coefficient is determined according to the information entropy of each feature.

[0010] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing of the present invention, wherein: the optimization goal is to minimize information redundancy and maximize feature complementarity; The minimization of information redundancy is achieved by calculating the mutual information between features, and the mutual information is calculated based on the joint probability distribution and marginal probability distribution of the features.

[0011] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing of the present invention, in which: in order to maximize feature complementarity, it is achieved by calculating the total amount of complementary information, which is determined by the information amount of each feature and the overlapping information amount between the features; The information content of the feature is calculated using information entropy, and the overlapping information content between features is calculated using joint entropy. Both the information entropy and the joint entropy are determined based on the probability distribution of the feature.

[0012] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing of the present invention, wherein: the risk assessment model calculates a risk score based on data within an observation time window, and the score is comprehensively calculated in a time-weighted manner, wherein a time decay coefficient is taken into account; During the calculation, the weight coefficient, time characteristics, deviation between the observed value and the reference value, time smoothing parameter and nonlinear mapping coefficient of each feature are considered separately, and the overall risk assessment value at a certain moment is obtained by weighted summation.

[0013] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing of the present invention, the optimal observation time window length is determined based on the risk assessment model and combined with historical data, and a fixed time window length is set. At the same time, a time decay coefficient is set, the reference mean and standard deviation of each feature are calculated, and the initial weight of the feature is set; At each evaluation, the data within a specific time window is used to convert the continuous integral into a discrete sum, obtain the latest feature observation value, and update the most recent measurement time of the feature; The risk function is calculated for each time point, and the risk score is obtained by time-weighted average. The risk level is determined based on the risk score to evaluate the patient's real-time status.

[0014] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing of the present invention, wherein: a self-correction algorithm is constructed based on the current risk assessment model, the assessment results are optimized, and the risk score is corrected by calculating a risk correction amount; The risk correction amount is calculated based on a time-varying learning rate function that takes into account the difference between the actual observed value and the predicted value and the correction attenuation coefficient; The time-varying learning rate function is determined by a basic learning rate, a learning rate adjustment coefficient, and a relative prediction error, wherein the relative prediction error is calculated based on the deviation between the actual observed value and the predicted value.

[0015] As a preferred embodiment of the risk identification method for pediatric intensive care unit nursing described in the present invention, the parameters are updated, including feature weights, time smoothing parameters and nonlinear mapping coefficients, and the marginal probability distribution is calculated based on the values of the features in the samples; the joint probability distribution is calculated based on the common values of the feature pairs in the samples; and the information content of the features is calculated based on the probability distribution of the features.

[0016] The present invention has the following beneficial effects: it combines static risk assessment with dynamic change trends to establish a refined hierarchical response mechanism. Patient risk is divided into three main levels based on risk values, and the warning levels are further subdivided according to the degree of change rate, achieving accurate assessment of patient status and timely warning. The evaluation indicators of the present invention are easy to obtain, the judgment criteria are clear, and the response measures are specific, making them easy for medical staff to master and implement. It has significant clinical application value and can effectively improve the accuracy of warnings, optimize the allocation of medical resources, reduce medical risks, and improve overall medical quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them: Figure 1 Flowchart of a risk identification approach for pediatric intensive care unit care. DETAILED DESCRIPTION

[0018] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0019] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0021] Example 1: Reference Figure 1 , which is a first embodiment of the present invention, provides a risk identification method for pediatric intensive care unit nursing, comprising: S1: Collect multimodal data, perform fusion processing, extract data features, and analyze potential risks.

[0022] Collect multimodal data, including continuous physiological indicator data, intermittent physiological indicator data and environmental data.

[0023] Continuous physiological index data includes heart rate index data, blood pressure index data, body temperature index data, blood oxygen saturation index data and respiratory rate index data; The sampling frequency of heart rate index data is 100Hz, and the measurement range is 30-300 beats / minute. The patient's ECG signal is collected through a multi-lead ECG monitor, and the heart rate value is calculated in real time. The sampling frequency of blood pressure index data is 50 Hz. The systolic and diastolic blood pressure of the children are collected through a non-invasive blood pressure monitoring device to determine the measurement range and measurement accuracy; Body temperature index data include axillary temperature, rectal temperature and skin surface temperature; The sampling frequency of blood oxygen saturation index data is 60Hz, and the measurement range is 70%~100%. Blood oxygen saturation is calculated through pulse waveform collection and perfusion index monitoring is used; The sampling frequency of respiratory rate index data is 50 Hz, the measurement range is 0-120 times / min, and it is obtained through respiratory waveform analysis.

[0024] Intermittent physiological index data include blood sugar level index data, electrolyte level index data and coagulation function index data; Blood glucose level index data is obtained by measuring fasting blood glucose values and postprandial blood glucose values, constructing a dynamic blood glucose monitoring curve, and analyzing blood glucose levels. The collection time is fixed, and in case of emergencies, the time is dynamically adjusted.

[0025] The electrolyte level indicator data include potassium ion concentration level, sodium ion concentration level, chloride ion concentration level and calcium ion concentration level, with fixed sampling time intervals.

[0026] Coagulation function index data included prothrombin time, activated partial thromboplastin time, international normalized ratio, and D-dimer level, with fixed sampling intervals.

[0027] Environmental data includes indoor temperature, relative humidity, ambient noise level, light intensity, CO2 concentration, oxygen concentration, and particulate matter concentration.

[0028] Time-synchronize the collected multimodal data to ensure that the collection timestamps of various data are synchronized. Specifically: When the system starts, all acquisition devices are synchronized with the reference clock for the first time, with the unified time format being a 64-bit UNIX timestamp, and the initial time deviation of each device is recorded; For continuous physiological indicator data, hardware timestamps are used, and each data packet carries precise time information; For intermittent physiological indicator data, record the timestamp of the actual sampling time; For environmental data, timestamps are loaded regularly according to the preset sampling period.

[0029] The timestamp of continuous physiological indicator data is used as the main reference, intermittent physiological indicator data are associated through timestamps, and environmental data are matched according to the principle of the latest timestamp.

[0030] Automatically calculate and compensate for network transmission delays for time-synchronized data, taking into account internal device processing delays and dynamically adjusting compensation parameters.

[0031] Determine the time alignment rules, including: aligning continuous physiological indicator data at the highest sampling frequency, maintaining the original timestamp of intermittent physiological indicator data, associating the most recent continuous physiological indicator data, and interpolating and aligning environmental data to the time point of continuous physiological indicator data.

[0032] Collect medical imaging data, including bedside X-ray data, ultrasound image data, and CT / MRI image data.

[0033] Based on the collected multimodal data and medical imaging data, key data features are extracted and data fusion is performed.

[0034] Specifically, multimodal data features include: time domain features, frequency domain features, statistical features, and morphological features; Time domain features include mean, variance, peak value and zero-crossing rate; frequency domain features include power spectrum density and main frequency component; statistical features include entropy value and correlation coefficient; morphological features include waveform feature points and trend characteristics; Extract data features of medical imaging data, including texture features, shape features, intensity features, and spatial features; Texture features include gray-level co-occurrence matrix and local binary pattern; shape features include area, perimeter and roundness; intensity features include gray-level statistical features; spatial features include edge features, corner features and local description features.

[0035] Principal component analysis is performed on the extracted data features to achieve data dimensionality reduction, and the data features are standardized to construct feature vectors; then, the weights of each feature are dynamically adjusted according to the data quality, and the weight parameters are continuously updated in combination with time series information to achieve adaptive optimization of the weights.

[0036] Multimodal data and medical imaging data are fused by linear weighting, which is expressed as: ; in, is the fused feature vector, For the The weight coefficient of the feature, For the feature vectors, .

[0037] Determine the weight coefficient based on the information entropy of the feature: ; in, For the The information entropy of a feature.

[0038] Determine the optimization goal of the fusion process to minimize information redundancy and maximize feature complementarity; Minimizing information redundancy is expressed as: ; in, Representation characteristics X and features Y The mutual information between Representation characteristics X and features Y The joint probability distribution of Representation characteristics X The marginal probability distribution of Representation characteristicsY The marginal probability distribution of .

[0039] In order to maximize feature complementarity, we first determine the total amount of complementary information, which is expressed as: ; in, is the total amount of complementary information, For the The amount of information of a feature, Features and features The amount of overlapping information between .

[0040] It can be expressed as: ; in, Features The probability distribution of Features The value of .

[0041] It can be expressed as joint entropy: ; in, Features The information entropy of Features The information entropy of is the joint entropy.

[0042] Further, and Respectively expressed as: ; ; in, Features The probability distribution of Features The value of Features and The probability of a common distribution.

[0043] It should be noted that when minimizing information redundancy, the probability distribution of discrete cases is considered, and when maximizing feature complementarity, the probability density of continuous cases is considered.

[0044] Based on the data fusion of spatiotemporal alignment and the selection of optimization targets, the effective fusion of multimodal medical data is achieved to obtain high-quality fused data.

[0045] S2: Build a personalized risk assessment model, combine historical health data, evaluate patient status in real time, and optimize prediction accuracy through self-correction algorithms.

[0046] A risk assessment model is constructed based on the fused high-quality data, which is expressed as: ; in, For the final risk score, T is the length of the observation time window, For time The risk function at time is the time decay coefficient.

[0047] Further, Expressed as: ; in, For the The weight coefficient of the feature, is the total number of features involved in risk assessment, Features The last updated time of is the time smoothing parameter, is the nonlinear mapping coefficient, Features In time The observed value of Features The reference mean, Features The standard deviation of .

[0048] Through all The features are summed, where From 1 to Each feature traversed. Each feature has its own weight , time characteristics , reference mean and standard deviation , the influence of all features is combined through the summation formula to obtain Overall risk assessment value .

[0049] Based on the risk assessment model, combined with historical data, the optimal observation time window length is determined, and then the time window length is set. T Fixed length, set time decay coefficient , for each feature Calculate reference mean and standard deviation , set the initial weight of the feature .

[0050] Each evaluation uses The data of the interval is converted into a discrete sum to obtain the latest Value, Update The time of the most recent measurement.

[0051] Calculate for each time point , calculate the time-weighted average to get Value, according to The value determines the risk level and evaluates the patient's real-time status.

[0052] A self-correction algorithm is constructed based on the current risk assessment model to optimize the assessment results. The correction model is expressed as: ; in, represents the risk correction amount, represents the adjusted risk score.

[0053] Further, Expressed as: ; in, is the time-varying learning rate function, is the actual observed risk value, is the risk value predicted by the model, is the correction attenuation coefficient.

[0054] Further, Expressed as: ; in, is the basic learning rate, is the learning rate adjustment coefficient, is the relative prediction error.

[0055] Further, Expressed as: ; in, is the actual observed risk value, is the risk value predicted by the model.

[0056] Furthermore, the parameters are updated, including feature weights, time smoothing parameters, and nonlinear mapping coefficients, which are expressed as: ; ; ; in, is the parameter update step size, is the updated feature weight, is the updated time smoothing parameter, is the updated nonlinear mapping coefficient.

[0057] Calculate the deviation between the predicted value and the actual observed value through the relative error Evaluate prediction accuracy and identify methodological biases and random errors; adaptively adjust the learning rate based on the size of the prediction error, reducing the learning rate to avoid overcorrection when the error is large, and maintaining moderate learning for continuous optimization when the error is small.

[0058] Taking into account the historical correction effects, the correction amount is time-weighted to ensure the smoothness and continuity of the correction. The correction amount is applied to the original score, and the final revised risk score is output to maintain the continuity and interpretability of the score.

[0059] S3: Monitor patient status in real time, provide intelligent warnings of abnormal situations, collect feedback from medical staff, and continuously optimize system performance.

[0060] Calculated at each time point , calculate the time-weighted average to get Value, right The value is corrected, and the The range of values sets a first threshold, a second threshold, and a third threshold, and the first threshold < the second threshold < the third threshold.

[0061] like ≤ the first threshold, indicating that the patient's indicators are within the normal range, vital signs are stable, and there are no obvious abnormal fluctuations. The adjusted risk score shows that the overall condition is good, and an 8-hour monitoring interval is adopted, which is in line with routine nursing standards; If the first threshold < If the value is less than or equal to the second threshold, it means that the patient's indicators have slightly deviated but have not reached the clinical intervention standard. The monitoring frequency needs to be increased to prevent the condition from worsening. A 4-hour monitoring interval can detect potential problems in a timely manner. Further testing ,like , it means that the risk is on the rise and the condition is getting worse, so you need to keep an eye on it. is the change in risk value after adjustment, Indicates a time interval.

[0062] like , it indicates that the risk is slowly decreasing and the condition is gradually improving, and the monitoring frequency needs to be maintained.

[0063] If the first threshold < ≤ the second threshold, and 0< If the value is ≤0.05, it indicates that the patient's condition is slightly abnormal. Increase the monitoring frequency to 4 hours / time, record the detailed changes in symptoms, and observe the trend of changes; If the first threshold < ≤ the second threshold, and 0.05< If the value is ≤0.1, it indicates that the patient's condition is slightly abnormal but the changes are accelerating. Increase the monitoring frequency to 2 hours / time, record the changes in various indicators in detail, and prepare possible intervention plans.

[0064] If the first threshold < ≤ the second threshold, and 0.1< , it indicates that the patient's condition changes rapidly. Increase the monitoring frequency to 1 hour / time, record the changes in vital signs in real time, and formulate an intervention plan.

[0065] If the second threshold < ≤ the third threshold, it indicates that multiple indicators are obviously abnormal and the monitoring frequency needs to be increased to 1 hour / time to ensure timely detection of changes.

[0066] If the second threshold < ≤ the third threshold, and 0< When the value is ≤0.1, it indicates that the patient's condition is obviously abnormal. Increase the monitoring frequency to 1 hour / time and record the changes in symptoms in detail.

[0067] If the second threshold < ≤ the third threshold, and 0.1< When the value is ≤0.15, it indicates that the patient's condition is rapidly deteriorating. The monitoring frequency should be increased to 30 minutes / time, and vital signs should be continuously recorded.

[0068] If the second threshold < ≤ the third threshold, and 0.15< When the patient's condition deteriorates rapidly, continuous bedside monitoring and emergency treatment measures should be initiated.

[0069] If the third threshold < , it indicates that the patient's condition is critical and continuous monitoring is required, and immediate medical intervention is required to ensure patient safety.

[0070] If the third threshold < , and 0< When the value is ≤0.05, it indicates that the patient is in critical condition. Continuous monitoring should be initiated immediately, vital signs should be recorded every 15 minutes, and intensive care plans should be implemented.

[0071] If the third threshold < , and 0.05< When the value is ≤0.1, it indicates that the patient's condition is rapidly deteriorating and a critical value response must be initiated immediately. Vital signs must be recorded every 5 minutes and emergency resources must be deployed immediately.

[0072] If the third threshold < , and 0.1< When the patient is in imminent danger of life, the hospital will initiate emergency rescue, continuously monitor vital signs, immediately carry out rescue measures, start special treatment plans, and mobilize emergency resources throughout the hospital.

[0073] In summary, the present invention uses the corrected risk value and its rate of change The patient early warning system combines static risk assessment with dynamic change trends to build a refined hierarchical response mechanism. The patient risk is divided into three main levels according to The system can be further subdivided into different warning levels based on the severity of the disease, achieving accurate assessment of the patient's condition and timely warning. The evaluation indicators of the present invention are easy to obtain, the judgment criteria are clear, and the response measures are specific, making it easy for medical staff to master and implement them. At the same time, the system can be promoted and used in medical institutions at all levels, with significant clinical application value, effectively improving the accuracy of warnings, optimizing the allocation of medical resources, reducing medical risks, and improving overall medical quality.

[0074] This embodiment also provides a computer device, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the risk identification method for pediatric intensive care unit care proposed in the above embodiment.

[0075] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0076] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the risk identification method for pediatric intensive care unit nursing proposed in the above embodiment is implemented.

[0077] Example 2: This example provides a risk identification method for pediatric intensive care unit nursing. In order to verify the beneficial effects of the present invention, a scientific demonstration is conducted through simulation experiments.

[0078] To fully verify the early warning effect of the present invention, data from 1,000 patients in the pediatric intensive care unit of a tertiary hospital from January to December 2023 were selected for comparative analysis. The patients were randomly divided into an experimental group (using the present method) and a control group (using the traditional MEWS scoring system), with 500 patients in each group.

[0079] The baseline characteristics of the subjects, including age, gender, underlying diseases, etc., were statistically tested and the distribution of baseline characteristics of the two groups of experiments is shown in Table 1: Table 1 Distribution of baseline characteristics feature Experimental group (n=500) Control group (n=500) P-value Age (years) 4.6±3.2 4.8±3.4 0.856 Gender (male / female) 272 / 228 268 / 232 0.912 Weight (kg) 16.4±8.2 16.8±8.4 0.842 Underlying disease (cases) 142 138 0.826 PRISM III score 8.6±4.2 8.8±4.4 0.892 When P>0.05, the difference between the two groups was not statistically significant. Table 1 was used to determine that the two groups of patients were similar, excluding the influence of the differences among the patients themselves on the research results.

[0080] Excluding the influence of patients, comparative experiments were conducted using the present invention and the prior art, and the experimental results are shown in Table 2: Table 2 Comparison of early warning effects Evaluation Metrics Experimental group (n=500) Control group (n=500) Early warning accuracy (%) 93.8 82.4 Missing reporting rate (%) 2.8 10.6 False alarm rate (%) 3.4 7.0 Average warning lead time (minutes) 52.6 32.4 Incidence of clinical worsening events (%) 4.8 8.6 PICU admission rate (%) 7.4 12.8 The early warning effectiveness comparison table comprehensively demonstrates the superiority of this invention over the traditional PEWS scoring system through six key indicators. First, in terms of early warning accuracy, the experimental group achieved an accuracy rate of 93.8%, significantly higher than the control group's 82.4%, an increase of 11.4 percentage points. Furthermore, the false alarm rate was significantly reduced to 2.8% (compared to 10.6% in the control group, a decrease of 7.8 percentage points), and the false alarm rate was also significantly reduced to 3.4% (compared to 7.0% in the control group, a decrease of 3.6 percentage points). This demonstrates that this invention offers greater accuracy and reliability in pediatric early warning.

[0081] In terms of early warning effectiveness, this invention can detect potential risks an average of 52.6 minutes in advance, 20.2 minutes earlier than traditional systems, providing a sufficient window for timely intervention by clinical medical staff. This precise and timely warning is directly reflected in clinical outcome indicators: the incidence of clinical deterioration events in the experimental group decreased to 4.8% (8.6% in the control group, a decrease of 3.8 percentage points), and the rate requiring PICU admission decreased to 7.4% (12.8% in the control group, a decrease of 5.4 percentage points).

[0082] Particularly noteworthy is the significant success of this invention in reducing the false negative rate. This significant decrease (from 10.6% to 2.8%) means that more children with potential risks can be identified promptly, which is crucial for preventing serious complications and improving prognosis. Furthermore, the reduction in the false positive rate (from 7.0% to 3.4%) significantly reduces the workload of medical staff and improves work efficiency.

[0083] In summary, the present invention demonstrates significant advantages in multiple dimensions, including early warning accuracy, timeliness, clinical effectiveness, and user experience, fully demonstrating its application value. These improvements not only enhance the quality of pediatric critical illness early warning but also provide strong support for improving pediatric medical safety.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A risk identification method for pediatric intensive care unit nursing, characterized by: include: Collect multimodal data, including continuous physiological indicator data, intermittent physiological indicator data and environmental data, and synchronize the collected data in time; Extract data features based on time-synchronized data and perform multimodal data feature fusion processing through linear weighting. Based on the preset optimization goals, perform data fusion processing on multimodal data and medical imaging data to obtain high-quality fused data. A risk assessment model is constructed based on the fused data. Combined with historical data, the risk value is calculated for each time point. After weighted average processing, the overall risk value is obtained and the risk level is determined. The prediction accuracy is optimized through a self-correction algorithm. Monitor patient status in real time, set warning levels, warn of abnormal situations based on changes in corrected risk values, collect feedback from medical staff, and continuously optimize system performance.

2. The risk identification method for pediatric intensive care unit nursing according to claim 1, characterized in that: The continuous physiological index data includes heart rate index data, blood pressure index data, body temperature index data, blood oxygen saturation index data and respiratory rate index data; The intermittent physiological index data include blood sugar level index data, electrolyte level index data and coagulation function index data; The environmental data includes indoor temperature, relative humidity, ambient noise level, light intensity, CO2 concentration, oxygen concentration, and particulate matter concentration.

3. The risk identification method for pediatric intensive care unit nursing according to claim 2, characterized in that: The time synchronization includes: For continuous physiological indicator data, hardware timestamps are used, and each data packet carries precise time information; For intermittent physiological indicator data, record the timestamp of the actual sampling time; For environmental data, timestamps are loaded regularly according to the preset sampling period; The timestamp of continuous physiological indicator data is used as the main reference, intermittent physiological indicator data are associated through timestamps, and environmental data are matched according to the principle of the latest timestamp; Determine the time alignment rules, including: aligning continuous physiological indicator data according to the highest sampling frequency, maintaining the original timestamp of intermittent physiological indicator data, associating the most recent continuous physiological indicator data, and interpolating and aligning environmental data to the time point of continuous physiological indicator data.

4. The risk identification method for pediatric intensive care unit nursing according to claim 3, characterized in that: The data fusion processing is a fusion processing of multimodal data and medical imaging data, including: Extracting data features of the multimodal data, including time domain features, frequency domain features, statistical features, and morphological features; Extracting data features of the medical image data including texture features, shape features, intensity features, and spatial features; Multimodal data and medical imaging data are fused through a linear weighted approach, where the feature weight coefficient is determined according to the information entropy of each feature.

5. The risk identification method for pediatric intensive care unit nursing according to claim 4, characterized in that: The optimization goal is to minimize information redundancy and maximize feature complementarity; The minimization of information redundancy is achieved by calculating the mutual information between features, and the mutual information is calculated based on the joint probability distribution and marginal probability distribution of the features.

6. The risk identification method for pediatric intensive care unit nursing according to claim 5, characterized in that: In order to maximize the complementarity of features, the total amount of complementary information is calculated, which is determined by the information content of each feature and the overlapping information content between features; The information content of the feature is calculated using information entropy, and the overlapping information content between features is calculated using joint entropy. Both the information entropy and the joint entropy are determined based on the probability distribution of the feature.

7. The risk identification method for pediatric intensive care unit nursing according to claim 6, characterized in that: The risk assessment model calculates a risk score based on the data within the observation time window, and the score is calculated comprehensively in a time-weighted manner, taking into account the time decay coefficient; During the calculation, the weight coefficient, time characteristics, deviation between the observed value and the reference value, time smoothing parameter and nonlinear mapping coefficient of each feature are considered separately, and the overall risk assessment value at a certain moment is obtained by weighted summation.

8. The risk identification method for pediatric intensive care unit nursing according to claim 7, characterized in that: Based on the risk assessment model, combined with historical data, the optimal observation time window length is determined, and a fixed time window length is set. At the same time, a time decay coefficient is set, the reference mean and standard deviation of each feature are calculated, and the initial weight of the feature is set; At each evaluation, the data within a specific time window is used to convert the continuous integral into a discrete sum, obtain the latest feature observation value, and update the most recent measurement time of the feature; The risk function is calculated for each time point, and the risk score is obtained by time-weighted average. The risk level is determined based on the risk score to evaluate the patient's real-time status.

9. The risk identification method for pediatric intensive care unit nursing according to claim 8, characterized in that: Build a self-correction algorithm based on the current risk assessment model, optimize the assessment results, and correct the risk score by calculating the risk correction amount; The risk correction amount is calculated based on a time-varying learning rate function that takes into account the difference between the actual observed value and the predicted value and the correction attenuation coefficient; The time-varying learning rate function is determined by a basic learning rate, a learning rate adjustment coefficient, and a relative prediction error, wherein the relative prediction error is calculated based on the deviation between the actual observed value and the predicted value.

10. The risk identification method for pediatric intensive care unit nursing according to claim 9, characterized in that: Update the parameters, including feature weights, time smoothing parameters, and nonlinear mapping coefficients, to calculate the marginal probability distribution based on the values of the features in the samples; Calculate the joint probability distribution based on the common values of feature pairs in the sample; The information content of a feature is calculated based on its probability distribution.

Images